reviews - nano medicine and nano biotechnology, vol.1, issue 1 (2009)

Wiley Interdisciplinary Reviews: Nanomedicine and Nanobiotechnology

Copyright © 2009 John Wiley & Sons, Inc.

TABLE OF CONTENTS

Volume 1 Issue 1 , Pages 1 - 148 (January/February 2009)

Editorial CommentaryA Hybrid Model for a Hybrid Science (p 1)Jim BakerPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.32

OpinionsThe public acceptance of nanomedicine: a personal perspective (p 2-5)David M. BerubePublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.30

Nanotechnology and orthopedics: a personal perspective (p 6-10)Cato T. Laurencin, Sangamesh G. Kumbar, Syam Prasad NukavarapuPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.25

OverviewProspects and developments in cell and embryo laser nanosurgery (p 11-25)Vikram Kohli, Abdulhakem Y. ElezzabiPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.29

Advanced ReviewsPharmacokinetics of nanomaterials: an overview of carbon nanotubes, fullerenes and quantum dots (p 26-34)Jim E. RivierePublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.24

Catalyst-functionalized nanomaterials (p 35-46)Yi Lu, Juewen LiuPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.21

Nanoparticle-based biologic mimetics (p 47-59)David E. Cliffel, Brian N. Turner, Brian J. HuffmanPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.20

Informatics approaches for identifying biologic relationships in time-series data (p 60-68)Brett A. McKinneyPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.12

Radioactive liposomes (p 69-83)William Thomas Phillips, Beth Ann Goins, Ande BaoPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.3

Magnetic resonance susceptibility based perfusion imaging of tumors using iron oxide nanoparticles (p 84-97)Arvind P. PathakPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.17

Optical nanoparticle sensors for quantitative intracellular imaging (p 98-110)Yong-Eun Koo Lee, Raoul KopelmanPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.2

Synthesis of poly(alkyl cyanoacrylate)-based colloidal nanomedicines (p 111-127)Julien Nicolas, Patrick CouvreurPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.15

Hydrogel mediated delivery of trophic factors for neural repair (p 128-139)Joshua S. Katz, Jason A. BurdickPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.10

Focus ArticleMesoporous silica-based nanomaterials for drug delivery: evaluation of structural properties associated with release rate (p 140-148)Maria Strømme, Ulrika Brohede, Rambabu Atluri, Alfonso E. Garcia-BennettPublished Online: Nov 20 2008 1:56PM DOI: 10.1002/wnan.13

Editorial Commentary

A Hybrid Model for a HybridScience

Those of you who know me personally will besurprised to see me on this masthead, as I have

said that I would not edit a ‘nanomedicine journal.’Despite initial appearances, I hope to convince youthat this publication is something different: a livingdocument, a hybrid that recognizes the diversity of theresearch that contributes to our field.

I have several concerns about the usefulnessof current journal formats for nanomedicine andnanobiotechnology. First, I feel that, because thisis such a cross-disciplinary field, traditional journalsoften are not able to serve their readers effectively. Bynecessity, most journals tend to be put together froma single perspective, be it medical, nanomaterials, ornanodevices, that does not address the needs of allinterested readers. Second, I am concerned that thestandard journal format does not offer an appropriateentree to students and professionals attempting toenter the field, again from a variety of disciplines. Thisis because most journals publish research or reviewpapers that are highly technical in nature, and sucharticles do not provide the type of perspective thata student, or a professional in an adjacent area ofresearch, needs to help them find their place in a field.In contrast, a textbook for this rapidly evolving fieldwould be extremely difficult to keep current.

So, if not a standard journal, what kindof publication is Wiley Interdisciplinary Reviews:Nanomedicine and Nanobiotechnology? This will bea unique endeavor that will combine the best featuresof online reference works (it is a comprehensive,authoritative, frequently updated resource comprisinginvited contributions from leading researchers) andreview journals (it offers high online visibility as wellas journal-type abstracting and indexing that willconfer appropriate professional credit on authors).This will be a truly living document that will exist pri-marily in electronic format but can also be producedin print versions periodically to provide access tothose without the skill or inclination to use electronicreferences. Most importantly, this work is meant toprovide an ongoing perspective on the developmentof the field. Several distinctive article types havebeen commissioned. Advanced reviews will surveyspecific areas in a citation-rich format suitable forgraduate students and researchers. Opinions on issues

DOI: 10.1002/wnan.032

related to nanomedicine and nanobiotechnologywill be invited from thought leaders in the field.Overviews will allow people not entirely familiarwith an area to rapidly gain a perspective on whatmight be important, new or innovative. Finally, wewill include shorter contributions, known here asFocus articles, in which authors will offer perspectiveson their own work in the field. This will providethem with an opportunity to place what they aredoing in the context of the long-term development ofnanomedicine and nanobiotechnology.

One of the key benefits of this publication is thatit will be written from multiple viewpoints rangingfrom medical to material science. Importantly, wehope to make this content accessible to individuals indiverse disciplines in such a way as to provide a singlesource and platform to discuss and define the field.We will develop an interactive online forum that willallow editors and contributors to discuss publishedarticles and offer input to the evolving knowledgebase. In addition, the highly structured format ofthe publication will allow instructors to create classtexts for different disciplines (such as medicine orbioengineering or material science) from the same setof articles. This will provide the novice with a meansto enter the field and gain familiarity with manyaspects that would have previously required diversecompilations of single articles from different sources.

Thus, I believe this evolving publication willnot only foster new initiatives in nanomedicineand nanobiotechnology but will also bring newinvestigators to the field. As with any new endeavor,I am sure there are going to be growing pains, andit will take an ongoing effort to try and achieve ourgoals. Wiley-Blackwell is firmly committed to thisconcept and believes it will fundamentally changethe separation between online journals and referenceworks. We look forward to your feedback and hopethat you find this new publication of interest, and staywith us to see how it evolves.

Jim BakerDirector of Michigan NanotechnologyInstitute for Medicine and theBiological Sciences ChiefUniversity of MichiganAnn Arbor, MI

Volume 1, January /February 2009 2008 John Wi ley & Sons, Inc. 1

Opinion

The public acceptance ofnanomedicine: a personalperspectiveDavid M. Berube∗

Limited understanding of a subject leads to limited perceptions, includingmisinformed biases and associations. In regard to the field of nanotechnology,prior biotechnologies have harmed public perception of nanotechnology throughassociation alone. While public bias is slow to convert toward truth, it is likely thatthe medical applications of nanotechnology will foster a renewed interest and trustin the field through the prolonged escape from death. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 2–5

This opinion piece is drawn from some work underconsideration by the Public Communication on

Science and Technology (PCOST) Project at NorthCarolina State University with which I am associated.It suggests lines of future research on the basis ofobservations from a review of the academic literaturein the fields of applied nanoscience and risk studies.

It is a blessing as well as a curse. Appliednanoscience or nanotechnology brings great opportu-nities, but at the same time it scares the bejebus out ofsome people. Exactly how and why many people areapprehensive about nanotechnology remain difficultto discern at this time. Given the inherent ambi-guity of the word ‘nanotechnology’ with its nearlyinfinite applications and the weaknesses of delibera-tive polling methodologies, any claim made regardingpublic attitude about nanotechnology should be crit-ically, if not skeptically, evaluated. Nevertheless, thishas not stopped some critics from speaking on thepublic’s behalf and voicing powerful concerns aboutprecaution and uncertainty.

The question surfaces whether patient con-sumers (read as the public) will be reluctant to embracenanomedicine. As argued below, there seems little like-lihood the field of medicine will have as much difficultyas nonhealth commercial industries. One of the rea-sons has to be the rigorous process drugs and devicesundergo before they are marketed. While some may

∗Correspondence to: David M. Berube, North Carolina StateUniversity, Raleigh, NC, USA.E-mail: [email protected]

DOI: 10.1002/wnan.030

criticize the FDA for their failures in this area, compar-ative to other potential applications, FDA regulation isreasonably acceptable. When something slips through,there are always the lawyers. Companies in the drugsand devices world understand the potential liabil-ity issues associated with the products they marketand generally act prudently. The public will embracenanomedicine due to the sense of ease drawn from theregulatory purview of the FDA and potential litigationagainst those who market unsafe medical products.Below I argue there is an additional motivator: thefear of death.

My second argument detailed below is this:there is some likelihood that advances in the fieldof nanomedicine and the diagnosis and treatment ofdisease will ease the way for nonhealth-related com-mercial applications. A variation of the psychologicalconcept of transference will be at work here.

First, we examine the biases associated withmedical developments and their introduction intocommerce. The feeling is generally called necrophobiaand it is the fear of death and dead things. Fear ofdeath is a dense concept and includes fear of the dyingprocess, of the dead, of being destroyed; for significantothers, of the unknown, of conscious death, for bodyafter death, and of premature death. While excessivefear of death may be socio-pathological, some fearof death is a survival mechanism for the species. Itranks first of all the phobias and is shared in somesense by almost everyone. This fear instigates care andcaution. For most of us, it animates our efforts toavoid unnecessary and excessively risky activities.

There is some evidence this fear may be intrinsicto our species, but more easily defended is the concept

2 2008 John Wi ley & Sons, Inc. Volume 1, January /February 2009

WIREs Nanomedicine and Nanobiotechnology The public acceptance of nanomedicine: a personal perspective

that we are taught to fear death. Most religionsseparate life and afterlife into distinct categories andthe transition from one to the other is mystical, if nottraumatic. We see death portrayed by the media as sadand grief-laden, and some of us experience epiphaniesfollowing a disturbing death—often by someone veryclose to us.

Fear of death is an important variable in healthcare delivery.1 There is even some evidence fear anddenial of death among medical professionals and thepublic is the basis of the belief that prolongation oflife is the predominant goal of medicine.2

The introduction of a new technology, exotic orfamiliar, which serves to prolong life taps into our fearof death and dying. Sating this apprehension makes theadoption of this technology, including nanomedicine,more likely against the same technology associatedwith some less essential application, like strongerautomobile bumpers and better performing baseballbats.

Second, consider a troubling phenomenon oftenassociated with the generation of polling data onpreferences toward applications of nanotechnology.More respondents opine about nanotechnology thancan define it.

This is not wholly surprising. We know fromperception studies that populations wishing to providepositive feedback to a survey will offer opinions theyfeel are likely to be anticipated by the survey question.This produces a range of false positive and negativeresults depending on the survey question and generalsubject of the survey itself.

Moreover, when respondents become befuddledwith a question, instead of admitting ignorance theydraw from a set of sensibilities they feel are relatedto the subject of the question, a form of reasoningby analogy. Some have suggested that when askedabout nanotechnology, respondents will equate itwith biotechnology and transfer their feelings aboutbiotechnology to nanotechnology. Others suggest theytransfer sensibilities about health and safety issues.Still others suggest environmental risks as a source forthe transference.3

This form of transference is fundamental in listsof important heuristics in perception. These phenom-ena are derivative of research on representativenessand set theory, as well as the availability and anchoringbiases. When making decisions, individuals over-relyon generalizing from classes, availability of instancesand scenarios, and adjustment from an anchor.4

Focalism is another name for this effect, andit is loosely related to the fallacy of composition inlogic. Some might want to call it simply a hastygeneralization.

Another phenomenon called contagion hasbeen a part of the debate over nanotechnologyfor some time. Those of us who carefully watchpublic pronouncements about the risks associatedwith nanotechnology have noticed a rhetorical devicesurfacing aside demands for more research about thehealth and safety implications of nanotechnologies.Speakers and commentators have added a powerfullynegative scenario. They claim the effect of a singleserious health and safety event might be sufficient toproduce a contagion effect. This effect assumes theentire gamut of businesses and industries engagedin nanotechnology applications will wither withattendant economic losses to a spate of stakeholders.

This phenomenon is incredibly dense andcomplex, whereby in some instances it is noticeablewhile in others it is not. In addition, there aresome examples testing the power of the contagioneffect. For example, contagion phenomena occurredneither with the release of Kleinman’s Magic Nano(which was responsible for over a hundred reportedcases of respiratory distress in Europe last year) norwith the Samsung SilverCare product line allegedlyrelated to waste treatment difficulties. However, bothof these examples did not receive substantial mediaattention, hence they were not amplified. While theJohnson & Johnson Tylenol tampering case from1982 is cited as a counter-example, the blame inthis instance was not the business but a miscreant.Recently, the food and personal care industries seem tohave weathered gales of their own: salmonella taintedfresh spinach, Escherichia coli-tainted green onions,melamine-tainted pet foods and fish feed, and ethylglycol (an antifreeze ingredient) tainted toothpastesin 2006 and 2007 as well, though in these instancesblame was equally unclear and dissipated across longsupply lines.

Generally, contagion refers to the spillover ofthe effects of shocks from one or more firms to otherfirms. Most studies of contagion limit their analysisto how shock affects firms in the same industry, or‘intra-industry’’’ contagion. Most of the studies oncontagion attempt to differentiate between a ‘pure’contagion effect and a signaling or information-basedcontagion effect. An example of a pure contagioneffect would be the negative effects of a bank failurespilling over to other banks regardless of the cause ofthe bank failure. An example of a signaling contagioneffect would be if a bank failure is caused by problemswhose revelation is correlated across banks, and thecorrelated banks are impacted negatively.

Evidence supporting intra-industry contagionis fairly common and comes from studies of thecredit default market5 and other financial institutions.


Opinion www.wiley.com/wires/nanomed

Evidence supporting inter-industry or extra-industrycontagion is mostly limited to the financial industry aswell. For example, studies on banks and life insurancecompanies6 demonstrate some cross-industry cascadephenomena. Evidence crossing industries as diverse ascosmetics and food production is nearly impossibleto find, but nanotechnology may be the exception.While a contagion event across a diverse industrymight be difficult to prove, it is clearly not impossible.Indeed, in terms of a newly emerging industry withunclear boundaries such as nanotechnology, it mightbe plausible.

We have already learned the deficit theoryof science literacy is fallacious. Providing moreinformation about science to a subject does notequate to a more positive feeling toward a scientificartifact, like nanotechnology per se. Noteworthy tosome, a recent study suggested that more informationactually reduces positive responses (see Kahan et al).3

However, there is an equally strong case thatinformation can affect feelings, especially whenmediated.7 While Kahan et al’s finding remainsvery consistent with the observation that merelymentioning risk of a phenomenon will increaseapprehensions by raising its saliency may be true;on some level mediated information can attenuatesome of the power of the effect.

One of the hypotheses we have been studyinginvolves a variation on the anchoring bias. The firstfeeling one develops about a phenomenon or artifacttends to anchor subsequent assessments. A negativeexperience or anchor is very difficult to erase orreduce. A positive anchor is difficult to degrade,though less so. Essentially, there is a bias towardthe negative. We argue when the public has a feeling,which is not based on understanding but drawn froman analog feeling (such as, nanotechnology is likebiotechnology so how I feel about biotechnology ismuch like how I should feel about nanotechnology),we might be able to rehabilitate the antecedent. Forthe example mentioned above, if the public developsa positive feeling about nanotechnology, it mighthelp rehabilitate a less than positive feeling aboutbiotechnology. By extension, we sense that a positiveexperience with nanotechnology in the present willaffect feelings about nanotechnology in the future aswell.

While this hypothesis needs to be tested, weanticipate that some of the first authentic applicationsof nanotechnology will make this case. Many, if notmost, of current product releases have involved theuse of nanoparticles in coatings, e.g., paint, and asreinforcement when associated with another media,e.g., carbon composites. We expect nanomedicine

will be perceived as actual nanotechnology and maybe willing to go as far as suggesting nanomedicineproducts will be perceived as archetypal for nan-otechnology based on public expectations of majorand breakthrough technological development in thefield of medicine in general.

Nanomedicine may help anchor public sen-timent positively. This should have some positiveeffects on feelings toward the subsequent introductionof products in nanomedicine (intra-industry conta-gion). Given the public positive response to exotichealth technologies, the positive public feelings fornanomedicine may transfer to subsequent applica-tions of nanotechnology in health and even outsideof health (inter-industry contagion). If the publicembraces applications in nanomedicine, follow-upapplications in the food industry may be positivelyaffected as well.

In final extension, if respondents to surveysabout nanotechnology opine without an understand-ing of nanotechnology and draw the warrant for theiropinion from an analog, then a positive feeling towardnanomedicine might help to rehabilitate the feelingstoward the relevant analog, such as biotechnology orenvironmental health and safety in general.

While this needs to be studied, there is someanecdotal evidence that this point of view is plausi-ble. Global warming concerns are rehabilitating ourfeelings toward nuclear fission power generation8,and advances in bioengineering less expensive phar-maceuticals may be rehabilitating our feelings towardgenetically modified foods. Opposition to ‘introducinggenetically modified foods into the US food supply’has declined from 58% in 2001 to 47% today, an11-point decrease.9 On the other hand, the public’sattitude toward genetically modified foods seem to beinversely related with the evolutionary ladder; henceefforts to genetically modify animals might exacerbatenegative attitudes toward genetically modified crops.9

In conclusion, we predict nanomedicine prod-ucts will be welcomed by the public as a stay againstthe fear of death and we find some solace in thisprognostication based on public responses towardbiomedicine, especially recent opinion shifts towardstem cell research.10 Furthermore, we envision a posi-tive feeling toward nanomedicine products will makelater introductions of nanomedicine products evenmore welcomed by the public. Finally, we hypothesizea positive feeling towards nanomedicine products maycarry over to nanoproduct lines from other industries,and a positive feeling toward nanomedicine productsmay rehabilitate past held beliefs toward analogicalproducts and other related phenomena.


WIREs Nanomedicine and Nanobiotechnology The public acceptance of nanomedicine: a personal perspective

REFERENCES

1. Seravalli EP. The dying patient, the physician, and thefear of death. N Engl J Med 1988, 319:1728–1730.

2. Meier DE, Morrison RS, Cassel CK. Improving pallia-tive care. Ann Intern Med 1997, 127:225–230.

3. Kahan DM, Slovic P, Braman D, Gastil J,Cohen GL. Affect, values, and nanotechnol-ogy risk perceptions: an experimental investiga-tion. Nanotechnology Risk Perceptions: The Influ-ence of Affect and Values. Washington, DC:Woodrow Wilson International Center for Schol-ars, Center Project on Emerging Nanotechnologies;2007, http: // www. nanotechproject .org / 108 / survey -finds-emotional-reactions-to-nanotechnology (accessedMarch 7, 2007).

4. Tversky A, Kahneman D. Judgment under uncertainty:heuristics and biases. Science 1974, 184:1124–1131.

5. Zhang G. Intra-Industry Credit Contagion: Evi-dence from the Credit Default Swap Market andthe Stock Market. EFMA 2004 Basel Meetings

Paper, 1995, March 1, Available at SSRN:http://ssrn.com/abstract=492682.

6. Brewer E, Jackson WE. Inter-Industry Contagionand the Competitive Effects of Financial DistressAnnouncements: Evidence from Commercial Banksand Life Insurance Companies. FRB of ChicagoWorking Paper No. 2002–23, 2002, December.Available at SSRN: http://ssrn.com/abstract=367180.doi:10.2139/ssrn.367180 (accessed July 13, 2007).

7. Pidgeon N, Kasperson RE, Slovic P, eds. The SocialAmplification of Risk. Cambridge: Cambridge Univer-sity Press; 2003.

8. Sweet W. Kicking the Carbon Habit—Global Warmingand the Case for Renewable and Nuclear Energy. NewYork: Columbia University Press; 2006.

9. http://pewagbiotech.org/newsroom/releases/112404.php3 (accessed, November 24, 1994).

10. Wooley M, Probst S. Public attitudes and perceptionsabout health-related research. J Am Med Assoc 2005,294:1380–1384.

RELATED ONLINE ARTICLES

Commercialization of nanotechnology.Nanoparticles in food as potential health hazards.Ethical issues in nanomedicine.


Opinion

Nanotechnology and orthopedics:a personal perspectiveCato T. Laurencin,1,2,3∗ Sangamesh G. Kumbar2,3 and Syam PrasadNukavarapu2,3

Bone is a nanocomposite material comprised of hierarchically arranged collagenfibrils, hydroxyapatite and proteoglycans in the nanometer scale. Cells areaccustomed to interact with nanostructures, thus providing the cells with a naturalbone-like environment that potentially enhance bone tissue regeneration/repair. Inthis direction, nanotechnology provides opportunities to fabricate as well as explorenovel properties and phenomena of functional materials, devices, and systems atthe nanometer-length scale. Recent studies have provided significant insights intothe influence of topographical features in regulating cell behavior. Topographicalfeatures provide essential chemical and physical cues that cells can recognizeand elicit desired cellular functions including preferential adhesion, migration,proliferation, and expression of specific cell phenotype to bring desired effects. Thecurrent article will address some of the nanotechnology implications in addressingissues related to orthopedic implants performance and tissue engineering approachto bone repair/regeneration. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 6–10

Nanotechnology, a new focus in the area ofbiomedical research, involves visualization,

manipulation, and fabrication of materials on thesmallest scales, in dimensions of 1 µm down to 10◦A.This diverse new field draws from various disciplines:biology, medicine, materials science, physics, andmanufacturing. At the nanoscale, materials possessseveral novel properties including extremely highsurface area to volume ratio, tunable optical emission,enhanced mechanical properties, and super paramag-netic behavior, which contrast the properties that aredeemed important when working with the bulk parentmaterials. Most of the current technical research inthis area is focused on fabrication and evaluation ofnanostructures, devices, and systems. Nanotechnology

∗Correspondence to: Cato T. Laurencin, Dean, School of Medicine,Distinguished Professor of Orthopaedic Surgery and Chemical,Materials and Biomolecular Engineering, The University ofConnecticut, Farmington, CT 06030–3800, USA.E-mail: [email protected], School of Medicine, University of Connecticut HealthCenter, Farmington, CT 06030–3800, USA2Department of Orthopaedic Surgery, University of Connecticut,Farmington, CT 06030, USA3Department of Chemical, Materials and Biomolecular Engineering,University of Connecticut, Farmington, CT 06269–3222, USA

DOI: 10.1002/wnan.025

is rapidly gaining momentum and attracting largeinvestments from both government and privatesectors. The National Nanotechnology Initiative(NNI), a multidisciplinary strategy for developmentof science and engineering fundaments, has proposeda total investment budget of $8.3 b for the year 2008.1

The global market for nanotechnology products, suchas nanoscale devices and molecular modeling systems,has seen an average annual growth rate of 27.5%,from $400 million in 2002 to $1.37 b in 2007.2

Moreover, nanomedicine, which is an offshootof nanotechnology, can be broadly defined as a tech-nology that uses molecular tools and knowledge ofthe human body for diagnosis, treatment, and pre-vention of diseases and traumatic injury.3 Healthcare applications of nanomedicine can be roughlyclassified into diagnostics, imaging, nanobiomateri-als, nanodevices/implants, novel drug delivery systems(NDDS) and issues related to toxicity. Applicationsand products dealing with NDDS dominate thenanomedicine market, and thus, there are currentlymore than 40 nanotechnology-based products avail-able in the market. Several biomaterials in the form ofself-assembled nanofibers/nanoparticles, electrospunnanofibers, nanocomposites, and hydroxyapatite arealso being used as integral parts of biomedical devicesto improve their in vivo performance ( Figure 1).


WIREs Nanomedicine and Nanobiotechnology Nanotechnology and orthopedics: a personal perspective

Nanotechnology:Nanobiomaterials

TopographySurface chemistry

Wettability

Cell compatibilityAdhesion

ProliferationDifferentiation

Orthopaedic applications Nano-patterned implants Nano composite fillers Nano featured Scaffolds Novel drug delivery Systems Diagnostics

Toxicity issuesInflammationCytotoxicity

Nanoparticle clearance

Improves

FIGURE 1 | Nanotechnology and orthopedic applications.

Nanohydroxyapatite (nHAp)-based products such asOstim, Vitoss, and Perossal are now commerciallyavailable for bone filling. Additionally, some of thenanoparticle composites available for dental fillingand repair applications include Filtek Supreme, CeramX duo, Tetric EvoCeram, Premise and Mondial. Sev-eral other nanotechnology approaches being exploredinclude coating of orthopedic implants with nHAp andbioactive signaling molecules to stimulate osteoblastproliferation and differentiation.

BONE: A NANOCOMPOSITEThe extracellular matrix (ECM) in natural bone tissueis principally composed of hierarchically arrangedcollagen fibrils, hydroxyapatite, and proteoglycansin the nanometer scale. The individual collagenhelical chains are 10 nm in length and self-assembleinto orientated collagen fibers measuring approxi-mately 500 nm in length. The mineral or inorganiccomponent of bone, hydroxyapatite, exists as plate-like nanocrystals, measuring 20–80 nm in length.Additionally, the triple helical structure of collagenprovides bone with a structural framework, hightensile strength, and flexibility, while crystallinehydroxyapatite accounts for the stiffness and highcompressive strength of bone. With a more in-depthinvestigation into the constituents of bone and theirproperties, researchers hope to develop scaffold andimplants that closely mimic the physicochemicalcharacteristics of natural bone.

Nanobiomaterials and Cellular RecognitionThe success of orthopedic implants and tissue-engineered constructs greatly depends on the

biocompatibility of the material. Furthermore, thebiocompatibility of an entity is largely determinedby its physicochemical properties. Many studies haveclearly demonstrated that cell behavior can be easilymanipulated by providing cells with suitable biochem-ical cues, surface topography and external stimuli.4

It has also been observed that minimal interactionsbetween implant surfaces and surrounding tissue oftenresult in poor tissue formation on the implant surfaceand can ultimately lead to failure of the treatment.

Researchers are currently investigating a varietyof methods/techniques to combine both the structuraland chemical components of natural ECM innovel materials for tissue engineering applications.The ECM, a self-assembled nanofibrillar complexthree-dimensional (3D) dynamic structure, plays avital role in determining the cell behavior. Thus, withinthe ECM, cells experience complex nanotopographyand encounter a variety of chemical cues in the form ofproteins and growth factors that regulate cell growth,differentiation and metabolic activity. Followingimplantation of a device, many proteins from bodilyfluids adsorb onto the surface of the implant andsubsequently control cell adhesion.5 In this case,the implanted materials are essentially playing arole similar to that played by the ECM. Initial celladhesions to implant surfaces are determined by thepresence of specific amino acid sequences availablefor binding to cell membrane integrin receptors. Forinstance, osteoblasts preferentially adhere to aminoacid sequences such as heparin-sulfate and arginine-glycine-aspartic acid (RGD) regions in the adsorbedproteins. In addition to chemical cues, surfacetopography significantly alters cell behavior and canbe responsible for changes in morphology, adhesion,motility, proliferation, endocytotic activity, and generegulation.4 Cells existing on surfaces having the sametopography but different chemistry,5 with differentconcentrations of protein adsorption also showed verysimilar cell behavior. Researchers, however, arguewhether the observed cell behavior can be attributed tosurface topography or differential protein adsorptionon the surface. Furthermore, it is apparent thatdetailed understanding of implant surface properties,protein adsorption, and cell behavior can potentiallycircumvent the problems associated with currentorthopedic implants.

NANOBIOMATERIALS

Nanotechnology has revolutionized many fields of sci-ence including fabrication and characterization of var-ious nanostructures. Several techniques for patterning



implant surfaces and for efficiently constructing scaf-folds for tissue engineering have emerged; some ofthese techniques include: lithography, polymer demix-ing, phase separation chemical etching, electrospin-ning, and molecular self-assembly. The nanofeaturescreated by these methods can provide the bulkmaterial with high surface area to volume ratio,tunable mechanical, electrical, optical emission, andsuper paramagnetic behavior, characteristics that havebeen successfully exploited for a variety of healthcare applications ranging from drug delivery tobiosensors.

TopographyNanotopographical features such as pores, ridges,grooves, fibers, nodes, and combinations of thesefeatures are known to influence cell behaviorsignificantly. The emphasis in current research ison developing implant surfaces with a suitablesurface topography to elicit desired cellular function.In one set of studies, cellular interaction withnanophase Ti, Ti6Al4V, and CoCrMo alloys resultedin increased osteoblast adhesion due to the presence ofmore particle boundaries compared to conventionalmetals.6 For instance, nanopatterned polystyrenegrooves with two different depths of 50 and 150 nm,with a periodicity of 500 ± 100 nm, showed a strongalignment and orientation of primary osteoblastsin the direction of the grooves.7 One other studyshowed that primary human osteoblasts migratedaway from Ti oxide surfaces patterned with 110-nm-high hemispherical protrusions of varying topographydensities.8 Another important finding reveals thatsensitivity and cell behavior on nanotopographicfeatures depends on the cell type. For example, with anincrease in either carbon nanofiber surface energy ora simultaneous change in carbon nanofiber chemistry,enhanced osteoblast adhesion was observed on carbonnanofibers while smooth muscle cell, fibroblast, andchondrocyte adhesion decreased.9 At this state, it isnot fully understood how cells detect and respondto nanofeatures. Thus, it is important to understandcell-nanotopography interactions, and that variabilityin results may exist among varying cell types.

Surface ChemistryImplant surface chemistry plays a critical rolein deciding the performance and success of thedevices. Proteins and other biomolecules dynamicallyadsorb to biomaterial surfaces upon implantation.These complex molecules can trigger nonspecificinflammatory responses characterized by foreign bodyreaction and fiber capsule formation. Thus nonspecific

inflammatory responses can limit integration ofthe device and influence in vivo performance.Limited success has been achieved through thedelivery of anti-inflammatory agents and nonfoulingof implant surfaces. Studies have demonstratedthat self-assembled monolayers (alkanethiols ongold), having well-controlled surface properties anddifferent terminal functionalities such as CH3,OH, COOH, and NH2, possess different affinitiesfor fibronectin adsorption, and thus differentiallyinfluence integrin binding and cell adhesion.10

The binding of monoclonal antibodies and α5 β1integrin to adsorbed fibronectin affinities were inthe order of OH > COOH NH2 > CH3 while α5integrin binding was in the order of COOH ��OH NH2 CH3, demonstrating the α5β1 integrinspecificity for fibronectin adsorbed onto the NH2 andOH.11 Differences in integrin binding differentiallyregulate focal adhesion assembly and signaling which,in turn, modulate cellular functions in biomaterial andimplant surfaces.

WettabilityIt is evident from the literature that the hydrophobic-ity or hydrophilicity of a surface can significantly altercell behavior. Moreover, the wettability of a materialcan allow for characterizing materials with regardsto the hydrophobic/hydrophilic categories. Implantor biomaterial surface composition, surface treat-ment, surface roughness, immobilization of variouschemical agents to the surface, and the presence ofnanofeatures on the surface alter the surface wettabil-ity and affect cell behavior. For instance, the surfacewettability of alumina can be improved by reduc-ing the alumina grain size from 167 to 24 nm.12 Inanother study, ultrafine titanium crystals producedby high pressure torsion provided a high degree ofsurface wettability, and thus, preosteoblasts showedenhanced attachment and proliferation rates.13 Addi-tionally, improved wettability enhances adsorptionof vitronectin and fibronectin on nanofeatures thatstimulate the osteoblast adhesion. It is importantto investigate the influence of surface topography,chemistry, and wettability of various biomaterials andimplant surfaces on protein adsorption and receptor-mediated cell adhesion. Optimization of these param-eters to improve adsorption of osteogenic proteins onthe implant surface and enhance implant exposuretime to cell integrin binding domains, may pro-vide opportunities to develop implant surfaces whichenhance the attachment, adhesion, and developmentalresponse of osteoblast precursors leading to acceler-ated osteointegration.


WIREs Nanomedicine and Nanobiotechnology Nanotechnology and orthopedics: a personal perspective

NANOCOMPOSITES

Bone is a natural composite of nHAp and organiccomponents such as collagen and proteoglycans.Hence, various tissue engineering strategies haveadopted composite scaffold approaches to moreclosely mimic the bone in structure and composi-tion. Nanoparticles including calcium triphosphate,bioactive glass, hydroxyapatite, and calcium-deficientHAp in combination with natural (PLGA, polycapro-lactone, and polyphosphazene) or synthetic (chitin,chitosan, and hyaluronic acid) biodegradable poly-mers have been fabricated into porous 3D scaffolds forbone repair/regeneration purposes.14 Not only doesthis approach allow for mimicking bone in composi-tion but incorporation of nanoceramics enhances thematerials mechanical strength and nanotopograph-ical features. The presence of a mineral phase onthe surface enhances the scaffold’s osteoconductivity,osteogenicity, and osteointegrative nature. Thoughmost of these scaffold types showed mechanical prop-erties in the range of human cancellous bone, fab-ricating scaffolds with mechanical performance closeto compact bone is still a persisting challenge. Ina bid to improve mechanical properties, researchershave begun to develop and evaluate composite poly-mer/ceramic matrices containing single or multiwalledcarbon nanotubes.15 In addition to the size, carbonnanotubes offer excellent properties, such as high ten-sile strength, high flexibility, and low density that canbe exploited to develop more successful orthopedicimplant materials.

NANOFIBERS

Cells are organized and trapped within organs andtissues by a nanofiberous 3D ECM. Electrospunnanofiber matrices closely mimic the structure ofnatural ECM and have shown great promise asscaffolds for tissue engineering applications. Polymericnanofiber scaffolds are characterized by ultrathincontinuous fibers, high surface-to-volume ratio,porosity and variable pore-size distribution, makingthem ideal for tissue regeneration efforts.16,17 Thenanofiber structure provides anchorage for cells whilehigh porosity provides a means for the supplyand removal of nutrients and metabolic wastes.16–18

Cell growth has been found to be significantlyenhanced on nanofiber matrices. Additionally, thenanofiber scaffold environment was found to drivemesenchymal stem cells to differentiate along anosteogenic lineage, resulting in the formation ofmineralized tissue.19 Current efforts in this area focuson coating biomedical implants with nanofibers to

provide high surface area, achieve tissue compatibilityand allow for selective delivery of bioactive agents.20

NANOSTRUCTURE CYTOTOXICITYISSUES

Degradation of nanocomposite or nanostructuredscaffolds result in erosion of nanoparticles that areeither retained or degraded and get excreted fromthe system. In the process of metabolism, nanopar-ticles pass though various organs such as blood,liver, and kidneys, and possibly cause oxidative stressand inflammation.21 Thus, it is important to estab-lish the clearance rate and the cytocompatability ofthese nanoparticles with hemocytes, stem cells, hepa-tocytes, and nephrocytes. Biodegradable nanoparticlescause less damage in contrast to biostable particles,provided the degradation products are physiologi-cally compatible. The cytotoxic effects of traditionallyused nHAp which are neither biodegradable nor sameas the native bone apatite can be mitigated whenreplaced with biodegradable carbonated hydroxyap-atite nanoparticles. Also, there are conflicting reportsin literature about the safety of carbon nanotubes forhuman use. However, their clearance can be improvedby functionalizing carbon nanotubes with hydrophilicgroups such as –COOH, –OH, and –NH2.

CONCLUSION

Nanotechnology has revolutionized many researchareas. Orthopedic research, in particular, has nowturned its focus to utilizing these developments toaddress the limitations associated with implant designand tissue regeneration.

A current fabrication and design issue fornanobased implants is optimization of physicochem-ical properties. It is also important to address issuesrelated to inflammatory response and toxicity. Acombined approach involving multifunctional nanoto-pographic features that incorporate bioactive factorsand a suitable cell population may be an alternativeto aid in the rapid development of engineered organs.

Recent studies have provided significant insightsinto the influence of topographic features in regulatingcell behavior, including preferential adhesion, migra-tion, proliferation, and expression of cell-specificphenotypes. It is critical to understand the molec-ular mechanisms governing cells and cell-materialinteraction to generate better scaffold and implantperformance. With the advent of nanofabricationtechniques several novel biomaterials can be fabri-cated into nanostructures that simulate the native



hierarchical structure of the bone. Results of the pre-liminary studies mentioned are quite encouraging, andthe next step should involve thorough evaluations of

these structures in suitable animal models to make thetransition to clinical use.

REFERENCES1. http://www.nano.gov/NNI 08Budget.pdf (accessed,

2007).

2. Edwards S: Biomedical Applications of NanoscaleDevices. Report ID: HLC031A, Norwalk, CT: BCCResearch, Sept, 2003.

3. Duncan R. Nanomedicines in action. Pharm J 2004,273:485–488.

4. Kumbar SG, Kofron MD, Nair LS, Laurencin CT. Cellbehavior toward nanostructured surfaces. In: Gon-salves KE, Laurencin CT, Halberstadt C, Nair LS: eds.Biomedical Nanostructures. New York: John Wiley &Sons; 2008, 261–295.

5. Curtis A, Britland S. Surface modification of bioma-terials by topographic and chemical patterning. In:Ogata N, Kim SW, Feijen J, Okano T, eds. AdvancedBiomaterials in Biomedical Engineering and DrugDelivery Systems. Tokyo: Springer-Verlag; 1996,158–167.

6. Webster TJ, Ejiofor JU. Increased osteoblast adhesionon nanophase metals: Ti, Ti6Al4V, and CoCrMo. Bio-materials 2004, 25(19):4731–4739.

7. Lenhert S, Meier MB, Meyer U, Chi L, Wiesmann HP.Osteoblast alignment, elongation and migration ongrooved polystyrene surfaces patterned by Langmuir-Blodgett lithography. Biomaterials 2005, 26:563–570.

8. Rice JM, Hunt JA, Gallagher JA, Hanarp P, Suther-land DS, et al. Quantitative assessment of the responseof primary derived human osteoblasts and macrophagesto a range of nanotopography surfaces in a sin-gle culture model in vitro. Biomaterials 2003,24(26):4799–4818.

9. Woo KM, Chen VJ, Ma PX. Nano-fibrous scaffold-ing architecture selectively enhances protein adsorptioncontributing to cell attachment. J Biomed Mater Res2003, 67A(2):531–537.

10. Keselowsky BG, Collard DM, Garcia AJ. Surfacechemistry modulates fibronectin conformation anddirects integrin binding and specificity to control celladhesion. J Biomed Mater Res 2003, 66A(2):247–259.

11. Dahmen C, Auernheimer J, Meyer A, Enderle A, Good-man SL, et al. Improving implant materials by coating

with nonpeptidic, highly specific integrin ligands.Angew Chem Int Ed 2004, 43(48):6649–6652.

12. Webster TJ, Ergun C, Doremus RH, Siegel RW,Bizios R. Specific proteins mediate enhanced osteoblastadhesion on nanophase ceramics. J Biomed Mater Res2000, 51(3):475–483.

13. Faghihi S, Azari F, Zhilyaev AP, Szpunar JA, Vali H,et al. Cellular and molecular interactions betweenMC3T3-E1 pre-osteoblasts and nanostructured tita-nium produced by high-pressure torsion. Biomaterials2007, 28(27):3887–3895.

14. Khan Y, El-Amin SF, Laurencin CT: In vitro and invivo evaluation of a novel polymer-ceramic compositescaffold for bone tissue engineering. Conference Pro-ceedings of the IEEE Engineering in Medicine andBiology Society, Vol. 1, New York, 2006, 529–530.

15. Harrison BS, Atala A. Carbon nanotube applica-tions for tissue engineering. Biomaterials 2007,28(2):344–353.

16. Li WJ, Laurencin CT, Caterson EJ, Tuan RS, Ko FK.Electrospun nanofibrous structure: a novel scaffoldfor tissue engineering. J Biomed Mater Res 2002,60:613–621.

17. Nukavarapu SP, Kumbar SG, Nair LS, Laurencin CT.Nanostructures for tissue engineering/regenerativemedicine. In: Gonsalves KE, Laurencin CT, Halber-stadt C, Nair LS, eds. Biomedical Nanostructures. NewYork: Wiley; 2008, 377–407.

18. Kumbar SG, Nukavarapu SP, Roshan R, Nair LS, Lau-rencin CT. Electrospun nanofiber scaffolds: engineeringsoft tissues. Biomed Mater 2008, 3:1–15.

19. Pelled G, Tai K, Sheyn D, Zilberman Y, Kumbar SG,et al. Structural and nanoindentation studies of stemcell-based tissue-engineered bone. J Biomech 2007,40:399–411.

20. Kumbar SG, Nair LS, Bhattacharyya S, Laurencin CT.Polymeric nanofibers as novel carriers for the deliveryof therapeutic molecules. J Nanosci Nanotechnol 2006,6:2591–2607.

21. Borm PJA. Particle toxicology: from coal mining tonanotechnology. Inhal Toxicol 2002, 14(3):311–324.

RELATED ONLINE ARTICLESAtomic Force Microscopy (AFM) and indentation force measurement of bone.Nanotechnology for bone materials.Nanofibers and nanofibrous composites for tissue engineering applications.


Overview

Prospects and developments incell and embryo laser nanosurgeryVikram Kohli∗ and Abdulhakem Y. Elezzabi1

Recently, there has been increasing interest in the application of femtosecond (fs)laser pulses to the study of cells, tissues and embryos. This review explores thedevelopments that have occurred within the last several years in the fields of celland embryo nanosurgery. Each of the individual studies presented in this reviewclearly demonstrates the nondestructiveness of fs laser pulses, which are used toalter both cellular and subcellular sites within simple cells and more complicatedmulticompartmental embryos. The ability to manipulate these model systemsnoninvasively makes applied fs laser pulses an invaluable tool for developmentalbiologists, geneticists, cryobiologists, and zoologists. We are beginning to seethe integration of this tool into life sciences, establishing its status amongmolecular and genetic cell manipulation methods. More importantly, severalstudies demonstrating the versatility of applied fs laser pulses have establishednew collaborations among physicists, engineers, and biologists with the commonintent of solving biological problems. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 11–25

Several studies have reported the application offemtosecond (fs) laser pulses as a precise scalpel

tool for performing cellular surgery.1–17 In each study,fs laser pulses were produced from a titanium sapphire(Ti:Sapphire) laser oscillator or amplifier (700–900nm) delivering a sub-10 fs to 250 fs pulse at arepetition rate of 76 MHz to 1 kHz. The fs laserpulses were coupled to a high numerical aperture (NA)microscope objective, NA = 0.95–1.4, and localizedto cellular and subcellular sites. Beam dwell timesranged from milliseconds to seconds and pulseenergies delivered to the sample were 0.03 to severalnanojoules per pulse (nJ/pulse). Model systems thathave been used in fs laser pulse mediated nanosurgeryinclude human metaphase chromosomes,4 Chinesehamster and canine kidney epithelial cells,1,2 plantchloroplasts,5 mitochondria in endothelial and HeLacells,6,7 yeast microtubules,8 the actin cytoskeletonin fixed 3T3 fibroblast and bovine endothelialcells,6,9 hamster ovary cells,10,17 Caenorhabditiselegans,11,12 Drosophila melanogaster,16 Sprague-Dawley rats and Danio rerio (zebrafish).13 Using these

∗Correspondence to: Vikram Kohli, University of Alberta, Edmon-ton, Alberta, Canada.E-mail: [email protected] of Electrical and Computer Engineering, University ofAlberta, Edmonton, Alberta, Canada

DOI: 10.1002/wnan.029

biological systems, intrachromosonal dissections,4

membrane surgery,1 cell isolation,1 cytoskeletal andmicrotubule ablation,6,8,9 knockdown of plastids,5

laser axotomy of neurons,11 intravascular disruptionof microvessels,13 cellular delivery of exogenous DNA,carbohydrates and quantum dots2,3,17 and the surgicalablation of Drosophila16 and zebrafish embryos3,15

have been demonstrated. In this paper, we present areview of current developments in fs laser mediatednanosurgery of cells and embryos with emphasis onthe fs laser as a tool able to induce ablation withhigh spatial resolution and with minimal transferof thermal and mechanical stresses to the materialinvestigated.

LASER INTERACTION WITHBIOLOGICAL MATERIALSFeatures that distinguish fs laser pulses from longerpulse durations (i.e., nanosecond pulses) include theability to localize cellular disruption to a sub-micronresolution, the low threshold energy needed to elicitablation and the lower conversion of energy intoshockwaves and cavitation bubbles, which are adverseside effects known to increase the spatial extentof cellular damage.18–22 When fs laser pulses arefocused to a high peak intensity of 1011–1013 W/cm2,optical breakdown occurs, resulting in the ablation of


Overview www.wiley.com/wires/nanomed

the biological material.19 The mechanism by whichthe material is ablated depends on the strength ofthe peak intensity, leading to quasi-ionized electrons(the electrons are not completely ionized from theatom, rather they occupy a higher energy state withinthe conduction band of the material) produced viamultiphoton absorption or tunneling ionization.19

The Keldysh parameter19,23,24

γ = ω

e

√cε0m�

4I(1)

determines the extent to which multiphotonabsorption or tunneling ionization governs the abla-tion process. In Eq. (1), ω, e, c, ε0, m, �, I representthe frequency of light, electron charge, speed of light,permittivity of free space, electron-hole reduced mass,bandgap energy, and peak intensity of the light pulse,respectively. When γ < 1, tunneling ionization dom-inates the laser-matter interaction process, while forγ > 1 multiphoton ionization dominates.19,23 As to avery good approximation we can consider biologicaltissue as water, Sacchi25 proposed that water shouldbe modeled with a bandgap energy of 6.5 eV. Thevalue of 6.5 eV arises from work conducted by Boyleet al.26 which examined the photolysis of liquid water.If we consider laser light produced from a Ti:Sapphirelaser oscillator with an emission spectrum centered at800 nm (1.55 eV) and a pulse duration of 100 fs, thenfor γ > 1, the biological tissue must simultaneouslyabsorb five photons to excite a valence electron to theconduction band. This conduction electron representsa ‘seed electron’ that undergoes free carrier linearabsorption by nonresonantly absorbing laser photonsthrough inverse bremsstrahlung.19,23 As the electronenergy increases, a condition is reached where theelectron undergoes impact ionization,19 defined as1.5� 24 where � is the effective ionization potentialgiven by19,24

� = 2π

�

√1 + γ 2

γE

(π

2, k

)(2)

where E(

π2 , k

)represents the elliptical integral of the

second kind with k = (1 + γ 2)−1/2.19 At 1.5� the elec-tron impact ionizes a valence electron, resulting in twoelectrons in the conduction band (seed electron andionized valence electron). Through linear absorptionof laser photons, these two electrons can participatein impact ionization, causing a rise in the densityof conduction band electrons. This cascade effectis properly termed avalanche ionization, where theionized electron density quickly rises to a critical valuewhere optical breakdown occurs.19,20,23 At optical

breakdown, the plasma frequency equals the laserfrequency and the critical electron density becomes20

Ncrit = ω2meε0

e2 (3)

where me is defined as the electron mass. At awavelength of 800 nm, Ncrit = 1021 cm−3, beyondwhich the plasma becomes highly reflective andabsorbing to laser light.19

As a consequence of using fs laser pulses forablation, seed electrons can be generated with anintensity value lower than the threshold intensityfor optical breakdown.19 With nanosecond laserpulses, no seed electrons are created by multiphotonionization for intensities below the threshold foroptical breakdown.19 Therefore, the requirement thatthe intensity must equal the threshold for opticalbreakdown to produce seed electrons indicates anincrease in the deposition of laser energy. However,this increased energy is funneled into shockwaves andcavitation bubbles, leading to a larger spatial disrup-tion of the material. In fact, Vogel et al. showed thatthe conversion of energy into cavitation bubbles for fslaser pulses was 6.8% versus 12.7% for nanosecondpulses.19 As a result, less energy is required to elicitablation of the material, which reduces the amount oftransient stresses such as shockwaves and cavitationbubble formation imparted to the sample.19

In fs laser-tissue interaction the mechanism ofablation between high and low repetition rate laseroscillators (i.e., 80 MHz vs. 1 kHz) is different.19

For instance, in 80 MHz ablation the pulse energyis below the threshold energy for optical breakdownwith each pulse producing a low density plasma.19

Ablation of the biological material occurs through theinteraction of multiple pulses through free electroninduced chemical decomposition of the material viabond-breaking.19 In contrast, with low repetition ratessuch as 1 kHz, the pulse energy for ablation is near orabove the breakdown threshold energy. Larger plasmadensities are created in comparison to 80 MHz withthe formation of minute cavitation bubbles. It has beensuggested that the cavitation bubbles are responsiblefor the dissection of the biological material.19

Using fs laser pulses, highly spatially localizedablation is achievable through the nonlinear multipho-ton ionization process (γ > 1). When a laser beam isfocused by a high NA objective, it is the electrondensity profile and not the irradiance profile that gov-erns the spatial extent of ablation. (The NA of themicroscope objective lens can alter the ablation pro-file. Using low NA objectives (i.e., NA < 0.9) plasmascan be generated ahead of the focus. Such a plasma


WIREs Nanomedicine and Nanobiotechnology Developments in cell and embryo laser nanosurgery

can effectively shield successive laser photons fromreaching the focus and induce plasma defocusing.19,27

Spatially asymmetric plasmas are created with lowNA objectives, with the observance of a high plasmadensity (before the geometrical focus) surrounded bya lower density region.27 In contrast, for NA ≥ 0.9,smaller symmetric plasmas are formed.27) The diffrac-tion limited laser spot size has transverse, dtrans, andlongitudinal, z, dimensions28

dtrans = 1.22λ

NAand z ≈ 2

(πw2

0

λ

)(4)

where, for example, NA = 1.3, λ = 800 nm and w0 =375 nm are the NA, wavelength of light and radius ofthe beam waist, respectively. The irradiance profilesalong the transverse and longitudinal direction are750 and 1104 nm, respectively. As the simultaneousabsorption of five photons is required to produce aseed electron, the ablation dimensions are effectivelyreduced by

√5,19 yielding 335 and 494 nm for the

transverse and longitudinal dimensions, respectively.(The transverse and longitudinal dimensions of theablation profile represent theoretical estimates. Theformation of cavitation bubbles, particularly withhigh repetition rate laser oscillators (i.e., MHz), canincrease these values.) Therefore, the ablation ofbiological tissue can be localized to a high spatialresolution, allowing key structures within biologicalmaterial to be removed or altered without affectingadjacent cellular sites. This unique property has madethe application of fs laser pulses a novel tool for thenondestructive study of biological materials.

Cellular and Subcellular NanosurgeryIn a study by Konig et al.,4 the authors reported thenanodissection of fixed air-dried human metaphasechromosomes using fs laser pulses (170-fs, 800 nm,80 MHz). Intrachromosonal dissections were madeby 500–2500 consecutive single line scans across thechromosome using an average laser power of 100mW (1.25 nJ/pulse).4 Dissection depths were analyzedusing scanning force microscopy and revealed a fullwidth at half maximum cut size of 170 ± 10 nm for500 consecutive scans.4 A reduction in the numberof laser line scans to 250 produced smaller cut sizeson the order of 85 ± 10 nm.4 It was also found thatthe cut sizes increased with an increasing number ofconsecutive scans, from 200 to 400 nm for 1000 to2500 scans. In addition to line scans, the authorsperformed stationary ablation of the chromosomeswith an average laser power of 15 mW (0.19 nJ/pulse)and varying beam dwell times.4

(a)

(c)

(b)

FIGURE 1 | Membrane surgery on a live MDCK cell. (a) Illustrates acell of 12 µm in length where three ∼ 800 nm incisions have beenmade. (b) When the sample is traversed along its long axis anadditional incision is made, with (c), two extra sub-micron surgicalincisions. The arrows in (a) indicate the ablated extracellular matrixsecreted by the cell. Unlike fibroblasts, MDCK cells are devoid of focaladhesions, and cell-substrate bonds anchor the cell to the substrate.With precise sample movement, the isolation of single MDCK cells canbe achieved when the laser traces the exterior contour of the cellmembrane. Laser parameters: pulse duration sub-10 fs; excitationwavelength 800 nm; oscillator repetition rate 80 MHz; pulse energy forsurgery 5 nJ/pulse; 0.95 NA 100× air microscope objective. (Reprinted,with permission, from Ref. 1. Copyright 2005 Wiley Periodicals, Inc.).

In the work conducted in our lab, Kohli et al.1

demonstrated the surgical dissection of biologicalmaterial using fs laser pulses. With a pulse energyof 5 nJ/pulse (sub-10 fs, 800 nm, 80 MHz)several dissection cuts were made in the plasmamembrane of live Madin-Darby Canine Kidney(MDCK) cells. 1 Figure 1 shows membrane surgery onthe mammalian cell, where the arrows represent theablated extracellular matrix. Post-laser surgery, thecell maintained normal morphology without evidenceof membrane re-orientation, cell collapse or blebformation, Figure 1. It was hypothesized that theabsence of cell disassociation after laser surgery waslikely as a result of coalescence of the dissected upperand lower plasma membrane.1 Further work by thisgroup used fs laser pulses as a novel tool for single cellisolation. Figure 2 depicts nanosurgical isolation of alive Chinese hamster fibroblast cell. Two fibroblastcells are shown initially tethered together by a focaladhesion. As shown in Figure 2(b–d), by scanning thecells along the dissection interface relative to the laser



(a) (b)

(d)(c)

FIGURE 2 | Live video observation of nanosurgical isolation of livefibroblast cells. (a) The arrows depict two fibroblast cells (V79-4), with atethered width of ∼1 µm. The dashed line represents the dissectioninterface the sample traverses relative to the fs laser spot. (b) Theapplication of focused laser pulses (1013 W/cm2/pulse), indicated by thearrow, nanosurgically ablates the focal adhesions adjoining the twofibroblast cells. (c) The surgery precisely isolates and detaches the cell,indicated by the dotted box. This is achieved without morphologicallycompromising the cell. (d) An in-focus image, depicted in the dottedbox, shows a live isolated folded fibroblast cell. Laser parameters: pulseduration sub-10 fs; excitation wavelength 800 nm; oscillator repetitionrate 80 MHz; pulse energy for surgery 5 nJ/pulse; 0.95 NA 100× airmicroscope objective. (Reprinted, with permission, from Ref. 1.Copyright 2005 Wiley Periodicals, Inc.).

spot, removal of the focal adhesion resulted in theisolation of a single fibroblast cell from its neighbor.1

Shen et al.6 ablated fluorescently labeled actincytoskeleton in fixed 3T3 fibroblast cells using apulse energy ranging from 1.5 to 3 nJ/pulse (100-fs, 1 kHz). By translating the cells relative to thelaser pulse, nanometer scale channels were made inthe cytoskeleton. The diameter of the ablated channelswas found to decrease as the pulse energy was lowered,with a threshold for actin ablation of 1.5 nJ/pulse.6

Confirmation that the cytoskeleton was ablated andnot photobleached was obtained by restaining the cellsafter laser irradiation.

Heisterkamp et al.9 also dissected fluorescentlylabeled actin in both fixed and live bovine capillaryendothelial cells using a pulse energy ranging from1.8 to 4.4 nJ/pulse (100-fs, 1 kHz). Similar to the

(a)

(b)

(5 µm)

0 s 2 s 5 s

FIGURE 3 | (a) Fluorescence microscope image of GFP-labeledmicrotubule network in an endothelial cell. (b) Time-lapse sequenceshowing rapid retraction of microtubule due to depolymerization. Thecross hair shows the position targeted by the laser; the arrows show theretracting ends of the microtubule. Laser parameters: pulse duration200–250-fs; excitation wavelength 790 nm; oscillator repetition rate 80MHz; pulse energy for dissection of GFP-labeled microtubule 1.5–1.8nJ/pulse; 1.4 NA oil immersion microscope objective. (Reprinted, withpermission, from Ref. 9. Copyright 2005 Optical Society of America).

observations of Shen et al., the width of the cuts wasfound to decrease as the pulse energy was reduced,from 600 to 240 nm for 4.4 and 2.2 nJ/pulse,respectively.9 Figure 3 shows the dissection of a singlegreen fluorescent protein (GFP) tagged microtubule ina live cell irradiated with 1000 pulses at a pulse energyof 1.5 nJ/pulse. It was found that within 2 s, the micro-tubule retracted because of depolymerization9 (Figure3(b)) (arrows). In addition to the ablation of actin,the authors also dissected the nucleus in a fixedendothelial cell. Transmission electron microscopyanalysis revealed that the nucleus could be ablatedwith a pulse energy as low as 1.8 nJ/pulse.9

Recently, Sacconi et al.8 demonstrated nano-surgery (100-fs, 80 MHz) of GFP-labeled micro-tubules in fission yeast cells. Individual mitotic spin-dles in anaphase B were irradiated with an averagelaser power of 4 mW (0.05 nJ/pulse) for 150 ms.8



(a)

(b)

(c)

(d)

0 255

FIGURE 4 | Emboss-filtered transmission (a),(c) andpseudo-color-coded autofluorescence images (b),(d) of chloroplasts inthe epidermal cell of E. densa before (a),(b) and 8 s after (c),(d) selectiveknock-out of part of a specific chloroplast (lightning symbol) with 800nm near infrared (NIR) fs laser pulses at a mean power of 30–50 mW inthe presence of the cell-impermeate fluorescent dye propidium iodide(PI). Note the active movement of the chloroplasts in the corticalcytoplasmic region of the target cell (arrowheads) as well as in theadjacent cells (arrows) after nanoprocessing. No PI fluorescence isdiscernible in the cytoplasm of the cells, indicating that the cells remainviable. Distinct cytoplasmic streaming in the cortical region of the cellswas invariably present even after 30 min. Scale bar = 50 µm. The insetpseudo-color-coded bar represents a pixel intensity profile between 0and 255 units. Laser parameters: pulse duration 170-fs; excitationwavelength 720 nm; oscillator repetition rate 80 MHz; pulse energy fornanodissection of chlorplast 0.38–0.63 nJ/pulse; beam dwell time forirradiation 13 ms. (Reprinted, with permission, from Ref. 5. Copyright2002 Blackwell Publishing).

Following nanosurgery, the spindles were bent andbroken into segments. In a similar experiment, theauthors determined the optimal average laser powerfor nanosurgery of cytoplasmic microtubules in inter-phase cells. For an average laser power below 2 mW(0.03 nJ/pulse),8 it was shown that the shape andlength of the microtubule remained unchanged afternanosurgery. However, above 2 mW, disassociationof the microtubules was observed, with the frequencyof breakage increasing with higher average laser pow-ers (4–8 mW). The optimal average laser power formicrotubule disassociation was found to be 4 mW,

20 µm(a) (b)

(c)

FIGURE 5 | Ablation of a single mitochondrion in a living cell.(a) Fluorescence microscopic image showing multiple mitochondriabefore fs laser irradiation. Target mitochondrion (marked by arrow)(b) before (c) after laser ablation with 2 nJ pulses. Laser parameters:pulse duration 100-fs; oscillator repetition rate 1 kHz; excitationwavelength 800 nm; pulse energy for the irradiation of mitochrondrion2 nJ/pulse; 1.4 NA oil immersion microscope objective. (Reprinted, withpermission, from Ref. 6. Copyright 2005 Tech Science Press).

yielding a 75% disassociation efficiency and 100%cell survival.8

Konig et al.10 demonstrated the nanodissectionof chromosomes in the nucleus of live Chinese hamsterovary cells, using a pulse energy of 0.4 nJ/pulse andan exposure time of 500 µs. At this pulse energy, chro-mosomes could be ablated without disruption to thenuclear envelope. However, at 0.63 nJ/pulse, dissec-tion of the chromosomes was accompanied by damageto the nuclear envelope and the outer cell membrane.10

In a study by Tirlapur and Konig,5 the authorsused fs laser pulses (170-fs, 720 nm, 80 MHz) for thenanodissection of plant cell walls and the partial andcomplete removal of chloroplasts in Elodea densa.Using an average laser power ranging from 30 to50 mW (0.38–0.63 nJ/pulse), lesions with a widthof< 400 nm were made in the plant cell wall.5 Figure4 depicts transmission and autofluorescence images ofthe chloroplast in E. densa before and after removalof this organelle. Figure 4(a, b) shows several chloro-plasts in the epidermal cell of the plant (arrows), wherethe lightning symbol identifies the chloroplast chosenfor removal. Using an average laser power of 30 mWand a beam dwell time of 13 ms, portions of thetargeted chloroplast were removed, (Figure 4(c, d)),without compromising the functionality or integrityof adjacent chloroplasts.5 To verify that adjacent plas-tids remained functional, phase-contrast transmissionmicroscopy was used to examine the cytoplasmicmovement of the organelles in the cortical region.Normal cytoplasmic movement was observed in allnonirradiated chloroplasts.5 To address whether sub-cellular removal of the plastids altered cell viability, anexamination of the presence of propidium iodide (PI)



0.60.50.40.4

0.5

0.6

0.7

0.8

0.9

1.0

0.3Time (s)

V/V

equi

l

0.20.10.0

(a) (b) (c)

FIGURE 6 | The response of a micropatterned MDCK cell suspended in 1.0 M sucrose when permeabilized by fs laser pulses. (a) MDCK cell beforepermeabilization. The arrow depicts the focused fs laser spot. Only one cell was chosen for permeabilization, demonstrating the precision of theprocess. (b) MDCK cell after permeabilization. The cell has increased in cellular size towards equilibrium volume. The arrow in (b) illustrates thepermeabilized cell. (c) Volumetric response of a micropatterned MDCK cell in a 0.2 M cryoprotectant sucrose solution. Initially the cell is in ashrunken state. Upon laser permeabilization, the cell quickly swells to equilibrium volume. The value of Vequil was taken to be the equilibriumvolume as measured using Image J analysis software. Scale bar in (a) and (b) is 40 µm. Laser parameters: pulse duration sub-10 fs; excitationwavelength 800 nm; oscillator repetition rate 80 MHz; pulse energy for permeabilization 3 nJ/pulse; 0.95 NA 100× air microscope objective.(Reprinted, with permission, from Ref. 2. Copyright 2005 Wiley Periodicals, Inc.).

Before

(a) (b) (c) (d) (e)

After 6 h 12 h 24 h

FIGURE 7 | Fs laser axotomy in Caenorhabditis elegans worms using 100 pulses of low energy (40 njJ) and short duration (200 fs) and a repetitionrate of 1 kHz. Fluorescence images of axons labeled with green fluorescent protein before, immediately after, and in the hours following axotomy.Arrow indicates point of severance. Scale bar, 5 µm. Laser parameters: pulse duration 200-fs; oscillator repetition rate 1 kHz; pulse energy for laseraxotomy 40 nJ/pulse; 1.4 NA 64× oil immersion microscope objective. (Reprinted, with permission, from Ref. 11. Copyright 2004 Nature PublishingGroup).

in the cytoplasm of the irradiated cells was performed.Using transmission and two-photon fluorescence, noaccumulation of PI was observed in the targeted cells,indicating that the cells remained viable.5

Shen et al.6 targeted a single fluorescentlylabeled mitochondrion in bovine adrenal capillaryendothelial cells with fs laser pulses (100-fs, 1 kHz).The purpose of the study was to elucidate theconnective properties of mitochondria to determinewhether this organelle forms a continuous networkor represents an independent structural unit. Afterstationary irradiation of the mitochondrion with afew hundred pulses at an energy of 2 nJ/pulse,surgical removal of the mitochondrion from theendothelial cell was accomplished without affectingneighboring mitochondria6 (Figure 5). Since only the

targeted mitochondrion was structurally damaged andremoved, (Figure 5(b,c) arrow), the authors claimedthat the absence of adjacent mitochondrial damageprovided direct evidence that this organelle exists asan independent unit.6

Watanabe et al.7 removed a mitochondrion ina human carcinoma cell line, HeLa, using a pulseenergy ranging from 2 to 7 nJ/pulse (150-fs, 1 kHz)and a beam dwell time of 250 ms. At 7 nJ/pulse,the removal of the mitochondrion was accompaniedby plasma membrane disruption indicated by cellularPI uptake.7 However, membrane disruption was notobserved following mitochondrial ablation for pulseenergies between 2 and 4 nJ/pulse, indicating thatthe cells remained viable after laser irradiation. Theauthors used confocal imaging to confirm that the



Hemorrhage

Intravascularclot

Extravasation

FIGURE 8 | Schematic of the three different vascularlesions that are produced by varying the energy and numberof laser pulses. At high energies, photodisruption produceshemorrhages, in which the target vessel is ruptured, bloodinvades the brain tissue, and a mass of red blood cells(RBCs) form a hemorrhagic core. At low energies, the targetvessel remains intact, but transiently leaks blood plasmaand RBCs forming an extravasation. Multiple pulses at lowenergy lead to thrombosis that can completely occlude thetarget vessel, forming an intravascular clot. Scale bars,50 µm. Laser parameters: pulse duration 100-fs; oscillatorrepetition rate 1 kHz; pulse energy for inducing vascularlesions 0.03–0.50 µJ/pulse; 0.8 NA 40× water immersionmicroscope objective. (Reprinted, with permission, fromRef. 13. Copyright 2006 Nature Publishing Group).

targeted mitochondrion was ablated, and that itsabsence (as detected by fluorescence microscopy) wasnot because of its diffusion out of the focal plane (bycytoplasmic streaming).7 Similar to the observationsof Shen et al.,6 Watanabe reported that neighboringmitochondria remained intact.

In a study by Tirlapur and Konig,14 the authorsemployed fs laser pulses (800 nm, 80 MHz) tointroduce DNA into Chinese hamster ovarian (CHO)cells and rat-kangaroo kidney epithelial cells (PtK2).With an average laser power ranging from 50 to 100mW (0.625–1.25 nJ/pulse), the cell membrane wasdisrupted in the presence of DNA plasmid vectorpEGFP-N1 encoding enhanced green fluorescentprotein (GFP). Disruption of the cell membrane after16 ms of irradiation resulted in the introduction ofDNA.14 Expression of the DNA construct was verifiedby two-photon fluorescence imaging.14

In similar work to that of Tirlapur and Konig,Stevenson et al.17 transfected CHO cells with fs laserpulses (120-fs, 800 nm, 80 MHz) using a pulse energyand beam dwell time ranging from 50 to 225 mW and10 to 250 ms, respectively. Contrary to the claim of100% transfection efficiency by Tirlapur and Konig,14

Stevenson measured an average transfection rate of50 ± 10% in 4000 laser-treated CHO cells.17 Thenonfluorescent dye, trypan blue, was used to confirmcell membrane viability.

In our lab, fs laser pulses were used to disrupt thecell plasma membrane for the purpose of introducingforeign substances into the cytoplasm of live MDCKcells. Kohli et al.2 showed that when fs laser pulseswere localized to the cell membrane, transient porescould be formed, exposing the extracellular spaceto the intracellular environment. Using a pulseenergy of 3 nJ/pulse (sub-10 fs, 800 nm, 80 MHz)cryoprotective disaccharides were cytoplasmically

Early fast phase (EFP)

0

(a) (b)

40 80Distance from targeted region (µm)

120

ControlPhoto-ablared

160

0.8

0.4

1.2

Fast phase (FP)

0 40 80Distance from targeted region (µm)

120

ControlPhoto-ablared

160

0.8

0.4Rat

e of

CF

I (µm

/min

)

Rat

e of

CF

I (µm

/min

) 1.2

FIGURE 9 | (a) CFI rate measured 1 min after ablation (during early fast phase (EFP), solid line) at different distances from the ablated region, andcomparison with control embryos (squares, N = 3). (b) Same measurement 15 min after photoablation (during fast phase (FP)). Scale bar: 20 µm.Laser parameters: pulse duration 130-fs; excitation wavelength 830 nm; oscillator repetition rate 76 MHz; pulse energy for embryo manipulation0.6–4 nJ/pulse; 0.9 NA water immersion microscope objective. (Reprinted, with permission, from Ref. 16. Copyright 2005 The National Academy ofScience of the USA).



(b)

(d)

(f)

Stage 5

Stage 5

Stage 5

Stage 7

Stage 7

Stage 7

Stage 7

(a)

Stage 7

Stage 7

(e)

(c)

FIGURE 10 | Multiphoton ablation allows quantified modulation ofspecific morphogenetic movements (a) and (b), control; (c) and (d),middorsal ablation, (e) and (f), postdorsal ablation). (a) Development ofan intact sGMCA embryo. Green represents images recorded at theequator. Red represents images recorded ≈ 20 µm under the surface.(c) Development of a sGMCA embryo after a 100 × 40 µm middorsalablation, resulting in disrupted lateral cell movements and no cephalicfurrow formation (gray arrowheads). (e) Development of a sGMCAembryo after 100 × 40 µm postdorsal ablation resulting in disruptedlateral cell movements only. (b),(d), and (f) Corresponding velocimetricanalysis for the same embryos at stage 7. Each experiment wasreproduced on five different embryos and gave similar results. Scale bar:100 µm. Black scale arrow, 5 µm/min. Laser parameters: pulse duration130-fs; excitation wavelength 830 nm; oscillator repetition rate 76 MHz;pulse energy for embryo manipulation 0.6–4 nJ/pulse; 0.9 NA waterimmersion microscope objective. (Reprinted, with permission, fromRef. 16. Copyright 2005 The National Academy of Sciences of the USA).

introduced through laser-induced transient pores forbiopreservation applications.2 When MDCK cells

were suspended in 1.0 m cryoprotective sucrose,the cells were found to swell to a new equilibriumvolume following transient pore formation as aresult of an intracellular accumulation of sucrose andwater (Figure 6(a, b)). The authors2 used volumetricanalyses to determine the longevity of the transientpore created in the cell membrane. Figure 6(c) depictsthe kinetics of the cell following permeabilization.Since the volumetric change was found to plateauwithin 200 ms (Figure 6(c)) it was hypothesized thatthis time corresponded to the lifetime of the laser-induced transient pore.2 The transient lifetime ofthe pore in varying molar concentrations was alsodetermined by the authors. A survival analysiswas performed using a membrane integrity assayconsisting of ethidium bromide and Syto 13. Inaddition, transport equations were used to estimatethe delivered intracellular concentration as a functionof the extracellular osmolarity.

In a recent study by Yanik et al.11 the authorsused fs laser pulses to perform laser axotomy of D-motor neurons in L4 larval-stage C. elegans. Severingof the D-neurons induced muscle contractions pre-venting backward locomotion. Figure 7 depicts timelapse images of the laser axotomy, where individualneurons were cut at the mid-body position using 100laser pulses at a pulse energy of 40 nJ/pulse (200-fs,1 kHz)11 (Figure 7(b)). The authors observed that thesevered neurons retracted following axotomy (Figure7(c, d)). Analysis of neuron regeneration revealed that54% of the laser-treated neurons (52 axons in 11worms) re-grew within 12–24 h (Figure 7(e)). A testof the motor neuron function showed that backwardlocomotion resumed within 24 h, with a functionalityapproaching that of wild type C. elegans.11

Chung et al.12 used fs laser pulses to studythe role of AFD neurons in C. elegans. Using apulse energy of 3 nJ/pulse (100-fs, 800 nm, 1kHz), individual dendrites within a bundle of amphiddendrites were severed.12 Severing of the dendriteswas accomplished without visible damage to adjacentdendrites. Similar to the observations made by Yaniket al., the ablated dendrites were found to retractfollowing ablation, with a retraction distance of5µm.12 To determine whether the dendrites re-grewafter laser dissection, the authors severed fluorescentlylabeled PHA and PHB sensory neurons and monitoredneuron growth for 24 h. In over 50 C. elegans, noneof the sensory neurons repaired, indicating that thecuts were permanent.12

Nishimura et al.13 used fs laser pulses tophotodisrupt microvessels in the parenchyma ofrat brains using a range of average laser powersfrom 0.03 to 0.5 µJ/pulse (100-fs, 1 kHz). Figure 8



FIGURE 11 | Middorsal ablationmodulates morphogenetic movements at theanterior pole, which are correlated with twistexpression. (a)–(f) Sequence of developmentat the anterior pole of control andphotoablated sGMCA embryos, showing thedisrupted movements of SP cells aftermiddorsal ablation. Approximate time afterthe onset of gastrulation is indicated inminutes (inverted contrast images). Blackscale arrow: 2 µm/min. Laser parameters:pulse duration 130-fs; excitation wavelength830 nm; oscillator repetition rate 76 MHz;pulse energy for embryo manipulation 0.6–4nJ/pulse; 0.9 NA water immersion microscopeobjective. (Reprinted, with permission, fromRef. 16. Copyright 2005 The NationalAcademy of Sciences of the USA).

(a) 8 min (stage 6)

Control

20 min (stage 6) (d) 8 min (stage 6)

(b) (e)

(c) (f)

CompressionExpansion

Photo-ablated

20 min (stage 6)

2PE

F

sequ

ence

Vel

ocim

etric

anal

ysis

Def

orm

atio

n an

alys

is

(b) (c)

B Y

(d)(a)

FIGURE 12 | (a) When sub-10 fs laser pulses were focused through the chorion, laser-induced transient pores were created at theblastomere–yolk interface or in individual blastomeres of zebrafish embryos. Transient pores were formed only at the focus, leaving the chorion layerundamaged. The pores were used to introduce foreign material into the embryonic cells. Three-dimensional movement of the laser focal spot allowedfor precise targeting of any location on or within the embryo. (b) An early 8-cell stage embryo was targeted for pore formation at the blastomere–yolkinterface (arrow). (c) A sub-micron (∼800 nm) transient pore was created at the interface dividing the blastomeres (B) and yolk (Y) (arrow). Thesub-micron pore is obscured by a laser-generated cavitation bubble. An energy of 3 nJ/pulse at a gated pulse train of 200–300 ms was used to formthe pore. (d) Depicts the developing embryo at 64/128-cell stage 45–60 min post-fs laser poration. Scale bar for (b),(d) and (c) represents 200 µm and5 µm, respectively. Laser parameters: pulse duration sub-10 fs; excitation wavelength 800 nm; oscillator repetition rate 80 MHz; pulse energy forembryo manipulation 3 nJ/pulse; beam dwell time 200–300 ms; 1.0 NA 60× water immersion microscope objective. (Reprinted, with permission,from Ref. 3. Copyright 2007 Wiley Periodicals, Inc.).

depicts three different vascular lesions that wereproduced using varying pulse energies and pulsedensities. These included laser induced hemorrhaging,extravasation and intravascular clot formation13

(Figure 8). At relatively high laser pulse energiesabove the threshold for extravasation of fluorescentlylabeled blood plasma (0.03 µJ), hemorrhage of theblood plasma and red blood cells from the targetedvessel was observed.13 Lowering the pulse energyresulted in more controlled vascular lesions. However,both extravasation of intact vessels with continuedblood flow and clot formation resulting in completevessel obstruction were observed.13 The authors also

measured the changes in adjacent and downstreamblood flow in the obstructed vessel following laser-induced clot formation.

Embryo NanosurgeryWhile the application of fs laser pulses has beenextensively used in the nanosurgery of simple cells,the study of complex multicompartmental biologicalsystems such as embryos remains a challenge. Theability to noninvasively manipulate the intracellularenvironment of individual embryonic cells has



(b)(a)

FIGURE 13 | Brightfield and fluorescence images (a),(b) of adechorionated embryo at 16-cell stage that was fs laser porated in theblastomere cells for introducing FITC. Direct poration of the cellsresulted in a stronger FITC signal than poration at the blastomere–yolkinterface. The concentration of FITC used was 0.02–0.03 mg/ml. Theembryo was porated using an energy of 0.5–0.6 nJ/pulse at a gatedpulse train of 200–500 ms. Scale bar represents 200 µm. Laserparameters: pulse duration sub-10 fs; excitation wavelength 800 nm;oscillator repetition rate 80 MHz; pulse energy for embryo manipulation0.5–0.6 nJ/pulse; beam dwell time 200–500 ms; 1.0 NA 60× waterimmersion microscope objective. (Reprinted, with permission, fromRef. 3. Copyright 2007 Wiley Periodicals, Inc.).

important implications for future developments inmedical and developmental biology.

In a recent study by Supatto et al.,16 theauthors used fs laser pulses to induce morphogeneticmovements in Drosophila embryos. The authorsdemonstrated that laser nanosurgery (130-fs, 830nm, 76 MHz) below the vitelline membrane couldbe achieved within the developing embryo withoutdisturbing cytoskeletal dynamics adjacent to theablated area.16 A series of dissection line cuts, 100 ×40 µm, were made 5–15 µm beneath the vitellinemembrane with varying pulse energies and pulsenumber densities (number of incident fs laser pulsesper area). The dissections were characterized byobserving endogenous fluorescence emission usingtwo-photon excited fluorescence.16 For pulse densitiesand pulse energies below 105 µm−2 and 4 nJ/pulse,respectively, no endogenous fluorescence emissionwas observed.16 However, with increasing pulsedensity, fluorescence emission was observed along thedissection cut with microexplosions in the perinuclearregion of the cytoplasm. With a pulse densityapproaching 106 µm−2, large cavitation bubbles inexcess of 5–6 µm in diameter were observed.16

Further work examined the in vivo modulationof cellularization front invagination (CFI) in embryosablated by fs laser pulses.16 Figure 9 depicts therate of CFI in control and laser ablated Drosophilaembryos.16 As shown in Figure 9(a), an increase inthe rate of CFI was observed for the early fastphase one min after laser ablation, relative to the

controls. Fifteen minutes after ablation, no differencein the CFI rate for the fast phase was observed,16

Figure 9(b). Despite the increase in CFI for the earlyfast phase, the authors reported that kymographanalysis showed that cellularization completed incells adjacent to the laser ablated area. In additionto monitoring changes in the cellularization rate,the in vivo morphogenetic movements in embryostargeted at dorsal ablation sites were quantified.16

Figure 10 depicts the morphogenetic movementsand velocimetric analysis of the ablated and controlembryos. In Figure 10(a, b), both cephalic furrowformation and lateral cell motions were clearlyobserved in control embryos (arrows). However,middorsal dissection, (Figure 10(c, d)), resulted inno cephalic furrow formation and the disruption oflateral cell movements.16 Ablation of the postdorsalregion was found to affect the lateral cell movementsonly, (Figure 10(e, f)), with furrow formationoccurring normally. The authors speculated that themechanism responsible for the modulation likely arosefrom the disruption of the motor region associatedwith the ablated area.16 Further investigationsexamined cell movement and twist expression afterlaser ablation.16 When embryos were targeted atthe middorsal site, the stomodeal primordium (SP)cell motions were affected16 (Figure 11). In controlembryos, expansion and compression of the SP cellswere readily observed (Figure 11(a–c)) however, thismovement was suppressed by middorsal ablationand the loss of furrow closure16 (Figure 11(d–f)).While ventral cells at the anterior pole in controlembryos were found to have forward movement, SPcells exhibited more backward directed motion inablated embryos (Figure 11(d–f)). At ablation sitesother than middorsal, no significant changes in thetwist expression were observed.16

Recently, in our lab Kohli et al.3 used fslaser pulses (sub-10 fs, 800 nm, 80 MHz) tointroduce exogenous material into early stage cellsof live developing embryos. The animal modelsystem chosen was the zebrafish (Danio rerio), anaquatic vertebrate organism that is genetically anddevelopmentally closer to humans than the commoninvertebrate Drosophila melanogaster.29,30 Presently,zebrafish are used in the study of genetics, drugmonitoring, human disease, cardiac function andblood disorders.31–37 Figure 12(a) depicts the methodused for targeting individual embryonic cells ofthe developing zebrafish.3 In both chorionated anddechorionated embryos, the authors focused fs laserpulses with a pulse energy ranging from 0.56 to 2.7nJ/pulse to a location near the blastomere-yolk (B–Y)interface for transient pore formation3 (Figure 12(a)).



FIGURE 14 | Fluorescence images 30 minpost-fs laser poration of developed (a) 32-cell,(d) 512/1K-cell, and (g) 128/256-cell stagechorionated embryos that were targetedbeyond the chorion for introducingperivitelline-FITC into the blastomeres. Thebrightfield embryos were laser porated at(b) 8-cell, (e) 128-cell, and (h) 32–64-cell stage.Uptake of perivitelline-FITC is evident asfluorescence in (c), (f), and (i), where individualblastomere cells are clearly visible. The arrowsin (c), (f), and (i) point to the location wheretransient pores were formed. Concentration ofFITC used was 0.02–0.03 mg/ml. All embryoswere fs laser porated using an energy of 3nJ/pulse at a gated pulse train of 200–300 ms.Embryos were dechorionated to eliminate theinterfering fluorescence signal originating fromthe perivitelline space. Scale bar represents200 µm. Laser parameters: pulse durationsub-10 fs; excitation wavelength 800 nm;oscillator repetition rate 80 MHz; pulse energyfor embryo manipulation 3 nJ/pulse; beamdwell time 200–300 ms; 1.0 NA 60× waterimmersion microscope objective. (Reprinted,with permission, from Ref. 3. Copyright 2007Wiley Periodicals, Inc.).

(b)

(e)

(h)

(a)

(d)

(g) (i)

(f)

(c)

The authors first addressed whether the applied laserpulses were deleterious to the development of theembryo. Fs nanosurgery was performed at the B–Yinterface (Figure 12(b, c) arrows), in early cleavageto early blastula (2-cell to 128-cell) stage embryosusing a pulse energy of 2.7 nJ/pulse and a beamdwell time of 200–500 ms. Figure 12(d) shows normaldevelopment of a laser treated 8-cell stage embryo,which has developed to 128-cell stage 45–60 min post-laser surgery.3 Other targeted embryos were found todevelop normally as compared to control embryos. Todetermine if laser surgery at the B–Y interface or onindividual blastomere cells lead to the formation of atransient pore, the authors suspended early cleavage toearly blastula (2-cell to 128-cell) stage dechorionatedembryos in a fluorescent reporter molecule, fluoresceinisothiocyanate (FITC), and examined fluorescenceuptake in the embryonic cells. Figure 13 depicts FITCfluorescence in the blastomere cells of a 16-cell stageembryo, confirming transient pore formation andexogenous material delivery. In 39 targeted embryos,a FITC loading efficiency of 87% was reported.3 Itwas conjectured that the distribution of the fluorescentprobe to adjacent blastomeres likely occurred throughblastomere bridges or gap junctions, depending on thedevelopmental stage.38–40

Figure 12(a) depicts the chorion, a proteinaceousmembrane surrounding the developing embryo, whichprovides protection from the environment. To showthat the applied laser pulses could still be focusedfor pore formation at the B–Y interface, the authorsfocused fs laser pulses beyond the structure of thechorion as shown in Figure 12(a). Early cleavage toearly blastula (2–cell to 128-cell) stage chorionatedembryos were suspended in the presence of FITC, andthe fluorescent probe was allowed diffuse into theperivitelline space (FITC was previously shown to beimpermeable to the blastomeres). Targeting the B–Yinterface with a pulse energy of 2.7 nJ/pulse and abeam dwell time of 200–300 ms, the authors foundthat they could introduce perivitelline FITC into theembryonic cells without compromising the structureof the chorion.3 This is evident in Figure 14(c, f,i), where after proteolytic digestion of the chorion(to remove the interfering fluorescent signal from theperivitelline region), fluorescence was observed in theindividual blastomere cells. In a total of 27 laser-treated embryos, a FITC loading efficiency of 78%was found.3

Exogenous material delivery was not limited toFITC, as the authors also demonstrated the delivery ofconjugated quantum dots and plasmid DNA. Quan-tum dots and DNA are important materials that have



(b)(a)

(c) (d)

(e) (f)

FIGURE 15 | (a) An early 2-cell stage dechorionated embryo thatwas fs laser porated in the blastomere cells for introducingStreptavidin-conjugated quantum dots. The quantum dots freelydiffused throughout the cells and remained fluorescent as the embryodeveloped. (b) Depicts the same embryo developed past germ ring.Concentration of the Streptavidin-conjugated quantum dot solution was0.3 µM. An energy of 1.5–2 nJ/pulse at a gated pulse train of 200–500ms was used, and 3–4 pores were created in each cell for introducingthe quantum dots. Fluorescence and brightfield images of 24 hpf larvae,(c)–(f), expressing the sCMV-EGFP construct that was introduceddirectly into the blastomere cells of an early to mid cleavage stage(2-cell to 8/16-cell) dechorionated embryo. (c),(d) Expression isobserved along the gut, as well as in the floor plate, and somites.(e),(f) Expression of sCMV-EGFP is seen throughout the tail of the larva,where expressing cells are those near the floor plate and somites.Concentration of the construct used was 170 µg/ml. An energy of0.5–0.6 nJ/pulse at a gated pulse train of 200–500 ms was used, with3–4 pores created per cell for introducing the plasmid (maximum of 2,2, 4, and 8 cells targeted per 2–, 4–, 8–, and 16-cell stage respectively).Laser parameters: pulse duration sub-10 fs; excitation wavelength 800nm; oscillator repetition rate 80 MHz; pulse energy for embryomanipulation 0.5–2 nJ/pulse; beam dwell time 200–500 ms; 1.0 NA60× water immersion microscope objective. (Reprinted, withpermission, from Ref. 3. Copyright 2007 Wiley Periodicals, Inc.).

potential uses for cell fate mapping and the develop-ment of stable transgenic fish lines.41,42 Using a pulse

energy of 1.5–2 nJ/pulse, streptavidin-conjugatedquantum dots (targeted near the B–Y inter-face) were introduced into 2-cell stage dechorionatedembryos3 (Figure 15(a, b)). Quantum dot fluorescencewas observed in the early embryonic cells (Figure 15a)while in later cell stages up to germ ring (Figure 15b)the quantum dots were visibly dispersed throughoutthe blastomeres.3 To determine if the applied laserpulses constituted a valid alternative method forDNA delivery, the authors laser transfected early tomid cleavage (2-cell to 8/16-cell) stage dechorionatedembryos in the presence of a circular plasmid, sCMV-EGFP, with a pulse energy of 0.56 nJ/pulse and a beamdwell time ranging from 200 to 500 ms.3 Expressionof the plasmid construct was observed in a 24-hpost-fertilization (hpf) larva (Figure 15(c, e)) withthe expression seen along the yolk-extension, floorplate, somites and tail cells of the larva.3 In over 45chorionated and dechorionated laser treated embryos,survival approached 90%, with embryo morphologyand behavior similar to the control sample.3

Kohli and Elezzabi15 further examined thedevelopment of zebrafish embryos after laser surgerywith the fs laser (sub-10 fs, 800 nm, 80 MHz). Usinga pulse energy of 0.56 nJ/pulse and a beam dwelltime of 100 ms, individual chorionated blastomerecells were surgically ablated at the early 2-cell stagein over 40 embryos.15 Each blastomere cell wasablated at three different locations with a total laserexposure time of 300 ms per targeted site. The authorsreared the embryos to 2 and 7 days post-fertilization(dpf) and used light microscopy (LM) and scanningelectron microscopy (SEM) to determine if the appliedlaser pulses induced morphological changes in thedevelopment of the embryos. Under LM, the bodyplans of control and laser-manipulated embryos wereinspected with emphasis on the development of thebody axis.15 Short-term survival (before 2 dpf), asdetermined by the above analysis, revealed a survivalpercentage of 93%.15 Viable larvae showed no differ-ences in developmental or hatching rates as comparedto the controls. SEM imaging showed key develop-mental structures including the caudal fin, dorsal fin,yolk sac extension, yolk sac and the olfactory pit tobe morphologically similar in laser-manipulated andcontrol larvae.15 As the laser-treated larvae aged, thepectoral fin buds lifted away from the yolk sac anddeveloped into mature pectoral fins along the lateralextent of the zebrafish body. This morphologicaldevelopment was consistent with control larvae.15

The authors concluded that no short-term effects ofthe laser on the development were observed.

While no short-term effects were observed,Kohli commented that the laser’s effect on embryonic



OP

OP

OC

OC

NC

NC

(a)

(c)

(d)

(b)

VF

VF

PF

PF

OV

OV

E

E

PM

PM

FIGURE 16 | Key developmental structures in a laser-manipulated and a control larva reared to 7 dpf. (a) Inverted whole body image of alaser-manipulated larva at 7 dpf. Structures indicated are the ventral fin (VF), notochord (NC), pectoral fin (PF), otic capsule (OC), otic vesicle (OV),eye (E), olfactory pit (OP), and the protruding mouth (PM). (b) Inverted whole body image of a control larva at 7 dpf. Similar developmental structuresobserved in (a) were also seen in (b). (c) Kinocilia projecting from the lateral crista of a laser-manipulated and (d) control larva. Scale bar for(a),(b) represent 200 µm and (c),(d) 1 µm, respectively. Laser parameters: pulse duration sub-10 fs; excitation wavelength 800 nm; oscillatorrepetition rate 80 MHz; pulse energy for embryo surgery 0.56 nJ/pulse; total beam dwell for laser-manipulation 300 ms; 1.0 NA 60× water immersionmicroscope objective.

development may not become apparent until laterdevelopmental stages.15 Using SEM, the authorsexamined control and laser-manipulated larvae rearedto 7 dpf. Developmental structures inspected includedthe protruding mouth, olfactory pit, pectoral fin,eye, otic capsule, otic vesicle, ventral fin, notochord,posterior forebrain, and dorsal midbrain.15 Nodifferences in the placement or patterning of thesestructures were observed between the samples. Figure16(a, b) depict mosaics of a laser-manipulated anda control larva with the developmental structuresindicated as mentioned above. High magnificationimages revealed that the olfactory pit in the controland laser-manipulated larvae was surrounded byepidermal cells, with the pit rims covered by longkinocilia.15 In the lumen of the eara crista was foundon the lateral wall with kinocilia projecting from thesensory epithelial, as seen in Figure 16 (c, d).15 Theauthors found no differences in neuromast patterning,with projecting kinocilia that were distributed alongthe lateral line of the zebrafish body.15 In controllarvae, neuromasts were found anterior to theolfactory pit, at the outer rim of the otic capsule,anterior to the diencephalon and adjacent to bothsides of the optic tectum (dorsal midbrain) anddiencephalons (posterior anterior-forebrain).15 Theneuromast patterning in laser-manipulated larvae wasfound to be identical to that seen in the controls. It wasconcluded that no long-term developmental effectscould be observed, thereby making the application offs laser pulses an important noninvasive tool for thestudy of live embryos. Further work is being conducted

by the authors to determine if any physiologicalresponses are induced following fs laser nanosurgery.

CONCLUSIONThe noninvasive nature of fs laser pulses and theirability to target subcellular sites with high spatialresolution are the major features that have made theseultrafast lasers an attractive tool for the study of livecells and embryos. This review article has exploreddevelopments in cell and embryo nanosurgery; eachreported study identified unique applications tobiology. These include the knockdown of subcellularorganelles, the opto-injection of exogenous materials,and functional analyses of laser-induced morpho-genetic and morphological changes in embryonicdevelopment. However, despite these advances, thefull potential of fs laser pulses has yet to be realized.Like the laser itself, the fs laser is a tool being devel-oped without having fully elucidated all of its potentialuses. In addition, the application of fs laser pulses isin some ways developing as a technique ‘in search of abiological problem’. It is through continued researchthat we will uncover novel applications that willundoubtedly benefit many biological disciplines. Weenvision that in the near future fs laser pulses will beused in the study of cell fate mapping to identify howindividual cells contribute to the overall embryonicdevelopment of organisms. It will be possible to cry-opreserve embryos with low solute permeabilities bydelivering impermeable and permeable cryoprotectiveagents. The generation of genetically modified



organisms will be possible as a result of the inter-ference of delivered exogenous nucleic acids, as willbe the potential development of stable transgenic celllines. The collaboration of physicists, engineers, and

cell and developmental biologists will enable the pur-suit of such applications, with the fs laser providing anew prospective for understanding essential biologicalsystems.

REFERENCES

1. Kohli V, Acker JP, Elezzabi AY. Cell nanosurgery usingultrashort (femtosecond) laser pulses: applications tomembrane surgery and cell isolation. Lasers Surg Med2005, 37:227–230.

2. Kohli V, Acker JP, Elezzabi AY. Reversible permeabi-lization using high-intensity femtosecond laser pulses:applications to biopreservation. Biotechnol Bioeng2005, 92(7):889–899.

3. Kohli V, Robles V, Cancela ML, Acker JP, Wask-iewicz AJ, et al. An alternative method for deliveringexogenous material into developing zebrafish embryos.Biotechnol Bioeng 2007, 98(6):1230–1241.

4. Konig K, Riemann I, Fritzsche W. Nanodissection ofhuman chromosomes with near-infrared femtosecondlaser pulses. Opt Lett 2001, 26(11):819–821.

5. Tirlapur UK, Konig K. Femtosecond near-infrared laserpulses as a versatile non-invasive tool for intra-tissuenanoprocessing in plants without compromising viabil-ity. Plant J 2002, 31(3):365–374.

6. Shen N, Datta D, Schaffer CB, LeDuc P, Ingber DE,et al. Ablation of cytoskeletal filaments and mitochon-dria in live cells using femtosecond laser nanoscissor.MCB 2005, 2(1):17–25.

7. Watanabe W, Arakawa N, Matsunaga S, Higashi T,Fukui K, et al. Femtosecond laser disruption of sub-cellular organelles in a living cell. Opt Express 2004,12:4203–4213.

8. Sacconi L, Tolic-Norrelykke IM, Antolini R,Pavone FS. Combined intracellular three-dimensionalimaging and selective nanosurgery by a non-linear microscope. J Biomed Opt 2005,10(1):):014002-1–014002-5.

9. Heisterkamp A, Maxwell IZ, Mazur E, Under-wood JM, Nickerson JA, et al. Pulse energy dependenceof subcellular dissection by femtosecond laser pulses.Opt Express 2005, 13(10):3690–3696.

10. Konig K, Riemann I, Fischer P, Halbhuber K-J. Intra-cellular nanosurgery with near infrared femtosecondlaser pulses. Cell Mol Biol 1999, 45(2):195–201.

11. Yanik MF, Cinar H, Cinar HN, Chisholm AD, Jin Y,et al. Functional regeneration after laser axotomy.Nature 2004, 432:822.

12. Chung SH, Clark DA, Gabel CV, Mazur E,Samuel AD. The role of the AFD neuron in C. ele-gans thermotaxis analyzed using femtosecond laserablation. BMC Neurosci 2006, 7:30.

13. Nishimura N, Schaffer CB, Friedman B, Tsai PS,Lyden PD, et al. Targeted insult to subsurface corticalblood vessel using ultrashort laser pulses: three modelsof stroke. Nat Methods 2006, 3(2):99–108.

14. Tirlapur UK, Konig K. Targeted transfection by fem-tosecond laser. Nature 2002, 418:290.

15. Kohli V, Elezzabi AY. Laser surgery of zebrafish (Daniorerio) embryos using femtosecond laser pulses: optimalparameters for exogenous material delivery, and thelaser’s effect on short- and long-term development.BMC Biotechnol 2008, 8(7):1–20.

16. Supatto W, Debarre D, Moulia B, Brouzes E, MartinJ-L, et al. In vivo modulation of morphogenetic move-ments in Drosophila embryos with femtosecond laserpulses. PNAS 2005, 102(4):1047–1052.

17. Stevenson D, Agate B, Tsampoula X, Fischer P,Brown CTA, et al. Femtosecond optical transfectionof cells: viability and efficiency. Opt Express 2006,14(16):7125–7133.

18. Oraevsky AA, Silva LBD, Rubenchik AM, Feit MD,Glinsky ME, et al. Plasma mediated ablation of bio-logical tissues with nanosecond-to-femtosecond laserpulses: relative role of linear and nonlinear absorption.IEEE J Quantum Elect 1996, 2(4):801–809.

19. Vogel A, Noack J, Huttman G, Paltauf G. Mechanismsof femtosecond laser nanosurgery of cells and tissues.Appl Phys B 2005, 81:1015–1047.

20. Niemz M. Laser-tissue Interactions: Fundamentals andApplications. Springer, ed. Berlin, Heidelberg, NewYork: Springer-Verlag; 2002.

21. Loesel FH, Fischer JP, Gotz MH, Horvath C, Juhasz T,et al. Non-thermal ablation of neural tissue with fem-tosecond laser pulses. Appl Phys B 1998, 66:121–128.

22. Noack J, Vogel A. Laser-induced plasma formationin water at nanosecond to femtosecond time scales:calculation of thresholds, absorption coefficients,and energy density. IEEE J Quantum Elect 1999,33(8):1156–1167.



23. Schaffer CB, Brodeur A, Mazur E. Laser-inducedbreakdown and damage in bulk transparent materi-als induced by tightly focused femtosecond laser pulses.Meas Sci Technol 2001, 12:1784–1794.

24. Kaiser A, Rethfeld B, Vicanek M, Simon G. Micro-scopic processes in dielectrics under irradiation bysubpicosecond laser pulses. Phys Rev B 2000,61(17):11437–11450.

25. Sacchi CA. Laser-induced electric breakdown in water.J Opt Soc Am B 1990, 8(2):337–345.

26. Boyle JW, Ghormley JA, Hochanadel CJ, Riley JF. Pro-duction of hydrated electrons by flash photolysis ofliquid water. J Phys Chem 1969, 73(9):2886–2890.

27. Arnold CL, Heisterkamp A, Ertmer W, LubatschowskiH. Computational model for nonlinear plasma for-mation in high NA micromachining of transpar-ent materials and biological cells. Opt Lett 2007,15(16):10303–10317.

28. Venugopalan V, Guerra A III, Nahen K, Vogel A. Roleof laser-induced plasma formation in pulsed cellularmicrosurgery and micromanipulation. Phys Rev Lett2002, 88(7):078103-1–078103-4.

29. Vogel G. Zebrafish earns its stripes in genetic screens.Science 2000, 288(5469):1160–1161.

30. van der Sar AM, Appelmelk BJ, Vandenbroucke-Grauls CMJE, Bitter W. A star with stripes: Zebrafishas an infection model. Trends Microbiol 2004,12(10):451–457.

31. Barut BA, Zon LI. Realizing the potential of zebrafishas a model for human disease. Physiol Genomics 2000,2:49–51.

32. Warren KS, Wu JC, Pinet F, Fishman MC. The geneticbasis of cardiac function: Dissection by zebrafish(Danio rerio) screens. Philos Trans R Soc Lond B2000, 355:939–944.

33. Dooley K, Zon LI. Zebrafish: A model system for thestudy of human disease. Curr Opin Genet Dev 2000,10:252–256.

34. Hill AJ, Teraoka H, Heideman W, Paterson RE.Review: zebrafish as a model vetebrate for investigatingchemical toxicity. Toxicol Sci 2005, 86(1):6–19.

35. Jagadeeswaran P, Sheehan JP. Analysis of blood coag-ulation in the zebrafish. Blood Cells Mol Dis 1999,25(15):239–249.

36. Nasevicius A, Ekker SC. Effective targeted gene ‘knock-down’ in zebrafish. Nat Genet 2000, 26:216–220.

37. Thisse C, Zon LI. Organogenesis-heart and blood for-mation from the zebrafish point of view. Science 2002,295:457–462.

38. Kimmel CB, Law RD. Cell lineage of zebrafish blas-tomeres, I. Cleavage pattern and cytoplasmic bridgesbetween cells. Dev Biol 1985, 108(1):78–85.

39. Kimmel CB, Ballard WW, Kimmel SR, Ulmann B,Schilling TF. Stages of embryonic development of thezebrafish. Dev Dyn 1995, 203:253–310.

40. Weinberg ES. Analysis of early development in thezebrafish embryo. In: Hennig W. ed., Early EmbryonicDevelopment of Animals. Berlin, Heidelberg: Springer-Verlag; 1992, 91–150.

41. Rieger S, Kulkarni RP, Darcy D, Fraser SE, Koster RW.Quantum dots are powerful multipurpose vital label-ing agents in zebrafish embryos. Dev Dyn 2005,234:670–681.

42. Higashijima S-I, Okamoto H, Ueno N, Hotta Y,Eguchi G. High-frequency generation of transgeniczebrafish which reliably express GFP in whole mus-cles or the whole body by using promoters of zebrafishorigin. Dev Biol 1997, 192:289–299.

FURTHER READING

Zipfel WR, Williams RM, Webb WW. Nonlinear magic: Multiphoton microscopy in the biosciences. NatBiotechnol 2003 21(11) 1369–1377.Debarre D, Supatto W, Pena A-M, Fabre A, Tordjmann T, et al. Imaging lipid bodies in cells and tissues usingthird-harmonic generation microscopy. Nat Methods 2006 3(1) 47–53.Watanabe W, Shimada T, Matsunaga S, Kurihara D, Fukui K, et al. Single-organelle tracking by two-photonconversion. Opt Express 2007 15(5) 2490–2498.


Nanosurgery in cancer therapy.Nano-imaging and surgery.


Advanced Review

Pharmacokineticsof nanomaterials: an overviewof carbon nanotubes, fullerenesand quantum dotsJim E. Riviere∗

A full understanding of the pharmacokinetic parameters describing nanomaterialdisposition in the body would greatly facilitate development of a firm foundationupon which risk assessment could be based. This review focuses on the dispositionof carbon based fullerenes and nanotubes, as well as quantum dots (QD) afterparenteral administration to primarily rodents. The common theme across allparticle types is that a major determinant of nanomaterial disposition is the degreeof interaction with the reticuloendothelial (RE) cell system. Small water-solubleparticles evading this system may be excreted by the kidney. Larger particlesand those with the proper surface charge may get targeted to RE cells in the liver,spleen and other organs. Most nanomaterial kinetics are characterized by relativelyshort blood half-lives reflecting tissue extraction and not by clearance from thebody. In fact, another common attribute to nanomaterial kinetics is retention ofparticles in the body. Finally, unlike many small organic drugs, nanomaterials maypreferentially be trafficked in the body via the lymphatic system that has obviousimmunological implications. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 26–34

The technological and biomedical advancementsinherent to the application of nanomaterials are

becoming increasingly evident. In the field of medicalapplications, a thorough understanding of their phar-macokinetic properties is crucial for their safe andefficacious application. From the perspective of char-acterizing potential adverse effects, these data are alsorequired for quantitative risk assessments to definethe toxicology of the material. Pharmacokinetics isdefined as the science of quantifying the rate and extentof the absorption, distribution, metabolism and elim-ination (ADME) of chemicals and drugs in the bodyusing mathematical modeling approaches. The aimof pharmacokinetics is to relate drug dose or chem-ical exposure to biological effect. Pharmacokineticshas developed distinct sets of models and parame-ters which have been useful to describe and predict

∗Correspondence to: Jim E. Riviere, Center for Chemical Tox-icology Research and Pharmacokinetics, North Carolina StateUniversity, Raleigh, NC 27606, USA. E-mail: Jim [email protected]

Center for Chemical Toxicology Research and Pharmacokinetics,North Carolina State University, Raleigh, NC 27606, USA

DOI: 10.1002/wnan.024

drug and chemical disposition, and also have specificmeanings and interpretations from the perspective ofthe regulatory agencies assigned to either approvedrugs or conduct occupational or environmental riskassessments of potentially adverse materials.

Most pharmacokinetic models and approacheshave been defined and applied to the dispositionof small organic drugs, chemicals or metals. Thephysiochemical processes driving disposition of smallmolecules in the body are related to diffusion andtransport or metabolism (biotransformation) by enzy-matic processes. Most of the models are defined bytransport through the blood. Are these processes andpathways also relevant to the disposition of largernanomaterials, some with unique physical properties?Pharmacokinetic analyses have been applied to pro-teins that in some cases are of dimensions equivalentto the manufactured nanomaterials. Viruses and lipidparticles circulating in the body also are of similarsizes. However, not known are the effects on phys-iologic disposition of the unique quantum physicalproperties that characterize manufactured nanoscaledmaterials of 1–100nm, which confer on them unique


WIREs Nanomedicine and Nanobiotechnology Pharmacokinetics of nanomaterials

physical attributes (fluorescence, electrical conductiv-ity, resistance to biological degradation etc.).

Physiological processes interacting with simi-lar sized compounds include many cellular recog-nition, opsinisation, adhesion, and uptake processesincluding phagocytosis, as well as lymphatic trans-port, that are not usually relevant to the dispositionof most small organic molecules. A central ques-tion of nanopharmacokinetics is whether these pro-cesses, which would be expected to apply to smallorganic molecules and biodegradable nanomaterials,also apply to the manufactured particles made frominert and degradation resistant novel materials suchas fullerenes, carbon nanotubes, QD or metallic parti-cles? What nanomaterial properties and characteristicsbest correlate to their disposition defined by the phar-macokinetic parameters of bioavailability, clearance,volume of distribution, or half-life? Do known nano-material issues of aggregation, binding, and particlesadministered with varying sizes modify interpretationof pharmacokinetic studies?

The focus of this manuscript is to review theliterature on the pharmacokinetics of nanomaterialsconducted to date and assess whether any overarchingfindings can be defined. Studies which would becharacterized as ‘classic’ for drugs or chemicals havenot been conducted for most of the manufacturednanomaterials. Therefore, an assessment will bemade of the basic principles of experimental designwhich should be followed if robust and precisepharmacokinetic parameters are to be obtained forthese materials. This review is restricted to carbonbased nanomaterials and QD since they have beenrelatively well studied. Focus is also on systemicpharmacokinetics after parenteral administration.Inhalational, dermal and oral routes of administrationhave not been included.

OVERVIEW OF PHYSIOLOGICALDISPOSITION

The goal of pharmacokinetic and absorption,distribution, metabolism and elimination studies isto assess the fraction of dose of a chemical, drug,or nanomaterial administered to an animal that isabsorbed into the systemic circulation and subse-quently distributed to tissues or is excreted from thebody. Figure 1 depicts the stages involved in ADMEprocesses as a function of route of administration,as well as subsequent elimination. Pharmacokineticmodels quantitate the rate and extent of a com-pound’s sojourn through the body determined by thesephysiological pathways by mathematically analyzing

blood (or plasma) concentrations over time. Fromsuch data, a number of descriptive parameters can bedetermined, including:

• Volume of Distribution (Vd): Proportion of drugdistributed in the body used to relate adminis-tered dose to observed blood concentrations. Vd= Concentration / dose.

• Clearance (Cl): Efficiency of removal of a com-pound from the blood. This can be determinedfor the whole body (ClB) or for specific organssuch as the kidney and liver.

• Half-life (T1/2): Time it takes for 50% ofa process (e.g., absorption, elimination) tobe completed. This often calculated parameterassumes first-order, linear behavior. T1/2 =[0.693 Vd] / Cl.

• Mean Residence Time (MRT): A parametersimilar to T1/2 denoting the average time acompound remains in the body.

• Bioavailability (F): The fraction of a doseabsorbed into the body and available for systemicdistribution. This is calculated from blood asthe ratio of the area under the curve (AUC)in blood seen after a specific route dividedby that seen after intravenous administration.Alternatively, other methods of assessing totalbody burden may be used in lieu of measuringblood concentrations.

The parameters above are often determined usingspecific mathematical models that account for morecomplex patterns of distribution or eliminationfrom the body (e.g., multidistribution compartmentsresulting in multiple Vds; non-linear pathways ofelimination, stochastic modeling approaches). Othermodeling approaches go into more realistic detail[e.g., physiological-based pharmacokinetic—(PBPK)models] but require more elaborate and data intensiveexperimental designs. A PBPK model describes thedisposition of a chemical based on a mathematicalmodel that mirrors the physiological structure ofthe body, with compartments linked by tissue bloodflow. Such models easily incorporate in vitro dataand define target doses of materials. Pharmacokinetictexts should be consulted for more detail on modelconstruction and data analysis.1–3

Most pharmacokinetic studies are conductedfrom the perspective of nanomaterial behavior in thesystemic circulation assessed by analysis of plasmaor blood concentrations. It is generally assumed thata compound must reach the systemic circulation in


Advanced Review www.wiley.com/wires/nanomed

Bile

Biotransformation

Urine

Tubularreabsorption

Secretions

Tissue binding

Tissues

Binding

Epidermis-dermis

TopicalLYMPHPO

IV

IN

Site of action

Site of toxicity Systemiccirculation

Portalcirculation

Proteinbinding

IM SC

Gut Feces

e.g. Bronchial prostate salivary

Liver

Liver

KidneyFIGURE 1 | Stages involved in absorption,distribution, metabolism and elimination (ADME).

order to be distributed to tissues and be availablefor excretion by the liver or kidney. When such anapproach is applied to drugs, a decay in blood con-centrations over time usually translates into excretionof the compound from the body via the urine, fecesor in the inhaled breath. Under these circumstances,T1/2 relates the length of time a drug is availablein the body. If extensive tissue distribution occurs,as is seen for may lipophilic contaminants such aschlorinated hydrocarbons, decay from blood may bevery slow as is seen above by the dependency of T1/2

on Vd. Elimination from the body may be inefficient(low Cl) further increasing T1/2. In fact, eliminationmay only occur by metabolism of the chemical to amore hydrophilic, and thus easily excretable chemicalform. However, for nanomaterials, decay in bloodconcentrations may be related to the compoundmovement into tissues where further excretion doesnot occur [e.g., trapped in reticuloendothelial (RE)system, bound to tissue proteins, postdistributionalaggregation]. In these cases, blood half-life mayparadoxically be relatively short despite the pro-longed body persistence. Nanomaterials may also betransported in the body via the lymphatic system, aphenomenon which complicates their pharmacoki-netic analysis based on blood sampling and alsoexposes lymphoid tissue to higher concentrations thanwould be seen secondary to distribution from blood.These idiosyncrasies of nanomaterials compared tomost drugs or small molecule xenobiotics requirecaution in using classic interpretations of the meaningof basic ADME parameters.

REVIEW OF THE LITERATURE

There have been a number of recent studiesthat have attempted to define basic dispositionand pharmacokinetic parameters for a number ofnanomaterials including fullerenes, carbon nanotubesand QD. A review of these studies presents areasonable perspective on the nature of researchconducted to date and a relatively consistent pictureof basic concepts that seem to apply to nanomaterialADME. All of these studies have been conducted inlaboratory animal species, primarily rodents. Humandata are not available.

Carbon-Based Nanomaterials—Fullerenes(C60) and NanotubesAn interesting literature has developed on thedisposition of carbon nanomaterials in laboratory ani-mals. Much of this work has resulted from attemptsto derivatize fullerenes or carbon nanomaterials toserve as vectors for delivery of specific therapeuticligands (e.g., antiviral compounds, anticancer drugs).This complicates data interpretation since nanoma-terial controls often do not exist due to solubilityissues, making it difficult to separate nanomaterialeffects from that of the attached ligand. There arelimited studies that independently alter size, surfacecharge or solubility to assess their effect on ADMEparameters. A number of different vehicles are oftenused to dose different particles, making vehicle effectsdifficult to identify. Finally, the nanomaterial is oftenmonitored using further molecular modifications tofacilitate assaying or tracking (e.g., fluorescent or



radioactive label) which may further alter ADMEproperties.

Rajagopalan et al.4 conducted a pharmacoki-netic study in Sprague–Dawley rats using 15 mg/kgMSAD-C60, a water-soluble C60 derivative withantiviral properties. Terminal blood T1/2 was approx-imately 7 h, MRT 11 h, and the Vd was 2 L/kg indicat-ing extensive distribution. There was no evidence ofurinary excretion and C60 was 99% bound to proteinin plasma. There was a great variability in disposition,not reflected in the T1/2 data, with 2/5 rats havingtwo-fold differences in Cl and Vd parameters. Theauthors attributed this to the extensive protein bind-ing. This observation underlines the weakness in usingT1/2 alone as the sole pharmacokinetic descriptor,since T1/2 is physiologically confounded by oppositechanges in Vd and Cl (recall above that T1/2 ∼= Vd/Cl).

Yamago et al.5 studied a 14C labeled(trimethylenemethane (TMM) derived) lipophilic yetwater-soluble C60 after IV and oral administration(dosed in ethanol/PEG/albumin vehicle) to mice andFischer rats. In both species, oral absorption wasminimal. After IV administration, only 5% of thecompound was excreted from the body, all by thefecal route. Most radiolabel was retained in the liverafter 30 h, primarily in Kupffer and perisinusoidalfat cells, not hepatocytes. Some C60 derivatives werealso located in the spleen, kidney and importantlythe brain. From 30 to 160 h, label in organs slowlydecreased without observable excretion from thebody. However, redistribution to skeletal muscle andhair was observed.

Qingnuan et al.6 reported that tissue distributionof 99mTc labeled C60(OH)x in mice and rabbitsafter IV dosing occurred primarily to the kidney,bone, spleen and liver with slow elimination fromthe body occurring after 48 h, except for bonewhich accumulated label. T1/2 in blood was 17 hin mice. Cagle et al.7 studied the biodistributionof endohedral metallofullerenes (166HoC82(OH)n) inmice and reported relatively rapid clearance fromblood over a few hours, bone accumulation, andliver localization with slow elimination. As with theC60(OH)x studies above, total body Cl was low withonly 20% of intact compound being excreted by 5days, a retention time much longer than when controlmetal chelates alone were administered. After 5 days,blood concentrations did not appreciably decay. Incontrast to other work, these C60(OH)x were excretedin urine, which may be a function of increased watersolubility compared to more lipophilic derivatives.

Bullard-Dillard et al.8 using 14C labeled particlesdosed to Sprague–Dawley rats showed longer persis-tence in the circulation of water-soluble ammonium

salt derivative C60 compared to very rapid clearanceof C60. Both primarily targeted the liver. For water-soluble C60 derivatives, Cl was low and material wasseen in spleen, lung and muscle, as well as the liver.Significantly, there was no radioactivity detected inurine and feces collected every 24 h for 5 days. Sim-ilarly, Qiang et al.9 studied 0.5 mg/kg intravenousC60(OH)24 disposition in mice and demonstratedorgan accumulation, as well as only 50% excretionafter 3 days, with over 50% excreted in feces andonly 4% in urine. Bone showed a pattern of continuedaccumulation. In tumor-bearing mice, tumor/muscleand tumor/blood averaged 2–6 fold depending on thetumor type. Studies of nanoparticle size distributionsin aqueous versus protein media showed different pat-terns indicating particle agglomeration in biologicalenvironments, a phenomenon which could impact thepattern of tissue kinetics seen.

Some workers have investigated the biodistribu-tion of carbon nanotubes in laboratory animals. Mostof these have used indirect measures of concentration(infrared),10 posicron emmission tomography,11 orparticles functionalized with specific tracers.12 Meth-ods using such modifications rely on radiolabel orfluorescent tags to both remain attached to the nano-material throughout its sojourn through the body, aswell as not to impart any different physiochemicalproperties that would alter ADME. The lack of a sen-sitive analytical assay for these materials has hinderedthis work. Studies have also been done in small num-bers of laboratory animals often with insufficient timepoints for proper pharmacokinetic analyses. Nan-otube length has not been rigorously controlled, nor insome cases even determined, in the studies reported todate. Nevertheless, some interesting patterns emerge.

Singh et al.12 studied various functionalized sin-gle walled nanotubes (SWNT), as well as multiwallednanotubes (MWNT) administered intravenously toBALB mice. This work demonstrated urinary excre-tion and accumulation in muscle, skin and kidneyfor neutral and positively charged SWNT, as well asMWNT. In contrast to other studies, extensive bodyaccumulation was not seen and clear urinary excretionwas evident. Guo et al.13 using labeled MWNT dosedintraperitoneally to mice showed a blood T1/2 of 5.5 h.Material was retained in the stomach and haircoat andwas excreted primarily in the feces. Cherukuri10 study-ing rabbits showed SWNT (1 × 300 nm) dispersedin pluronic F108 preferentially accumulated in liverafter 24 h. IR spectra suggested that the SWNT disso-ciated from the pluronic dosing media and interactedwith proteins, making the SWNT actually a studyin nanotube-protein disposition. This is consistentwith the findings of Dutta et al.14 who demonstrated



that protein adsorption onto SWNT altered biologi-cal interactions, although binding to albumin couldbe prevented by pretreatment with Pluronic F127surfactant. Liu et al.11 studied phospholipid-coatedSWNT in mice using PET. Tissue distribution (liver,spleen) and blood T1/2 (0.5–2 h) were dependentupon phospholipid substituents. Significant body bur-den persisted after a 24 h sacrifice. Finally, Denget al.15 studied the disposition of taurine functional-ized MWNT after intravenous dosing to mice and alsoobserved distribution primarily to liver, lung, spleenand heart, but not brain, stomach, muscle nor bone.As reported for many materials, liver deposition wasprimarily to Kupffer cells. These studies establish apattern of distribution of larger material to tissuesthat compose the RE system in the liver as Kupffercells, spleen, lymph nodes and bone marrow. As willbe seen, this theme is also played out with other typesof nanomaterials.

Quantum DotsThere are systemic disposition studies reported usinginherently fluorescent QD derivatized for medicalimaging. As seen with the carbon nanomaterials, moststudies do not rigorously determine particle concen-trations using analytical techniques as is required inpharmacokinetic studies. However, general patternsof particle distribution can often be assessed. Thesestudies suggest that QDs, tagged with homing pep-tides, can be targeted to specific tissues (e.g., lung,vessels) after intravenous administration of ≈ 10mg/kg to mice. Similar to carbon materials reviewedabove, they accumulate in liver and spleen.16 Coat-ing QD with polyethylene glycol allows particles toescape detection by RE tissues (liver, spleen, lymphnodes). Imaging studies in mice clearly show that QDsurface coatings alter their disposition and pharma-cokinetic properties.17 Plasma T1/2 was less than 12min for amphiphilic poly (acrylic), short chain (750Da) methoxy-PEG or long chain (3400 Da) carboxy-PEG QD, but over an hour for long-chain (5000 Da)methoxy-PEG QD. These coatings determined the pat-tern of in vivo tissue localization, with retention ofsome QDs occurring up to 4 months.

Fischer et al.18 studied the pharmacokinetics ofCdSe/ZnS QD after intravenous administration toSprague–Dawley rats, either coated with bovine serumalbumin [(BSA)-QD; hydrodynamic radius of 80 nm]or bound to mercaptoundecanoic acid crosslinkedto lysine (LM-QD; hydrodynamic radius of 25 nm).Blood clearance of BSA-QD was 1.23 compared to0.59 mL/min-kg for LM-QD. BSA-QD T1/2 was 39min versus 58 min for LM-QD. The Vd for both

was approximately 65 mL/kg. By 90 min, the liverhad accumulated 40 and 99% of LM-QD and BSA-QD, respectively. Electron microscopy located QDprimarily to Kupffer cells within the liver. No QD ofeither form were detected in urine or feces for up to10 days. This study nicely illustrates the uncouplingof blood decay to elimination from the body or tissuedistribution as represented by Vd estimates, as thesepharmacokinetic parameters do not reflect the natureof tissue distribution, nor are sensitive to irreversibletissue binding of substances such as nanomaterials.

Several very recent studies using QD furtherextend these observations. Schipper et al.19 studiedQD with various coatings after intravenous adminis-tration to mice. As seen in other studies, liver was aprimary target of uptake, with smaller amounts goingto spleen, bone and lung. Within the ranges of sizesstudied (12—21 nm), size had no influence on biodis-tribution. There was no evidence of clearance fromthe body through 36 h. Yang et al.20 demonstrateda complete lack of excretion after 28 days for QD705 after IV injection to mice. Although plasma T1/2was short (18.5 h), there was continued redistributionfrom body sites to liver and kidney over 28 days. Asseen with the carbon-based material, decay in bloodis not necessarily associated with clearance from thebody.

As can be appreciated from this research, mostinvestigators have failed to detect excretion fromsystemically dosed QD. Choi et al.21 reported in aseries of studies that QD with zwitterionic or neutralorganic coatings prevented QD adsorption to serumproteins that kept the QD hydrodynamic radius lessthat 5.5 nm. Under this size, QD could be excreted bythe kidney. Larger sizes of QD were not excreted fromthe body. This phenomenon followed what would beexpected based on size cutoffs for renal filtration ofvarious peptides and proteins. Globular proteins withdiameters of 5–6 nm are regularly excreted, whilelarger proteins are not. This work is also consistentwith 5 nm colloidal gold nanoparticle studies byBalogh et al.22 whereby positive charged particleswere excreted by the kidney, whereas negative chargedor neutral small particles and 22 nm particles of allsurface charges were targeted to the liver, spleen orlung presumed secondary to opsinisation and RE cellremoval.

In order to begin exploring the biodistributionkinetics of nanomaterials, our laboratory infusedthree concentrations of PEG-coated or COOH-coated QD621 into an isolated perfused porcine skinflap preparation and modeled arterial-venous (AV)extraction quantified by fluorescent intensity andvalidated using inductive coupled plasma emission



spectroscopy for cadmium.23 Perfusion media wasa Krebs Ringer albumin containing buffered solution.Data was analyzed based on a pharmacokinetic modelpreviously optimized for platinum chemotherapeuticcompound distribution in infused skin, a compoundclass also marked by irreversible tissue binding.24

COOH-QD 621 (negative charge) uptake into skinwas 2–3 fold greater than for PEG-QD621 (neutral).This is consistent with increased tissue distributionof negative charged QD discussed above. However,the data for both QDs were also marked by astatistically significant periodicity (period ≈ 90 min)in vascular uptake, a phenomenon never previouslyseen with infused drugs (cisplatin, carboplatin,lidocaine or testosterone). This finding appearsunique for nanomaterials and suggests that the tissuedistribution of QD may be different than other organicmolecules. In fact, such periodicity in QD tissueextraction from blood is consistent with the tissueredistribution seen for QD in mice,20 as well asan erratic pattern of early QD deposition in skinreported by Ballou et al.17 The charge selectivity fortissue distribution seen in the perfused skin study isimportant as RE cells are not present in this isolatedorgan model.

OTHER NANOMATERIALSIt is beyond the scope of this review to discussbiodistribution studies conducted using nanoparticlescomposed of other materials, as the above studiesusing carbon and QD are representative. However,a few studies are worth noting. Kreyling andco-workers have conducted a number of elegantbiodistribution studies using defined sizes of 192Irradio-labeled iridium nanoparticles after inhalationaland parenteral administration.25,26 These studiessuggest a size-dependent pattern of pulmonarydeposition after inhalation. However, their work alsoclearly demonstrate that a small amount of inhalediridium (< 1%) is translocated to other organs,including the liver, spleen, heart and brain. Similar toother materials discussed above, liver concentrationscontinued to increase and minimal excretion was seenafter parenteral dosing. Smaller particles (15 vs 80nm) had significantly increased tissue translocationafter inhalational exposure.

Many studies have been conducted using phar-maceutical ‘nanoformulations’. The materials areoften polymers of drug or carrier substances pre-viously used in parenteral drug administration butnow formulated as nano-scale material (particles,liposomes, nanocapsules, micelles, dendrimers andnanoplexes).27,28 The bulk of the applications in this

area are formulations (e.g., Nanoedge and NanoCrys-tal platforms for increasing oral delivery) where theextreme surface area to mass ratios of nanomaterialresult in refined control of drug delivery, or nanofor-mulation surfaces improve solubility or permeabil-ity in biological systems. In other cases, nano-sizeddrugs appear to have enhanced and/or targeted tissuebiodistribution and have surface properties designedto reduce RE cell uptake. Some nano-based phar-maceutics are already on the US market, examplesbeing Doxil—doxorubicin HCL liposome injectionand Abraxane—paclitaxel protein-bound particles forinjectable suspension. Manufactured materials suchas 60 nm bioconjugated quantum rods have beendeveloped for receptor-mediated transport across theblood brain barrier.29 The fundamental differencebetween most of these pharmaceutical materials com-pared to the manufactured materials discussed ear-lier is that pharmaceutics are specifically constructedof materials that are biodegradable, and thus not-persistent in the body, Because of this property,excretion ultimately occurs and standard pharma-cokinetic approaches have been used to describe theirdisposition. In contrast, the manufactured nanoma-terials are not biodegradable and thus once withincertain tissue, may accumulate. Finally, the uniquesurface chemistry of these ‘hardened’ materials mayfurther alter disposition or bioactivity.

LYMPHATIC TRANSPORT

Although most ADME and pharmacokinetic studiesfocus on the blood circulatory system, animals andhumans also have another system which trafficscells and large lipophilic molecules and proteinsthroughout the body, the lymphatic system, acomponent of the RE system. The lymphatics havebeen extensively studied relative to their role inabsorption of particulates and protein therapeuticsfor molecules with molecular weights greater than16 KDa.30–32 After absorption in local lymphaticvessels, a compound moves to regional lymph nodesand ultimately re-enters the systemic circulationvia the thoracic duct. An important toxicologicalimplication of this pathway is that all such transportedmaterial has the potential for interaction with theimmune system resident in regional lymph nodes.Nanomaterials, being relatively large and dependingon surface modifications, are ideal candidates forlymphatic transport. In support of this hypothesis,studies have shown that after subcutaneous QD



injection, some nanomaterials end up in draininglymph nodes.33 Solid lipid nanoparticles designedfor magnetic resonance imaging have been shown toenter lymph after duodenal administration to rats.34

In other imaging studies conducted in pigs, intra-dermal injection of 400 pMoles of fluorescent QDtargeted sentinel lymph nodes,35 a finding relevant todermal absorption of even minute fractions of topi-cally applied or orally dosed nanomaterials. In fact,it has been suggested that nanocapsules, ultrafine oilydroplet-coated polymeric drug substances, may beone of the most promising candidates for lymphatictargeting.36 Similarly, QD injected into porcine lungparenchyma was used to map lymphatic drainage andvisualize lymph nodes during surgery.37 This preferen-tial uptake of some nanomaterials by lymph coupledwith the tendency of some materials to also interactwith the RE system makes this aspect of nanomaterialdeposition different than most small organic drugs.

CONCLUSIONSThere are a few conclusions that can be drawn fromthe available literature on nanomaterial biodistribu-tion and kinetics based on study of carbon basedmaterials and QDs. First, most nanomaterials tend toaccumulate in the liver, potentially because of RE celltrapping. However, particles also distribute to othertissues, including the kidney, depending on the surfacecharacteristics and size. The effect of size across dif-ferent nanomaterials has only begun to be evaluated.There is some consensus beginning to develop thatpossibly particles with hydrodynamic radii less than5–6 nm may be eliminated from the kidney. However,if they are larger and have specific surface character-istics (e.g., negative charge), they may interact withthe RE system or become protein bound and not beexcreted in the kidney.

Secondly, all classes of particles also have exten-sive tissue retention, a property of potential toxi-cological significance since tissue accumulation andpersistence in the body may occur.38 This happens in

the face of relatively short blood T1/2 since these par-ticles are effectively being ‘cleared’ into tissue depotsrather than excreted from the body into the urine orfeces as most drugs are. Note that as tissue depositionoccurs, Vd may also increase resulting in prolongationof T1/2. What is the driving force for tissue redistribu-tion between tissues observed in some studies?

A third issue relates to how preferential transportby the lymphatic system affects interpretation of clas-sic pharmacokinetic parameters. What is the relation-ship between persistence of nanomaterial absorbedinto the lymphatic system relative to redistributionback to the central blood circulation?

A fourth issue is the exact state the nanomaterialexists once deposited in a tissue. Do surface coat-ings persist after translocation from the blood? Theassociation of a carbon nanotube with a surfactantsuch as sodium dodecyl sulfate (SDS) is pH dependentas is the actual surface area of a nanotube coveredby the compound.39 For example, what is the natureof a nanomaterial ‘sequestered’ in a lysosome char-acterized by a very acidic pH? Subtle differences inthe physiological milieu of the nanotube (e.g., localion concentration) would also be expected to altersubsequent particle disposition as well as toxicity.

Finally, studies are difficult to directly compare(different species, doses, vehicles, different approachesto functionalizations, lack of common characteriza-tion techniques) both within and across nanomate-rials, making interpretation problematic at best anddefinitely not adequate to begin risk assessment anal-yses for manufactured materials. Longer-term studieswith complete particle characterization before dosingand after tissue deposition are required. Are nano-materials deposited into tissues stable over long timeframes? Studies in nonrodent species with body massand thus physiological time-clocks closer to that ofhumans are required. Finally, there is a need for moreclassic pharmacokinetic studies to be conducted sothat physicochemical parameters across nanomateri-als can be correlated to the parameters of disposition.

REFERENCES

1. Gibaldi M, Perrier D. Pharmacokinetics. 2nd ed. NewYork: Marcel Dekker; 1982.

2. Riviere JE. Comparative Pharmacokinetics: Principles,Techniques, and Applications. Ames, IA: Iowa StatePress; 2003.

3. Ette EI, Williams PJ. Pharmacometrics: the Science ofQuantitative Pharmacology. New York: Wiley; 2007.

4. Rajagopalan P, Wudl F, Schinazi RF, Boudinot FD.Pharmacokinetics of a water-soluble fullerene in rats.Antimicrob Agents Chemother 1996, 40:2262–2265.

5. Yamago S, Tokuyama H, Nakamura E, Kikuchi K,Kananishi S, et al. In vivo biological behavior of awater-miscible fullerene: 14C labeling, absorption,distribution, excretion and acute toxicity. Chem Biol1995, 2:385–389.



6. Qingnuan L, Yan X, Xiaodong Z, Ruili L, Qieqie D,et al. Preparation of 99mTc-C60(OH)x and its biodis-tribution studies. Nucl Med Biol 2002, 29:707–710.

7. Cagle DW, Kennel SJ, Mirzadeh S, Alford JM, Wil-son LJ. In vivo studies of fullerene-based materialsusing endohedral metallofullerene radiotracers. ProcNatl Acad Sci USA 1999, 96:5182–5187.

8. Bullard-Dillard R, Creek KE, Scrivens WA, Tour JM.Tissue sites of uptake of 14C-labeled C60. Bioorg Chem1996, 24:376–385.

9. Qiang Z, Sun H, Wang H, Xie Q, Liu Y, et al. Biodis-tribution and tumor uptake of C60(OH)x in mice.J Nanopart Res 2006, 8:53–63.

10. Cherukuri P, Gannon CJ, Leeuw TK, Schmidt HK,Smalley RE, et al. Mammalian pharmacokineticsof carbon nanotubes using intrinsic near-infraredflurorescence. Proc Natl Acad Sci USA 2006,103:18882–18886.

11. Liu Z, Cai W, He L, Nakayama N, Chen K, et al. Invivo biodistribution and highly efficient tumour tar-geting of carbon nanotubes in mice. Nat Nanotechnol2006, 2:47–52.

12. Singh R, Pantarotto D, Lacerda L, Pastorin G,Klumpp C, et al. Tissue biodistribution and bloodclearance rates of intravenously administered carbonnanotubes radiotracers. Proc Natl Acad Sci USA 2006,103:3357–3362.

13. Guo J, Zhang X, Li Q, Li W. Diodistribution of func-tionalized multiwall carbon nanotubes in mice. NuclMed Biol 2007, 34:579–583.

14. Dutta D, Sundaram SK, Teeguarded JG, Riley BJ,Fifield LS, et al. Adsorbed proteins influence the biolog-ical activity and molecular targeting of nanomaterials.Toxicol Sci 2007, 100:303–315.

15. Deng X, Jia G, Wang H, Sun H, Wang X, et al.Translocation and fate of multi-walled carbon nan-otubes in vivo. Carbon 2007, 45:1419–1424.

16. Akerman ME, Chan WCW, Laakkonen P, Bhatia SN,Ruoslahti E. Nanocrystal targeting in vivo. Proc NatlAcad Sci USA 2002, 99:12617–12621.

17. Ballou B, Lagerholm BC, Ernst LA, Bruchez MP, Wag-goner AS. Noninvasive imaging of quantum dots inmice. Bioconjug Chem 2004, 15:79–86.

18. Fischer HC, Liu L, Pang KS, Chan CW. Pharmacoki-netics of nanoscale quantum dots: in vivo distribution,sequestration and clearance in the rat. Adv Funct Mater2006, 16:1299–1305.

19. Schipper ML, Cheng Z, Lee ZC, Bentolila LA, Iyer G,et al. MicroPET-based biodistribution of quantum dotsin living mice. J Nucl Med 2007, 48:1511–1518.

20. Yang RSH, Chang LW, Wu JP, Tsai MH, Wang HJ,et al. Persistent tissue kinetics and redistribution ofnanoparticles, quantum dot 705, in mice. EnvironHealth Perspect 2007, 115:1339–1343.

21. Choi HS, Liu W, Misra P, Tanaka E, Zimmer JP, et al.Renal clearance of quantum dots. Nat Biotechnol 2007,25:1165–1170.

22. Balogh L, Nigavekar SS, Nair BM, Lesniak W,Zhang C, et al. Significant effect of size on the in vivobiodistribution of gold composite nanodevices in mousetumor models. Nanomedicine 2007, 3:281–296.

23. Lee HA, Imran M, Monteiro-Riviere NA, Colvin VL,Wu W, et al. Biodistribution of quantum dot nanopar-ticles in perfused skin: evidence of coating dependencyand periodicity in arterial extraction. Nano Lett 2007,7:2865–2870.

24. Williams PL, Riviere JE. Definition of a physiologicpharmacokinetic model of cutaneous drug distributionusing the isolated perfused porcine skin flap (IPPSF).J Pharm Sci 1989, 78:550–555.

25. Kreyling WG, Semmler M, Erbe F, Mayer P, Take-naka S, et al. Translocation of ultrafine insoluble irid-ium particles from lung epithelium to extrapulmonaryorgans is size dependent but very low. J Toxicol EnvironHealth A 2002, 65:1513–1530.

26. Kreyling WG, Semmler-Behnke M, Moller W. Healthimplications of nanoparticles. J Nanopart Res 2006,8:543–562.

27. Devalapally H, Chakilam A, Amiji M. Role of nan-otechnology in pharmaceutical product development.J Pharm Sci 2007, 96:2547–2565.

28. Lehr CM. Nanoscaled carriers: nanomedicine for theimproved delivery of drugs across biological barriers.Drug Deliv Technol 2007, 7:34–39.

29. Xu G, Yong KT, Roy I, Mahajan SD, Ding H, et al.Bioconjugated quantum rods as targeted probes forefficient transmigration across an in vitro blood-brainbarrier. Bioconjug Chem 2008, 19:1179–1185.

30. Porter CJH, Edwards GA, Charman SA. Lymphatictransport of proteins after s.c. injection: implicationsof animal model selection. Adv Drug Deliv Rev 2001,50:157–171.

31. McLennan DN, Porter CJH, Edwards GA, Brumm M,Martin SW, et al. Pharmacokinetic model to describethe lymphatic absorption of r-metHu-Leptin aftersubcutaneous injection to sheep. Pharm Res 2003,20:1156–1162.

32. McLennan DN, Porter CJH, Edwards GA, Martin SW,Heatherington AC, et al. Lymphatic absorption is theprimary contributor to the systemic availability ofepoetin alfa following subcutaneous administration tosheep. J Pharmacol Exp Ther 2005, 313:345–351.

33. Gopee NV, Roberts DW, Webb P, Cozart CR, Siito-nen PH, et al. Migration of intradermally injectedquantum dots to sentinel organs in mice. Toxicol Sci2007, 98:249–257.



34. Peira E, Marzola P, Podio V, Aime S, Sbarbati A, et al.In vitro and in vivo study of solid lipid nanoparti-cles loaded with superparamagnetic iron oxide. J DrugTarget 2003, 11:19–24.

35. Kim S, Lim YT, Soltesz EG, De Grand AM, Lee J, et al.Near-infrared fluorescent type II quantum dots forsentinel lymph node mapping. Nat Biotechnol 2004,22:93–97.

36. Nishioka Y, Yoshino H. Lymphatic targeting withnanoparticulate system. Adv Drug Deliv Rev 2001,47:55–64.

37. Soltesz EG, Kim S, Laurence RG, DeGrand AM,Parungo CP, et al. Intraoperative sentinel lymph nodemapping of the lung using near-infrared fluorescentquantum dots. Ann Thorac Surg 2005, 79:269–277.

38. Riviere JE, Tran L. Pharmacokinetics of nanomaterials.In: Monteiro-Riviere NA, Tran L, eds. Nanotoxicol-ogy. New York: Informa; 2007, 127–152.

39. Ke PC, Qiao R. Carbon nanomaterials in biologicalsystems. J Phys Condens Matter 2007, 19:373101, 25pages.


Toxicology of nanomaterials.Human health implications of nanomaterial exposure.Characterization of nanomaterials for toxicity assessment.


Advanced Review

Catalyst-functionalizednanomaterialsYi Lu∗ and Juewen Liu1

With rapid development in both nanotechnology and biotechnology, it is nowpossible to combine these two exciting fields to modulate the physical properties ofnanomaterials with the molecular recognition and catalytic functional propertiesof biomolecules. Such research efforts have resulted in a larger number ofsensors that can detect a broad range of analytes ranging from metal ions,small molecules, and nucleic acids down to proteins. These sensors will findimportant applications in nanomedicine. In this article, the design of sensorswith four classes of nanomaterials (metallic, semiconductor, magnetic, and carbonnanotube nanoparticles) is reviewed. Metallic nanoparticles possess distance-dependent optical properties and are useful for designing colorimetric sensors.Semiconductor nanoparticles or quantum dots (QDs) appear to be superioralternatives to traditional organic fluorophores in many aspects, such as broadexcitation range, narrow emission peaks, and high photo stability. QD sensorsbased on either energy transfer or charge transfer are summarized. Furthermore,magnetic nanoparticles are shown to be useful as smart magnetic resonanceimaging (MRI) contrast agents. Finally, some carbon nanotubes show near-IRemission properties, and thus, are potentially useful for in vivo sensing. Sensorsbased on either tuning the emission intensity or wavelength are discussed. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 35–46

With the recent focus of research on nanoparticlesynthesis, functionalization, characterization,

and application, many physical and chemical proper-ties of these materials have been extensively explored,such as optical, magnetic, electronic, thermal, andcatalytic properties.1,2 The ability to modulate theseproperties in response to external chemical envi-ronment allows us to use nanomaterials as sensorcomponents.1–6 By definition, nanoparticles are inthe scale of 1–100 nanometers, which is compara-ble to many important biological macromolecules,such as proteins and nucleic acids. It is known thatthese biopolymers possess highly specific molecularrecognition abilities. For example, a protein enzymecatalyzes the turnover of only certain substrates, whilea piece of single-stranded DNA can bind its comple-mentary strand specifically. Therefore, conjugation ofthese biopolymers to nanoparticles may transduce spe-cific molecular recognition and catalytic properties of

∗Correspondence to: Yi Lu, Department of Chemistry, Universityof Illinois at Urbana, Champaign, IL, USA. E-mail: [email protected] of Chemistry, University of Illinois at Urbana,Champaign, IL, USA

DOI: 10.1002/wnan.021

proteins and nucleic acids into the change of physicalproperties of nanoparticles, and one major applica-tion of these catalyst-functionalized nanomaterials issensing. Because of their small sizes, these sensorscould be used as probes inside a cell. Currently, sincethe toxicity of nanomaterials has not been fully estab-lished, only in vitro applications of these sensors, suchas dipstick tests of serum samples, are reviewed here,which may serve as a basis for their future biomedicalapplications in the human body.

Many biopolymers can be used as molecu-lar recognition elements in sensor design, such asantibodies, enzymes, DNA, and RNA. This chapterfocuses on molecules with catalytic properties includ-ing protein and nucleic acid enzymes. For sensingapplications, sensitivity could be increased with cat-alytic turnovers of enzymes. In the enzyme world,protein has long been the only player. This situationwas changed in the early 1980s with the discov-ery of catalytic RNAs or ribozymes.7,8 In 1994, thefirst catalytic DNA (also known as deoxyribozyme orDNAzyme) molecule was isolated.9–15 Compared toRNA, DNA has much higher stability and it is alsorelatively cost-effective to chemically synthesize DNA.



Many DNAzyme-based sensors have been developedrecently.16 Biopolymer-based sensors with other prop-erties, such as ligand binding, will also be discussed.This chapter will be divided into several sections andeach section focuses on a particular class of nanoma-terials, including metallic, semiconductor, magnetic,and carbon nanotube nanoparticles.

CATALYST-FUNCTIONALIZEDMETALLIC NANOPARTICLES ANDTHEIR APPLICATIONS INCOLORIMETRIC SENSING

Metallic Nanoparticles as Color-ReportingAgentsNoble metallic nanoparticles such as gold andplatinum nanoparticles possess a number of uniqueproperties including size and distance-dependentoptical properties, electric conductivity, and catalyticproperties. Gold nanoparticles (AuNPs) measuringfrom several nanometers to below 100 nm, forexample, are red in the dispersed state. The colorgradually changes to purple or blue when AuNPsare aggregated due to surface plasmon coupling. Inaddition to this distance-dependent optical property,AuNPs have extinction coefficients three to fiveorders of magnitude higher than traditional organicchromophores.17,18 Therefore, highly sensitive sen-sors can be constructed with minimal consumptionof materials. The chemistry of bioconjugation to goldsurfaces has been well established. Proteins can becovalently linked to AuNPs through the thiol groupon a cysteine and DNA can be attached throughchemically introduced thiol groups during solid phasesynthesis. Compared to other signaling methods, col-orimetric detection requires no analytical instruments,which makes on-site and real-time detection possible.

Heavy Metal Detection withDNAzyme-Functionalized AuNPsThe use of DNA-functionalized AuNPs for sens-ing applications was first reported by Mirkin andcoworkers.19,20 It was demonstrated that these AuNPscan self-assemble in the presence of a complementaryDNA with a vivid red-to-blue color transition. Withappropriate signal amplification methods, the sensi-tivity of such nanoparticle-based nucleic acid sensorsrivals that of polymerase chain reaction (PCR).21

Introducing catalysts that require certain cofac-tors for their activities allows design of colorimetricsensors for analytes beyond nucleic acids. We firstemployed a Pb2+-dependent DNAzyme to assembleAuNPs.22–25 The secondary structure of the DNAzyme

is shown in Figure 1(a). It contains an enzyme strand(in green) and a substrate strand (in black).26,27 Thesubstrate has a single RNA linkage (rA) that serves asthe cleavage site. In the presence of Pb2+, the enzymestrand cleaves the substrate into two pieces (Fig-ure 1(b)). To incorporate AuNP binding functions, thesubstrate strand was extended on both ends with theextended fragments being complementary to the DNAattached to AuNPs (Figure 1(c)). In each assembledAuNP aggregate, there are hundreds to thousands ofAuNPs. For the clarity of the figure, only two particlesare drawn to show the linkage between AuNPs. Addi-tion of Pb2+-induced cleavage of the substrate and dis-assembled the AuNPs, accompanied by a blue-to-redcolor change.24,25 Because the DNAzyme is selectivefor Pb2+, addition of other metal ions did not causecleavage or color change. As a result, this DNAzyme-linked nanostructure is useful for colorimetric detec-tion of lead ions. Alternatively, dispersed AuNPs weremixed with the DNAzyme. Addition of Pb2+ cleavedthe substrate, and therefore, inhibited the assembly ofAuNPs.23 Both approaches have been demonstratedand a representative thin layer chromatography (TLC)plate spotted with the sensor solution reacted withdifferent metal ions is shown in Figure 1(d). Increas-ing Pb2+ concentration resulted in a color progressionfrom bluish purple to red, and all other competing ionsshowed only background color, suggesting the highselectivity of the DNAzyme for Pb2+ has been main-tained. Under optimized conditions, the sensor canchange color in ∼5 min in the presence of Pb2+ witha detection limit around 100 nM. This method can begenerally applied to other DNAzyme-functionalizedAuNPs in response to different chemical or biolog-ical stimuli. By replacing the lead DNAzyme witha uranium-specific DNAzyme,28 colorimetric sensorsfor uranium has also been obtained.29 To improve thesensitivity even further, a label-free DNAzyme-AuNPsystem was designed to take advantage of the salt-induced aggregation of AuNPs for colorimetric sens-ing, reducing the detection limit of Pb2+ to 3 nM.29,30

Biosensors with Tunable Detection RangesThe DNAzyme/AuNP-based colorimetric sensors havea unique and useful feature of a tunable detectionrange.22 This is important because the safety level(or maximum contamination levels, MCL) for certaintoxic chemicals such as many heavy metal ions aredifferent, depending on the media to be tested. Forexample, the MCL for lead in water is 75 nM whilethe MCL for lead in blood is 500 nM. A sensorwith tunable dynamic ranges can apparently matchdifferent MCL requirements of various applications.As shown in Figure 2(a), there is a GT wobble base


WIREs Nanomedicine and Nanobiotechnology Catalyst-functionalized nanomaterials

C C C CCCG GG

G CC

T T TT TA

G G G GGA A AAT T C CC A AT A T TCC

G

G

GG

= adenosine

GGGT

TC

GGGGGTA

AA A

A

GTA

AA

AA

G GA T GTAT

TCG

r3′

5′

5′ 5′

3′ 3′Enzyme: 17E

Substrate: 17S Cleavage site

A(a)

A C U G A G C U

(b)

(c)

(d) (h)

(g)

(f)

(e)

(j)

(i) C C C CCCG GG

G CC

T T TT TA

G G G GGA A AAT T C C CAT

TTG G

GGG

G GGC

GC

T

T

GAAAA

AA AAA

A

A TA

AA

AA

G GA T GTAT

TCG

r3′

5′

5′

3′

Cleavage site

A

C C CAA T

G

GGGGTCG

GT GTAC C C CCCGGG

GG C C

C

C

T T TT T

T

TTT

T

GG

GG

G

A

G G G GGA A AA AT T T

AA AA

A

A

A

A

AAA

A

A

A

A

rA

Pb2+

Pb2+

+

−

0 0.3 0.5 1 2 3 4 5 µM Pb2+

Mg2+ Ca2+ Mn2+ Co2+ Ni2+ Cu2+ Zn2+ Cd2+ (5 µM)

+ −

FIGURE 1 | Colorimetric sensing with functional nucleic acids and gold nanoparticles (AuNPs). (a) The secondary structure of a Pb2+-specificDNAzyme. (b) Schematic presentation of cleavage of the substrate by the enzyme in the presence of Pb2+. (c) In the presence of Pb2+, theDNAzyme-assembled AuNPs are dispersed with a blue-to-red color transition. (d) A TLC plate with the sensor solutions spotted. The sensor shows ared color only in the presence of Pb2+. (e) Schematics of the adenosine aptamer binding to its target. (f) Schematic presentation that the aptamer canbe either a random coil or in a complexed structure upon binding to its target. (g) The adenosine aptamer-linked AuNPs change color from blue to redin the presence of adenosine through a structure-switching process. (h) Color of the sensor in the presence of different nucleosides. (i) Anadenosine-dependent aptazyme constructed on the basis of the Pb2+-specific DNAzyme and the adenosine aptamer. The aptazyme is active only inthe presence of adenosine. (j) Color of the aptazyme-based adenosine sensor spotted on a TLC plate.

pair close to the cleavage site in the Pb2+-specificDNAzyme. Changing the T base in the enzymestrand to a C base completely abolishes theenzyme activity31 (Figure 2(b)). Although inactive,the mutated DNAzyme can still assemble AuNPs,and the extinction spectra of DNAzyme-assembledAuNPs in the presence or absence of Pb2+ are shownin Figure 2(c) and (d) for the native and mutatedDNAzymes, respectively. The red curves are thespectra of AuNPs in the presence of Pb2+, whilethe blue curves are those in the absence of Pb2+. Asignificant increase in the 522 nm plasmon peak and adrop in the 700 nm region were observed for the native

DNAzyme in the presence of Pb2+ (Figure 2(c)), whileno change was observed for the mutant (Figure 2(d)).The extinction ratio at these two wavelengths wereused for quantifying the color of the system, with ahigh ratio associated with dispersed AuNPs of redcolor and a low ratio with aggregated particles ofblue color. By using all native DNAzyme, a detectionrange from 0.1 to 2µM was obtained (solid squares,Figure 2(e)). When 5% of native DNAzyme were usedalong with 95% mutant, the detection range shiftedto 10–200µM (open squares, Figure 2(e)). Recently, anew method of using pH to adjust the detection rangeis also reported.29

(c)

(d)

(a)

(b)

(e)

FIGURE 2 | The secondary structure of the native (a) and mutated (b) DNAzyme. The position of mutation is shown in blue. The extinction spectraof DNAzyme-assembled gold nanoparticles (AuNPs) in the presence (red) or absence (blue) of Pb2+ for the native DNAzyme (c) and the mutatedDNAzyme (d). (e) Pb2+-dependent color change of AuNPs with 100% native DNAzyme and with only 5% native DNAzyme and 95% mutatedDNAzyme .



Small Organic Molecule Detection withAptamer-Functionalized AuNPsIn addition to catalytic functions, DNA has also beenshown to be a useful ligand to bind many moleculesof choice with high affinity and specificity, and thesebinding DNAs are known as DNA aptamers.32,33

Similar to antibodies, aptamers can be selected toessentially any chemical or biological species of choice,ranging from metal ions, small organic molecules, andproteins, to whole cells.34–36 Similar to DNAzymes,the binding properties of aptamers have also beenemployed to control the assembly state of AuNPsfor sensing applications. Shown in Figure 1(e) is anaptamer for adenosine,37 which is a piece of shortsingle-stranded DNA. This DNA aptamer can adopteither random coil structures or a folded structure insolution. The equilibrium between these two statescan be shifted by addition of adenosine, becauseadenosine can associate with the aptamer to induceits folding38–41 (Figure 1(f)). To bind AuNPs, theaptamer was extended on one of its ends. One AuNP(with DNA in pink color) was associated with afraction of the extension (in gray, Figure 1(g)) anda fraction of the aptamer sequence (in blue). In thepresence of adenosine, the aptamer part folded tobind adenosine, and therefore, the number of basepairs left linking to this AuNP decreased from 12to 5, leading to the dissociation and disassembly ofAuNPs.42–44 As a result, the color of the solutionchanged from blue/purple to red. As shown inFigure 1(h), only adenosine induced a red color whileother ribonucleosides all gave purple colors. Therate of color change was very fast and it tookonly seconds to observe the color transition to redin the presence of target molecules. This method isalso generally applicable to constructing colorimetricsensors for other chemical targets, such as cocaine,potassium ions and their combinations.42,44 Aptamershave also been directly functionalized onto AuNPs.Some macromolecules, such as thrombin and platelet-derived growth factor, possess multiple aptamer-binding sites. Therefore, these protein moleculescan crosslink aptamer-functionalized AuNPs to formaggregates and change the color of the systemsto blue.45,46 Recently, the use of AuNP stabilitydifferences in high salt conditions to probe aptamerbinding has also been demonstrated. In these systems,AuNPs aggregated due to colloidal instability to saltinstead of DNA crosslinking.47,48

Aptazyme-Based DetectionsA combination of DNAzymes and aptamers resultsin an interesting allosteric molecule, referred to asallosteric DNAzymes or aptazymes.49,50 For example,

by inserting the adenosine aptamer into one of the sub-strate binding arms of the Pb2+-dependent DNAzyme,an adenosine-dependent aptazyme was designed51

(Figure 1(i)). Binding of adenosine strengthenedformation of the Pb2+ binding site in the DNAzymeand allowed cleavage, whereas in the absence ofadenosine, Pb2+ binding was disrupted. Based onthis aptazyme and similar methods developed forcolorimetric detection of Pb2+, a colorimetric assayfor adenosine was demonstrated.52

Dipstick Tests with Aptamer-FunctionalizedAuNPsEven though AuNP-based colorimetric sensors caneliminate the use of analytical instruments fordetections, there still is room for further improvingtheir user friendliness. For example, a sensor in theformat of a pH paper or a pregnancy test strip canmake it more readily accepted by users in medicaldiagnostics. One of the most useful methods to convertantibody-based assays to user friendly test kits isthe lateral flow technology. The application of thistechnology in nucleic acid-based detections, however,is rarely reported.53,54 We explored the feasibility ofusing lateral flow devices and developed aptamer-based sensors that could be used as simple dipsticks.55

The lateral flow device contained four overlap-ping pads placed on a backing (Figure 3(a)). The fourpads were (from top to bottom): absorption pad,membrane, glass fiber conjugation pad, and wickingpad. Aptamer-linked AuNPs were spotted on the con-jugation pad while streptavidin was applied on themembrane as a thin line (Figure 3(a), left strip). To becaptured by streptavidin, some AuNPs were labeledwith biotinylated DNA (black stars in Figure 3(a)).When the wicking pad of the device was dipped intoa solution, the solution moved up along the deviceand rehydrated the AuNPs. In the absence of adeno-sine, the rehydrated AuNP aggregates migrated to thebottom of the membrane where they stopped becauseof their large size (Figure 3(a), middle strip). In thepresence of adenosine, the AuNPs were disassembleddue to binding of the aptamer to adenosine.42 Thedispersed AuNPs then migrated along the membraneand were captured by streptavidin to form a red line(Figure 3(a), right strip).

To carry out detection, the adenosine-sensitivedevices were dipped into buffers containing variousnucleoside species at different concentrations (Fig-ure 3(b)). No red band was observed in the absence ofadenosine. With increasing adenosine concentrations,intensified red bands were observed, and the detectionlimit was ∼20 µM. No red bands were observed with1 mM cytidine or uridine, suggesting that the high



0 0.01 0.02 0.05 0.1 0.5 1 1 1

A A A A A A A C U Coc Coc Coc Coc Coc Coc Ade

(mM) 0 0.1 0.2 0.5 1 2 2 (mM)Absorptionpad

Membrane= Adenosine

biotin

Wickingpad

Conjugationpad

− +

(a) (b) (c)

FIGURE 3 | Aptamer and gold nanoparticles (AuNP)-based lateral flow device. (a) Left: adenosine-induced disassembly of AuNP aggregates intored-colored dispersed particles. Biotin is denoted as a black star. Right: lateral flow devices loaded with the aggregates (on the conjugation pad) andstreptavidin (on the membrane in cyan color) before use (left strip), in a negative (middle strip), or a positive (right strip) test. (b) Test of theadenosine lateral flow device with varying concentrations of nucleosides. A, adenosine; C, cytidine; U, uridine. (c) Test of the cocaine lateral flowdevice with varying concentrations of cocaine in undiluted human blood serum. Coc, cocaine; Ade, adenosine.

selectivity of the aptamer was maintained. Similarly,cocaine-sensitive strips were also prepared and thepossibility of using such devices to detect analytes inhuman blood serum was tested. As can be observedfrom Figure 3(c), a distinct red line was observed whenthe serum contained 0.2 mM cocaine, and the colorintensity increased with increasing cocaine concentra-tion; while adenosine failed to produce a red line.These results demonstrate that the device is compati-ble with biological samples, making its application inmedical diagnosis possible.

Nanoparticles as Enzyme CarriersIn the examples illustrated above, the role ofAuNPs was mainly as color-reporting groups. Inother cases, metallic nanoparticles were also usedas enzyme carriers for signal amplification. Shownin Figure 4(a) is the proposed secondary structureof a DNAzyme56,57 that can bind hemin and thecomplex has high peroxidase activities. For example,it can catalyze the conversion of luminol to generatechemiluminescence. Willner and coworkers haveextended this DNAzyme on both ends with oneend attached to a AuNP and the other end beinga probe to detect another piece of DNA.58 Thetarget DNA acted as a linker between this pieceof DNA extension and a DNA immobilized on asolid support (Figure 4(b)). In the presence of thetarget DNA, chemiluminescence could be detected onthe solid support. In this case, AuNPs acted as notonly as a support for DNAzyme immobilization, butalso as a means of signal amplification because eachAuNP could bind ∼96 DNA molecules. Compared todirectly immobilized DNAzyme, the AuNP/DNAzymesystem gave at least a 10-fold increase in sensitivity.

OHOH

O

O

NH2

NHNH

O

Luminol

Target

Thrombin

H2O2

H2O

Apo-GOx

Glucose

Gluconic acid

ee

AT

TTG G

T

G

GG

G

G

T

GG

T

TT

TT

G

G

G

G

G

G

G

G

C

G

G

G

Light

O

NH2

3′5′

5′3′

(a)

(c)

(e)

(d)

(b)

Au55 Au55FAD FAD

Au

Pt

G

FIGURE 4 | (a) Secondary structure of a DNA aptamer that can bindhemin (denoted as the diamond shape). This aptamer/hemin complex isa DNAzyme that can catalyze conversion of luminal. (b) Schematicpresentation of DNA detection with DNAzyme-functionalized goldnanoparticles (AuNPs). (c) Secondary structure of a thrombin-bindingDNA aptamer. (d) Schematic presentation of thrombin detection with Ptnanoparticles acting as a catalyst for signal generation. (e) Schematicpresentation of AuNP as wiring for electron relay redrawn from Ref. 59.

Nanoparticles as Catalysts for SignalGeneration and AmplificationNoble metal nanoparticles, such as gold and platinum,have high catalytic activities for many chemical



transformations. Coupled with biopolymers, thisproperty has also been utilized for sensor design.60

Thrombin is an important serine protease in theblood coagulation cascade. The secondary structureof a thrombin DNA aptamer is shown in Figure 4(c).The authors indicated that there are two aptamer-binding sites in each thrombin molecule. Therefore,by attaching one thrombin aptamer on a solid supportand another on a Pt nanoparticle, the presence ofthrombin can link the nanoparticle to the solid surface(Figure 4(d)). The catalysis of H2O2 to H2O was usedas a way to electrochemically signaling the presenceof thrombin.61

Nanoparticles as Wiring for Electron RelayAuNPs also possess excellent electric conductingproperties. Because their sizes can be made to be com-parable to protein and nucleic acids, these nanopar-ticles can be used as a nanowire to conduct electronsfrom the enzyme-active site to bulk electrodes, andthus facilitate the detection of enzymatic reactionsby electrochemical methods. Willner and coworkersimmobilized 1.4 nm AuNPs on an Au electrode witha dithiol linkage.59 N-aminoethyl flavin adenine din-ucleotide (FAD) was attached to the AuNPs throughthiol linkages. FAD is the cofactor for glucose oxidase(GOx). Apo-GOx was prepared and reconstituted onthe immobilized FAD, which allowed electrocatalyticoxidation of glucose. Higher concentration of glucoseproduced stronger currents, which enabled sensitiveglucose detection. This method has also been appliedto reconstitute glucose dehydrogenase.

CATALYST-FUNCTIONALIZEDSEMICONDUCTOR NANOPARTICLESAND THEIR APPLICATIONS INFLUORESCENT SENSING

Semiconductor nanoparticles are usually calledquantum dots (QDs), many of which possess usefuloptical properties, such as broad excitation bandsand narrow emission peaks.1,62,63 The emission wave-lengths can be controlled by adjusting size, shape,or composition of the QDs. Compared to traditionalorganic fluorophores, QDs have much higher photostability and are resistant to photobleaching. Theemission properties of QDs can be strongly affectedby the external environment, such as the presence ofquenchers or molecules that can perturb the energystate of QDs. Therefore, QDs are increasingly used asan alternative for organic fluorophores in a numberof applications, including sensing.

Detection Based on Fluorescence ResonanceEnergy TransferFluorescence resonance energy transfer (FRET) is awidely used technique in biology and biophysics.64

A FRET system contains a donor and an acceptor.The donor should be fluorescent and its fluorescencespectrum should have some overlaps with theabsorption spectrum of the acceptor. The acceptorcan be either fluorescent or nonfluorescent. In thelatter case, it is usually called a quencher. When thedonor and acceptor are close to each other (usuallywithin 10 nm), energy can transfer from the exciteddonor to the acceptor. FRET efficiency depends onthe distance between the two fluorophores, and ashorter distance gives stronger energy transfer. MostFRET experiments were performed with organicfluorophores. Recently, nanoparticles (includingQDs) have been used both as FRET donors andacceptors for sensing applications.

Medintz et al. functionalized 530 nm QDs(emission peaks at 530 nm) with Cy3-labeledmaltose binding proteins (MBP)65 (Figure 5(a)).Cy3.5-conjugated β-cyclodextrin specifically binds tothe pocket of MBP. Therefore, there are two FRETpairs in this system. For the QD/Cy3 pair, QD is thedonor, and for the Cy3/Cy3.5 pair, Cy3 is the donor.Exciting the 530 nm QD led to the energy relay toCy3 and then to Cy3.5. In the presence of maltose,the Cy3.5 labeled β-cyclodextrin was displaced,which decreased the FRET from Cy3 to Cy3.5, andincreased the Cy3 emission. This system is thereforeuseful for maltose detection.

In addition to separating the donor and acceptorby the ligand displacement reaction as mentionedabove, the two can also be separated by cleavagereactions. Proteases are protein enzymes that cancleave proteins and peptides. Assays for proteaseactivities are important for understanding a numberof bioprocesses and diseases including cancer andinfectious diseases. Medintz et al. prepared severalquencher-labeled peptide sequences and attached thesepeptides to QDs (Figure 5(b)). These peptides weresubstrates for various proteases including caspase-1,thrombin, collagenase, and chymotrypsin. Initiallythe QD emission was quenched. In the presence ofthe proteases, the peptide was cleaved, releasing thequencher and unmasking the QD emission.66 In aslightly earlier report, Chang et al. employed AuNPsas quenchers, which were linked to QDs by a substratepeptide for collagenase. Similarly, enhanced emissionwas observed in the presence of the protease.67

This method is not limited to protease assays. Raoand coworkers attached a Cy5-labeled substrate forβ-lactamase to QDs, and the QD emission was



QD

Thrombin aptamerThrombin aptamer

Quencher

FRET

Thrombin

QD

Ex Em Em

Maltose

QD

FRETEx

Cy3

Cy3.5

MBP

EmEm

QD

FRETFRETEx

EmEm

Cy3

Cy3.5

MBP

β-cyclodextrin-Cy3.5

(a)

(b)

(c)

ProteaseQD

Ex Em Em

Dye or

quencher

QD

FRET

Cleavage site

Ex

FIGURE 5 | (a) Maltose detection based on a displacement reaction with Cy3.5 labeled β-cyclodextrin. MBP, maltose binding protein.(b) Detection of proteases by conjugating a quencher-labeled peptide to QDs. (c) Detection of thrombin based on the structure-switching property ofa thrombin aptamer. A quencher-labeled DNA was released due to aptamer binding to thrombin.

quenched up to 95% by Cy5. Addition of β-lactamasecleaved the substrate linker and released Cy5, resultingin enhanced emission.68

Ellington and coworkers prepared throm-bin aptamer-functionalized QDs. Quenchers werebrought close to the QD surface by a small piece DNAthat was complementary to a fraction of the aptamersequence (Figure 5(c)). In the presence of thrombin,the aptamer switched its structure and bound throm-bin, releasing the quencher-labeled DNA to produceenhanced emission.69 One of the main advantages ofQDs is that under the same excitation light, emis-sions of different wavelengths can be generated inthe same solution, which is useful for informationencoding. We recently encoded adenosine and cocaineaptamer-linked AuNPs with QDs that emitted at 525and 585 nm, respectively. This system was capable ofdetecting both analytes in one pot.70

Detection Based on Charge TransferIn addition to energy transfer, charge transfer has alsobeen applied to modulate the emission properties ofQDs. Benson and coworkers immobilized MBP on theQD surface71 (Figure 6(a)). A ruthenium compoundwas used as an electron donor. In the absence of mal-tose, the ruthenium compound was close to the QD

Maltose

Thrombin

e−

Near IRQD

Ex Em Em

Thrombin aptamerThrombin aptamer

Near IRQD

Ex Em Em

(NH 3

) 4RuQD

Ex

e−

MBP

EmEm

QD

Ex

(NH3)4Rue−

EmEm

MBP

(a)

(b)

FIGURE 6 | (a) Detection of maltose based on modulation of chargetransfer from a labeled ruthenium compound to quantum dots (QDs).(b) Detection of thrombin based on the charge transfer from thrombinto QDs.

surface and the charge transfer efficiency was high,inhibiting the QD emission. However, in the presenceof maltose, the MBP underwent a conformationchange and the ruthenium compound was separatedfrom the QD surface, leading to increased emission.A similar strategy has also been used to couple the



same ruthenium compound to an intestinal fatty acid-binding protein.72 Upon palmitate binding, the envi-ronment around the ruthenium compound becamemore hydrophobic, which enhanced electron transferand led to decreased emission. Strano and coworkersfound that thrombin aptamer-functionalized PbSQDs showed thrombin-dependent emission decrease(Figure 6(b)), while other proteins such as bovineserum albumin (BSA), streptavidin, proteinase K, orlysozyme did not have such an effect. They attributedthis drop in emission to the charge transfer from thefunctional groups on thrombin to QDs.73

Magnetic Nanoparticles as MagneticResonance Imaging Contrast AgentsIn addition to unique optical properties, certain metaland metal oxide nanoparticles possess useful mag-netic properties.74,75 Depending on the aggregationstate, for example, superparamagnetic iron oxidenanoparticles can affect the magnetic relaxation ofthe surrounding water proton, and thus, are useful ascontrast agents in magnetic resonance imaging (MRI).We have prepared adenosine aptamer-functionalizediron oxide nanoparticles similar to those describedin Figure 1(g), except that AuNPs were replacedby iron oxide nanoparticles.76 In the presence ofadenosine, aggregated nanoparticles disassembledinto individual ones. In a second example, thrombinaptamer-functionalized magnetic nanoparticles wereassembled upon addition of thrombin to formaggregated structures, taking advantage of multipleaptamer-binding sites in each thrombin molecule.77

In both cases, altered water proton relaxation (T2)was observed, suggesting potential applications ofthese functionalized magnetic nanoparticles as smartmagnetic contrast agents.

CATALYST-FUNCTIONALIZEDNANOTUBES AND THEIRAPPLICATIONS IN NEAR-IRFLUORESCENT SENSINGCarbon nanotubes have recently been found to beuseful materials for sensing applications becauseof their near-IR fluorescence property. Near-IRsensors are very useful in molecular diagnosticsand nanomedicine. Similar to QDs, nanotubes arealso highly photostable. However, it is relativelydifficult to functionalize nanotubes for bioconjugationwhile still maintaining its useful optical property.Other than optical detection, nanotubes can also beincorporated into microelectronic devices such as fieldeffect transistors (FETs). Attaching biopolymers tothese nanotubes allows other modes of detection.

Sensing with Change of Emission IntensityStrano and coworkers reported that electroactivespecies such as K3Fe(CN)6 could irreversibly adsorbon the surface of single-walled carbon nanotubesand act as a quencher for nanotube emission witha quenching efficiency up to 83.3%.78,79 K2Fe(CN)6,on the other hand, quenched the emission only by27.4% under identical conditions. With a dialysismethod, glucose oxidase was immobilized on thenanotube surface through van der Waals interactions.This enzyme converted β-d-glucose into gluconic acidand hydrogen peroxide. The latter could react withK3Fe(CN)6 to produce K2Fe(CN)6, thus reducingquenching (Figure 7(a)). With this method, a detectionlimit of 34.7µM glucose was achieved. Detectionof glucose in blood under physiological conditionswas also demonstrated. Other glucose oxidaseimmobilization methods, such as through a DNAinterlayer, were also reported to be successful.80 Withthe unique IR-emitting property of carbon nanotubes,the sensors may be useful for in vivo applications.

Sensing with Shift of Emission WavelengthIn addition to emission intensity-based sensing,modulation of the chemical environment of thenanotubes to generate shifted emission wavelengthshas also been shown to be a useful way of makingsensors. Strano and coworkers demonstrated thatDNA hybridization on the surface of solution-suspended single-walled carbon nanotubes inducedshifted emission wavelength through a band gapfluorescence modulation. Unmodified 24-mer DNAwas immobilized on nanotube surface with a dialysismethod. Upon hybridization to its complementarystrand, the nanotube emission wavelength showeda blue shift with an energy of ∼2 meV, whilenoncomplementary DNA did not show such a shift.The system had a detection limit of 6 nM. The time

(b)(a) Gluconic acid

Z-DNA

B-DNA

Glucose

NanotubeGox

[Fe(CN)6]3− [Fe(CN)6]4−

H2O2 Hg2+

FIGURE 7 | (a) Glucose detection with glucose oxidase-immobilizedcarbon nanotubes. [Fe(CN)6]3– strongly quenches nanotube emission,while [Fe(CN)6]2– has much lower quenching efficiency. (b) Hg2+-induced DNA B-Z conformational change for Hg2+ detection.Nanotubes emit at different wavelengths depending on the DNAconformation. (Some parts of the figure are adapted from Ref. 76).



required for the detection was relatively long, and ittook ∼13 h to reach the steady state.81

The same group has also demonstrated othermeans of shifting the nanotube emission wavelengths.For example, native double-stranded DNA are right-handed double helix, or known as the B-form DNA.Certain DNA sequences under certain conditions canswitch conformation to a left-handed Z-form (Fig-ure 7(b)). When such a B-Z transition takes placeon the surface of a single-walled carbon nanotube,the emission wavelength of the nanotubes will alsochange. Hg2+ was found to be particularly effectivein inducing such B-Z transitions and a sensor forHg2+ detection was designed based on this observa-tion. Again, because of the near-IR emission propertyof nanotubes, Hg2+ could be detected even in wholerooster blood and black dye solutions.82

Nanotubes Functionalized with DNAzymesand AptamersIn a collaborative research effort between the Lu,Kane, and Dordick groups, the Pb2+-dependentDNAzyme was immobilized onto multiwalled car-bon nanotubes.83 The immobilized DNAzyme formeda highly stable hybrid with nanotubes and main-tained high activity. In the presence of Pb2+, over400 turnovers were observed for each DNAzyme.In addition to sensing, such high activity may allowmany applications ranging from directed assemblyof nanotubes to cellular therapeutics. DNA aptamersfor both thrombin and IgE have also been assembled

in FETs and analyte-dependent current changes wereobserved.84 In the case of IgE, the detection limitreached 250 pM with a dynamic range up to 20 nM,which covered the clinically relevant level of 10 nMfor patients with high IgE levels.85

CONCLUSION

In this review, we have summarized some recent devel-opments in the design of catalyst-functionalized nano-materials and their applications in sensing relevant tonanomedicine. Nanoparticles and other nanomaterialspossess useful physical properties that can be modu-lated by changing their chemical environment. Bio-catalysts and biopolymers, on the other hand, possesshighly specific target recognition properties. Conjuga-tion of these biomolecules to nanomaterials can there-fore provide a useful means of making biosensors forbiomedical applications. Sensors based on the modu-lation of optical properties of metallic nanoparticles,semiconductor nanoparticles, magnetic nanoparticles,and carbon nanotubes have all been realized. Thechanges in the optical properties were either inducedby different assembly states, or by changes in localchemical environment. Most of the examples arefocused on in vitro applications of the sensors. Futureworks are likely to move towards in vivo detection. Toachieve this goal, however, the toxicity of these nano-materials has to be investigated to make sure they aresafe for medical applications. Future research effortswill also focus on demonstration of sensors for clini-cally relevant targets under physiological conditions.

NOTES

We wish to thank other Lu group members and collaborators for contributing experimental work described inthis article, and Dr. Zehui Cao for proofreading the manuscript and for his helpful suggestions. The Lu groupresearch described in this article has been generously supported by the US Department of Energy, NationalScience Foundation, Department of Defense, National Health Institute, and Department of Housing and UrbanDevelopment.

REFERENCES

1. Alivisatos AP, Gu W, Larabell C. Quantum dots as cel-lular probes. Annu Rev Biomed Eng 2005, 7:55–76.

2. Rosi NL, Mirkin CA. Nanostructures in biodiagnos-tics. Chem Rev 2005, 105(4):1547–1562.

3. Katz E, Willner I. Nanobiotechnology: integratednanoparticle-biomolecule hybrid systems: synthesis,properties, and applications. Angew Chem Int Ed 2004,43(45):6042–6108.

4. Feldkamp U, Niemeyer CM. Rational design of DNAnanoarchitectures. Angew Chem Int Ed 2006,45(12):1856–1876.

5. Lu Y, Liu J. Functional DNA nanotechnology: emerg-ing applications of DNAzymes and aptamers. CurrOpin Biotechnol 2006, 17(6):580–588.

6. Lu Y, Liu J. Smart nanomaterials inspired by biology:dynamic assembly of error-free nanomaterials in



response to multiple chemical and biological stimuli.Acc Chem Res 2007, 40:315–323.

7. Kruger K, Grabowski PJ, Zaug AJ, Sands J,Gottschling DE, et al. Self-splicing RNA: autoexcisionand autocyclization of the ribosomal RNA interveningsequence of tetrahymena. Cell 1982, 31(1):147–157.

8. Guerrier-Takada C, Gardiner K, Marsh T, Pace N, Alt-man S. The RNA moiety of ribonuclease P is thecatalytic subunit of the enzyme. Cell 1983, 35(3 Pt2):849–857.

9. Breaker RR, Joyce GF. A DNA enzyme that cleavesRNA. Chem Biol 1994, 1(4):223–229.

10. Breaker RR. DNA enzymes. Nat Biotechnol 1997,15(5):427–431.

11. Sen D, Geyer CR. DNA enzymes. Curr Opin ChemBiol 1998, 2(6):680–687.

12. Joyce GF. Directed evolution of nucleic acid enzymes.Annu Rev Biochem 2004, 73:791–836.

13. Achenbach JC, Chiuman W, Cruz RPG, Li Y.DNAzymes: from creation in vitro to application invivo. Curr Pharm Biotechnol 2004, 5:312–336.

14. Silverman SK. In vitro selection, characterization, andapplication of deoxyribozymes that cleave RNA.Nucleic Acids Res 2005, 33(19):6151–6163.

15. Peracchi A. DNA catalysis: potential, limitations, openquestions. Chembiochem 2005, 6(8):1316–1322.

16. Navani NK, Li Y. Nucleic acid aptamers and enzymesas sensors. Curr Opin Chem Biol 2006, 10:272–281.

17. Yguerabide J, Yguerabide EE. Light-scattering submi-croscopic particles as highly fluorescent analogs andtheir use as tracer labels in clinical and biolog-ical applications. I. Theory. Anal Biochem 1998,262(2):137–156.

18. Jin R, Wu G, Li Z, Mirkin CA, Schatz GC. Whatcontrols the melting properties of DNA-linked goldnanoparticle assemblies? J Am Chem Soc 2003,125(6):1643–1654.

19. Mirkin CA, Letsinger RL, Mucic RC, Storhoff JJ. ADNA-based method for rationally assembling nanopar-ticles into macroscopic materials. Nature 1996,382(6592):607–609.

20. Elghanian R, Storhoff JJ, Mucic RC, Letsinger RL,Mirkin CA. Selective colorimetric detection of polynu-cleotides based on the distance-dependent opti-cal properties of gold nanoparticles. Science 1997,277(5329):1078–1080.

21. Nam J-M, Stoeva SI, Mirkin CA. Bio-bar-code-basedDNA detection with Pcr-like sensitivity. J Am ChemSoc 2004, 126(19):5932–5933.

22. Liu J, Lu Y. A colorimetric lead biosensor usingDNAzyme-directed assembly of gold nanoparticles.J Am Chem Soc 2003, 125(22):6642–6643.

23. Liu J, Lu Y. Accelerated color change of gold nanopar-ticles assembled by DNAzymes for simple and fast

colorimetric Pb2+ detection. J Am Chem Soc 2004,126(39):12298–12305.

24. Liu J, Lu Y. Stimuli-responsive disassembly of nanopar-ticle aggregates for light-up colorimetric sensing. J AmChem Soc 2005, 127(36):12677–12683.

25. Liu J, Lu Y. Design of asymmetric DNAzymes fordynamic control of nanoparticle aggregation states inresponse to chemical stimuli. Org Biomol Chem 2006,4(18):3435–3441.

26. Santoro SW, Joyce GF. A general purpose RNA-cleaving DNA enzyme. Proc Natl Acad Sci USA 1997,94(9):4262–4266.

27. Li J, Zheng W, Kwon AH, Lu Y. In vitro selection andcharacterization of a highly efficient Zn(Ii)-dependentRNA-cleaving deoxyribozyme. Nucleic Acids Res 2000,28(2):481–488.

28. Liu J, Brown AK, Meng X, Cropek DM, Istok JD, et al.A catalytic beacon sensor for uranium with parts-per-trillion sensitivity and millionfold selectivity. Proc NatlAcad Sci USA 2007, 104(7):2056–2061.

29. Wang Z, Lee JH, Lu Y. Label-free colorimetric detec-tion of lead ions with a nanomolar detection limit andtunable dynamic range by using gold nanoparticles andDNAzyme. Adv Mater 2008, 20(17):3263–3267.

30. Wei H, Li B, Li J, Dong S, Wang E. DNAzyme-basedcolorimetric sensing of lead (Pb2+) using unmodi-fied gold nanoparticle probes. Nanotechnology 2008,19:1–5.

31. Brown AK, Li J, Pavot CMB, Lu Y. A lead-dependentDNAzyme with a two-step mechanism. Biochemistry2003, 42(23):7152–7161.

32. Tuerk C, Gold L. Systematic evolution of ligandsby exponential enrichment: RNA ligands to bac-teriophage T4 DNA polymerase. Science 1990,249(4968):505–510.

33. Ellington AD, Szostak JW. In vitro selection of RNAmolecules that bind specific ligands. Nature 1990,346(6287):818–822.

34. Wilson DS, Szostak JW. In vitro selection of functionalnucleic acids. Annu Rev Biochem 1999, 68:611–647.

35. Jayasena SD. Aptamers: an emerging class of moleculesthat rival antibodies in diagnostics. Clin Chem 1999,45(9):1628–1650.

36. Shangguan D, Li Y, Tang Z, Cao Z, Chen H, et al.Aptamers evolved from live cells as effective molec-ular probes for cancer study. Proc Natl Acad Sci USA2006, 103(32):11838–11843.

37. Huizenga DE, Szostak JW. A DNA aptamer thatbinds adenosine and Atp. Biochemistry 1995,34(2):656–665.

38. Nutiu R, Li Y. Structure-switching signaling aptamers:transducing molecular recognition into fluorescence sig-naling. Chem Eur J 2004, 10(8):1868–1876.

39. Nutiu R, Li Y. Structure-switching signaling aptamers.J Am Chem Soc 2003, 125(16):4771–4778.



40. Nutiu R, Li Y. In vitro selection of structure-switchingsignaling aptamers. Angew Chem Int Ed 2005,44(7):1061–1065.

41. Nutiu R, Li Y. Aptamers with fluorescence-signalingproperties. Methods 2005, 37(1):16–25.

42. Liu J, Lu Y. Fast colorimetric sensing of adenosine andcocaine based on a general sensor design involvingaptamers and nanoparticles. Angew Chem Int Ed 2006,45(1):90–94.

43. Liu J, Lu Y. Preparation of aptamer-linked goldnanoparticle purple aggregates for colorimetric sensingof analytes. Nat Protoc 2006, 1(1):246–252.

44. Liu J, Lu Y. Smart nanomaterials responsive to mul-tiple chemical stimuli with controllable cooperativity.Adv Mater 2006, 18:1667–1671.

45. Pavlov V, Xiao Y, Shlyahovsky B, Willner I. Aptamer-functionalized Au nanoparticles for the amplified opti-cal detection of thrombin. J Am Chem Soc 2004,126(38):11768–11769.

46. Huang C-C, Huang Y-F, Cao Z, Tan W, Chang H-T.Aptamer-modified gold nanoparticles for colorimetricdetermination of platelet-derived growth factors andtheir receptors. Anal Chem 2005, 77(17):5735–5741.

47. Wang L, Liu X, Hu X Song S, Fan C. Unmodified goldnanoparticles as a colorimetric probe for potassiumDNA aptamers. Chem Commun 2006, 36:3780–3782.

48. Zhao W, Chiuman W, Brook MA, Li Y. Simple andrapid colorimetric biosensors based on DNA aptamerand noncrosslinking gold nanoparticle aggregation.Chembiochem 2007, 8(7):727–731.

49. Breaker RR. Engineered allosteric ribozymes as biosen-sor components. Curr Opin Biotechnol 2002,13(1):31–39.

50. Hesselberth J, Robertson MP, Jhaveri S, Ellington AD.In vitro selection of nucleic acids for diagnostic appli-cations. Rev Mol Biotechnol 2000, 74(1):15–25.

51. Wang DY, Lai BHY, Sen D. A general strategy foreffector-mediated control of RNA-cleaving ribozymesand DNA enzymes. J Mol Biol 2002, 318(1):33–43.

52. Liu J, Lu Y. Adenosine-dependent assembly ofaptazyme-functionalized gold nanoparticles and itsapplication as a colorimetric biosensor. Anal Chem2004, 76(6):1627–1632.

53. Glynou K, Ioannou PC, Christopoulos TK, Syri-opoulou V. Oligonucleotide-functionalized goldnanoparticles as probes in a dry-reagent strip biosensorfor DNA analysis by hybridization. Anal Chem 2003,75(16):4155–4160.

54. Famulok M, Mayer G. Chemical biology: aptamers innanoland. Nature 2006, 439(7077):666–669.

55. Liu J, Mazumdar D, Lu Y. A simple and sensitive ‘‘Dip-stick’’ test in serum based on lateral flow separation ofaptamer-linked nanostructures. Angew Chem Int Ed2006, 45(47):7955–7959.

56. Li Y, Geyer CR, Sen D. High-specificity recognition ofanionic porphyrins by DNA aptamers. Biochemistry1996, 35(21):6911–6922.

57. Travascio P, Bennet AJ, Wang DY, Sen D. A ribozymeand a catalytic DNA with peroxidase activity: activesites versus cofactor-binding sites. Chem Biol 1999,6(11):779–787.

58. Niazov T, Pavlov V, Xiao Y, Gill R, Willner I.DNAzyme-functionalized Au nanoparticles for theamplified detection of DNA or telomerase activity.Nano Lett 2004, 4(9):1683–1687.

59. Xiao Y, Patolsky F, Katz E, Hainfeld JF, Willner I.‘‘Plugging into Enzymes’’: nanowiring of redoxenzymes by a gold nanoparticle. Science 2003,299(5614):1877–1881.

60. Bock LC, Griffin LC, Latham JA, Vermaas EH,Toole JJ. Selection of single-stranded DNA moleculesthat bind and inhibit human thrombin. Nature 1992,355(6360):564–566.

61. Polsky R, Gill R, Kaganovsky L, Willner I. Nucleicacid-functionalized Pt nanoparticles: catalytic labels forthe amplified electrochemical detection of biomolecules.Anal Chem 2006, 78(7):2268–2271.

62. Han M, Gao X, Su JZ, Nie S. Quantum-dot-taggedmicrobeads for multiplexed optical coding ofbiomolecules. Nat Biotechnol 2001, 19(7):631–635.

63. Medintz IL, Uyeda HT, Goldman ER, Mattoussi H.Quantum dot bioconjugates for imaging, labelling andsensing. Nat Mater 2005, 4(6):435–446.

64. Clegg RM. Fluorescence resonance energy transfer andnucleic acids. Methods Enzymol 1992, 211:353–388.

65. Medintz IL, Clapp AR, Brunel FM, Tiefenbrunn T,Uyeda HT, et al. Self-assembled nanoscale biosensorsbased on quantum dot fret donors. Nat Mater 2003,2(9):630–638.

66. Medintz IL, Clapp AR, Brunel FM, et al. Proteolyticactivity monitored by fluorescence resonance energytransfer through quantum-dot-peptide conjugates. NatMater 2006, 5(7):581–589.

67. Chang E, Miller JS, Sun J, Yu WW, Colvin VL, et al.Protease-activated quantum dot probes. Biochem Bio-phys Res Commun 2005, 334(4):1317–1321.

68. Xu C, Xing B, Rao J. A self-assembled quantum dotprobe for detecting B-lactamase activity. Biochem Bio-phys Res Commun 2006, 344(3):931–935.

69. Levy M, Cater SF, Ellington AD. Quantum-dotaptamer beacons for the detection of proteins. Chem-biochem 2005, 6(12):2163–2166.

70. Liu J, Lee JH, Lu Y. Quantum dot encoding ofaptamer-linked nanostructures for one pot simultane-ous detection of multiple analytes. Anal Chem 2007,79:4120–4125.

71. Sandros MG, Gao D, Benson DE. A modularnanoparticle-based system for reagentless small



molecule biosensing. J Am Chem Soc 2005,127(35):12198.

72. Aryal BP, Benson DE. Electron donor solvent effectsprovide biosensing with quantum dots. J Am Chem Soc2006, 128(50):15986–15987.

73. Choi JH, Chen KH, Strano MS. Aptamer-cappednanocrystal quantum dots: a new method for label-free protein detection. J Am Chem Soc 2006,128(49):15584–15585.

74. Perez JM, Josephson L, O’Loughlin T, Hoegemann D,Weissleder R. Magnetic relaxation switches capable ofsensing molecular interactions. Nat Biotechnol 2002,20(8):816–820.

75. Perez JM, O’Loughin T, Simeone FJ, Weissleder R,Josephson L. DNA-based magnetic nanoparticle assem-bly acts as a magnetic relaxation nanoswitch allowingscreening of DNA-cleaving agents. J Am Chem Soc2002, 124(12):2856–2857.

76. Yigit MV, Mazumdar D, Kim H-K, Lee JH,Odintsov B, et al. Smart ‘‘Turn-on’’ magnetic reso-nance contrast agents based on aptamer-functionalizedsuperparamagnetic iron oxide nanoparticles. Chem-biochem 2007, 8(14):1675–1678.

77. Yigit MV, Mazumdar D, Lu Y. Mri detection of throm-bin with aptamer functionalized superparamagneticiron oxide nanoparticles. Bioconjug Chem 2008,19:412–417.

78. Barone PW, Parker RS, Strano MS. In vivo fluores-cence detection of glucose using a single-walled carbon

nanotube optical sensor: design, fluorophore proper-ties, advantages, and disadvantages. Anal Chem 2005,77(23):7556–7562.

79. Barone PW, Baik S, Heller DA, Strano MS. Near-infrared optical sensors based on single-walled carbonnanotubes. Nat Mater 2005, 4(1):86–92.

80. Karachevtsev VA, Glamazda AY, Leontiev VS,Lytvyn OS, Dettlaff-Weglikowska U. Glucose sensingbased on nir fluorescence of DNA-wrapped single-walled carbon nanotubes. Chem Phys Lett 2007,435(1–3):104–108.

81. Jeng ES, Moll AE, Roy AC, Gastala JB, Strano MS.Detection of DNA hybridization using the near-infraredband-gap fluorescence of single-walled carbon nan-otubes. Nano Lett 2006, 6(3):371–375.

82. Heller DA, Jeng ES, Yeung T-K, Martinez BM,Moll AE, et al. Optical detection of DNA conforma-tional polymorphism on single-walled carbon nan-otubes. Science 2006, 311(5760):508–511.

83. Yim T-J, Liu J, Lu Y, Kane RS, Dordick JS. Highlyactive and stable DNAzyme-carbon nanotube hybrids.J Am Chem Soc 2005, 127:12200–12201.

84. So H-M, Won K, Kim YH, Kim B-K, Ryu BH, et al.Single-walled carbon nanotube biosensors usingaptamers as molecular recognition elements. J AmChem Soc 2005, 127(34):11906–11907.

85. Maehashi K, Katsura T, Kerman K, Takamura Y, Mat-sumoto K, et al. Label-free protein biosensor based onaptamer-modified carbon nanotube field-effect transis-tors. Anal Chem 2007, 79(2):782.


Uses of nanowire sensors in medicine and biology.Magnetic nanoparticle biosensors.


Advanced Review

Nanoparticle-based biologicmimeticsDavid E. Cliffel,∗ Brian N. Turner1 and Brian J. Huffman1

Centered on solid chemistry foundations, biology and materials science havereached a crossroad where bottom-up designs of new biologically importantnanomaterials are a reality. The topics discussed here present the interdisciplinaryfield of creating biological mimics. Specifically, this discussion focuses on mimicsthat are developed using various types of metal nanoparticles (particularlygold) through facile synthetic methods. These methods conjugate biologicallyrelevant molecules, e.g., small molecules, peptides, proteins, and carbohydrates,in conformationally favorable orientations on the particle surface. These newproducts provide stable, safe, and effective substitutes for working withpotentially hazardous biologicals for applications such as drug targeting,immunological studies, biosensor development, and biocatalysis. Many standardbioanalytical techniques can be used to characterize and validate the efficacyof these new materials, including quartz crystal microbalance (QCM), surfaceplasmon resonance (SPR), and enzyme-linked immunosorbent assay (ELISA).Metal nanoparticle–based biomimetics continue to be developed as potentialreplacements for the native biomolecule in applications of immunoassays andcatalysis. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 47–59

THE IMPORTANCE OF BIOMIMICS

The accurate mimicking of biologically importantmaterials in a benign form is critical for the

development of drug carriers, sensors, and catalysts.The use of whole or modified pathogens presentsmany challenges to researchers in terms of personalsafety, facility requirements, and overall time andcost. Additionally, while the inactivated or killedform of a given pathogen can be used, thereare always risks such as conformational changesor losses during inactivation, or a specimen thatremains partially active. These challenges requirethe development of a surrogate that circumventsthe need for active biological systems. Biomimeticnanoparticles offer an easy way to present theactive part of a biomolecule with better stabilityand without the harmful payload. Additionally,nanoparticles provide a way to modify a surfacewith multiple functional groups because of theirhigh surface area. All these attributes have led

∗Correspondence to: David E. Cliffel, Vanderbilt University,Nashville, TN, USA. E-mail: [email protected] University, Nashville, TN, USA.

DOI: 10.1002/wnan.020

to nanoparticles becoming a diverse platform forbiomimicking.

Since the development of water-soluble, ligand-capped nanoparticles almost 15 years ago,1 the useof nanoparticles in biological systems has increaseddramatically. This is due, in part, to the fact thatthey can be chemically modified to mimic an antigenor biological marker of interest. Unlike growing cellcultures or working with live animals, which is timeconsuming and expensive, nanoparticle synthesis isrelatively straightforward and can be carried out ona larger scale. The chemistry to conjugate functionalligands and macromolecules to nanoparticles has beenwell developed (especially place exchange2 and amidelinkage3) and can be adapted to fit a myriad ofsystems, for example, antigen/antibody interaction,via different synthetic routes. Nanoparticles offer amethod whereby a surface can be multifunctionalizedto create a broad spectrum of functionality, whetherpresenting multiple epitopes of the same antigenor two different reactive species from a catalyst.This review will discuss the creation, modification,characterization, and uses of nanoparticle-basedbiological mimics, and the tools that can be usedto validate their biological activity.



HAuIIICl4 + HSR80% MeOH/ 20% CH3COOH

(AuISR)nNaBH4/H2O Aux (SR)y FIGURE 1 | Modified Brust reaction scheme for polar

ligands.

NANOPARTICLE SYNTHESISAND FUNCTIONALIZATION

The scientific study of colloidal metal particles datesback to Faraday in the mid-19th century.4 Thesynthesis and characterization, notably by electronmicroscope, of water ‘soluble’ gold colloids as smallas 18 nm was completed by Turkevich and coworkersin 1951.5 Schiffrin and Brust, 43 years later, reportedmetal particles stabilized by alkanethiols. Murrayand coworkers termed these ‘monolayer-protectedclusters’ (MPCs) and defined them as differing frommetal colloids because they can be repeatedly driedas well as isolated from and redissolved in commonsolvents without decomposing or aggregating.6

MPCs are synthesized using a bottom-up approach,suggesting that a wide variety of nanomaterials ispossible from a small number of building blocks.7

Nanoparticles are created with a variety ofcore types and capping ligands to create water- ororganic-soluble products with desired functions. Bothmetallic and nonmetallic starting materials are usedin the creation of nanoparticles, such as MPCs,1,6,8–12

organic polymers,13–16 virus-like particles (VLPs),17–22

protein particles,23 colloidal particles,5,24,25 andsemiconductor quantum dots.26 Thiol-capped MPCshave received more focus because of their ease ofcreation, water and air stability, electrochemical andoptical properties, and their ability to be surface-functionalized by the addition of biologically relevantligands, such as peptide sequences of epitopes. GoldMPCs can range in size from 1 to 10 nm, containingapproximately 55–1000 gold atoms with molecularweights between 10 and 200 kDa.27

We acknowledge that a broad spectrum ofnanometer-sized materials is present in the literatureas previously mentioned. However, the focus of this

review is stable, water-soluble gold-core MPCs andtheir targeted use in biological mimetics.

Synthetic RoutesWater solubility of MPCs is best accomplishedby using a thiolated, polar protecting ligand in amodified Brust reaction1,9 as seen in Figure 1. In theBrust reaction, tetrachloroauric acid is reduced fromAu3+ to Au1+ in the presence of the thiol cappingligand, yielding a colorless gold-thiol solution. Thisis either composed of a gold-thiol polymer6 ortetramer.28 Following the initial reduction, the goldis further reduced to Au0 in the presence of sodiumborohydride (NaBH4), yielding a black to darkbrown solution. Other potent reducing agents, suchas lithium aluminum hydride (LiAlH4) or lithium tri-ethylborohydride, have been used to reduce differentmetal cores such as palladium and platinum.11,12

Key examples of thiolate ligands that havebeen used to produce water-soluble and long-term(months) air- and water-stable clusters are tiopronin,9

glutathione,29 4-mercaptobenzoic acid,301-thio-β-D-glucose,31 and N, N, N-trimethyl(mercaptounde-cyl)ammonium (TMA)32 as depicted in Figure 2.

Functionalization of Monolayer-ProtectedClustersTransformation of water-soluble MPCs into biolog-ical mimics has been accomplished using a varietyof synthetic functionalization strategies. However, themost widely used, straightforward method is the thiolplace-exchange reaction seen in Figure 3.

Solution-Phase Place ExchangeIn the place-exchange reaction, an incoming ligand,such as a thiol-containing biomolecule, replaces one

SHH

O

O

OHHO

O

NH2

OSH

O

OH

OHO

SH

(a) (b)

SH

(e)

(c)

O

HO

HO

HO

OH

SH

(d)

N N H

HN

N+

FIGURE 2 | Examples of thiolate ligands used inthe synthesis of water-soluble Aumonolayer-protected clusters (MPCs). (a) tiopronin,(b) glutathione, (c) 4-mercaptobenzoic acid,(d) 1-thio-β-D-glucose, (e) N, N, N-trimethyl(mercaptoundecyl)ammonium (TMA).


WIREs Nanomedicine and Nanobiotechnology Nanoparticle-based biologic mimetics

Aux(SR)y + n HSR′ Aux(SR)y−n(SR′)n + n HSR

FIGURE 3 | Scheme of the thiol place-exchange reaction. Thestoichiometry of the incoming to exiting ligand is 1:1.

of the original capping ligands in a 1:1 ratio. Placeexchange on nanoparticles was first described byMurray and coworkers who used alkanethiolateclusters with ω-functionalized thiols in toluene.2

This reaction has since been expanded to aqueoussolutions and can also be carried out in aqueousbuffer solutions.10 Multiple research groups havestudied the dynamics by which place exchange occursin ligand solutions. According to Murray’s work,the rate of ligand exchange depends upon boththe concentration of incoming and exiting ligands,implying an associative (Sn2-like) mechanism.33

Lennox and coworkers, on the other hand, report thatthe reaction is zero-order with respect to the incomingligand.34 Zerbetto’s lab found that the associativemechanism is accurate, but that the newly introducedligand interacts with multiple existing ligands on thecluster.35 These interactions cause the kinetics tochange as the reaction proceeds. Nevertheless, whilethe exact mechanism for place-exchange reactionsmay be complicated, the utility of place-exchangefor functionalizing MPCs results from its simplicity.

Reaction rates also play an important role in theplace-exchange dynamics. The reaction rate increaseswith smaller-sized entering ligands and shorter chainlength of the protecting ligand.33 Consequently, itis thermodynamically favorable to place-exchangea large biomolecule, such as a peptide or proteinfragment, with a small protecting ligand such astiopronin. Further, it is important to consider thatsubtle differences in the structure of the incomingligand, such as branching, can have a significanteffect on both the rate of place exchange and thestability of the monolayer.36 Additionally, it shouldbe noted that the reaction proceeds more favorablyat different sites on the core: vertex sites > edgesites > near-edge sites > terrace sites,33 as depictedin Figure 4.

The variations in reactivity due to thermody-namics and kinetics originate from the differences inelectron density37 and steric accessibility38 of thesesites. This unique property of nanoparticles leadsto some degree of predictability, and therefore con-trol, in where the place-exchanged functional groupswill anchor on the core. The rate of exchange isalso increased by oxidative electronic charging of thecore by electrochemical means39 or in the presence ofdioxygen.40 The extent of reaction can be enhanced

Interiorterracesites

Solutionthiol

Slow

Solutionthiol

Near-Edgeterracesites

Slow

Solutionthiol

Edgesites

Fast

Solutionthiol

Vertexsites

Fast

FIGURE 4 | Chart of the different rates at which place exchangeand (possibly) migration occur. (Reprinted, with permission, fromRef. 33. Copyright 1999 American Chemical Society.)

by increasing the incoming ligand concentration, butit should be noted that the extent of exchange rarelyapproaches 100%, owing to the difficulty of exchangeat terrace sites.33

It is also important to realize that the rateof ligand place exchange on MPCs becomes sloweras the particles age, probably because of a slowrearrangement of the ligands on the surface.41

Unfortunately, no kinetic or mechanistic study ofplace exchange has considered the new findings aboutthe presence of gold-thiolate tetramer rings on thesurface of MPCs as reported by the Cliffel group usingmass spectrometry28 and Hakkinen and coworkersin a theoretical paper.42 Most recently, Kornberget al. determined the specific crystal structure of ap-mercaptobenzoic acid MPC.43 The X-ray structureshowed surface bridging interactions between goldatoms and the thiol groups of the protecting ligands.Also, the structure contains conformational featuresspecific to the phenyl ligands, for example, phenylstacking, T stacking, and sulfur–phenyl interactions.All the recent findings show nuances in surfacestructure that could help to better explain thecomplexities of the place-exchange mechanism.

Solid-Phase Place ExchangeAs an alternative to the solution-phase place exchangediscussed above, Huo and coworkers have stud-ied solid-phase place-exchange reactions. In theiroriginal report of this reaction, they employeda polystyrene Wang resin with acetyl-protected6-mercaptohexanoic acid attached via an ester bond.44

The thiol groups were deprotected and allowed toundergo place exchange with butanethiolate-protectedgold nanoparticles, followed by washing away ofunexchanged product and cleaving of the exchangedparticles. Their results showed that they could place-exchange one ligand on to a particle surface. Thiswas proven using coupling chemistry to make dimernanoparticle complexes, rather than trimers or largeraggregates that would result from multiple exchangedligands.



The same group has compared this solid-phaseapproach to the solution-phase approach and foundthe solid phase approach to be advantageous in termsof controlling the number of ligands attached percluster and preserving the order of ligands on thesurface.45 Recently, the same group has reported asolid-phase approach using a noncovalent interactionof the incoming ligand with silica gel,46 which isdepicted in Figure 5. This strategy employs milderreaction conditions, thereby making it amenable toa wider class of molecules, such as large biologicallyrelevant functional groups.

Adapting Biological Functional Groups to PlaceExchangeMacromolecules often contain thiols, for example,cysteine residues in proteins or adenosyl phospho-thioate residues in DNA oligonucleotides.47 Thesegroups make biomolecules readily amenable to theplace-exchange reaction. Strategies to introduce thiolgroups into macromolecules include, but are cer-tainly not limited to, the use of Traut’s reagent(2-iminothiolane)48 in the case of proteins, the inclu-sion of terminal cysteine residues during the synthesisof peptides, and the conversion of phosphates tophosphorothioates using 3H-1,2-benzodithiole-3-one1,1-dioxide.49 It is also possible to introduce ligandsinto the MPC monolayer which undergo electrostaticinteractions with biomolecules, for example, the useof biotin–streptavidin interaction or biotin–anti-biotininteraction.50 All these routes provide straightforwardmethods to ready a group for place exchange andcreate functional nanoparticles.

11-mercaptoundecanoic acid

5% acetic acid

HS= COOH

Amine-functionalizedpolymer beads

Gold nanoparticles

−+

+++

+++

++ + +

+

++

++

+++

++

−

−

−−

−

−

−

−

−+

++ + +

+

++

++

++

+

+

++

+ + ++

+

++

+−

−

−

−

−

−

−

−

FIGURE 5 | The noncovalent interaction based place-exchangereaction. (Reprinted, with permission, Wiley Periodicals, Inc.)

Direct Monolayer FunctionalizationThere are other strategies to functionalize MPCs. Animportant class of these strategies, which is gainingpopularity, is the use of simple organic reactions onligands already bound to the MPC. Examples includetriazole cycloaddition to a bromine functionality,51

direct functionalization of a hydroxyl group,52 amidecoupling, and ester coupling.3 All these methodsenable post-exchange reaction chemistry to occur,allowing a surface to be modified in a controlledfashion.

Characterization Methodsfor Monolayer-Protected ClustersA thorough characterization of MPCs and func-tionalized MPCs is critical before applying them inbiological uses. Determination of core size is easilyaccomplished via transmission electron microscopy(TEM).1 Core sizes have also been determined by massspectrometric methods.53,54 Thermogravimetric anal-ysis (TGA)55 provides an easy method to determinethe ratio of organic ligand to inorganic core mate-rial. Combining the core size and organic : inorganicratio data yields an approximate average molecu-lar formula for homofunctionalized MPCs.54 Nuclearmagnetic resonance (NMR) spectroscopy is useful fordetermining the structure and composition of the pro-tecting monolayer. Protecting ligands have broadenedpeaks in both 1H and 13C spectra due to spin–spinrelaxational (T2) broadening, dispersity in bindingsites, and dipolar broadening due to packing densitygradients.56,57

Once functionalized as a biomolecular mimic, itis critical to characterize MPCs for two attributes.The first is the quantity of biologically relevantfunctional groups attached per cluster, generallyreported as an average of all the clusters in a sample.The second factor is the secondary structure of thebiomimics post conjugation. 1H NMR is a simpleway to semiquantitatively determine the numberof antigen peptides per cluster via integration ofknown protecting ligand peaks versus new broadenedbiomolecule peaks. The accuracy of this method canbe enhanced through the use of I2-induced MPCdecomposition (termed the ‘death reaction’), whichleads to sharper peaks with less overlap.58

Secondary structure determination has provento be more challenging. Drobny and coworkersdescribe the use of novel solid-state NMR techniquesto investigate the secondary structure of peptidesimmobilized on gold MPCs via amide coupling.59 Fortheir experiments, they used cross-polarization magicangle spinning (CPMAS) and double-quantum dipolar



recoupling with a windowless sequence (DQDRAWS).They showed that a peptide maintained a helicalstructure upon conjugation, but with a slight changein backbone torsion angle. Mandal and Kraatzrecently described similar characterizations of peptidesplace-exchanged onto MPCs using Fourier transforminfrared spectroscopy (FT-IR) and Fourier transformreflection absorption spectroscopy (FT-RAIRS).60

Using amide I bands, they observed that the secondarystructure of a leucine-rich peptide bound to goldtransitions from α-helical to β-sheet with greatersurface curvature. Results showed that free peptides,2-D self-assembled monolayers (SAMs) on gold, andpeptides on 20-nm gold MPCs showed α-helicalstructure because of less surface curvature. However,10-nm and particularly 5-nm gold MPCs showedincreasing amounts of β-sheet conformation due tothe increased surface curvature. Understanding thenature of primary and secondary structure becomescritical as functionalized nanoparticles are used forpractical applications.

ANTIGENIC VALIDATION USINGIMMUNOASSAYS

One of the most effective ways to validate function-alized nanoparticles’ ability to mimic a biologicalantigen is to look for recognition from a specific anti-body. Since antibodies are generally targeted for oneepitope of interest, they allow for specificity and serveas the keystone in many bioanalytical techniques. Thissection will quickly highlight some selected analyticaltools used to detect antigen mimics in a sensitive andspecific fashion.

Enzyme-Linked Immunosorbent AssayELISA describes a family of techniques used for thevalidation of antigen–antibody interaction throughthe detection of antigen or antibody. This techniquewas first described by Engvall and Perlmann in1971.61 ELISA generally involves the adsorption ofan antigen onto a plastic substrate, followed byrecognition with a primary antibody. Then, detectablesecondary antibody, specific to the primary, isincubated. Secondary antibodies use tags: for example,horseradish peroxidase or alkaline phosphatase thatgive a detectable signal upon activation with a specificsubstrate.

Quartz Crystal MicrobalanceA powerful tool used by our lab and others to detectantigens against specific antibodies is the quartz crystal

Au0

SSSS

SS

SS

Protein A

FIGURE 6 | The quartz crystal microbalance (QCM) biosensorshowing an antibody recognizing a functionalized nanoparticle (antigenmimic). (Reprinted, with permission, Wiley Periodicals, Inc.)

microbalance (QCM).47,62–68 This technique is basedon a piezoelectric oscillator that changes frequencywith the addition of a mass load, specifically, theantigen. This frequency shift is then converted to amass load, so the instrument acts as a highly sensitivemass balance. Depicted in Figure 6 is a cartoonrepresentation of a QCM biosensor, with a genericantibody–antigen system. The detection limit of QCMtechnology is continuously increasing as higher-frequency crystals are developed, reaching easily to thenanogram level and down to hundreds of picograms.A convenient reason to use QCM for measuringbiomimic binding is the built-in amplification ofusing nanoparticles. Since QCM is essentially a massdetection method, the large molecular weight ofthe gold nanoparticle improves the sensitivity forbiomimic studies.

Surface Plasmon ResonanceAnother popular tool for bioanalytical measure-ments is the optical technique, surface plasmonresonance (SPR).25,67–69 This technique detects therefractive change of an incident laser source, whichthen translates into the on and off rates of theantigens. SPR utilizes commercially available goldsurfaces with prefabricated substrates. These pre-fabricated substrates allow easier surface function-alization to create the biosensor. With subnanogramdetection limits, SPR is another powerful tool forbioassays.



NANOPARTICLE-BASED MIMETICS

Protein-Functionalized NanoparticlesProtein A–Coated NanoparticlesNanoparticles are capable of being functionalizedwith whole proteins, while still undergoing the samebiomolecular recognition events as the free proteins.Recently, Rosenzweig and Thanh demonstratedthe viability of biomolecular recognition of wholeprotein–coated gold nanoparticles in the developmentof an aggregation-based assay.70 They were ableto detect anti-protein A in serum by aggregatingprotein A-coated gold nanoparticles and observingan absorbance change at 620 nm.

Antibody Fragment–Conjugated NanoparticlesKornberg and coworkers described single-chain Fv(scFv) antibody fragments conjugated to glutathionegold MPCs.71 The scFv were rigidly coupled andexhibited specificity in binding to the antigen protein.By eliminating the flexible regions present in the wholeantibody, rigidity was accomplished. Conjugationwas accomplished by attaching a cysteine-terminatedC-terminal affinity tag (FLAG) to the scFv. To assistthe place exchange of the glutathione with scFv, theyused oxidative charging of the metal core. Usingcryo-electron microscopy, they were able to verifythe antibody activity by observing the attachment offour Au71–scFv–glutathione units to single tetramericinfluenza N9 neuraminidase units. In both these cases,the nanoparticle was used to aid in the detection ofantibody–antigen binding, without actually using it asa biomimetic building block.

Peptide-Functionalized Gold NanoparticlesPeptide-Functionalized Particles for BiologicalRecognitionNanoparticles can be surface-functionalized by par-ticle assembly and stabilization with a peptide or byplace-exchange with the ligand after particle assem-bly. The first example of biologically relevant particlesis the synthesis of nanoparticles with a protectingpeptide from the histidine-rich protein II (HRP-II)of Plasmodium falciparum.72 Using standard fmocprocedures, Wright and coworkers recreated this pep-tide from HRP II and used it as a stabilizing ligandon different metal core particles: ZnS, Au0, Ag0,TiO2, and AgS. The biological significance comesfrom the recognition of the particle by a monoclonalantibody specific for P. falciparum. They were ableto detect the peptide-encapsulated particles as theywould the whole protein. This antibody–nanoparticlerecognition shows that their particle mimics the nativeepitope.

Recently, Cliffel and coworkers developedseveral MPCs that mimic antigens of interest.The first was a glutathione (GSH)-passivated goldcluster (GSH-MPC) that was then detected witha polyclonal anti-GSH antibody.63 The antibodyvery specifically recognized the GSH-MPC versusa standard tiopronin-passivated nanoparticle, eventhough both surface ligands only differ by about oneamino acid, seen in Figure 7. While glutathione is not atraditional antigen, it serves as a proof of concept thatan MPC can be functionalized with a surface peptide,and then specifically recognized via its antibody.

Another MPC this group synthesized containsan epitope from the hemagglutinin (HA) protein ofinfluenza,64 termed an HA-MPC. The 10-amino acidpeptide was again synthesized with standard fmocprocedures with a terminating cysteine residue topromote place-exchange chemistry. This peptide wasselected because it is a neutralizing site for influenzaand there was a commercially available monoclonalantibody specific for this epitope on HA. Also, thisexperiment compared 2-D SAMs to 3-D nanoparticlesas depicted in Figure 8. It was shown that the HA-MPC was more efficient in presenting the peptide tothe antibody, resulting in a higher ratio of antibody topeptide binding when compared to the 2-D surface.

Another novel feature of epitope presentingMPCs is that they can be size-separated. Using aspecific sized particle, the peptide is forced intoadopting a conformation closer to the native structure.Previous work by Murray and coworkers had shownthat ligands are dynamically attached to the surfaceand will therefore migrate across the MPC to findthe most stable conformation possible.3,33,73 Cliffel’sresearch group applied this concept to their work onthe protective antigen (PA) of Bacillus anthracis.66

The PA protein is one of three precursors of theanthrax toxin. PA was selected because it precedes the

O OHS

HN

NH

O

HO

O

OH

NH2

SH

NH

O

O

OH

(a)

(b)

FIGURE 7 | Comparison of the glutathione (a) and tiopronin(b) ligands used to functionalize MPCs. Tiopronin is a truncated form ofthe 3-amino acid glutathione, with overlap shown in red. (Reprinted,with permission, from Ref. 62. Copyright 2005 American ChemicalSociety.)



CysCys

Cys

Cys

CysCys

Cys

Cys C

ysC

ys

Cys

Cys

Cys

Cys

Cys

Cys

(a) (b) (c)

Au Au

FIGURE 8 | Different nanostructures used: (a) the 3-D GSH-MPC, (b) the 3-D 10-amino acid hemagglutinin monolayer-protected clusters(HA-MPC), and (c) the same 10-amino acid sequence for HA as a 2-D self-assembled monolayer (SAM) with tiopronin spacers. (Reprinted, withpermission, from Ref. 63. Copyright 2005 American Chemical Society.)

other two proteins (edema factor and lethal factor)in their transport for infection, which makes it anideal target for neutralizing antibodies. Specifically,the C-terminus and two loops of the PA protein wereidentified as cell-receptor sites, making them the bestcandidates for their work.

Again, tiopronin MPCs were used and place-exchanged with the relevant peptide for the regionson PA. Since some of the PA epitopes selectedwere loop regions, the peptide was designed so thatit could mimic its native conformation by puttingcysteine residues on both the N- and C-termini. Thisallowed bidentate attachment across the nanoparticlesurface to reconstruct the natural loop. Shown inFigure 9 is the stepwise process in the creation of theconformational mimic.

For comparison, a second cluster was createdthat only had a cysteine on the C-terminus formonodentate attachment. This creates two types ofclusters, both with the proper primary structure, butonly one with the secondary structure closer to thenative conformation, as illustrated in Figure 10.

A QCM-based antibody–antigen binding studyrevealed that the loop-presenting cluster was more

Monolayer-protected

cluster

Conformationalmimic

Cys

Cys

Syntheticpeptideepitope

Au0

FIGURE 9 | Step-by-step creation of the conformational mimicnanoparticle using the synthetic protective antigen (PA) peptide.

DCys

Cys

Cys

Cys

BA

C

Cys

CysCys

FIGURE 10 | Loop presenting MPCs (left) shown compared to thenative protein (middle) and the linear presenting monolayer-protectedcluster (MPC) (right). (Reprinted, with permission, Wiley Periodicals,Inc.)

strongly recognized than the linear epitope cluster.More specifically, the loop epitope had a higheraffinity constant (Ka) for this particular antibody thanthe linear epitope, especially at physiological salineconcentrations. This data shows that the bidentatestructure was better recognized and bound moretightly. This suggests that the commercial antibodymay have a conformational paratope. The QCM wasused to detect the antibody binding to the MPC, whichwas electrostatically held to the sensor.

Peptides are not only limited to use as thefunctional group. Naik and coworkers used peptidesin a novel and sophisticated way. Multifunctionalpeptides were used as the reducing agent, gold-protecting ligand, and presenting epitope.74 PeptideA3 was selected from a phage peptide display libraryand found to both bind to gold and reduce it. Flg,



a peptide commonly used in tagging proteins with abiomolecular recognition domain, was also found toreduce gold. They were able to produce Flg-A3 andA3-Flg gold nanoparticles in a one-pot synthesis withgood monodispersity and were capable of binding toanti-Flg IgG on glass slides.

As an extension of peptide-epitope-protectedgold MPCs, a collaboration of the Cliffel and Wrightresearch groups synthesized tiopronin MPCs function-alized with either the flag epitope (flag-MPC), HA epi-tope (HA-MPC), flag and HA epitope (flag/HA-MPC),or no epitope.75 The peptide epitopes were attachedto the cluster via a cysteine-terminated polyethy-lene glycol (PEG) hexamer using place exchange.The PEG linker provides enhanced accessibility bymoving the epitope away from the particle’s sur-face. QCM immunosensors, as previously described,using either anti-flag or anti-HA IgG were used toevaluate the immunological activity of the mimics syn-thesized. They were able to detect the HA-MPC andHA/flag-MPC using the anti-HA immunosensor, andthe flag-MPC and HA/flag-MPC using the anti-flagimmunosensor. Neither one detected the tioproninMPCs without peptide epitopes. In all these trials,biological recognition serves as a quick means tovalidate peptide nanoparticles and determine bindingconstants.

Peptide-Functionalized Gold Nanoparticlesfor Biological TargetingBiomimetic nanoparticles have shown promise as atool for targeted cell entry. Targeted entry is complex,but the small size of gold nanoparticles and thefunctionality available from synthetic peptides makethis delicate task a possibility. Inspired by viruses,Tkachenko and coworkers conjugated peptides tobovine serum albumin (BSA) via an ester linker, andthen conjugated the BSA to gold nanoparticles.76

The four peptides they used were from viral cellentry/targeting proteins, and they were able to achievetargeted entry of the gold nanoparticles into thenucleus of HepG2 cells. Furthermore, it should benoted that the cells were still viable after entry of thegold nanoparticles.76

Cell-Binding Studies of FunctionalizedNanoparticlesGold nanoparticles, as previously mentioned, can befunctionalized with many different ligands. Resultsfrom Rotello’s group show that the charge of thecapping ligand can affect how the particle binds tocell surfaces.77 Positively charged ligands, such asTMA, cause an attraction between the particle andthe negatively charged cell wall. The increased binding

leads to higher toxicity and cell lysis. Conversely, thesame negatively charged cell wall has little attractionto a carboxylate nanoparticle, leading to less celllysis. Cell walls with no overall charge, however,lysed slightly more with negatively charged particles.These findings present interesting considerations whenconducting studies at the cellular level.

Further work by Schmid and coworkers hasshown that very small gold nanoparticles (Au55 cores,1.4 nm) can actually bind to DNA in the cell. This isdue to gold’s preference for the negatively chargedbackbone of DNA, which partially removes theprotecting ligand group.78 Au55 clusters entrenchedin the DNA grooves are depicted in Figure 11. Cellentry and specific targeting can serve as tools to eithermark cells for imaging or cause controlled cell death.

Peptide-Functionalized Gold Nanoparticles forBiological CatalysisThe first report of using multifunctionalized goldMPCs for catalysis was by Frigeri and cowork-ers earlier this decade.79 Using N-methylimidazole-functionalized gold nanoparticles, they were able tocatalyze the hydrolysis of an activated ester. Scriminand coworkers have created water-soluble gold MPCsplace-exchanged with histidine-phenylalanine dipep-tides that are capable of mimicking hydrolyticenzymes.80 These two examples represent stepstoward gold nanoparticle–based enzyme mimicsthat inspired Scrimin and coworkers to term them‘nanozymes’. More recently, Morse and coworkerswere able to use gold nanoparticles to mimic thecatalytic activity of an enzyme in the sponge Tethyaaurantia responsible for producing silica needles bysimply conjugating organic molecules to the protect-ing monolayer of gold nanoparticles.81 The catalyticsite of the aforementioned enzyme in T. aurantia uses anucleophilic -OH group interacting with a hydrogen-bonding imidazole group to accomplish hydrolysis of

FIGURE 11 | Model of Au55 clusters irreversibly bound to thegrooves of DNA. (Reprinted, with permission, Wiley Periodicals, Inc.)



FIGURE 12 | Hydroxy- andimidazole-functionalizednanoparticles workingtogether to catalyze silicaformation. The ligands used tofunctionalize the particles areshown in (a), while theinteraction between particlesin (b) and (c). Part (d) showsthe stepwise synthetic route ofthe ligands. (Reprinted, withpermission, Wiley Periodicals,Inc.)

N NH

AuS

N

O

AuS O

H O H

N

HN

N(CH2)N

O

SAu

Si

OEt

OEt

OEt

EtO

AuS

(CH2)N

AuS

(CH2)N

O

N

HN

N(CH2)N

O

SAu

Si

OEt

OEtEtO

HOEt

H2O

AuS

(CH2)N O N

HN

N(CH2)N

O

SAu

Si

OEt

OEtEtO H

O

H

AuS (CH2)N O H

N

HN

N(CH2)N

O

SAu

Si

OEt

OEt

OEt

EtO

+

(b)

(a) (d)

(c)

a silicon alkoxide precursor and subsequent polycon-densation to silica. Hydroxy-terminated nanoparticleswere afforded simply by using 11-mercaptoundecanolas the protecting ligand in a Brust synthesis. Theimidazole-terminated nanoparticles were obtainedby using amide coupling of an imidazole func-tionality to 11-mercaptoundecanoic-protected goldnanoparticles. This idea is illustrated in Figure 12.

Carbohydrate-Functionalized NanoparticlesMany important processes in biology rely oncarbohydrate–protein interactions, and it may becomeconvenient to functionalize gold nanoparticles withcarbohydrates instead of proteins or peptides. Thiswas first accomplished by Penades and coworkerswhen they used carbohydrate-functionalized goldnanoparticles to mimic glycocalyx, the sticky filmfound on the outside of many different cells.82

As a further example of non-protein-relatedgold nanoparticle biomimetics, Chen and cowork-ers observed high affinity and specificity for bindingof carbohydrate-encapsulated nanoparticles to con-canavalin A.83 Carbohydrates were attached to the

gold core using a thiol linker in a place-exchange reac-tion. Interaction with concanavalin A was monitoredusing SPR.

CONCLUSIONThe exciting field of nanoparticle-based biologicalmimetics is rapidly developing. The results willcontinue to grow as new discoveries are made in thesecondary structure of proteins, opening the door foreven more material to incorporate into an expandingtoolbox. Also, new techniques are being created tobetter characterize the interface between biologicalrecognition and nanoscale structures.

The field has brought out many interestingresults concerning the synthesis and conjugation ofnanoparticle-based mimetics that can compete withtheir native analogs in functional complexity. Thefuture of this field will be focused on creatingmore sophisticated nanoparticle-based biomimetics,centered on biological recognition, catalysis, andtargeting. Metal nanoparticle–based drug targets maywork well under ideal conditions for analytical meritand in vitro studies, but their effect in complex



biological systems is only beginning to unravel. Thefield of biological mimetics will continue to thrive,

with nanoparticles playing a critical role in research,diagnostics, and therapeutics.

REFERENCES

1. Brust M, Walker M, Bethell D, Schiffrin DJ,Whyman R. Synthesis of thiol-derivatized goldnanoparticles in a two-phase liquid-liquid system.Chem Commun 1994, 7:801–902.

2. Hostetler MJ, Green SJ, Stokes JJ, Murray RW. Mono-layers in three dimensions: synthesis and electrochem-istry of omega-functionalized alkanethiolate-stabilizedgold cluster compounds. J Am Chem Soc 1996,118(17):4212–4213.

3. Templeton AC, Hostetler MJ, Warmoth EK, Chen S,Hartshorn CM, et al. Gateway reactions to diverse,polyfunctional monolayer-protected gold clusters. J AmChem Soc 1998, 120:4845–4849.

4. Faraday M. Experimental relations of gold (and othermetals) to light. Philos Trans R Soc Lond 1857,147:145–181.

5. Turkevich J, Stevenson PC, Hillier J. A study of thenucleation and growth processes in the synthesis ofcolloidal gold. Discuss Faraday Soc 1951, 11:55–75.

6. TempletonAC,WuelfingWP, MurrayRW. Monolayer-protected cluster molecules. Acc Chem Res 2000,33:27–36.

7. Niemeyer CM. Nanoparticles, proteins, and nucleicacids: biotechnology meets materials science. AngewChem Int Ed 2001, 40:4128–4158.

8. Rowe RP, Plass KE, Kim K, Kurdak C, Zellers ET,et al. Single-phase synthesis of functionalized goldnanoparticles. Chem Mater 2004, 16:3513–3517.

9. Templeton AC, Chen S, Gross SM, Murray RW.Water-soluble, isolable gold clusters protected by tio-pronin and coenzyme a monolayers. Langmuir 1999,15:66–76.

10. Templeton AC, Cliffel DE, Murray RW. Redoxand fluorophore functionalization of water-soluble,tiopronin-protected gold clusters. J Am Chem Soc 1999,121:7081–7089.

11. Yee C, Scotti M, Ulman A, White H, Rafailovich M,et al. One-phase synthesis of thiol-functionalized plat-inum nanoparticles. Langmuir 1999, 15:4314–4316.

12. Yee CK, Jordan R, Ulman A, White H, King A, et al.Novel one-phase synthesis of thiol-functionalized goldpalladium and iridium nanoparticles using superhy-dride. Langmuir 1999, 15:3486–3491.

13. Croce TA, Hamilton SK, Chen ML, Muchalski H,Harth E. Alternative o-chinodimethane crosslinkingprecursors for intramolecular chain collapse nanopar-ticles. Macromolecules 2007, 40:6028–6031.

14. Ahmed F, Photos PJ, Discher DE. Polymersomes asviral capsid mimics. Drug Dev Res 2006, 67:4–14.

15. Joralemon MJ, Smith NL, Holowka D, Baird B, Woo-ley KL. Antigen-decorated shell cross-linked nanopar-ticles: synthesis, characterization, and antibody inter-actions. Bioconjug Chem 2005, 16:1246–1256.

16. Taubert A, Napoli A, Meier W. Self-assembly of reac-tive amphiphlic block copolymers as mimetics forbiological membranes. Curr Opin Chem Biol 2004,8:598–603.

17. Bosio CM, Moore BD, Warfield KL, Ruthel G,Mohamadzadeh M, et al. Ebola and Marburg virus-like particles activate human myeloid dendritic cells.Virology 2004, 326:280–287.

18. Noad R, Roy P. Virus-like particles as immunogens.Trends Microbiol 2003, 11(9):438–444.

19. Ogasawara Y, Amexis G, Yamaguchi H, Kajigaya S,Leppla SH, et al. Recombinant viral-like particles ofparvovirus B19 as antigen carriers of anthrax protectiveantigen. In vivo 2006, 20:319–324.

20. Raja KS, Wang Q, Finn MG. Icosahedral virus par-ticles as polyvalent carbohydrates display platforms.Chembiochem 2003, 4:1348–1351.

21. Szecsi J, Boson B, Johnsson P, Dupeyrot-Lacs P,Matrosovich M, et al. Induction of neutralising anti-bodies by virus-like particles harbouring surface pro-teins from highly pathogenic H5N1 and H7N1influenza viruses. Virol J 2006, 3:1–7.

22. Wang Q, Lin T, Tang L, Johnson JE, Finn MG. Icosa-hedral virus particles as addressable nanoscale buildingblocks. Angew Chem Int Ed 2002, 41:459–462.

23. Jahanshahi M, Williams S, Lyddiatt A, Shojaosa-dati SA. Preparation and purification of synthetic pro-tein nanoparticles. IEE Proc Nanobiotechnol 2004,151(5):176–182.

24. Gole A, Vyas S, Phadtare S, Lachke A, Sastry M. Stud-ies on the formation of bioconjugates of endoglucanasewith colloidal gold. Colloids Surf B 2002, 25:129–138.

25. Lyon LA, Pena DJ, Natan MJ. Surface plasmon res-onance of au colloid-modified au films: particle sizedependence. J Phys Chem B 1999, 103:5826–5831.

26. Matsuya T, Tashiro S, Hoshino N, Shibata N,Nagasaki Y, et al. A core-shell type fluorescentnanosphere possessing reactive poly(ethylene glycol)tethered chains on the surface of zeptomole detectionof protein in time-resolved fluorometric immunoassay.Anal Chem 2003, 75:6124–6132.



27. Whetten RL, Shafigullin MN, Khoury JT, Schaaff TG,Vezmar I, et al. Crystal structures of molecular goldnanocrystal arrays. Acc Chem Res 1999, 32:397–406.

28. Gies AP, Hercules DM, Gerdon AE, Cliffel DE. Elec-trospray mass spectrometry study of tioproninmonolayer-protected gold nanoclusters. J Am ChemSoc 2007, 129(5):1095–1104.

29. Schaaff TG, Knight G, Shafigullin MN, Borkman RF,Whetten RL. Isolation and selected properties of a 10.4kDa gold: glutathione cluster compound. J Phys ChemB 1998, 102(52):10643–10646.

30. Johnson SR, Evans SD, Brydson R. Influence of aterminal functionality on the physical propertiesof surfactant-stabilized gold nanoparticles. Langmuir1998, 14(23):6639–6647.

31. Ackerson CJ, Jadzinsky PD, Kornberg RD. Thiolateligands for synthesis of water-soluble gold clusters.J Am Chem Soc 2005, 127(18):6550–6551.

32. Cliffel DE, Zamborini FP, Gross SM, Murray RW.Mercaptoammonium-monolayer-protected, water-soluble gold, silver, and palladium clusters. Langmuir2000, 16(25):9699–9702.

33. Hostetler MJ, Templeton AC, Murray RW. Dynam-ics of place-exhange reactions on monolayer-protected gold cluster molecules. Langmuir 1999,15:3782–3789.

34. Kassam A, Bremner G, Clark B, Ulibarri G,Lennox RB. Place exchange reactions of alkyl thi-ols on gold nanoparticles. J Am Chem Soc 2006,128(11):3476–3477.

35. Montalti M, Prodi L, Zaccheroni N, Baxter R,Teobaldi G, et al. Kinetics of place-exchange reac-tions of thiols on gold nanoparticles. Langmuir 2003,19(12):5172–5174.

36. Hong R, Fernandez JM, Nakade H, Arvizo R,Emrick T, et al. In situ observation of place exchangereactions of gold nanoparticles. Correlation of mono-layer structure and stability. Chem Commun 2006,(22):2347–2349.

37. Hakkinen H, Barnett RN, Landman U. Electronicstructure of passivated Au-38(SCH3)(24) nanocrystal.Phys Rev Lett 1999, 82(16):3264–3267.

38. Donkers RL, Song Y, Murray RW. Substituent effectson the exchange dynamics of ligands on 1.6 nmdiameter gold nanoparticles. Langmuir 2004,20(11):4703–4707.

39. Song Y, Murray RW. Dynamics and extent of lig-and exchange depend on electronic charge ofmetal nanoparticles. J Am Chem Soc 2002,124(24):7096–7102.

40. Myung N, Bae Y, Bard AJ. Effect of surface pas-sivation on the electrogenerated chemiluminescenceof CdSe/ZnSe nanocrystals. Nano Lett 2003,3(8):1053–1055.

41. Chechik V. Reduced reactivity of aged Au nanoparticiesin ligand exchange reactions. J Am Chem Soc 2004,126(25):7780–7781.

42. Hakkinen H, Walter M, Gronbeck H. Divide and pro-tect: capping gold nanoclusters with molecular gold-thiolate rings. J Phys Chem B 2006, 110:9927–9931.

43. Jadzinsky PD, Calero G, Ackerson CJ, Bushnell DA,Kornberg RD. Structure of a thiol monolayer-protectedgold nanoparticle at 1.1 A resolution. Science 2007,318:430–433.

44. Worden JG, Shaffer AW, Huo Q. Controlled func-tionalization of gold nanoparticles through a solidphase synthesis approach. Chem Commun 2004,(5):518–519.

45. Shaffer AW, Worden JG, Huo Q. Comparison study ofthe solution phase versus solid phase place exchangereactions in the controlled functionalization of goldnanoparticles. Langmuir 2004, 20(19):8343–8351.

46. Liu X, Worden JG, Dai Q, Zou JH, Wang JH, et al.Monofunctional gold nanopartictes prepared via anoncovalent-interaction-based solid-phase modifica-tion approach. Small 2006, 2(10):1126–1129.

47. Bardea A, Dagan A, Ben-Dov I, Amit B, Will-ner I. Amplified microgravimetric quartz-crystal-microbalance analyses of olignucleotide complexes: aroute to a Tay-Sachs biosensor device. Chem Commun1998, 839–840.

48. Ghosh SS, Kao PM, McCue AW, Chappelle HL. Use ofmaleimide-thiol coupling chemistry for efficient synthe-ses of oligonucleotide-enzyme conjugate hybridizationprobes. Bioconjug Chem 1990, 1:71–76.

49. Iyer RP, Egan W, Regan JB, Beaucage SL. 3H-1,2-benzodithiole-3-one 1,1-dioxide as an improved sul-furizing reagent in the solid-phase synthesis ofoligodeoxyribonucleoside phosphorothioates. J AmChem Soc 1990, 112:1253–1254.

50. Connolly S, Fitzmaurice D. Programmed assembly ofgold nanocrystals in aqueous solution. Adv Mater 1999,11(14):1202–1205.

51. Fleming DA, Thode CJ, Williams ME. Triazolecycloaddition as a general route for functional-ization of Au nanoparticles. Chem Mater 2006,18(9):2327–2334.

52. Tan H, Zhan T, Fan WY. Direct functionalizationof the hydroxyl group of the 6-mercapto-1-hexanol(MCH) ligand attached to gold nanoclusters. J PhysChem B 2006, 110(43):21690–21693.

53. Tai Y, Murakami J, Saito K, Ikeyama M, Tajiri K,et al. Plasma desorption mass spectroscopy of thiol-passivated gold nanoparticles. Eur Phys J D 2003,24(1–3):261–263.

54. Whetten RL, Khoury JT, Alvarez MM, Murthy S,Vezmar I, et al. Nanocrystal gold molecules. Adv Mater1996, 8(5):428.



55. Terrill RH, Postlethwaite TA, Chen CH, Poon CD,Terzis A, et al. Monolayers in three dimensions: NMR,SAXS, thermal, and electron hopping studies of alka-nethiol stabilized gold clusters. J Am Chem Soc 1995,117(50):12537–12548.

56. Badia A, Cuccia L, Demers L, Morin F, Lennox RB.Structure and dynamics in alkanethiolate monolayersself-assembled on gold nanoparticles: A DSC, FT-IR,and deuterium NMR study. J Am Chem Soc 1997,119(11):2682–2692.

57. Hostetler MJ, Wingate JE, Zhong CJ, Harris JE,Vachet RW, et al. Alkanethiolate gold cluster moleculeswith core diameters from 1.5 to 5.2 nm: core and mono-layer properties as a function of core size. Langmuir1998, 14(1):17–30.

58. Templeton AC, Hostetler MJ, Kraft CT, Murray RW.Reactivity of monolayer-protected gold clustermolecules: steric effects. J Am Chem Soc 1998,120:1906–1911.

59. Bower PV, Louie EA, Long JR, Stayton PS,Drobny GP. Solid-state NMR structural studies ofpeptides immobilized on gold nanoparticles. Langmuir2005, 21(7):3002–3007.

60. Mandal HS, Kraatz H-B. Effect of the surface curva-ture on the secondary structure of peptides adsorbedon nanoparticles. J Am Chem Soc 2007, 129:6356–6357.

61. Engvall E, Perlmann P. Enzyme-linked immunosorbentassay (Elisa) quantitative assay of Immunoglobulin-G.Immunochemistry 1971, 8(9):871.

62. Patolsky F, Ranjit K, Lichtenstein A, Willner I.Dendritic amplification of DNA analysis byoligonucleotide-functionalized Au-nanoparticles.Chem Commun 2000, 1025–1026.

63. Gerdon AE, Wright DW, Cliffel DE. Quartz crystalmicrobalance detection of glutathione-protected nan-oclusters using antibody recognition. Anal Chem 2005,77:304–310.

64. Gerdon AE, Wright DW, Cliffel DE. Hemagglutininlinear epitope presentation on monolayer-protectedclusters elicits strong antibody binding. Biomacro-molecules 2005, 6:3419–3424.

65. Gerdon AE, Wright DW, Cliffel DE. Quartz crystalmicrobalance characterization of nanostructures assem-blies in biosensing. In: Kumar CSSR: ed. NanosystemCharacterization Tools in the Life Sciences, Vol. 3:Wiley-VCH;2006, 109–144.

66. Gerdon AE, Wright DW, Cliffel DE. Epitope mappingof the protective antigen of B. Anthracis by using nan-oclusters presenting conformational peptide epitopes.Angew Chem Int Ed 2006, 45:594–598.

67. Janshoff A, Galla H-J, Steinem C. Piezoelectric mass-sensing devices as biosensors - an alternative tooptical biosensors? Angew Chem Int Ed 2000, 39:4004–4032.

68. Koßlinger C, Uttenthaler E, Drost S, Aberl F, Wolf H,et al. Comparison of the QCM and the SPR method forsurface studies and immunological applications. SensActuators B 1995, 24-25:107–112.

69. Hutter E, Fendler JH, Roy D. Surface plasmon reso-nance studies of gold and silver nanoparticles linkedto gold and silver substrates by 2-aminoethanethioland 1,6-hexanedithiol. J Phys Chem B 2001,105:11159–11168.

70. Thanh NTK, Rosenzweig Z. Development of anaggregation-based immunoassay for anti-proteinA using gold nanoparticles. Anal Chem 2002,74(7):1624–1628.

71. Ackerson CJ, Jadzinsky PD, Jensen GJ, Kornberg RD.Rigid, specific, and discrete gold nanoparticle/antibodyconjugates. J Am Chem Soc 2006, 128(8):2635–2640.

72. Slocik JM, Moore JT, Wright DW. Monoclonal anti-body recognition of histidine-rich peptide encapsulatednanoclusters. Nano Lett 2002, 2(3):169–173.

73. Ingram RS, Hostetler MJ, Murray RW. Poly-hetero-ω -fuctionalized alkanethiolate-stabilized gold clustercompounds. J Am Chem Soc 1997, 119:9175–9178.

74. Slocik JM, Stone MO, Naik RR. Synthesis of goldnanoparticles using multifunctional peptides. Small2005, 1(11):1048–1052.

75. Miller SA, Keil RG, Hiatt LA, Cliffel DE, Wright DW.Nanoparticle simulant for influenza virus. Nano Lett2007, In Submission.

76. Tkachenko AG, Xie H, Coleman D, Glomm W,Ryan J, et al. Multifunctional gold nanoparticle-peptide complexes for nuclear targeting. J Am ChemSoc 2003, 125(16):4700–4701.

77. Goodman CM, McCusker CD, Yilmaz T, Rotello VM.Toxicity of gold nanoparticles functionalized withcationic and anionic side chains. Bioconjug Chem 2004,15:897–900.

78. Tsoli M, Kuhn H, Brandau W, Esche H, Schmid G.Cellular uptake and toxicity of Au55 clusters. Small2005, 1(8–9):841–844.

79. Pasquato L, Rancan F, Scrimin P, Mancin F, Frigeri C.N-methylimidazole-functionalized gold nanoparticlesas catalysts for cleavage of a carboxylic acid ester.Chem Commun 2000, (22):2253–2254.

80. Pasquato L, Pengo P, Scrimin P. Functional goldnanoparticles for recognition and catalysis. J MaterChem 2004, 14(24):3481–3487.

81. Kisailus D, Najarian M, Weaver JC, Morse DE. Func-tionalized gold nanoparticles mimic catalytic activity ofa polysiloxane-synthesizing enzyme. Adv Mater 2005,17(10):1234.

82. Barrientos AG, de la Fuente JM, Rojas TC, Fernan-dez A, Penades S. Gold glyconanoparticles: syntheticpolyvalent ligands mimicking glycocalyx-like surfacesas tools for glycobiological studies. Chem Eur J 2003,9(9):1909–1921.



83. Lin CC, Yeh YC, Yang CY, Chen GF, Chen YC,et al. Quantitative analysis of multivalent interactionsof carbohydrate-encapsulated gold nanoparticles

with concanavalin A. Chem Commun 2003,(23):2920–2921.


Metal nanoparticles for DNA and antigen detection.Using nanoparticles to push the limits of detection.


Advanced Review

Informatics approaches foridentifying biologic relationshipsin time-series dataBrett A. McKinney∗

A vital goal of the genomic era is to identify biologic relationships betweengenes and gene products and to understand how these relationships influencephenotypes. Time course data contain a vast amount of causal and mechanisticinformation about complex systems, but experimental and informatics challengesmust be overcome to produce and extract this information from biologic systems.Mathematical modeling and bioinformatics methods are being developed inanticipation of experiments involving the coordinated measurement of cellularand molecular quantities at various spatial and temporal scales. Experimentalmethods that probe at the nanoscale will facilitate the exploration of biologicsystems at the single-cell and single-molecule level, but will also introduce specialchallenges for mathematical modeling because events at nanoscale concentrationsare subject to the influence of intrinsic noise. This review addresses the progress,challenges, and frontiers in the field of time-series informatics. The ultimate goalof time-series informatics is to move beyond descriptive relationships and towardpredictive models of emergent, or systemic, behaviors of biologic systems asa whole. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 60–68

In recent years, high-throughput, genome-wideexperiments have led to the vigorous develop-

ment of new bioinformatics tools and algorithmsthat identify genes and gene products associated withphenotypic variables and that model causal interac-tions in gene networks. These large-scale experimentshave created vast amounts of data, but most arelimited to a single snapshot in time, allowing onlya coarse approximation of the underlying dynamicsystem. Because of the lack of time-rich biologicdata, most bioinformatics efforts have focused ona static viewpoint of biology. For example, statis-tical methods to discriminate between phenotypicclasses from microarray data are reaching maturity.However, technologies such as kinetic reverse tran-scription polymerase chain reaction1,2 will soon allowfor the coordinated measurement of dense time-series

∗Correspondence to: Brett A. McKinney, Department of Genetics,University of Alabama School of Medicine, Birmingham, AL 35294,USA. E-mail: [email protected]

Department of Genetics, University of Alabama School of Medicine,Birmingham, AL, 35294 USA.

DOI: 10.1002/wnan.012

involving concentrations and expression levels of bio-logically active molecules. Bioinformatics approachesare under development in anticipation of the availabil-ity of more time-enriched data sets. This article reviewsthe promising developments and challenges in the areaof time-series bioinformatics. These challenges includemodel structure inference and parameter estimation,the identification of models that predict phenotypicoutcomes, and understanding the mechanisms of noiseregulation in biologic systems.

Nanobased approaches are better suited thanconventional methods to probe biologic systems atscales relevant to molecular mechanisms, in particularat the single-cell level.3,4 Processes at this scale involvestatistical fluctuations that may be large, and biologicsystems have necessarily evolved to function in thepresence of such noise. In fact, there is evidence thatbiologic networks exploit noise, but if not properlycontrolled, fluctuations may lead to inappropriatesystemic behavior of the network and possibly toan increased susceptibility to disease. Modeling timeseries at the nanoscale, where noise effects becomemore pronounced, is particularly challenging but mayhold important keys to understanding the etiology


WIREs Nanomedicine and Nanobiotechnology Biologic relationships in time-series data

phenotypes that have eluded standard statisticalanalysis.

GRAPH-BASED NETWORKSIn this article, a model-free network is defined as agraph whose nodes represent genes and gene productsand whose edges represent physical or functionalinteractions. The structure, or topology, of a graphprovides insight into the organizing principles ofcellular systems as well as into the functional motifsand interactions between genes and gene products.5

Statistical clustering is a potentially useful way to infergraph edges by identifying the patterns of correlationbetween profiles in time series, but inferring networkconnections by correlation or other metrics may beliethe underlying molecular interactions that are implicitin the time-series profiles. For example, consider thesimple hypothetical time series shown in Figure 1.Profiles for quantities y and z are the most correlatedand correlation-based clustering would predict a closeconnection between y and z, with x more distantlyrelated. In fact, y and z co-inhibit x; however,clustering cannot capture all of the complexity ofthe actual model shown in Figure 2. Specifically, thereis directionality and information flow to this negativefeedback loop with x activating y, y activating z, zsuppressing x, and each gene product degrading inproportion to its own concentration.

Clustering has limited the ability to identify thedirectional interactions displayed in Figure 2, but maybe useful for organizing information as input for moremechanistic algorithms discussed later in this review.For clustering time series, it is important to recognize

0 1 2 3 4 5 6 70

0.2

0.4

0.6

0.8

1

1.2

1.4

Time

Exp

ress

ion

(arb

itrar

y un

its)

xy

z

FIGURE 1 | Time-series profiles for hypothetical gene regulatorysystem described in Figure 2. Parameters used for the simulation areκ = (0.9, 1.0, 0.6), γ = (1.0, 0.6, 0.8), and θ = 0.9.

Gene 1

Repressor z Enzyme y

mRNA x k1q + z

g1xx⋅

y⋅

z⋅

−=

k2x g2y−=

k3y g3z−=

FIGURE 2 | Hypothetical single-gene regulatory network, simulatedin Figure 1, involving a negative feedback loop with measured geneproducts x, y, and z. A single gene with mRNA concentration xproduces an enzyme with concentration y. Enzyme y catalyzes areaction step leading to metabolite z, which inhibits the gene thatcodes for the enzyme. Parameters κ and γ are the production anddegradation constants and θ modulates the inhibitory Hill function.

that the time-series profile of each biomarker is not acollection of independent and identically distributed(iid) observations. To properly cluster time series, thedynamics should be taken into account explicitly, asfor example in Ref. 6, where the authors use a linearapproximation of the dynamics in the form of anautoregressive (AR) model to guide the clustering ofgene expression profiles. This method should be morereliable than metric-based clustering, but AR has thedisadvantage of being linear and univariate.

Graph theoretic methods facilitate the visual-ization of biologic relationships, but the lack of anunderlying model hinders graphical methods frommaking experimentally verifiable predictions from sys-tem perturbations. The nodes of a realistic biologicnetwork should represent time-varying activity andedges should represent flux through the network. Atthe other extreme of formalism complexity, nonlineardifferential equations, which will be discussed in thefollowing section, have been used to derive detailedmodels of gene networks, but the computationalcomplexity of numerical integration and parameterestimation currently limits the size of the network thatcan be analyzed. The rest of the review will focus onmechanistic and data-driven modeling approaches.

MODEL-BASED NETWORKSA Bayesian network (BN) combines probabilitytheory and graph theory by constructing directedacyclic graphs (DAGs) that represent the dependenciesbetween variables through probabilistic models. BNsprovide a probabilistic framework for network graphsand have been used for gene network analysis of staticgene expression data.7 However, BNs are designedfor time-independent data, and their basis in acyclicgraphs prevents them from handling feedback loops,which are necessary to model real biologic networks.



In contrast to graph-based networks, which arequalitative, static representations of cellular processesand pathways, dynamic model-based networks aregoverned by an underlying model that allows themto generate experimentally testable output. Numerousformalisms have been used to model kinetic data, butthis article will focus on more quantitative formalismsbased on differential equations. For a broad overviewof model formalisms, see Ref. 8.

A more reductionist approach to understandingcellular networks would be to model the noncovalentbonding and enzymatic reactions of each macro-molecule. However, this would not be feasible evenif precise quantitative biologic data were available;thus, in practice mathematical approximations to thephysical system are used. The simplest ordinary dif-ferential equation (ODE) that may describe biologictime-series panel data is a linear system:

dyi

dt=

n∑j=1

Aijyj, i = 1, . . . , n (1)

where y is a vector of functions describing the timevariation of each molecule i and the constant matrix Asummarizes the coupling strengths between moleculesi and j. The matrix A lends itself to visualization asa graphical network, but fails to accurately modelthe nonlinear dynamics of a real biologic system.Equation (1) can be expressed naturally as a dynamicBayesian network (DBN), which is a generalization ofBNs to time-series data. DBN is a promising inferencealgorithm that has been applied to the analysis of genenetworks from time series.9–12 A DBN can also beformulated as a Kalman filter (KF),13 a widely usedtool in engineering for tracking and estimation. Usinga linear system of ODEs like Eq. (1), the KF has beenused to estimate DBN parameters for gene networkinference.14 The KF is a Bayesian method in the sensethat it provides a way to incorporate prior informationto update the current state of the system. The KF canbe extended to nonlinear models by replacing thematrix in Eq. (1) with a vector of nonlinear functionsas in Eq. (2) below, but the model becomes moregeneral than a network, and the system might be betterdescribed as a dynamic Bayesian model or generalizedDBN (GDBN). The unscented Kalman filter (UKF)is an accurate and computationally efficient methodfor estimating parameters of nonlinear dynamicsystems.15,16 It has been used to model in vivo proteintime-series with noise and nonlinearities17,18 and hasthe ability to model unobserved state components, yetcomputational limitations still must be overcome forhigh-dimensional systems.

Numerous classes of ODE systems have beenproposed from mathematical biology to modelnonlinearity in biologic systems. For example, theoperon model19,20 with nonlinear Hill functions wasused to simulate the profiles shown in Figure 1. Thegeneral form of a nonlinear differential equation is

dyi

dt= fi[y(t), λ, εi], i = 1, . . . , n (2)

where the model f is a vector of nonlinear functionsof vector y with model parameter vector λ, andnoise vector ε. Popular nonlinear ODEs for modelingbiologic systems include generalized mass action(GMA) and synergistic-systems (S-systems).21–24 TheS-system preserves some of the interpretability of apurely graph-based approach while having the abilityto model realistic nonlinearities. In addition, theS-system has a bounded number of parameters,making it more efficient for analyzing cellular andmolecular networks than GMA. The canonical formof the S-system without noise is the following power-law system of nonlinear differential equations

Yi = αi

N∏k=1

Ygikk − βi

N∏k=1

Yhikk , i = 1, . . . , n.

Each equation for the time rate of changeof biochemical Yi is composed of a term fornet production from metabolic biomolecules thatcontribute to the increase of Yi with rate αi and aterm representing net degradation of Yi from catabolicbiomolecules with rate βi. The Ys may represent amolecule, cell, protein, or other gene product withinthe system. Kinetic order parameters gik and hik, on thereal number line, represent the regulatory influence ofYk on Yi. In principle, the S-system reduces structureidentification to parameter estimation; however, inpractice, the number of parameters is too largefor nonlinear systems-identification algorithms. Apromising approach to reduce the computationalexpense of parameter estimation is to decouple theODE system into independent algebraic equations.25

Parameters estimated in this way may be used as initialguesses to speed up other, more computationallyintensive, estimators.

It is rare that the biologic mechanism of a givenprocess is completely described; thus, one of the goalsof bioinformatics is to develop data-driven algorithmsto automate the identification of the model structureand parameters from time series. The enormity ofthe search space of possible model structures callsfor heuristic search methods such as evolutionaryalgorithms.17,18,26,27 When learning the structure of



a model, it is often necessary to include parsimonyconstraints in the objective function. A typical choicefor objective function involves some variant of leastsquares deviation of the model prediction from thetime-series panel data. It is often useful to divide theterms in the least squares sum by the correspondingdata value to prevent variables with extreme valuesfrom dominating the objective function. To penalizehigh-connectivity models that over-fit the data, onemay add a parsimony or complexity term that isusually a function of the number of parameters in themodel.28

SAMPLING FREQUENCY

In classical model inference, the model structure isfixed and a parameter is unidentifiable if it cannot beestimated from the data, no matter how large the sam-pling frequency is. The least squares definition of iden-tifiability is often used because it takes measurementerror into account.29 Identifiability is more difficultto assess for dynamic network inference from bio-logic time series because the model structure is oftennonlinear and/or unknown. For an experimenter, amore practical quantity is the minimum sampling fre-quency—or the number of time points sampled forthe duration of the experiment—needed to unequiv-ocally identify a model. If the experimental system isinsufficiently sampled, the system is underdetermined,meaning multiple models may fit the data. The prob-lem is analogous to a sample size calculation to achievea desired statistical power in a clinical trial involvingmultiple regression.30 If the structure of the modelis fixed, one can minimize the reduced chi-squarestatistic (i.e., the maximum likelihood parameter esti-mation) χ2/ν, where ν is the number of degrees offreedom, and then the level of significance can be esti-mated in terms of the incomplete gamma function.31

Of course, biologic model identification typicallyinvolves the identification of the model structure aswell as its parameters. It is an open research questionas to how to rigorously calculate the minimum sam-pling frequency for biologic network identification;however, the minimum number of measurements willdepend on the measurement error, the variation inthe profile curvatures, the number of biomarkers inthe network, and the sparseness of the connectivity ofthe network. For a sparse network of Boolean func-tions with K regulatory inputs per gene, the minimumnumber M of sampling points needed to identify anetwork of N biomarkers was shown to be of order2K[K + log(N)].32 This value for M was derived underideal conditions but represents a reasonable lower

bound for modeling with a more complex continuousformalism such as nonlinear differential equations.

To overcome low-frequency sampling, one coulduse interpolation, random effects regression, orsmoothing; however, these methods could be prob-lematic for systems with high levels of noise that causethe system to deviate from smooth profiles. A recur-sive approach using the UKF has been successful forparameter estimation of dynamic biologic models.17,18

Figure 3 shows the results of this parameter estimationapproach for data (filled circles) simulated based onthe hypothetical model shown in Figure 2, disturbedby a large measurement noise and sparsely sampled(only five time points). The recursive steps are depictedas multiple lines for each variable shown in Figure 3at the end of each UKF pass through the time series. Inthe absence of prior information on the parameters, allparameters are initialized to zero, resulting in an initialsystem with constant solutions. The predicted param-eters at the end of each pass through the time seriesare used as input for the next recursion step. The UKFis insensitive to the initial choice of parameters for thismodel and converges to the correct parameters after 11recursive steps. Recursion can help overcome sparselysampled systems but also leads to increased compu-tation time. Fewer loops through the time series andimproved performance may be achieved by using high-quality initial guesses33 or more qualitative guessesbased on known pathway connectivity.

For proper experimental design, simulationsshould be performed to determine the necessary

0 1 2 3 4 5 6 7

0.6

0.8

1

x

0 1 2 3 4 5 6 70.6

0.8

1

1.2

y

0 1 2 3 4 5 6 70

0.20.40.60.8

Time

z

FIGURE 3 | Recursive parameter estimation with unscented Kalmanfilter for extremely sparse, noisy data (filled circles) simulated withmodel shown in Figure 2. In each panel, each overlaid predicted timecurve corresponds to the recursive steps through the time series.Parameters converge after 11 steps.



sampling frequency to reduce the false-positive modelrate. A true-positive detection of a dynamic modelis not an all or nothing prospect; one can correctlyidentify parts of the model and misidentify others.18

For stochastic system-identification algorithms, ratherthan taking the top-scoring system as the final model,a useful strategy to detect false-positive model compo-nents might be to inspect the set of top-scoring modelsto identify consensus model components and compo-nents that show more inter-model variation, and thenrun the algorithm again, this time focusing on theuncertain model components, which are more likely tobe false. The best way to reduce false-positive modelsthat all reasonably describe the available experimentaldata is to make computational predictions to designa new experiment that can discriminate betweenthe hypothetical models.34 The time-series samplingfrequency need not be uniform. In a more rapidlyvarying domain, it is advantageous to use a highersampling frequency in order to capture detailed fea-tures of the profile. Another practical challenge froman experimental standpoint is anticipating when sucha rapid variation will occur for a given biomarker.35

For example, transcription occurs on the scale ofhours while metabolic reactions occur on the scaleof minutes. Knowledge of such scales as well as timedelays can aid experimental and algorithm design.

SUPERVISED MODELINGThe next frontier in time-series informatics is toidentify global properties and predict global states ofdynamic network models. How does the perturbationof individual inputs of a noisy dynamic networkaffect properties of the network as a whole? Suchsystemic properties might be disease susceptibilityor drug/vaccine response. It is not currently clearhow to predict such a property directly from adynamic network, but in other areas of bioinformatics,involving time-independent data, supervised statisticallearning and data-mining algorithms have been used topredict the state of a phenotypic variable from multipleinput variables.36 A similar approach has been usedwith knowledge-driven dynamic models to simulatetime-series output, which is used to train a decisiontree to predict the state of a selected output variablefrom perturbations in initial concentrations.37 Byitself, this approach does not predict a globalphenotype of an individual; however, coupled withother bioinformatics to identify gene productsassociated with the phenotype, the final decision leavescould predict some phenotypically relevant functionalof the simulated gene product outputs. Among otherthings, a potential application of this technique might

be the rational design of preventative and therapeuticinterventions. The complexity of biologic networksposes many challenges to model-driven therapeuticdesign strategies due to interconnected clusters intranscription networks and the evolutionary evidenceof network rewiring.38 The robustness of genenetworks to noise (Ref.39 and next section) may alsomake them robust to external manipulation, or maygive rise to adverse side effects. Thus, a multivariatestrategy is necessary to design combination therapies,which may be the best treatment strategy for manydiseases.40

Introduced here is an integrative, supervisedstrategy for vaccine improvement using aspects ofthe dynamic model simulation method described inRef.37. In machine learning, a problem is supervisedif there is an outcome/class variable, such as aphenotype, which typically is used for classification.For rational vaccine development or improvementof existing vaccines, the goal is to maximizeimmunogenicity while minimizing reactogenicity.Step 1 of the proposed strategy would involve high-throughput screening to identify target cytokinesassociated with adverse events (e.g., Ref.41) anda parallel analysis of antibody titers to identifytarget cytokines associated with protective immunity.Assuming that a dynamic model exists—eitherknowledge or data driven—Step 2 involves thegeneration of a large artificial data set with randominitial perturbations of cytokines and other signalingand regulatory molecules of the model as theindependent variables, and the response variable isa functional involving the target-molecule (found inStep 1) expression levels at the initial and final timepoint. The functional acts as the outcome variable forthe set of perturbations. An example functional thatmeasures the ratio of immunogenic to reactogenicexpression change from the initial to final time pointsfor a given combination of initial perturbations p is ofthe form

Fp =

∑i∈immunogenic

[ypi (tfinal) − yp

i (tinitial)]

∑r∈reactogenic

[ypr (tfinal) − yp

r (tinitial)](3)

The next step is to find a combination of molecularperturbations that maximizes Eq. (3). Hence, Step 3uses clustering or other methods to discretize Fp acrossthe random Step 2 simulations into ‘high’, ‘medium’,and ‘low’ states, and then a decision tree is trained withthe goal of identifying multivariate tree paths fromthe root node to ‘high’ output leaves. This strategycould generate hypotheses for improving protectiveimmunity while reducing adverse events, and identify



network perturbations that lead to unstable behaviorof the system. The success of this strategy is contingentupon an accurate model of the immune regulatorysystem and the availability of time-series data to tunesuch a model. An additional challenge to realizing therational design of therapeutics is the possibly naiveassumption that vaccine immune response kinetics canbe modeled by the same model structure or kineticparameters for all individuals; that is, the modelsmay show genetic heterogeneity. A related goal willbe to identify dynamic network motifs or modules42

associated with a given phenotype and to target thesemotifs rationally to achieve the desired outcome.

PHENOTYPIC EFFECT OF NOISEAT NANOSCALEWhen using the differential equation formalism topredict network outputs from perturbed inputs, itis commonly assumed that the concentration ofeach molecule or expression of each gene productvaries smoothly. In reality, the expression of eachmolecule depends on the number and state of othermolecules, which are subject to random fluctuations.These external fluctuations, or extrinsic noise, causethe number of molecules to change abruptly fromtime point to time point. Furthermore, isogenic,identically prepared populations of cells may varyin expression level for a particular gene due to theorder of the cascade of microscopic events leadingto that gene’s expression. The conditions for thedominant effect of intrinsic noise in gene regulatorynetworks have been created in multicolor fluorescenceexperiments43,44 and aspects have been modeled withdetailed simulations.45–47

For the measurement of a molecule across Mcells, the variance of the measurements is oforder σ 2

i /M, where σ 2i is the variance of a single

measurement.48 Thus, if the measurements are aver-aged over a large number of cells, then one expectslow noise effects and smooth time-series profiles. Atthe other extreme, abrupt changes in profiles may bemagnified at nanoscale concentrations, where thereis low copy number or low concentration. Kalmanfilters using the differential equation formalism canhandle limited noise effects, but when statisticalfluctuations of concentrations become very large, apurely stochastic formalism may be more suitable.The Gillespie algorithm has become a popular methodto directly simulate the stochastic mechanisms ofa dynamic system.49 Under typical experimentalconditions, it is sufficient to model with deterministicdifferential equations and is preferred for largersystems due to the computational cost of stochastic

simulation. However, for sparsely sampled time seriesit may be difficult to estimate the noise strength,making it difficult to determine whether a time-seriesprofile is merely random.

Biologic networks have evolved to function inthe presence of noise and in most cases they behave insuch a way as to reduce the effect of noise; however,in certain situations noise may be magnified to createheterogeneity in cell populations or to allow cells toadapt to a fluctuating environment.45,47,50 Thus, cellsand networks may exploit noise, but it is conceivablethat this flexibility, if not properly controlled, may alsolead to adverse systemic behavior such as disease. Apossible model for the pathogenesis of some diseasesthat have eluded the reductionist approach to theprediction of disease susceptibility directly from thegenome may be the failure of a network motifto properly regulate the intrinsic noise. Obviouslyenvironmental factors also contribute to disease, butin certain cases a disease phenotype may be a rare,emergent property of a stochastic network caused bythe network’s lifetime exposure to noise. For example,a stochastic mechanism has been hypothesized forhaploinsufficiency diseases in which one allele indiploid cells is insufficient to assure normal function.51

In this model, the decrease in gene dose to oneallele leads to an increased susceptibility to stochasticinterruptions in gene expression. This interruptionmay lead to the increased probability of a drop in geneexpression below a critical threshold and consequentlyto an increase in lifetime disease susceptibility. Thiseffect of increased noise in haploinsufficiency mayalso be found in tumor suppressors.52 Another sourceof noise may arise from epigenetic factors, such asDNA methylation, which can modify transcriptionalactivity stochastically.53 The way gene and cellularnetworks deal with noise and its potential role in theetiology of disease phenotypes, particularly late onset,is far from understood and represents an opportunityand challenge for time-series bioinformatics.

CONCLUSIONBiologic time series contains considerably morecausal information than gene expression or proteinabundance measured at a single time point; thus,the development of high-throughput technologies togather data with high temporal information contentis eagerly anticipated. However, the identification ofmathematical models for these dense data poses manypractical and fundamental challenges. A challengebeyond the scope of this article is the role of spatialeffects as they could be important in situations suchas modeling protein activity, which depends on the



protein’s location within the cell. This adds anotherlevel of complexity to the computational challengesdiscussed above, and tools such as partial differentialequations will be needed to model networks when suchspatiotemporal data become more readily available.Another challenge to reaching a systems-level under-standing of organisms will be to integrate other datatypes—genomic, proteomic, structural, environmen-tal, clinical, and phenotypic—into time-series model-ing.

A dynamic model is just that: a phenomenolog-ical model of the true underlying system. However,an accurate model can reveal insight into biologicrelationships and may act as an in silico experimentaltool to generate testable hypotheses. Possibly the mostambitious time-series bioinformatics research frontieris to predict global/systemic properties, such as dis-ease susceptibility, of a biologic system from dynamic

network inputs. The reductionist approach has beenvery successful at identifying susceptibility genes formany phenotypes, but many common multifactorialphenotypes that have eluded this reductionist strategymay be an emergent property of the entire system, asopposed to a property that is possessed by any iso-lated part of the system. In future predictive dynamicnetwork models, the phenotype may be an emergentproperty of the model or perhaps may be modeled asa hidden variable that describes the state of the wholesystem. Such a model may need to include genomic,proteomic, environmental, and epigenetic factors aswell as the lifetime effect of intrinsic noise expressedin regulatory networks. Such a global model of geneticand cellular networks may also lead to improved pre-ventative and therapeutic interventions40 by indicatingways to modulate multiple targets and simultaneouslyreduce adverse side effects.

NOTES

The java software used in this paper for recursive parameter estimation of generalized dynamic Bayesiannetworks with the unscented Kalman filter is available from the author upon request.This work was supported by NIH Grant: K25 AI-64625.

REFERENCES

1. Baranzini SE, Mousavi P, Rio J, Caillier SJ, Stillman A,et al. Transcription-based prediction of response toIFNbeta using supervised computational methods.PLoS Biol 2005, 3:e2.

2. Wen X, Fuhrman S, Michaels GS, Carr DB, Smith S,et al. Large-scale temporal gene expression mapping ofcentral nervous system development. Proc Natl AcadSci U S A 1998, 95:334–339.

3. Banerjee B, Balasubramanian S, Ananthakrishna G,Ramakrishnan TV, Shivashankar GV. Tracking oper-ator state fluctuations in gene expression in single cells.Biophys J 2004, 86:3052–3059.

4. Elowitz MB, Levine AJ, Siggia ED, Swain PS. Stochas-tic gene expression in a single cell. Science 2002,297:1183–1186.

5. Barabasi AL, Oltvai ZN. Network biology: under-standing the cell’s functional organization. Nat RevGenet 2004, 5:101–113.

6. Ramoni MF, Sebastiani P, Kohane IS. Cluster analysisof gene expression dynamics. Proc Natl Acad Sci U S A2002, 99:9121–9126.

7. Friedman N, Linial M, Nachman I, Pe’er D. UsingBayesian networks to analyze expression data. Annual

Conference on Research in Computational Molecu-lar Biology (RECOMB). Tokyo: ACM Press; 2000,127–135.

8. de Jong H. Modeling and simulation of genetic regula-tory systems: a literature review. J Comput Biol 2002,9:67–103.

9. Kim SY, Imoto S, Miyano S. Inferring gene networksfrom time series microarray data using dynamicBayesian networks. Brief Bioinform 2003, 4:228–235.

10. Ong IM, Glasner JD, Page D. Modelling regulatorypathways in E. coli from time series expression profiles.Bioinformatics 2002, 18(suppl 1):S241–S248.

11. Smith VA, Jarvis ED, Hartemink AJ. Evaluating func-tional network inference using simulations of com-plex biological systems. Bioinformatics 2002, 18(suppl1):S216–S224.

12. Smith VA, Jarvis ED, Hartemink AJ. Influence of net-work topology and data collection on network infer-ence. Pac Symp Biocomput 2003, 11:164–175.

13. Kalman RE. A new approach to linear filtering andprediction problems. J Basic Eng 1960, 82:35–45.

14. Perrin BE, Ralaivola L, Mazurie A, Bottani S, Mal-let J, et al. Gene networks inference using dynamic



Bayesian networks. Bioinformatics 2003, 19(suppl2):II138–II148.

15. Julier S, Uhlmann J, Durrant-Whyte HF. A newapproach for the nonlinear transformation of meansand covariances in filters and estimators. IEEE TransAutom Control 2000, 45:477–482.

16. Voss HU, Timmer J, Kurths J. Nonlinear dynamicalsystem identification from uncertain and indirect mea-surements. Int J Bifurcat Chaos 2004, 14:1905–1933.

17. McKinney BA, Crowe JE, Voss HU, Crooke PS, Bar-ney N, et al. Hybrid grammar-based approach to non-linear dynamical system identification from biologicaltime series. Phys Rev E Stat Nonlin Soft Matter Phys2006, 73:021912.

18. McKinney BA, Tian D. Grammatical immune systemevolution for reverse engineering dynamic Bayesianmodels. Cancer Informatics 2008, 6:433–447.

19. Goodwin BC Temporal Organization in Cells. NewYork: Academic Press; 1963.

20. Tyson JJ, Othmer HG. The dynamics of feedback con-trol circuits in biochemical pathways. Prog Theor Biol1978, 5:1–62.

21. Savageau MA. Biochemical systems analysis. I. Somemathematical properties of the rate law for thecomponent enzymatic reactions. J Theor Biol 1969,25:365–369.

22. Savageau MA. Parameter sensitivity as a criterion forevaluating and comparing the performance of biochem-ical systems. Nature 1971, 229:542–544.

23. Savageau MA Biochemical Systems Analysis: a Studyof Function and Design in Molecular Biology. Reading,MA: Addison-Wesley; 1976.

24. Voit EO Computational Analysis of Biochemical Sys-tems: a Practical Guide for Biochemists and MolecularBiologists. Cambridge: Cambridge University Press;2000.

25. Voit EO, Almeida J. Decoupling dynamical systems forpathway identification from metabolic profiles. Bioin-formatics 2004, 20:1670–1681.

26. Kikuchi S, Tominaga D, Arita M, Takahashi K,Tomita M. Dynamic modeling of genetic networksusing genetic algorithm and S-system. Bioinformatics2003, 19:643–650.

27. Shin A, Iba H. Construction of genetic network usingevolutionary algorithm and combined fitness function.Genome Inform 2003, 14:94–103.

28. Crampin EJ, Schnell S, McSharry PE. Mathematicaland computational techniques to deduce complex bio-chemical reaction mechanisms. Prog Biophys Mol Biol2004, 86:77–112.

29. Chavent G Identifiability of parameters in the outputleast square formulation. In: Walter E: ed. Identifiabil-ity of Parametric Models. Tarrytown, NY: Pergamon;1987.

30. Cohen J Statistical Power Analysis for the BehavioralSciences. 2nd ed. Mahwah, NJ: Lawrence ErlbaumAssociates; 1988.

31. Bevington PR, Robinson DK Data Reduction andError Analysis for the Physical Sciences. 2nd ed. NewYork: McGraw-Hill; 1992.

32. D’haeseleer P, Liang S, Somogui R. Genetic networkinference: from co-expression clustering to reverse engi-neering. Bioinformatics 2000, 16:707–726.

33. Veflingstad SR, Almeida J, Voit EO. Priming nonlinearsearches for pathway identification. Theor Biol MedModel 2004, 1:8.

34. Kremling A, Fischer S, Gadkar K, Doyle FJ, Sauter T,et al. A benchmark for methods in reverse engineeringand model discrimination: problem formulation andsolutions. Genome Res 2004, 14:1773–1785.

35. Nicholson JK, Holmes E, Lindon JC, Wilson ID. Thechallenges of modeling mammalian biocomplexity. NatBiotechnol 2004, 22:1268–1274.

36. Hastie T, Tibshirani R, Friedman J The Elements ofStatistical Learning: Data Mining, Inference, and Pre-diction: New York: Springer-Verlag; 2003.

37. Hua F, Hautaniemi S, Yokoo R, Lauffenburger DA.Integrated mechanistic and data-driven modelling formultivariate analysis of signalling pathways. J R SocInterface 2006, 3:515–526.

38. Ihmels J, Bergmann S, Gerami-Nejad M, Yanai I,McClellan M, et al. Rewiring of the yeast transcrip-tional network through the evolution of motif usage.Science 2005, 309:938–940.

39. von Dassow G, Meir E, Munro EM, Odell GM. Thesegment polarity network is a robust developmentalmodule. Nature 2000, 406:188–192.

40. Kubinyi H. Drug research: myths, hype and reality. NatRev Drug Discov 2003, 2:665–668.

41. McKinney BA, Reif DM, Rock MT, Edwards KM,Kingsmore SF, et al. Cytokine expression patterns asso-ciated with systemic adverse events following smallpoximmunization. J Infect Dis 2006, 194:444–453.

42. Hartwell LH, Hopfield JJ, Leibler S, Murray AW.From molecular to modular cell biology. Nature 1999,402:C47–C52.

43. Isaacs FJ, Blake WJ, Collins JJ. Molecular biology.Signal processing in single cells. Science 2005,307:1886–1888.

44. Rosenfeld N, Young JW, Alon U, Swain PS,Elowitz MB. Gene regulation at the single-cell level.Science 2005, 307:1962–1965.

45. Kaern M, Elston TC, Blake WJ, Collins JJ. Stochastic-ity in gene expression: from theories to phenotypes.Nat Rev Genet 2005, 6:451–464.

46. Kepler T, Elston TC. Stochasticity in transcriptionalregulation: origins, consequences, and mathematicalrepresentations. Biophys J 2001, 81:3116–3136.



47. McAdams HH, Arkin A. Stochastic mechanisms ingene expression. Proc Natl Acad Sci U S A 1997,94:814–819.

48. Wolkenhauer O, Ullah M, Kolch W, Cho K-W. Model-ing and simulation of intracellular dynamics: choosingan appropriate framework. IEEE Trans Nanobiosci2004, 3:200–207.

49. Gillespie DT. A rigorous derivation of the chemicalmaster equation. Physica A 1992, 188:404–425.

50. McAdams HH, Arkin A. It’s a noisy business! Geneticregulation at the nanomolar scale. Trends Genet 1999,15:65–69.

51. Cook DL, Gerber AN, Tapscott SJ. Modeling stochas-tic gene expression: implications for haploinsufficiency.Proc Natl Acad Sci U S A 1998, 95:15641–15646.

52. Kemkemer R, Schrank S, Vogel W, Gruler H,Kaufmann D. Increased noise as an effect ofhaploinsufficiency of the tumor-suppressor geneneurofibromatosis type 1 in vitro. Proc Natl AcadSci U S A 2002, 99:13783–13788.

53. Whitelaw NC, Whitelaw E. How lifetimes shapeepigenotype within and across generations. Hum MolGenet 2006, 15(Spec No 2):R131–R137.


Noise in biological circuits.


Advanced Review

Radioactive liposomesWilliam Thomas Phillips,∗ Beth Ann Goins1 and Ande Bao1

Many methods of labeling liposomes with both diagnostic and therapeuticradionuclides have been developed since the initial discovery of liposomes 40 yearsago. Diagnostic radiolabels can be used to track nanometer-sized liposomes inthe body in a quantitative fashion. This article reviews the basic methods ofsingle photon emission computed tomography (SPECT) and positron emissiontomography (PET) imaging and labeling of liposomes with single photon and dualphoton positron emission radionuclides. Examples of the use of these diagnosticimaging agents will be shown. The ability to track the uptake of liposomes inhumans and research animals on a whole body basis is providing researcherswith an excellent tool for developing liposome-based drug delivery agents. Theattachment of therapeutic radionuclides to liposomes also has great promise incancer therapy. Recent developments in the use of liposomes carrying therapeuticradionuclides for cancer therapy will also be reviewed. Many of the radiolabelingand tracking technologies developed for nanosized liposomes will also be usefulfor the imaging and tracking of other nanoparticles. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 69–83

Liposomes are spontaneously forming lipid bilay-ers that enclose an aqueous space. Very soon

after their initial discovery more than 40 years ago,they were recognized as promising drug carriers fordiagnostic and therapeutic agents.1,2 Composed ofnaturally occurring components, liposomes have amyriad of possible compositions and modifications,making them extremely flexible drug carriers. Lipo-somes range in size from 50 nm to several microme-ters, with the most stable and useful size being in thelipid nanoparticle size range of 90–250 nm. In partic-ular, several liposome formulations in the size rangeof 90–100 nm are clinically approved for the deliveryof antifungal, antiparasitic, and anticancer agents.3,4

Incremental, continuous progress in the field of lipo-some technology has greatly increased the viabilityof liposomes as clinically useful drug carriers. Theseachievements include the development of (1) methodsto produce homogeneously sized liposomes in thenanometer size range, (2) methods for large-scale lipo-some manufacture, (3) methods to stably encapsulatesufficient quantities of therapeutic agents within theliposomes, (4) methods to prolong the circulation time

∗Correspondence to: William Thomas Phillips, University of TexasHealth Science Center at San Antonio, TX, USA.E-mail: [email protected] Department, University of Texas Health Science Centerat San Antonio, TX, USA.

DOI: 10.1002/wnan.003

of liposomes in the blood,5–7 (5) methods to activelytarget liposomes in vivo by surface modification withtargeted ligands,8–11 and (6) effective methods of non-invasively tracking and quantifying the distribution ofliposomes in the body.12–16 This article will review theuses of radioactive liposomes for in vivo applications.Basic methods of labeling liposomes with radionu-clides suitable for detection using nuclear imagingcameras placed outside the body and in vivo imag-ing methods with these radioactive liposomes will bedescribed. Examples of practical imaging applicationsof radioactive liposomes for research and potentialclinical application will also be reviewed. Finally, lipo-somes labeled with therapeutic radionuclides will bediscussed.

SCINTIGRAPHIC IMAGINGOne of the most effective methods of trackingand quantitatively determining the distribution ofliposomes in the body is through the scintigraphicimaging of radiolabeled liposomes. Scintigraphicimaging, also known as gamma photon imaging, isa common clinically used technology that involves thedetection of gamma photons emitted from radioactivemolecules. Clinical imaging of injected radioactivemolecules or molecular composites is widely usedto detect the in vivo behavior of radionuclides andindirectly the in vivo behavior of these molecules, or



liposomes, and other particle molecular compositesin the body that cannot be detected by moreanatomically based imaging methods such as standardX-ray and X-ray computed tomography (CT) imaging.Technological advances over the last 30 years haveresulted in great improvements in the resolutionand sensitivity of the cameras used for scintigraphicimaging.17–19 Over this same period, a variety ofmethods have been developed for the labeling ofliposomes with various radionuclides.

TWO TYPES OF SCINTIGRAPHICIMAGING

There are two types of gamma photon imaging thatdetect photons (γ -rays) resulting from the radioac-tive decay of radionuclides. These are based on eithersingle photon image detection or dual-annihilation-photon image detection. Single photon imaging witha gamma camera can either be planar projectionimaging, or tomographic three-dimensional imagingknown as single photon emission computed tomog-raphy (SPECT) imaging. A second type of imagingknown as positron emission tomography (PET) imag-ing is based on the detection of two annihilationphotons that are simultaneously emitted at approx-imately 180◦ from each other. These annihilationphotons are generated from the annihilation reactionbetween an emitted positron and a shell electron. PETimaging is always acquired for display of tomographicimages, and planar projection imaging is not generallyfeasible.

Single Photon ImagingFor diagnostic applications with single photonimaging, the liposomes are labeled with a radionuclidethat emits γ -rays with energies ranging from 100 to200 keV (Table 1). These energies are high enough topenetrate a human body while low enough to be easilycollimated by lead and be detected with a scintillationcrystal. In single photon gamma camera imaging,100–200 keV photons are emitted from liposomalradionuclides that have been previously administeredinto the body and allowed sufficient time to distributein particular locations. In general, owing to theirparticulate nature, radiolabeled liposomes localize byphysiologic processes responsible for clearing particlesfrom the body; however, their blood clearanceprofile can be modified by adjusting particle sizesor attaching certain molecules, such as polyethyleneglycol (PEG). Liposomes can also be molecularlytargeted to specifically localize in a disease processor organ of interest.11,20 The emitted photon travels

through body tissue and exits the body whereuponit is detected by a crystal inside the gamma camera.A lead collimator in front of the crystal allows onlythe photons from predictable projections to traversethrough the collimator and thus provides spatiallocalization of the single photon emission. Followingsuccessful passage through the collimator, the photoninteracts with the crystal detector and generates ascintillation that is detected by photon amplifiers,and the position of the scintillation on the crystal isdetermined.

Positron Emission Tomography ImagingIn PET imaging, a proton-rich radionuclide decays,resulting in emission of a positron (positively chargedelectron). When this positron reaches almost zerokinetic energy, it immediately reacts with a shellelectron, resulting in the transition of these twoelectrons into the emission of two annihilationphotons that are emitted at approximately 180◦ fromeach other. These dual photon emissions can belocalized along a line by the simultaneous detection ofthe two 180◦ emitted photons by crystal scintillationdetectors that are part of a ring of detectorssurrounding the body. Multiple emissions can beiteratively localized in the body, providing a three-dimensional tomographic image of the distributionof the positron radionuclide. Collimators are notrequired for this type of imaging, although fairly thickcrystals are required to detect the high-energy 511 keVphotons that are emitted from all positron emissionradioisotopes. The most commonly used radionuclidesfor PET imaging have relatively short half-life, suchas oxygen-15 (2 min), carbon-11 (20 min), andfluorine-18 (110 min) (Table 1). PET radionuclideswith longer half-lives, such as copper-64 (64Cu)(12.7 h) and iodine-124 (124I) (4.2 days), may beparticularly promising for tracking the distributionof long-circulating nanoparticles such as liposomes.

METHODS OF LABELING LIPOSOMESWITH RADIOACTIVE MOLECULES

Commonly used radionuclides for the radiolabelingof liposomes for imaging studies with a gammacamera are technetium-99m (99mTc), indium-111(111In), and gallium-67 (67Ga) (Table 1). Theseradionuclides are all widely available from localclinical radiopharmacies. 99mTc has the most idealproperties of the readily available single photonemission diagnostic imaging radionuclides due toits optimal imaging characteristics which include anideal photon energy of 140 keV and the fact that


WIREs Nanomedicine and Nanobiotechnology Radioactive liposomes

it is relatively inexpensive because it can be eluteddaily from a commercially available molybdenum-99 (99Mo)/99mTc generator. Because 67Ga, 111In, and123I are cyclotron products, these agents are moreexpensive and not always available in every nuclearmedicine department.

Several detailed reviews have been recentlywritten describing the methodology for radiolabelingliposomes22–24 (Table 2). When designing a liposomelabeling method, important factors to consider areease of preparation, and in vitro and in vivo stability.Ideally, liposomes should be labeled just prior to theinitiation of experiments by using premanufacturedstock liposomes. This situation is ideal because theradionuclides used in clinical imaging and researchstudies are fairly short-lived (on the order of hoursto days) (Table 1). This type of postmanufacturelabeling is known as ‘afterloading’ or ‘remote labeling’in which the preformed liposomes are labeled justprior to the start of the experiment. Also, afterradiolabeling the radioactive liposomes should bestable, with a high percentage of the radionuclideremaining with the liposomes. Any significant releaseof the radionuclide from the liposomes during orafter injection will obviously lead to complications ininterpreting biodistribution results.

Table 2 summarizes several different methodsthat have been reported for labeling liposomes withboth single photon and dual photon emission (PET)radioactive molecules. A relatively simple method oflabeling liposomes is to incubate the premanufac-tured liposomes with a lipophilic radiolabel. Thisresults in association of the label within the lipid

bilayer. This approach generally yields very unsta-ble radiolabeling25 and is therefore not preferable. Incontrast, two afterloading approaches have provento yield radiolabeled liposomes that have a high effi-ciency and good radiochemical stability. With theseapproaches, the radionuclide is either (1) trapped inthe enclosed aqueous phase of liposomes, or (2) boundto a lipid-conjugated chelator incorporated in the lipidbilayer of preformed liposomes.

One type of afterloading method that has beendeveloped for labeling liposomes uses radionuclidechelators attached to the liposomal surface.15,49,50

One method that has been found to be stablefor the labeling of liposomes with 111In uses themetal chelator diethylenetriamine pentacetic acid(DTPA) conjugated to phosphatidylethanolamine (PE)lipid. A detailed description of the preparation andapplication of radiolabeled DTPA–PE liposomes hasbeen previously published.50

DTPA–PE is not very effective for labeling lipo-somes with 99mTc.51 For this purpose, a new chela-tion method based on the technetium chelator, N-hydroxysuccinimidyl hydrazino nicotinate hydrochlo-ride (HYNIC), has been developed by Lavermanet al.15 With this method, the HYNIC ligand isconjugated to the amino group of distearoyl phos-phatidylethanolamine (DSPE) and subsequently incor-porated into the lipid bilayer during the liposomepreparation. Just prior to the imaging study, the lipo-somes are incubated with 99mTc. This HYNIC labelingmethodology has been investigated as a method tolabel liposomes for detection of infection.15 It hasalso been used to investigate the relatively short blood

TABLE 1 Physical Characteristics of Common Diagnostic Single Photon and Dual Photon Positron EmittingRadionuclides Used for Labeling Liposomes

Radionuclide (abbreviation) Half-life Type of Photon Method of Production Photons (keV), (Abundance (%))a

Gallium-67 (67Ga) 3.3 days Single photon Cyclotron 93 (38)

185 (21)

300 (17)

Technetium-99m (99mTc) 6.01 h Single photon 99Mo/99mTc generator 141 (89)

Indium-111(111 In) 2.8 days Single photon Cyclotron 171 (91)

245 (94)

Iodine-123(123 I) 13.2 h Single photon Cyclotron 159 (83)

Carbon-11 (11C) 20 min Dual photon PET Cyclotron 511

Copper-64 (64Cu) 12.7 h Dual photon PET Cyclotron 511

Copper-62 (62Cu) 3.5 h Dual photon PET Cyclotron 511

Fluorine-18 (18F) 109 min Dual photon PET Cyclotron 511

Iodine-124 (124I) 4.2 days Dual photon PET Cyclotron 511

Oxygen-15 (15O) 2 min Dual photon PET Cyclotron 511aData derived from Ref. 21.



TABLE 2 Methods of Labeling Liposomes with Diagnostic Radionuclides

Radionuclide Labeling Mode Labeling Method Cite

Single photon radionuclides

Gallium-67 (67Ga) After-loading Oxine with NTA-liposomes 26

After-loading Oxine with DF-Liposomes 13

Indium-111 (111 In) Surface chelation DTPA-fatty acid containing liposomes 27,28

After-loading Oxine with NTA-liposomes 29

After-loading Ionophore A23187 with NTA-liposomes 30,31

After-loading Oxine with DF-liposomes 32–34

After-loading Oxine with DTPA-liposomes 35

Iodine-123 (123 I) Encapsulation Sodium iodide or iodinated phospholipid 36

After-loading Bolton-Hunter reagent with arginine-containing liposomes 37

Technetium-99m (99mTc) Surface labeling 99mTc-pertechnetate 38,39

Surface chelation DTPA-phospholipid containing liposomes 40,41

Surface chelation HYNIC-phospholipid containing liposomes 15

After-loading HMPAO with GSH-liposomes 42

After-loading BMEDA with cysteine/GSH-liposomes 43

After-loading BMEDA with ammonium sulfate gradient 12

Positron-emitting radionuclides

Fluorine-18 (18F) Encapsulation 18F-FDG during manufacture 44

Encapsulation 18F attached to lipid molecule, followed by liposome manufacture. 45

Iodine-124 (124I) After-loading Bolton-Hunter reagent with arginine-containing liposomes 37

Oxygen-15 (15O) After-loading Binding to hemoglobin in pre-formed liposomes 46–48

circulation times that occur when very low doses oflipids are used, as well as the greatly shortened cir-culation time of PEG-coated liposomes when they arereinjected at frequent intervals.52,53

In another type of afterloading method, aradionuclide is coordinated with a lipophilic chelatorand then mixed with an aliquot of liposomesencapsulating a second molecule. Once the lipophilicchelator carries the radionuclide across the lipidbilayer, the second molecule interacts with theradionuclide complex and traps the radionuclidewithin the interior of the liposome. Several methodshave been developed that use this second afterloadingapproach to label liposomes. This approach to labelingpreformed liposomes has an important advantage inthat it does not leave the radiolabel on the surface ofthe liposomes where it might interfere with specifictargeting molecules or where it would have theopportunity to interact with natural metal-bindingmolecules in the body.

This approach has been used with 111In, 99mTc,and 67Ga.16,30,43,54–56 One method using the radionu-clide 111In has been widely adopted for liposomeimaging studies.57–59 With this method, a clinically

available 111In complex, 111In-oxine, is used for lipo-some labeling. The 111In-oxine is incubated withpremanufactured liposomes encapsulating DTPA.60

111In-oxine migrates into the aqueous interior ofa liposome where it is trapped by trans-chelationonto DTPA. These 111In-labeled liposomes are stableand can be used for long-term tracking of liposomesbecause of the long half-life of 111In (68 h).

A method for labeling liposomes with 99mTcuses the lipophilic complex 99mTc–hexamethylpropy-leneamine oxime (99mTc–HMPAO).16 With thismethod, lipophilic 99mTc–HMPAO enters the lipo-some where it interacts with pre-encapsulated glu-tathione and converts to the hydrophilic form, andthus is trapped in the liposome. This method has beensuccessfully used by many investigators.16,61–69

Another similar afterloading method forradiolabeling liposomes with 99mTc uses 99mTc–N, N-bis(2-mercaptoethyl)-N’,N’-diethyl-ethylenediamine(99mTc–BMEDA).43 This method has been shown tohave good in vitro and in vivo stability with a varietyof preformed liposome formulations. Comparedwith 99mTc–HMPAO, an additional advantage ofthe BMEDA-labeling method is that it is effective



for labeling liposomes with β-particle-emittingtherapeutic radionuclides, rhenium-186 (186Re) orrhenium-188 (188Re).70 A second advantage of thisBMEDA method is that it can be used to directlylabel commercially available liposome formulationssuch as liposomal doxorubicin (Doxil) via sharingthe pH gradient mechanism for loading drugs intothe liposomes.12 BMEDA for research purposesis available from a commercial vendor (ABX,Radeburg, Germany).

A new method has recently been described tolabel preformed liposomes with radiohalogenatedagents such as iodine-123 (123I) and iodine-124(124I).37 With this method, a Bolton–Hunter reagent,in the form of an activated ester, crosses the liposomemembrane and reacts with arginine that has beenpreviously encapsulated in the liposome. Use of 123Imay have some advantages because it is a singlephoton emitter with a longer half-life of 13.2 h, ascompared to 6 h for 99mTc. This radionuclide is readilyavailable from the local radiopharmacy in manylocations, although it is fairly expensive because of thefact that it must be produced by a cyclotron. This samemethod could also be used to label liposomes with 124I,which is a long-lived positron emitter with a half-life of 4.2 days. The labeling of liposomes with thislong-lived PET radionuclide may have useful researchapplications. However, the possibility of a high patientdose and its challenge in radiation protection mayneed to be considered with 124I owing to its relativelyhigh β-particle energy and a high-energy γ -photonemission (1.69 MeV γ -ray, 10.4%).

A new method has recently been describedfor labeling liposomes with the PET radionuclide,fluorine-18 (18F).45 In this method, 18F is incorpo-rated into the dipalmitoylglycerol lipid molecule bynucleophilic substitution of the p-toluenesulfonyl moi-ety. This procedure has a decay-corrected yield of43%. Following rapid production of the 18F-lipid,long-circulating liposomes are labeled by adding the18F-lipid during a rapid manufacturing procedure, andthen used for PET imaging. Although this procedure,in which the liposome is labeled as part of the man-ufacturing process, is less ideal than the afterloadingmethod of labeling preformed liposomes, the PETimage produced by a very high-resolution state-of-the-art microPET camera is of excellent quality. Thishigh-quality image demonstrates the future potentialfor PET imaging of radioactive liposomes, which willprovide improved anatomical localization of labeledliposomes and other nanoparticles labeled with PETimaging radionuclides.45 A disadvantage with the useof 18F is that it has a short half-life of 109 min so thatit would be useful for tracking liposomes only for a

short time period of no more than 8 h; however, thisshort time of imaging may still prove useful for certainapplications.

SPECIFIC USES OF SCINTIGRAPHICIMAGING FOR LIPOSOME DRUGDEVELOPMENTScintigraphic imaging of liposomes has been success-fully used in the last 15 years for studies of drugdelivery and as potential diagnostic imaging agents.Several comprehensive reviews have been writtendescribing in detail these diagnostic uses of radioac-tive liposomes.22,71–73 In the following section somespecific examples of the use of radioactive liposomeswill be provided with emphasis on research carriedout in the last 3 years.

Noninvasive Determination of DrugConcentrationsThe tracking of radiolabeled liposomes by scinti-graphic imaging can be used as a method to non-invasively and quantitatively determine how muchof a drug is taken up by the targeted site. A val-idation of this approach has recently been shownby Kleiter et al.69 In this research, liposomes werelabeled with 99mTc using the HMPAO method.16

The purpose of the study was to determine whethernoninvasive imaging could be used to predict theamount of doxorubicin that accumulated in a ratfibrosarcoma tumor following the administration oflong-circulating PEG liposomes containing doxoru-bicin (Doxil) without the need to biopsy the tumor.Scintigraphic images were obtained at 18 h follow-ing injection of the 99mTc-labeled liposomes, and thetumors were removed for radioactivity counting andfor measurement of the doxorubicin concentration.The results of this study demonstrated that there wasa significant positive correlation between the intratu-moral radioactivity and the amount of doxorubicinin the tumor. This provides good support for the useof this labeling and imaging methodology to nonin-vasively track the distribution of drugs delivered byliposome carriers. This report was also interesting inthat these studies were performed on rats that hadreceived local hyperthermia directed at the tumor.The images clearly demonstrated the increased uptakein the hyperthermia-treated tumor as compared withthe tumors that were not treated with hyperthermia.Hyperthermia increased the uptake of the radiola-bel associated with the liposomes by approximatelyfourfold, and it increased the uptake of measureddoxorubicin concentrations in the tumor by a slightlysmaller increment that ranged from 2.6- to 3-fold.69



Infection and Inflammation Targeting ofLiposomesLiposomes have been shown to accumulate heavilyat sites of inflammation and infection by passivetargeting.14,62,74,75 Microscopic studies have shownthat the liposomes accumulate by enhanced extrava-sation and become mainly localized in macrophagesand to a lesser extent in endothelial cells in the regionof the infection/inflammation.76 Liposomes labeledwith 99mTc have been effective tools in quantitat-ing and localizing the site of uptake in the infection.Planar whole rat images shown in Figure 1 clearlydemonstrate the uptake of liposomes in induced coli-tis in a rat.62 At 24 h, rats with colitis had 13.5% ofthe injected dose of liposomes located in the inflamedcolon as compared to 0.1% in the colon of control nor-mal animals.62 These planar images demonstrate thecapability of scintigraphic imaging for noninvasivelyquantitating the percentage of injected dose that accu-mulates in a targeted region of the body. Liposomescontinue to be very promising carriers for deliveryof drugs to inflamed regions of the body, although,to date, no clinical products have specifically takenadvantage of the inflammatory targeting of liposomes.

FIGURE 1 | Planar whole rat images.

Liposomes for Treating ArthritisLiposome imaging has been used in the preclinicaldevelopment of liposomes containing prednisolonefor the therapy of rheumatoid arthritis in an ani-mal model.58 Long-circulating PEG-coated liposomescontaining prednisolone as well as DTPA werelabeled with 111In-oxine using the method describedpreviously.60 These liposomes were found to accumu-late in the inflamed joints by imaging. The increaseduptake was very obvious on the images, and the111In labeling of the liposomes was used to determinethe clearance from the blood and the biodistributionin the tissues. The inflamed hind paws had sevenfoldincreased activity in comparison with the hind paws ofnormal rats. A single dose of liposome-encapsulatedprednisolone resulted in complete remission of theinflammatory response in 1 week.

A similar use of imaging has also beenrecently reported for liposomes that carry superoxidedismutase (SOD) on their surface.59 These liposomeswere used to treat experimental arthritis. The useof liposomes as a carrier for SOD greatly improvedthe pharmacokinetic behavior of SOD, allowing thispowerful antioxidant to reach the site of inflammationin the joint. A comparison of the therapeutic efficiencywas made between liposomes that had no SOD ontheir surface versus liposomes in which the SOD wascarried on their surface. Imaging using the 111In-oxine/DTPA method demonstrated that even though bothliposome formulations had similar accumulations inthe inflamed joints, the liposomes with the SOD ontheir surface were significantly more effective for thetreatment of arthritis.59 The investigators have namedthis type of liposome, an ‘enzymosome’.

Bone Marrow–Targeted LiposomesA recent paper has reported that liposomes witha special surface modification have very highuptake in bone marrow.68 With this special surfacemodification, rabbits injected with these liposomes,also referred to as vesicles, had a very highuptake in the bone marrow as revealed by imagingwith the 99mTc-HMPAO method as demonstratedin Figure 2. The special modification to the liposomesurface was the addition of a negative charge tothe liposome surface by adding a nonphospholipidanionic amphiphile, l-glutamic acid, N-(3-carboxy-1-oxopropyl)-1,5-dihexadecyl ester [succinic acid (SA)],to the liposome composition. The addition of a smallamount of PEG to this liposome surface also appearedto modestly enhance the already high uptake ofthe liposomes in the marrow. Imaging was usedto determine the most ideal concentration of PEG



FIGURE 2 | Imaging with 99mTc-HMPAO method.

to have on the surface of these SA-liposomes fora maximized bone marrow uptake. On the basisof imaging studies, the uptake in the marrow wasvery rapid, with more than 60% of the infused dosetaken up by the marrow at 6 h post intravenousadministration.68 Although the precise mechanismby which these liposomes accumulate in marrow isunknown, electron microscopy studies as well asfluorescently labeled liposome confocal microscopystudies have demonstrated that the liposomes inthe marrow are located in marrow macrophages.It is believed that the special surface modificationof the liposomes results in specific targeting of thescavenger receptors on the surface of the bone marrowmacrophages. The specific targeting of therapeuticagents to bone marrow using these SA-liposomes maybe a promising approach for the delivery of drugs tothe bone marrow.

Intracavitary LiposomesLiposomes with high retention in body cavities andin the lymph nodes that drain from these cavitieshave been developed for drug delivery applicationsusing imaging as a tool.77–80 These liposomes containbiotin on their surface, and when these biotin-coatedliposomes are injected into a body cavity such asthe peritoneum or the pleural space within 2 h ofthe injection of avidin into the same cavity, theliposomes will have high retention in this cavity. Themechanism by which this retention occurs is thoughtto be due to the aggregation of the liposomes by the

high affinity that the multivalent avidin has for thebiotin liposomes. Liposome imaging using the 99mTc-HMPAO methodology has been used to monitor theretention in the cavity and determine the best timingfor the injection of the avidin.77–80

When the avidin/biotin–liposome methodologyis used, most of the liposomes administered inthe body cavity are retained in that cavity for aprolonged time, whereas the same biotin–liposomeformulation without the avidin is rapidly clearedfrom the cavity over a 4–6 h period during whichtime the liposomes return to the blood from thelymphatics that drain the cavity. Figure 3 depicts a3-D volume reconstruction SPECT image of a nuderat with an intraperitoneal ovarian cancer xenograft,acquired 4 h after intraperitoneal injection of 99mTc-biotin–liposomes and 3.5 h after an intraperitonealinjection of avidin. The SPECT image is superimposedon a bone window CT image. As can be seen from thefigure, much of the dose of 99mTc-biotin–liposomes istrapped in the peritoneum. Without the avidin/biotintrapping system, nearly all liposomes would havebeen cleared from the peritoneum by 4 h postintraperitoneal administration.

This avidin/biotin–liposome approach has greatpotential for the intracavitary retention of therapeuticagents encapsulated in liposomes. One potential usefor these liposomes is in the treatment of ovariancancer in which the cancer cells are generallydisseminated into the peritoneal cavity and lymph

FIGURE 3 | 3-D -volume reconstruction SPECT image of a nude ratwith an intraperitoneal ovarian cancer xenograft.



nodes that drain this cavity at the time of diagnosis.Another use could be in the treatment of lungcancer in which the cancer frequently drains intothe mediastinal nodes. Imaging studies using theavidin/biotin–liposome procedure have shown that alarge liposome dose can be targeted to the mediastinalnodes following administration of biotin liposomesalong with avidin into the pleural cavity.

Studies of Liposomes for Cancer TherapyBiodistribution studies in rats of the clinical formula-tion of Doxil labeled using the previously described99mTc-BMEDA method have been performed by theauthors.12 The ability to directly label commerciallyavailable liposome products may have important clin-ical uses.

Imaging has also been used to study theefficacy of specifically targeted liposomes. Severalstudies have used imaging to show the promise ofmodifying the surface of liposomes with moleculesspecifically targeted for the treatment of cancer.8,81

In these studies, liposomes were formed to containboth a polychelating amphiphilic polymer (PAP)and a specific targeting antibody on their surface.The specific polychelating polymer was composed ofhydrophilic blocks carrying multiple side chains of themetal chelating agent, DTPA. For specific targeting,a monoclonal antibody against nucleosomes was alsoattached to these liposomes. Tumor-to-muscle ratiofor the liposomes with the specific antinucleosomeantibodies was 13.9 at 24 h versus lower ratiosdetermined for the nonspecifically targeted liposomesthat were either modified with nonspecific polyclonalantibodies (4.3) or liposomes with an unmodifiedsurface (3.0) in rats that had Lewis lung carcinomasimplanted on their thighs. This comparison withboth types of control liposomes provides convincingevidence of the specificity of this antibody for targetingthe tumor. This same antibody was also shown tohave a significantly increased tumor uptake in ratsthat had been implanted with colon carcinoma cells,demonstrating that antinucleosome antibodies mayenhance the targeting of liposomes to a wide varietyof cancer types.

Liposomes as Intratumorally AdministeredCarriersInvestigation of the feasibility of directly injectingliposome therapeutics into solid tumors was originallystimulated by failed attempts to introduce liposomesas gene carriers systemically.82–84 Clinical trials usingcationic liposomes carrying the E1A gene were

performed to treat squamous cell carcinoma usingan intratumoral injection technique.82,83

Imaging has been used to better understandthe distribution of liposomes after intratumoralinjection.85 From these imaging studies of 99mTc-liposomes, it appears that the particulate nature ofliposomes offers significant advantages for direct intra-tumoral administration. The behavior of liposomesfollowing intratumoral injection differs substantiallyfrom that of the free unencapsulated drug. Whenfree 99mTc-BMEDA was injected, it appears to bequickly cleared into the blood supply of the tumorinterstitial space, with the remaining activity isolatedin a less diffusive manner in the tumor. Liposomesappear to diffuse through the interstitial space of atumor with a prolonged high retention. The degreeof diffusion may depend on the characteristics of theparticular liposome formulation injected. Liposomeintratumoral diffusion should result in improved solidcancer therapy owing to a more homogeneous distri-bution throughout the tumor and a high intratumoralretention.

A significant advantage of liposomes for use inintratumoral injection is that they have the potentialto move into the lymphatic vessels that drain fromthe solid tumors where they can deliver anticancertherapy to lymph nodes and other lymphatics thatdrain from the tumor. It is these lymph nodes thatoften contain tumor micrometastasis at the time thetumor is diagnosed.

LIPOSOMES AS CARRIERS OFTHERAPEUTIC RADIONUCLIDES FORTUMOR THERAPYLiposomes are also promising carriers of therapeuticradionuclides for treatment of cancer. They can carryeither therapeutic β-emitting or α-particle-emittingradionuclides for cancer therapy.86 Several therapeuticradionuclides have similar chemical features withdiagnostic radionuclides and therefore can share thesimilar radiolabeling methods [for example, 111In andyttrium-90; 67Ga and holmium-166; 99mTc and 188Re,186Re; radioactive halogens such as 123I, 124I, 131I andastatine-211, (211At)]. Table 3 contains the physicalcharacteristics of β-emitting therapeutic radionuclidesthat have been used for labeling liposomes.

Table 4 contains a list of therapeutic radionu-clides and methods that have been reported forlabeling liposomes with therapeutic radionuclides. Incertain cases, the methods currently used to label lipo-somes with diagnostic radionuclides could easily beadapted for the labeling of liposomes with therapeuticradionuclides. To date, this avenue of cancer therapy



TABLE 3 Physical Characteristics of Therapeutic Radionuclides Described for Radiolabeling Liposomes

Radionuclide(abbreviation)

Half-lifea Average Energy (MeV)a Mean Range (mm)a X/γ -ray Energies (Abun-dance) (keV)a,

Production

Iodine-131 (131I) 8.0 days 0.182 0.91 284 (5.8) Reactor

364 (82)

637 (6.5)

Lutetium-177(177Lu)

6.7 days 0.133 0.67 113 (7) Reactor

208 (11)

Rhenium-186(186Re)

3.8 days 0.362 1.8 137 (9) Reactor

Rhenium-188(188Re)

16.9 h 0.764 3.5 155 (15) Reactor andgenerator

Yttrium-90 (90Y) 2.7 days 0.935 3.9 None Reactor andgenerator

aData derived from Ref. 87–89.

with liposomes is mostly theoretical, with few studiesreported in the literature using liposomes as carriersof therapeutic radionuclides in cancer therapy.86,90–92

Previous theoretical dosimetry studies haveaddressed the potential use of radiotherapeutic lipo-somes for treatment of tumors via intravenous90,91,98

and intraperitoneal injection.99 There are some sig-nificant advantages of using the intratumoral deliveryroute for liposomes containing therapeutic radionu-clides compared to intravenous injection, such asthe much lower radiation doses delivered to liver,spleen, kidneys and other normal tissues, as well asthe potential of simultaneous treatment of metastaticlymph nodes that drain from the region of thetumor.100 Another significant advantage of intratu-morally administered radiotherapeutic liposomes isthat perfect homogeneity of distribution within the

solid tumor is not required, as the β-particle pathlength treats a field of tumor cells surrounding thelocation of the radiolabeled liposomes. This meansthat the therapeutic radionuclides carried with lipo-somes do not necessarily have to come directly intocontact with every cancer cell. In addition, release ofthe radionuclide from the liposomes is not requiredfor effective therapy. This field effect of β-particleemission is illustrated in Figure 4.

One possible use of radiotherapeutic liposomesis to treat residual tumor in the intraoperativesituation. In many cases, the surgeon is unable toremove all of the cancer during surgery so thatthe margins of the resected tumor are positive. Thisgenerally means that there is cancer remaining at theoperative site, which severely compromises patientsurvival. This positive margin can frequently be

TABLE 4 Methods of Labeling Liposomes with Therapeutic Radionuclides

Radionuclide (abbreviation) Labeling Mode Labeling Method Cite

Iodine-131 (131I) Encapsulation Sodium iodide or iodinated phospholipid 36

After-loading Iodinated activated esters with arginine-containing liposomes 37

Lutetium-177 (177Lu) Bilayer intercalation 3-cholesteryl NTA hexyl ether 93

Rhenium-186 (186Re) Encapsulation Rephos chelator 94

Bilayer intercalation

After-loading BMEDA with cysteine/GSH-liposomes 43

After-loading BMEDA with ammonium sulfate gradient 12

Rhenium-188 (188Re) Encapsulation Rephos chelator 94

Bilayer intercalation

After-loading BMEDA with ammonium sulphate gradient 95,96

Yttrium-90 (90Y) Surface chelation DTPA-phospholipid containing liposomes 92

After-loading Ionophore A23187 with NTA-liposomes 97



FIGURE 4 | Field effect of β-particle emission.

determined during the operation. Radiotherapeuticliposomes that target residual tumor could be injectedin the region of the positive tumor margin to sterilizethe surgical margin of tumor cells.105 Because theliposomes will drain through the lymph nodes, theywould also have the potential to treat micrometastasisin those nodes. These liposomes could contain atherapeutic radionuclide for radionuclide therapy, achemotherapeutic drug, or a combination of both.Intraoperatively applied liposomes could, therefore,provide an additional tool for the surgeon, particularlywhen the margins of the tumor are positive.

LIPOSOMES LABELED WITHRHENIUM-186 AND RHENIUM-188Our group has developed a novel method of labelingliposomes with the radionuclides of rhenium. Thismethod uses BMEDA to post-load 99mTc, 188Re,or 186Re into liposomes.70 A second research grouphas also used this same BMEDA chemistry to labelliposomes with 188Re.95,96

One of the advantages of rhenium therapeuticradionuclides compared with some other therapeuticradionuclides is that they emit a low ratio of photonsin the ideal range of gamma camera imaging. Forevery 10 and 6.5β-emissions, both 186Re and 188Reradionuclides emit a γ -photon. This is an ideal ratioof β-to γ -emissions. The photon energy of bothrhenium radionuclides is in the range of the photonenergy of 99mTc (140 keV) so that the radiolabeledliposomes can be tracked continuously over timewhile performing therapy. This ability to image thedistribution of 186Re-liposomes provides a significant

advantage as compared to other commonly usedtherapeutic radionuclides such as yttrium-90 (90Y)and phosphorus-32 (32P), which are pure β-emittingradionuclides without γ -photons for imaging. Asecond significant advantage of 188Re and 186Recompared with many other therapeutic radionuclidessuch as 90Y and 32P is that they have almostno affinity for bone. Bone uptake occurs whenthe radioisotope eventually becomes separated fromits chelator during metabolism of its carrier inthe body. High bone uptake, which results in thesuppression of hematopoietic cells in bone marrow,is generally the dose-limiting factor with currentantibody-based radionuclide therapies. In contrast,rhenium radionuclides are mainly cleared through theliver and kidneys in a similar manner as 99mTc.

Previous theoretical dosimetry studies haveaddressed the potential use of radiotherapeuticliposomes for treatment of tumors via intravenousinjection.101–103 In a recent experimental study withlong-circulating pegylated liposomes radiolabeledwith 188Re using the BMEDA method, 188Reliposomes had much higher uptake in the tumoras compared with 188Re-BMEDA alone. 188Re-liposomes were found to have a 7.1-fold highertumor-to-muscle ratio as compared to intravenouslyadministered unencapsulated 188Re-BMEDA in a C26murine colon carcinoma solid tumor animal model.95

In addition to this intravenous investigationof 188Re-liposomes, our group has investigated thepotential use of 186Re-liposomes for intratumoraltherapy.70 High-resolution SPECT/CT images revealthe intratumoral distribution of therapeutic liposomesin Figure 5. The SPECT image was obtained from the



FIGURE 5 | High-resolution SPECT/CT images revealing intratumoraldistribution of therapeutic liposomes.

137 keV γ -photons emitted by 186Re. The fusion ofthe SPECT image with the higher-resolution CT imageprovides improved anatomic detail of the intratu-moral distribution of the 186Re-liposomes. The abilityto image the precise location of radionuclides duringtherapy can provide image-based feedback for further

image-guided delivery of 186Re-liposomes to ensurecomplete tumor coverage of therapeutic liposomes.There are some significant advantages of using theintratumoral delivery route for 186/188Re-liposomescompared to the intravenous injection route, suchas the much lower radiation dose delivered to liver,spleen, kidneys and other normal tissues, and thepotential of simultaneous targeting of metastaticlymph nodes that drain from the region of the tumoras we just addressed.100

Recent initial results with intratumorally admin-istered 186Re-liposomes by our group have demon-strated effective local control of 1.5-cm-diametersolid tumors in a rat head-and-neck xenograft tumormodel.104

CONCLUSION

The ability to track the uptake of liposomes in humansand research animals on a whole body basis is provid-ing researchers with an excellent tool for developingtargeted liposome-based drug delivery agents. Theattachment of therapeutic radionuclides to liposomesalso has great promise in cancer therapy. Many ofthe radiolabeling and noninvasive imaging technolo-gies developed for nanosized liposomes will also beuseful for the tracking of other nanoparticle sys-tems under development as diagnostic and therapeuticagents.

REFERENCES

1. Bangham AD. Liposomes: The Babraham connection.Chem Phys Lipids 1993, 64:275–285.

2. Gregoriadis G, Ryman BE. Liposomes as carriers ofenzymes or drugs: A new approach to the treatmentof storage diseases. Biochem J 1971, 124:58.

3. Barratt G, Legrand P. Comparison of the efficacy andpharmacology of formulations of amphotericin B usedin treatment of leishmaniasis. Curr Opin Infect Dis2005, 18:527–530.

4. Gabizon A, Shmeeda H, Barenholz Y. Pharmacoki-netics of pegylated liposomal doxorubicin: Review ofanimal and human studies. Clin Pharmacokinet 2003,42:419–436.

5. Torchilin VP, Papisov MI. Why do polyethyleneglycol-coated liposomes circulate so long? J LiposomeRes 1994, 4:725–739.

6. Phillips WT, Klipper RW, Awasthi VD, Rudolph AS,Cliff R, et al. Polyethylene glycol-modified liposome-encapsulated hemoglobin: a long circulating red

cell substitute. J Pharmacol Exp Ther 1999,288:665–670.

7. Allen TM, Hansen C, Martin F, Redemann C, Yau-Young A. Liposomes containing synthetic lipidderivative of poly(ethylene) show prolonged circu-lation half-lives in vivo. Biochim Biophys Acta 1991,1066:29–36.

8. Erdogan S, Roby A, Torchilin VP. Enhanced tumorvisualization by gamma-scintigraphy with 111In-labeled polychelating-polymer-containing immunoli-posomes. Mol Pharmacol 2006, 3:525–530.

9. Gupta B, Torchilin VP. Monoclonal antibody 2C5-modified doxorubicin-loaded liposomes with signifi-cantly enhanced therapeutic activity against intracra-nial human brain U-87 MG tumor xenografts innude mice. Cancer Immunol Immunother 2007,56:1215–1223.

10. Allen TM, Sapra P, Moase E. Use of the post-insertionmethod for the formation of ligand-coupled lipo-somes. Cell Mol Biol Lett 2002, 7:889–894.



11. Sapra P, Tyagi P, Allen TM. Ligand-targeted lipo-somes for cancer treatment. Curr Drug Deliv 2005,2:369–381.

12. Bao A, Goins B, Klipper R, Negrete G, Phillips WT.Direct 99mTc labeling of pegylated liposomal doxoru-bicin (Doxil) for pharmacokinetic and non-invasiveimaging studies. J Pharmacol Exp Ther 2004,308:419–425.

13. Gabizon A, Huberty J, Straubinger RM, Price DC,Papahadjopoulos D. An improved method for in vivotracing and imaging of liposomes using a gallium67-deferoxamine complex. J Liposome Res 1988–89,1:123–135.

14. Goins B, Klipper R, Rudolph AS, Cliff RO,Blumhardt R, et al. Biodistribution and imaging stud-ies of technetium-99m labeled liposomes in rats withfocal infection. J Nucl Med 1993, 34:2160–2168.

15. Laverman P, Dams ETM, Oyen WJG, Storm G,Koenders EB, et al. A novel method to label liposomeswith Tc-99m via the hydrazino nicotinyl derivative. JNucl Med 1999, 40:192–197.

16. Phillips WT, Rudolph AS, Goins B, Timmons JH,Klipper R, et al. A simple method for producing atechnetium-99m-labeled liposome which is stable invivo. Nucl Med Biol 1992, 19:539–547.

17. Townsend DW, Carney JP, Yap JT, Hall NC. PET/CT today and tomorrow. J Nucl Med 2004, 45(Suppl1):4S–14S.

18. O’Connor MK, Kemp BJ. Single-photon emissioncomputed tomography/computed tomography: Basicinstrumentation and innovations. Semin Nucl Med2006, 36:258–266.

19. Zaidi H. Recent developments and future trends innuclear medicine instrumentation. Z Med Phys 2006,16:5–17.

20. Torchilin VP. Targeted pharmaceutical nanocarriersfor cancer therapy and imaging. Am Assoc Pharm Sci2007, 9:E128–E147.

21. Barnes, WE. In: Henkin RE, Boles Ma, Dillehay Gl,Halama Jr, Karesh Sm, et al. eds., Basic Physics ofNuclear Medicine. St Louis, MO: Mosby; 1996,43–63.

22. Laverman P, Boerman OC, Storm G. Radiolabeling ofliposomes for scintigraphic imaging. Methods Enzy-mol 2003, 373:234–248.

23. Goins BA, Phillips WT. In: Gregoriadis G. eds. Meth-ods for Tracking Radiolabeled After Injection inthe Body. New York: Informa Healthcare; 2007,191–209.

24. Laverman P, Storm G, Phillips WT, Bao A, Goins B.In: Gregoriadis G. eds. Radiolabeling of Liposome forScintigraphic Imaging. New York: Informa Health-care; 2007, 169–185.

25. Love WG, Amos N, Williams BD, Kellaway IW.Effect of liposome surface charge on the stability

of technetium (99mTc) radiolabelled liposomes. AnnRheum Dis 1989, 6:105–113.

26. Ogihara I, Kojima S, Jay M. Differential uptake ofgallium-67-labeled liposomes between tumors andinflammatory lesions in rats. J Nucl Med 1986,27:1300–1307.

27. Holmberg E, Maruyama K, Litzinger DC, Wright S,Davis M, et al. Highly efficent immunoliposomes pre-pared with a method which is compatible with variouslipid compositions. Biochem Biophys Res Commun1989, 165:1272–1278.

28. Klibanov A, Maruyama K, Torchilin VP, Huang L.Amphipathic polyethylene glycols effectively prolongthe circulation time of liposomes. FEBS Lett 1990,268:235–237.

29. Hwang KJ, Merriam JE, Beaumier PL, Luk KS.Encapsulation, with high efficiency, of radioactivemetal ions in liposomes. Biochim Biophys Acta 1982,716:101–109.

30. Mauk MR, Gamble RC. Preparation of lipid vesi-cles containing high levels of entrapped radioactivecations. Anal Biochem 1979, 94:302–307.

31. Proffitt R, Williams L, Presant C, Tin G, Uliana J,et al. Tumor-imaging potential of liposomes loadedwith In-111-NTA: biodistribution in mice. J NuclMed 1983, 24:45–51.

32. Awasthi VD, Goins B, Klipper R, Phillips WT. Dualradiolabeled liposomes: biodistribution studies andlocalization of focal sites of infection in rats. NuclMed Biol 1998, 25:155–160.

33. Boerman OC, Storm G, Oyen WJG, van Bloois L, vander Meer JWM, et al. Sterically stabilized liposomeslabeled with indium-111 to image focal infection.J Nucl Med 1995, 36:1639–1644.

34. Gabizon A, Chisin R, Amselem S, Druckmann S,Cohen R, et al. Pharmacokinetic and imagingstudies in patients receiving a formulation ofliposome-associated adriamycin. Br J Cancer 1991,64:1125–1132.

35. Harrington KJ, Mohammadtaghi S, Uster PS, GlassD, Peters AM, et al. Effective targeting of solid tumorsin patients with locally advanced cancers by radiola-beled pegylated liposomes. Clin Cancer Res 2001,7:243–254.

36. Hardy JG, Kellaway IW, Rogers J, Wilson CG. Thedistribution and fate of 131I-labelled liposomes. JPharm Pharmacol 1980, 32:309–313.

37. Mougin-Degraef M, Jestin E, Bruel D, Remaud-LeSaec P, Morandeau L et al. High-activity radio-iodinelabeling of conventional and stealth liposomes. J Lipo-some Res 2006, 16:91–102.

38. Barratt GM, Tuzel NS, Ryman BE. In: Gregoriadis G.eds., The Labeling of Liposomal Membranes withRadioactive Technetium.Boca Raton, FL: CRC Press;1984, 93–106.



39. Richardson VJ, Jeyasingh K, Jewkes RF, Ryman BE,Tattersall MHN. Properties of [99mTc] technetium-labelled liposomes in normal and tumour-bearing rats.Biochem Soc Trans 1977, 5:290–291.

40. Tilcock C, Ahkong QF, Fisher D. 99mTc-labeling oflipid vesicles containing the lipophilic chelator PE-DTPA: effect of tin-to chelate ratio, chelate contentand surface polymer on labeling efficiency and biodis-tribution behavior. Nucl Med Biol 1994, 21:89–96.

41. Ahkong QF, Tilcock C. Attachment of 99mTc tolipid vesicles containing the lipophilic chelatedipalmitoylphosphatidylethanolamine-DTTA. NuclMed Biol 1992, 19:831–840.

42. Phillips WT, Rudolph AS, Goins B, Timmons JH,Klipper R. et al. A simple method for producing atechnetium-99m-labeled liposome which is stable invivo. Int J Rad Appl Instrum B 1992, 19:539–547.

43. Bao A, Goins B, Klipper R, Negrete G, Mahin-daratne M, et al. A novel liposome radiolabel-ing method using 99mTc-‘‘SNS/S’’ complexes: invitro and in vivo evaluation. J Pharm Sci 2003,92:1893–1904.

44. Oku N, Tokudome Y, Tsukada H, Okada S. Real-time analysis of liposomal trafficking in tumor-bearing mice by use of positron emission tomography.Biochim Biophys Acta 1995, 1238:86–90.

45. Marik J, Tartis MS, Zhang H, Fung JY, Kheirolo-moom A, et al. Long-circulating liposomes radiola-beled with [(18)F]fluorodipalmitin ([(18)F]FDP). NuclMed Biol 2007, 34:165–171.

46. Phillips WT, Lemen L, Goins B, Rudolph AS,Klipper R, et al. Use of oxygen-15 to measure oxygen-carrying capacity of blood substitutes in vivo. Am JPhysiol 1997, 272:H2492–H2499.

47. Goins B, Klipper R, Martin C, Jerabek PA,Khalvati S, et al. Use of oxygen-15-labeled molec-ular oxygen delivery studies of blood and bloodsubstitutes. Adv Exp Med Biol 1998, 454:643–652.

48. Awasthi V, Yee SH, Jerabek P, Goins B, Phillips WT.Cerebral oxygen delivery by liposome-encapsulatedhemoglobin: a positron-emission tomographic evalu-ation in a rat model of hemorrhagic shock. J ApplPhysiol 2007, 103:28–38.

49. Hnatowich DJ, Friedman B, Clancy B, Novak M.Labeling of preformed liposomes with Ga-67 andTc-99m by chelation. J Nucl Med 1981, 22:810–814.

50. Torchilin VP, Weissig V, Martin FJ, Heath TD. In:Torchilin VP, Weissig V. eds., Surface Modification ofLiposomes. Oxford: Oxford University Press; 2003,193.

51. Tilcock C, Ahkong QF, Fisher D. 99mTc-labeling oflipid vesicles containing the lipophilic chelator PE-DTTA: effect of tin-to-chelate ratio, chelate contentand surface polymer on labeling efficiency and biodis-tribution behavior. Nucl Med Biol 1994, 21:89–96.

52. Dams ET, Laverman P, Oyen WJ, Storm G, Scher-phof GL. et al. Accelerated blood clearance andaltered biodistribution of repeated injections of ster-ically stabilized liposomes. J Pharmacol Exp Ther2000, 292:1071–1079.

53. Laverman P, Brouwers AH, Dams ET, Oyen WJ,Storm G. et al. Preclinical and clinical evidence fordisappearance of long-circulating characteristics ofpolyethylene glycol liposomes at low lipid dose. JPharmacol Exp Ther 2000, 293:996–1001.

54. Gabizon A, Huberty J, Straubinger RM, Price DC,Papahadjpulos D. An improved method for in vivotracing and imaging of liposomes using a gallium-67-desferoxamine complex. J Liposome Res 1989,1:123–135.

55. Hwang KJ, Merriam JE, Beaumier PL, Luk KF.Encapsulation, with high efficiency, of radioactivemetal ions in liposomes. Biochim Biophys Acta 1982,716:101–109.

56. Corvo ML, Boerman OC, Oyen WJ, Van Bloois L,Cruz ME, et al. Intravenous administration of super-oxide dismutase entrapped in long circulating lipo-somes. II. In vivo fate in a rat model of adjuvant arthri-tis. Biochim Biophys Acta 1999, 1419:325–334.

57. Harrington KJ, Mohammadtaghi S, Uster PS, GlassD, Peters AM, et al. Effective targeting of solid tumorsin patients with locally advanced cancers by radiola-beled pegylated liposomes. Clin Cancer Res 2001,7:243–254.

58. Metselaar JM, Wauben MH, Wagenaar-Hilbers JP,Boerman OC, Storm G. Complete remission of exper-imental arthritis by joint targeting of glucocorticoidswith long-circulating liposomes. Arthritis Rheum2003, 48:2059–2066.

59. Gaspar MM, Boerman OC, Laverman P, Corvo ML,Storm G, et al. Enzymosomes with surface-exposedsuperoxide dismutase: in vivo behaviour and ther-apeutic activity in a model of adjuvant arthritis. JControl Release 2007, 117:186–195.

60. Harrington KJ, Rowlinson-Busza G, Syrigos KN,Uster PS, Vile RG, et al. Pegylated liposomes havepotential as vehicles for intratumoral and sub-cutaneous drug delivery. Clin Cancer Res 2000,6:2528–2537.

61. Goins B, Klipper R, Rudolph AS, Cliff RO, BlumhardtR, et al. Biodistribution and imaging studies oftechnetium-99m-labeled liposomes in rats with focalinfection. J Nucl Med 1993, 34:2160–2168.

62. Awasthi V, Goins B, McManus L, Klipper R,Phillipsa WT. [99mTc] liposomes for localizing exper-imental colitis in a rabbit model. Nucl Med Biol 2003,30:159–168.



63. Lee CM, Choi Y, Huh EJ, Lee KY, Song HC. et al.Polyethylene glycol (PEG) modified 99mTc-HMPAO-liposome for improving blood circulation and biodis-tribution: the effect of the extent of PEGylation.Cancer Biother Radiopharm 2005, 20:620–628.

64. Dagar S, Krishnadas A, Rubinstein I, Blend MJ,Onyuksel H. VIP grafted sterically stabilized lipo-somes for targeted imaging of breast cancer: in vivostudies. J Control Release 2003, 91:123–133.

65. Erdogan S, Ozer AY, Ercan MT, Hincal AA. Scinti-graphic imaging of infections with 99m-Tc-labelledglutathione liposomes. J Microencapsul 2000,17:459–465.

66. Oyen WJ, Boerman OC, Storm G, van Bloois L,Koenders EB, et al. Detecting infection and inflamma-tion with technetium-99m-labeled Stealth liposomes.J Nucl Med 1996, 37:1392–1397.

67. Tilcock C, Yap M, Szucs M, Utkhede D. PEG-coatedlipid vesicles with encapsulated technetium-99m asblood pool agents for nuclear medicine. Nucl MedBiol 1994, 21:165–170.

68. Sou K, Goins B, Takeoka S, Tsuchida E, Phillips WT.Selective uptake of surface-modified phospholipidvesicles by bone marrow macrophages in vivo. Bio-materials 2007, 28:2655–2666.

69. Kleiter MM, Yu D, Mohammadian LA, Niehaus N,Spasojevic I, et al. A tracer dose of technetium-99m-labeled liposomes can estimate the effect of hyperther-mia on intratumoral doxil extravasation. Clin CancerRes 2006, 12:6800–6807.

70. Bao A, Goins B, Klipper R, Negrete G, Phillips WT.186Re-liposome labeling using 186Re-SNS/S com-plexes: in vitro stability, imaging, and biodistributionin rats. J Nucl Med 2003, 44:1992–1999.

71. Goins BA. In: Kumar MNVR. eds. RadiolabeledLipid Nanoparticles for Cancer Diagnosis and Treat-ment. American Scientific Publishers; 2008, 65–82.

72. Goins BA, Phillips WT. In: Torchilin VP, Weissig V.eds., Radiolabelled Liposomes for Imaging andBiodistribution Studies.Oxford: Oxford UniversityPress; 2003, 319–336.

73. Phillips WT, Goins B. Assessment of liposome deliv-ery using scintigraphic imaging. J Liposome Res 2002,12:71–80.

74. Morgan JR, Williams KE, Davies RL, Leach K,Thomson M, et al. Localisation of experimen-tal staphylococcal abscesses by 99mTc-technetium-labelled liposomes. J Med Microbiol 1981,14:213–217.

75. Laverman P, Boerman OC, Oyen WJ, Dams ET,Storm G, et al. Liposomes for scintigraphic detectionof infection and inflammation. Adv Drug Deliv Rev1999, 37:225–235.

76. Laverman P, Dams ET, Storm G, Hafmans TG,Croes HJ, et al. Microscopic localization of PEG-liposomes in a rat model of focal infection. J ControlRelease 2001, 75:347–355.

77. Phillips WT, Medina LA, Klipper R, Goins B. A novelapproach for the increased delivery of pharmaceuticalagents to peritoneum and associated lymph nodes. JPharmacol Exp Ther 2002, 303:11–16.

78. Medina LA, Calixto SM, Klipper R, Phillips WT,Goins B. Avidin/biotin-liposome system injected inthe pleural space for drug delivery to mediastinallymph nodes. J Pharm Sci 2004, 93:2595–2608.

79. Medina LA, Klipper R, Phillips WT, Goins B. Phar-macokinetics and biodistribution of [111In]-avidinand [99mTc]-biotin-liposomes injected in the pleuralspace for the targeting of mediastinal nodes. NuclMed Biol 2004, 31:41–51.

80. Zavaleta CL, Phillips WT, Soundararajan A, GoinsBA. Use of avidin/biotin-liposome system forenhanced peritoneal drug delivery in an ovarian cancermodel. Int J Pharm 2007, 337:316–328.

81. Elbayoumi TA, Torchilin VP. Enhanced accumula-tion of long-circulating liposomes modified with thenucleosome-specific monoclonal antibody 2C5 in var-ious tumours in mice: gamma-imaging studies. Eur JNucl Med Mol Imaging 2006, 33:1196–1205.

82. Ueno NT, Bartholomeusz C, Xia W, Anklesaria P,Bruckheimer EM, et al. Systemic gene therapy inhuman xenograft tumor models by liposomal deliveryof the E1A gene. Cancer Res 2002, 62:6712–6716.

83. Villaret D, Glisson B, Kenady D, Hanna E, Carey M,et al. A multicenter phase II study of tgDCC-E1A forthe intratumoral treatment of patients with recurrenthead and neck squamous cell carcinoma. Head Neck2002, 24:661–669.

84. Nishikawa M, Hashida M. Pharmacokinetics of anti-cancer drugs, plasmid DNA, and their delivery systemsin tissue-isolated perfused tumors. Adv Drug DelivRev 1999, 40:19–37.

85. Bao A, Phillips WT, Goins B, Zheng X, Sabour S,et al. Potential use of drug carried-liposomes for can-cer therapy via direct intratumoral injection. Int JPharm 2006, 316:162–169.

86. Kostarelos K, Emfietzoglou D. Liposomes as carriersof radionuclides: from imaging to therapy. J LiposomeRes 1999, 9:429–460.

87. Kowalsky RJ, Falen SW. Radiopharmaceuticals inNuclear Pharmacy and Nuclear Medicine. 2nd ed.Washington, DC: American Pharmacists Association;2004.

88. Zweit J. Radionuclides and carrier molecules for ther-apy. Phys Med Biol 1996, 41:1905–1914.

89. Hoefnagel CA. Radionuclide cancer therapy. AnnNucl Med 1998, 12:61–70.



90. Kostarelos K, Emfietzoglou D. Tissue dosimetryof liposome-radionuclide complexes for internalradiotherapy: Toward liposome-targeted therapeu-tic radiopharmaceuticals. Anticancer Res 2000,20:3339–3346.

91. Kostarelos K, Emfietzoglou D, Stamatelou M.Liposome-mediated delivery of radionuclides totumor models for cancer radiotherapy: a quantitativeanalysis. J Liposome Res 1999, 9:407–424.

92. McQuarrie S, Mercer J, Syme A, Suresh M, Miller G.Preliminary results of nanopharmaceuticals used inthe radioimmunotherapy of ovarian cancer. J PharmPharmacol Sci 2005, 7:29–34.

93. Bard DR, Knight CG, Page-Thomas DP. Effect of theintra-articular injection of lutetium-177 in chelatorliposomes on the progress of an experimental arthritisin rabbits. Clin Exp Rheumatol 1985, 3:237–242.

94. Hafeli U, Tienfenauer LX, Schubiger PA, Weder HG.A lipophilic complex with 186Re/188Re incorporatedin liposome suitable for radiotherapy. Nucl Med Biol1991, 18:449–454.

95. Chang YJ, Chang CH, Chang TJ, Yu CY, Chen LC,et al. Biodistribution, pharmacokinetics andmicroSPECT/CT imaging of 188Re-bMEDA-liposome in a C26 murine colon carcinomasolid tumor animal model. Anticancer Res 2007,27:2217–2225.

96. Chen LC, Chang CH, Yu CY, Chang YJ, Hsu WC,et al. Biodistribution, pharmacokinetics and imag-ing of (188)Re-BMEDA-labeled pegylated liposomesafter intraperitoneal injection in a C26 colon car-cinoma ascites mouse model. Nucl Med Biol 2007,34:415–423.

97. Utkhede D, Yeh V, Szucs M, Tilcock C. Uptake ofyttrium-90 into lipid vesicles. J Liposome Res 1994,9:407–424.

98. Emfietzoglou D, Kostarelos K, Sgouros G. An ana-lytical dosimetry study for the use of radionuclide-liposomes conjugates in internal radiotherapy. J NuclMed 2001, 42:499–504.

99. Syme AM, Kirkby C, Riauka TA, Fallone BG,McQuarrie S. Monte carlo investigation of single cellbeta dosimetry for intraperitoneal radionuclide ther-apy. Phys Med Biol 2004, 49:1959–1972.

100. Phillips WT, Klipper R, Goins B. Novel method ofgreatly enhanced delivery of liposomes to lymphnodes. J Pharmacol Exp Ther 2000, 295:309–313.

101. Emfietzoglou D, Kostarelos K, Sgouros G. An analyticdosimetry study for the use of radionuclide-liposomeconjugates in internal radiotherapy. J Nucl Med 2001,42:499–504.

102. Kostarelos K, Emfietzoglou D. Tissue dosimetryof liposome-radionuclide complexes for internalradiotherapy: Toward liposome-targeted therapeu-tic radiopharmaceuticals. Anticancer Res 2000,20:3339–3345.

103. Kostarelos K, Emfietzoglou D, Papakostas A, YangWH, Ballangrud A, et al. Binding and interstitialpenetration of liposomes within avascular tumorspheroids. Int J Cancer 2004, 112:713–721.

104. Bao A, Phillips W, Goins B, Otto RA. Tumorbrachytherapy using intratumoral injection of beta-emitting therapeutic radionuclides carried withinnanoparticles. Med Phys 2006, 33:2236–2237.

105. Wang SX, Bao A, Herrera SJ, Phillips WT, Goins B,Santoyo C, Miller FR, Otto RA. Intra operative186Re - liposome radionuclide therapy in a headand neck squamous cell carcinoma xenograft pos-itive surgical margin model. Clin Canc Res 2008,14(12):3975–3983.

FURTHER READING

Goins BA, Phillips WT. Radiolabeled liposomes for imaging and biodistribution studies. In: Torchilin VP,Weissig V. eds., Liposomes. A Practical Approach. 2nd ed. Oxford University Press; 2003.Laverman Peter, Phillips WT, Bao A, Storm Gert, Goins BA. Radiolabeling of liposomes for scintigraphicimaging. In: Gregoriadis G. eds., Liposome Technology. 3rd ed. Boca Raton, FL: CRC Press; 2007.


Liposome activated delivery.Liposomes as pharmaceutical nanocarriers.Liposome- and immunoliposome-based cancer therapeutics.


Advanced Review

Magnetic resonance susceptibilitybased perfusion imaging of tumorsusing iron oxide nanoparticlesArvind P. Pathak1,2∗

Abundant preclinical and preliminary clinical data have convincingly supportedantiangiogenic therapy as an effective strategy for the inhibition of tumor growth.This has led to an acute need for developing biological markers (biomarkers) ofvascular remodeling that can be monitored in vivo, at repeated intervals in largenumbers of patients with a variety of tumors in a noninvasive manner. Recently,magnetic resonance (MR) perfusion imaging with iron oxide nanoparticleshas demonstrated the potential to be such a surrogate endpoint, that is, abiomarker intended to substitute for a clinical endpoint and predictive of clinicalbenefit. Consequently, both US Food and Drug Administration (FDA) and theNational Cancer Institute (NCI) have major initiatives underway to improvethe development of cancer therapies and the outcomes for cancer patients viabiomarker development and evaluation. The biophysical principles, physiologicalrelevance and range of imaging techniques underlying the success of susceptibilitybased contrast MR perfusion imaging with iron oxide nanoparticles as such abiomarker, are the subject of this review. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 84–97

The dependence of tumor growth and metastasis onthe vasculature has made angiogenesis a promis-

ing target for therapy.2,3 Strategies for inhibitingangiogenesis comprise a broad spectrum of antiangio-genic agents which include endothelial toxins directlytargeted to endothelial antigens, growth factor antag-onists that silence angiogenic growth factor signal-ing, protease inhibitors that inhibit proteases crucialto tumor invasion, and endogenous antiangiogeniccompounds.4 As classified by Siemann et al., strate-gies for abolishing extant tumor vessels comprise anarray of vascular disrupting agents that include tumorendothelium targeted antibodies conjugated to variousmoieties that induce thrombosis, and small molecule

∗Correspondence to: Arvind P. Pathak, Russell H. MorganDepartment of Radiology and Radiological Science, Johns HopkinsUniversity In Vivo Cellular Molecular Imaging Center (JHUICMIC), Baltimore, MD, USA. E-mail: [email protected] H. Morgan Department of Radiology and RadiologicalScience, Johns Hopkins University In Vivo Cellular MolecularImaging Center (JHU ICMIC), Baltimore, MD, USA.2Department of Oncology, The Johns Hopkins University School ofMedicine, Baltimore, MD, USA.

DOI: 10.1002/wnan.017

drugs that induce extensive vascular collapse by dis-rupting the cytoskeleton of proliferating endothelialcells.5 Several of these agents are currently in variousphases of clinical trials.6 Recent results from the van-guard of antiangiogenic therapy such as the demon-stration of the clinically significant normalization oftumor vessels in recurrent glioblastoma patients bydaily administration of AZD2171, an oral tyrosinekinase inhibitor of vascular endothelial growth factor(VEGF) receptors,7 and the enhanced progression-free survival in chemotherapy-naıve metastatic orrecurrent breast cancer patients treated with beva-cizumab, a VEGF-specific antibody, in conjunctionwith chemotherapy compared to patients treatedwith standard chemotherapy alone,8 have heraldedan urgent need for biomarkers to guide antiangio-genic monotherapy as well as combination therapy.9

These developments have collectively made it imper-ative to identify in vivo biomarkers of angiogenesisto: (1) direct the dosing and scheduling of antian-giogenic agents, (2) obtain early measurable signs oftheir therapeutic efficacy, (3) aid in the identificationof the tumor vascular ‘normalization’ window, and(4) enable the testing of novel therapeutic regimensand agents. ‘Vascular normalization’ was described in


WIREs Nanomedicine and Nanobiotechnology MR susceptibility based perfusion imaging

1972 by Le Serve and Hellmann10 and more recentlydefined by Jain as the restoration of the blood vesselarchitecture (i.e., vessel density and caliber, fractaldimensions, pericyte coverage or basement membranedevelopment) and consequently restoration of vascu-lar function assessed in terms of a decrease in bloodvessel permeability to macromolecules (and thus alle-viated interstitial fluid pressure) and hypoxia, andimproved blood perfusion,11 all of which resulted inenhanced drug delivery to the tumor and eventuallyimproved patient outcome.11

The plethora of available contrast mechanisms,in conjunction with its superior dynamic functionalrange, bestow on magnetic resonance imaging (MRI)the potential to be a noninvasive, in vivo biomarkerthat can circumvent the drawbacks of traditionalbiomarkers of angiogenesis, which include tissuebiopsies and microvessel density (MVD).12 Althoughwidely used, both biopsies and MVD only providestatic ‘snap-shots’ and do not lend themselves todynamic in situ assessments of the status of thetumor microvasculature, nor can they be used fornoninvasive, in vivo monitoring at repeated intervalsin large cohorts of patients.13 Although there exista wide array of MR contrast mechanisms, theexquisite sensitivity of susceptibility based MRI tothe underlying vascular architecture gives it thepotential to be a formidable tool in the noninvasive,in vivo assessment of tumor angiogenesis.14 Thisreview briefly highlights the basic concepts underlyingthe susceptibility based contrast mechanism, with aspecific focus on perfusion MRI of cancer usingiron oxide nanoparticles. It also calls attention tosome of the novel techniques investigators haveemployed for assessing tumor angiogenesis, efficacyof antiangiogenic therapy and vascular normalization,and concludes with caveats to bear in mind whenemploying these techniques.

SUSCEPTIBILITY BASED MAGNETICRESONANCE CONTRASTA material’s tendency to interact with and distort anapplied magnetic field is quantified in terms of its mag-netic susceptibility. Specifically, the magnetization(M) induced in a sample is related to the applied field(H) by the susceptibility (χ ) of the material which isgiven by: M = χH. The majority of tissues relevant tohuman MRI are diamagnetic (i.e., χ < 0), while para-magnetic and ferromagnetic contrast agents exhibitsusceptibilities of χ > 0 and χ � 1, respectively.15

Often, distortions in the applied magnetic field such asthose caused by surgical staples, biomedical implantsand internal susceptibility differences at various

tissue interfaces, lead to unwanted artifacts in MRI.Then again, susceptibility is also an intrinsic tissueproperty and regional variations in susceptibilitycan be exploited to extract important physiologicalinformation about healthy and pathological tissueby the judicious use of different MR pulse sequencesand/or MR contrast agents.15

Since Rosen et al.16,17 demonstrated in the early1990s that a bolus of a high dose of gadolinium(Gd)-chelated contrast agent produces a transientdecrease in signal intensity that can be converted intoa concentration-time curve from which the relativecerebral blood volume (rCBV) can be computedusing tracer kinetic principles,18 there has been adramatic increase in the use and development ofsusceptibility contrast-based MRI. This has not onlyresulted in the development of elegant models forelucidating the biophysical phenomena involved,19–23

but has also resulted in the development of novelimaging strategies24 and contrast agents,25 includingsuperparamagnetic iron oxide (SPIO) nanoparticles.26

Specifically, when a paramagnetic contrast agentis restricted to the vascular compartment, each bloodvessel perturbs the local magnetic field that waterprotons experience (Figure 1). As a consequence, eachdiffusing water proton experiences a slightly differentmagnetic field, and thus resonates at a differentfrequency. As protons diffuse through the microscopicfield heterogeneities they lose phase coherencebecause of their random Brownian motion throughthese heterogeneous field distributions, eventuallyresulting in the attenuation of the MR signal20,21

(Figure 1). It should be pointed out that even withoutthe diffusive movement of water, there exists aheterogeneity of resonant frequencies because ofthe presence of microscopic field inhomogeneitieswithin an imaging voxel, which in turn affects theMR signal intensity (in gradient echo images) bycausing intravoxel dephasing. When the passage ofa (super)paramagnetic contrast agent bolus is trackeddynamically using either T2- or T∗

2-weighted MR pulsesequences,16 it is known as dynamic susceptibilitybased contrast (DSC) MRI, which is discussed in anensuing section.

The effect o magnetic field inhomogeneities ontransverse relaxation can be characterized as:19,20

1T∗

2= 1

T2+ 1

T′2

(1)

The relaxation rate 1/T∗2 (R∗

2) is the rate at whichthe gradient-echo (GE) signal amplitude decays orthe ‘effective’ T2 and 1/T2 (R2) is the rate atwhich the spin-echo (SE) amplitude decays or the



(a)Capillary

χ1 χ2

χ1 ≈ χ2

χ1 χ2

χ1 > χ2

Capillary

Distance (µm)Distance (µm)

(b) (c)

0.4

0.3

0.2

0.1

0

7060

5040

3020

100 0

1020

3040

5060

70

−0.1

−0.2

−0.3

−0.4

∆B/B

0∆χ

B0

FIGURE 1 | Schematic illustrating the origins of susceptibility based Magnetic resonance (MR) contrast. (a) In the absence of any susceptibilitydifference between blood (χ1) and the surrounding tissue (χ2), no microscopic magnetic field gradient is set up and diffusing water protonsexperience the same local magnetic field. (b) When a susceptibility difference (�χ ) arises between the intravascular space and the surroundingtissue, say as a result of the presence of superparamagnetic iron oxide (SPIO) contrast agent, a microscopic field gradient (---) is set up that perturbsthe local magnetic field, and diffusing water protons experience different local magnetic fields, leading to loss of phase coherence, and MR signalattenuation that can be followed dynamically using either T2- or T∗

2 -weighted MR pulse sequences. This constitutes the basis of dynamicsusceptibility based contrast (DSC) magnetic resonance imaging (MRI). (c) Surface plot illustrating the three-dimensional aspects of mathematicallysimulated microscopic magnetic field gradients induced around a microvessel. The orientation of the applied field (B0) and the axis along which thenormalized field change (�B/B0�χ ) is plotted are shown in the inset. (Reprinted, with permission, from Ref.14. Copyright 2004 Elsevier).

‘true’ T2. The relaxation rate, 1/T′2 (R′

2), is therelaxation rate contribution attributable to magneticfield inhomogeneities. In the presence of a magneticfield perturber, that is, contrast agent bearing tumorvessel, the relative R2 and R∗

2 relaxation rates dependon the diffusion coefficient (D) of spins in the vicinityof the induced magnetic field inhomogeneity, theradius (R) of the magnetic field perturber (i.e., tumorvessel diameter), and the variation of the Larmorfrequency at the surface of the perturber.19,20,22,27 Thetwo physical characteristics R and D can be collapsedinto the proton correlation time τD:

τD = R2

D(2)

The variation in Larmor frequency (dω), at the surfaceof the tumor vessel is given by:

dω = γ (�χ )B0 (3)

where γ is the proton gyromagnetic ratio (42.58MHz/T), �χ the susceptibility difference between thetumor vessel and its background tissue (∼ 1–10 ppmfor SPIO15,28), and B0 the strength of the appliedmagnetic field. The relative magnitudes of τD and dω

determine the magnitude of the susceptibility inducedrelaxation effects or contrast, which can broadlybe classified into three regimes.17,19,20,22,27 The firstregime is one in which the rate of diffusion (1/τD)

is substantially greater than the frequency variation(dω), that is τDδω � 1. In this ‘fast exchange’ regime,the fast diffusion rate of the spins causes them toexperience a similar range of field inhomogeneitieswithin an echo time (TE), resulting in minimalloss of phase coherence and similar loss of phasecoherence between gradient- and SE sequences (Figure2(a)). This is also called the ‘motional averaged’regime as the susceptibility induced local magneticfield gradients are averaged out.19–21 The secondregime is one in which the rate of diffusion (1/τD)is substantially smaller than the frequency variation(dω), that is, τDδω � 1. In this ‘slow exchange’ regime,the phase that a proton accumulates as it passes aperturber is large, that is the effect is the same as itwould be for the case of static field inhomogeneities.Because of the absence of motion averaging, theGE relaxation rate tends to be greater than the SErelaxation rate (Figure 2(a)). There will be no signalattenuation on a T2-weighted scan because the 180◦

pulse during the SE sequence refocuses static magneticfield inhomogeneities, while intravoxel dephasing stilloccurs in a GE sequence (because of the absence of asimilar refocusing RF pulse). The third regime is one inwhich τD δω ∼ 1, that is, water diffusion is neither fastenough to be in the motionally narrowed regime, norslow enough to be approximated as linear gradientcharacteristic of the slow regime, making descriptionsof the susceptibility induced contrast more complex.



Fast

(a) (b)

Intermediate Slow

4.0

GE, TE = 60 ms

SE, TE = 100 ms

3.0

In(S

)/T

E (

l/s)

2.0

1.0

0.01.0 10.0

Radius (µm)

100.0

∆R2

Vessel diameter

∆R2,

∆R

2* (

sec−1

)

∆R2*

FIGURE 2 | (a) Schematic illustrating the three different regimes of susceptibility induced relaxation effects and the differential sensitivity ofgradient-echo (GE) (�R∗

2 ) and spin-echo (SE) (�R2) relaxation rates to vessel caliber. This sensitivity to vessel size constitutes the basis for imagingmacro- and microvascular blood volume as well as imaging vessel-size. (b) Size dependence of �R∗

2 and �R2 for fractional volume = 2% and�χ = 1 × 10−7. �R2 peaks for microvessels, while �R∗

2 > �R2 for all radii and plateaus for macrovessels. (Reprinted, with permission, fromRefs.14,21).

In this ‘intermediate exchange’ regime, SE relaxationis maximum and the GE relaxation is similar to whatit would be in the slow exchange regime (Figure 2(a)).In this regime, analytic solutions to estimate signalloss in the presence of diffusion become invalid andnumerical simulations are required.19–21,27

From this phenomenological description ofsusceptibility based contrast, it is apparent that SEand GE sequences have greatly differing sensitivitiesto the size and scale of the field inhomogeneities,resulting in a differential sensitivity to tumor vesselsize. Monte Carlo simulations have demonstrated that(Figure 2(b)) the SE relaxation rate change (�R2)increases, reaches a maximum for capillary-sizedvessels (∼ 4–5 µm), and then decreases inversely withvessel radius.21 The GE relaxation rate change (�R∗

2)increases and then plateaus to remain independentof vessel radius beyond capillary-sized vessels (Figure2(b)). A consequence of this result is that the SErelaxation rate changes are maximally sensitive to themicrovascular blood volume, while the GE changesare more sensitive to the total blood volume. Thisresulting contrast has been exploited in different waysby various investigators as will become apparent inthe sections that follow.

Iron Oxide Nanoparticles in MagneticResonance ImagingWhen the magnetic dipole moments of closelypacked neighboring atoms interact and preferen-tially align parallel to each other, such materialsdemonstrate a magnetic ordering that results inregions or ‘domains’ that are always spontaneouslymagnetized.29 If one decreases the size of such mul-tidomain ferro- or ferrimagnetic particles, eventually

one ends up with a single magnetic domain particlethat can exhibit unique magnetic properties knownas superparamagnetism25,29 (Figure 3(a)). Superpara-magnetic particles have very high magnetic suscep-tibilities and can be magnetized to saturation evenin weak applied magnetic fields (Figure 3). However,when this external field is removed, the magneticdipole moments of individual superparamagnetic par-ticles become randomly oriented as a result of thermalmotions and no net magnetization is retained by thesample, unlike in ferro- or ferrimagnetic materialswhich exhibit remnant magnetization29,30 (Figures3(b–c)). As classified in an outstanding review byWang et al.,25 iron oxide particles used in MRI belongto the class of SPIO contrast agents that include largeor oral SPIO agents, standard SPIO agents, ultrasmallSPIO (USPIO) agents and monocrystalline iron oxidenanoparticle (MION) agents, of which the first threeare in various stages of clinical trials (Tables 1 and2). These superparamagnetic contrast agents shortenthe T1 and T2/T∗

2 relaxation times and exhibit R1

and R2 relaxivities (4 × 104 and 16 × 104 sec−1 M−1,respectively at 20 MHz for AMI-25) an order ofmagnitude greater than that of conventional paramag-netic chelates like diethylenetriamine penta-acetic acid(Gd-DTPA) (0.45 × 104 and 0.57 × 104 sec−1 M−1,respectively at 20 MHz)31 and because of their smallersize and surface properties they usually exhibit longercirculation times. Their eventual effect on the MR sig-nal depends on several factors such as the MR pulsesequence employed, size of the particles, contrast agentcomposition and concentration.31,32 At a given tem-perature, ferro- and superparamagnetism are deter-mined by particle size, e.g., magnetite comprised ofcrystals ∼ 5–30 nm is superparamagnetic but becomes



MS

MR

F

= Magnetic domains

(a) (b)

2000

1000

S

B0

50 1000(c)

H (Oe)

M (

cmu/

cm3 )

FIGURE 3 | Schematic illustrating: (a) random orientations of the magnetic domains in a superparamagnetic iron oxide (SPIO) particle in theabsence of any applied magnetic field, and (b) application of an external magnetic field B0 causing the magnetic domains of the SPIO particle toorient along B0. (c) Superparamagnetic particles (S) (open arrows) have a very high magnetic susceptibility and can be magnetized (M) to saturation(MS = saturation magnetization) even in weak external magnetic fields (H). Unmagnetized ferromagnetic and ferrimagnetic materials (F) (solidarrows) also become magnetized along this curve. However, once magnetized, ferromagnetic and ferrimagnetic materials retain their magnetization(MR = remnant magnetization), even if the external field is reduced to zero. However, at ambient temperatures superparamagnetic materials do notretain their magnetization in the absence of an applied field. (Reprinted, with permission, from Ref. 25 Copyright 2001 Springer and Ref. 30 Copyright1972 Addison-Wesley).

ferromagnetic when made up of larger crystals.31

Josephson et al. also demonstrated that the iron incolloids made from para-, ferro- and SPIOs had vastlydifferent effects on both R1 and R2 relaxivities.31 Theyobserved that the larger the average iron oxide coresize, the greater the R2/R1 ratio, and thus proposeda classification scheme for MR contrast agents onthe basis of the relative magnitudes of R1 and R2,rather than the susceptibility of the agent. Accordingto their classification, a type I agent would be onewith 1 � R2/R1 � 2, such as paramagnetic chelates; atype II agent would be one with 2< R2/R1 � 25, suchas dispersed superparamagnetic oxides e.g., AMI-25;and a type III agent would be one with R2/R1 > 25,such as ferromagnetic iron oxides.31

In most cases the susceptibility induced MRsignal attenuation produced by SPIOs is exploitedto obtain strong T2- and T∗

2-weighted contrast.However, when particles of smaller size are employedin conjunction with T1-weighted MR pulse sequences,strongly T1-weighted images can be obtained.26

In addition to the abovementioned factors, imagecontrast in SPIO enhanced MRI is not only dependenton the biodistribution of the agent but can also beprofoundly affected by the spatial distribution orclustering of these particles.31,33

Several different MRI applications of SPIOs havebeen realized depending upon the size and compo-sition (or coating) of the nanoparticles employed.These applications have been summarized in an excel-lent review by Corot et al. and include imagingmacrophage uptake in a gamut of disease modelsranging from stroke to infection, metastatic lymph

node imaging, MR angiography, molecular imaging,and cellular label imaging.24 The ensuing sections willbriefly describe the different techniques in which SPIOperfusion imaging has been employed in the study ofcancer and the phenomenon of tumor angiogenesis.

THE SIGNIFICANCE OF ASSESSINGPERFUSION IN CANCER

Over a century ago, Virchow recognized that thetumor stroma exhibited a distinct capillary network.34

Since then, studies of tumor vascular morphologyhave identified a variety of structural and functionaldifferences between tumor and normal vasculature(a comprehensive review is given in Ref.35). Tumor-induced blood vessels are typically sinusoidal, fragileand highly permeable with discontinuous basementmembranes. Other characteristics of the tumorvasculature include (1) spatial heterogeneity anddisorganized branching hierarchies, (2) arterio-venousshunts, (3) acutely and transiently collapsing vessels,(4) poorly differentiated and leaky vessels lacking insmooth muscle cell lining, and (5) an inability to matchthe metabolic demand of rapidly proliferating cancercells, resulting in areas of hypoxia and necrosis.

The structural anomalies of the tumor vascu-lature not only result in areas of hypoxic tissuewithin the tumor mass but also in altered hemo-dynamics, blood rheology, and blood flow36 withprofound consequences on conventional treatmentmodalities, pharmacokinetics, radiosensitivity, prolif-eration rate, invasive and metastatic potential, andthe metabolic microenvironment (pO2, pH, energy



TABLE 1 Classification of Superparamagnetic Iron Oxide (SPIO) Agents

Agent Particle Size Class of Agent Trade Name Developer FDA Status

AMI-227 300 nm Oral SPIO Lumirem (EU) Guerbet S.A. Approved

Gastromark (US) Advanced Magnetics Approved

OMP 3.5 µm Oral SPIO Abdoscan Amersham Health (nowGE Healthcare)

Approved

AMI-25 80–150 nm SSPIO Endorem (EU) Guerbet Approved

Feridex IV (US) Bayer HealthCarePharmaceuticals

Approved

SHU-555A 62 nm SSPIO Resovist (EU) Bayer Schering PharmaAG

Phase III com-plete

AMI-277 20–40 nm USPIO Sinerem (EU) Guerbet S. A. Discontinued

Combidex (US) AMAG Pharmaceuticals Phase III

NC100150 20 nm USPIO Clariscan (US) Amersham Health (nowGE Healthcare)

Discontinued

– 2–5 nm MION – – Experimental

Adapted from Ref.25 with permission

TABLE 2 Pharmacokinetic Properties ofSuperparamagnetic Iron Oxide (SPIO) Agents

Type of Surface Half-life BiodistributionAgent Coatings

Oral SPIO Siloxane,polystyrene

Oral Gut

SSPIO Dextran,carbodextran

2–6 h Liver and spleen

USPIO Dextran,carbohydrate-polyethyleneglycol

24–36 h Lymph nodes

MION Dextran ∼ 4 h Targeted imag-ing

Parameters for this table obtained from Refs 24,25

status).14 This makes it imperative to have noninva-sive, in vivo indices of tumor perfusion and of theremodeled tumor vascular geometry, to enable us tooptimize tumor perfusion related parameters and facil-itate successful chemo-, immuno-, radio-, thermo- ortargeted-therapy.

In addition, as mentioned above the dependenceof tumor growth and metastasis on the vasculaturehas made angiogenesis a promising target for therapy.Strategies for inhibiting angiogenesis or normalizingthe extant vasculature are currently in various phasesof clinical trials with recent results indicating anurgent need for biomarkers to guide such therapies.As some of the pre-clinical and clinical approachesdescribed below will indicate, by providing an indexof changes in both vascular structure (vessel size index,

permeability and blood volume) and function (bloodflow or perfusion), susceptibility contrast (SPIO)enhanced perfusion MRI demonstrates the potentialto be such a biomarker.

TYPES OF PERFUSION-RELATEDMAGNETIC RESONANCE IMAGINGPerfusion MRI with SPIOs can be broadly classi-fied into three categories: (1) dynamic susceptibilitycontrast MRI, (2) steady-state susceptibility contrastMRI, and (3) vessel-size imaging. A brief descriptionof each of these applications follows.

Dynamic Susceptibility Contrast MagneticResonance ImagingSeveral investigators have acquired rCBV maps fromfirst-pass DSC studies, with good spatiotemporalresolution.16,37,38 With this technique, preliminaryresults indicate that MRI-derived rCBV may betterdifferentiate histologic tumor types than conventionalMRI39,40 and provide information to predict tumorgrade.41 To quantitatively measure rCBV or cerebralblood flow (CBF), regional changes in signal intensityversus time need to be converted into concentrationversus time curves. As mentioned above, bothempirical data and modeling indicate that fora given TE, the T∗

2 rate change (�R∗2 = 1/T∗

2 =1/T∗

2postcontrast − 1/T∗2precontrast) is proportional to the

brain tissue concentration:20,21

�R∗2 = 1

T∗2postcontrast

− 1T∗

2precontrast= k[conc.] (4)



where k is a tissue-specific, MR pulse sequenceand field strength dependent, calibration factor.16,42

Assuming monoexponential signal decay, the signalintensity change following SPIO injection is

S(t) = S0e−TE(�R∗2(t)) (5)

yielding:

−1TE

ln(

S(t)S0

)= kC(t) (6)

where S0 is the signal intensity before administrationof the contrast agent, S(t) the tissue signal withcontrast, TE the echo time, and C(t) the concentration-time curve of the tissue. Without knowledge of thearterial input function (AIF) it is still possible todetermine relative values of CBV and the mean transittime (MTT), which in turn yield relative values ofthe CBF.43 For example, one can determine rCBVby computing the area under the concentration-timecurve, that is, �R∗

2(t) curve by directly integrating ona voxel-wise basis:

rCBV =t∫

0

�R∗2(τ )dτ = −1

TE

t∫0

lnS(τ )S0

dτ (7)

The relative MTT (rMTT) is then given as:

rMTT =

t∫0

τ�R∗2(τ )dτ

t∫0

�R∗2(τ )dτ

(8)

And the relative (rCBF) as:

rCBF = rCBVrMTT

(9)

However to quantify the absolute blood volume andperfusion, one must know the concentration of thetracer in the arterial blood supply Ca. As outlined byOstergaard,44 for an infinitely short bolus of contrastagent, the tissue concentration as a function of timemay then be written as

Ct(t) = CBF × Ca × R(t) (10)

where R(t) is the tissue residue function that describesthe fraction of tracer present in the voxel at time t postinjection. Thus, the residue function is a decreasingfunction of time, wherein R(0) = 1, and if the tracerdoes not bind to the vasculature, R(∞) = 0. In Eq. 10,

the product CBF × R(t) is known as the tissue impulseresponse function. However, since most injections areof finite duration, the observed concentration-timecurve is the convolution of the ideal tissue-transitcurve with the AIF Ca(t):44

Ct(t) = CBF × Ca(t) ⊗ R(t) (11)

In order to extract the absolute CBF from Eq. 11,the tissue impulse response needs to be determinedby deconvolution, essentially fitting CBF×R(t) fromthe experimental data. Typically, Eq. 11 is solved forCBF ×R(t) using a transform approach or by a linearalgebraic approach.44 Briefly, in the Fourier transform(FT) approach, the convolution theorem of the FT isutilized, that is, F(f ⊗ g) = F(f ).F(g). Hence Eq. 11can be solved as follows:

F{Ct(t)} = F{CBF × R(t) ⊗ Ca(t)}= F{CBF × R(t)} × F{Ca(t)}⇒ F{Ct(t)}

F{Ca(t)} = F{CBF × R(t)}

⇒ F−1{

F{Ct(t)}F{Ca(t)}

}= CBF × R(t) (12)

where F and F−1 denote the discrete and inversediscrete FTs, respectively. Since R(0) = 1, the CBFcan be determined from the initial height of the tissueimpulse response function, and the CBV from:

CBV =

t∫0

Ct(τ )dτ

t∫0

Ca(τ )dτ

(13)

For an instantaneous bolus injection, the centralvolume principle states that CBF = CBV/MTT, whereMTT is the mean transit time of contrast agentthrough the vascular network.18 A detailed expositionof the principles of cerebral perfusion imaging, thelinear algebraic solution to Eq. 11 and technical issuesassociated with the solutions of Eq. 11, have beenpresented in an excellent review by Ostergaard.44 Inorder to reduce reperfusion effects on assessmentsof CBV and CBF, Ca(t) and Ct(t) are corrected forrecirculation, which is usually accomplished by fittingthem to a gamma-variate function with a recirculationcut-off.45 An obstacle to the application of the centralvolume principle for the calculation of blood flowis the direct measurement of the MTT. Weisskoffet al. have demonstrated that MTT is not the firstmoment of the concentration-time curve for MR of



Normal30

25

20

No.

of o

bser

vatio

ns

15

10

5

00 40 80

(a)

(b)

∆R2*/∆R2

120 160

Tumor

Normal60

50

40

No.

of o

bser

vatio

ns

30

20

10

00 10 20

Diameter (µm)

30 40 50 60 70

Tumor

(c) (d)

FIGURE 4 | (a) Post-Gd MR image of a 9L gliosarcoma bearing rat brain illustrating the tumor ROI (yellow hatched ellipse).(b) Post-monocrystalline iron oxide nanoparticle (MION) ratio �R∗

2 /�R2 map of a 9L gliosarcoma bearing rat brain wherein one can clearly see theelevated ratio values in the angiogenic tumor rim (yellow hatched ellipse). (c) A histogram of the �R∗

2 /�R2 data showing that for the tumor ROI (overall slices), �R∗

2/�R2 is shifted to the right i.e., larger caliber vessels with respect to the contralateral brain ROI (over all slices) �R∗2/�R2, a difference

that is also apparent from (d) the histogram of the stereologically calculated vessel radii. (Reprinted, with permission, from Ref.62. Copyright 2001).

intravascular tracers, but does provide a useful relativemeasure of flow.46

There have been several DSC studies of tumorxenograft models using SPIOs that employ theprinciples described above. These include a study thatdeveloped the appropriate stereologic correlates forhistologically validating MR-derived cerebral bloodvolume maps in a brain tumor model using MION.38

Another demonstrated the efficacy of DSC MRI intracking the morphological and functional changesinduced by the antiangiogenic agent SU11657 in abrain tumor model.47 As mentioned above, the SErelaxation rate changes are maximally sensitive to themicrovascular blood volume, while the GE changesare more sensitive to the total blood volume. Onthe basis of this observation, SE sequences have beenused in many tumor studies with the assumption thattumor angiogenesis is primarily characterized by anincrease in the microvasculature.39 However, giventhe large (> 20 µm) tortuous vessels usually foundin tumors,38 whether SE or GE methods are mostappropriate remains to be determined.

Steady-State Susceptibility ContrastMagnetic Resonance ImagingAs demonstrated in the section above, measurement ofperfusion necessitates the use of a dynamic MRI pulsesequence that can sample the first-pass of the SPIOwith the appropriate temporal resolution. However,more recently it has been demonstrated by severalinvestigators that the rCBV can be mapped under‘steady-state’ conditions, that is after the first-pass of

the contrast agent and without prior knowledge ofthe AIF.42,48,49 This has only been made possible withthe development of ‘blood pool’ contrast agents withvery long intravascular half lives such as USPIOs andMION.26,50 Pre- and post-contrast T2- or T∗

2-weightedimages can be acquired at high-resolution with a highsignal-to-noise ratio from which the steady-state �R2or �R∗

2 can be computed reliably, because unlike DSCprotocols the temporal resolution requirements forsteady-state susceptibility MRI are less stringent, andUSPIOs and MION have very strong susceptibilityeffects. When an intravascular contrast agent suchas USPIO or MION achieves steady-state in theintravascular space after mixing, the rCBV can becalculated on a voxel-wise basis according to:51

rCBV ∝ �R∗2 = 1

TEln

Sprecontrast

Spostcontrast(14)

where TE is the echo time, Sprecontrast the GE sig-nal before administering the SPIO contrast agent andSpostcontrast the GE signal after the contrast agent hasachieved steady-state throughout the blood pool. Asimilar relationship can be used to calculate the rCBVfrom the SE signal. As before, the differential sensitiv-ity of the gradient- and SE sequences can be exploitedto give a ‘macrovascular’ rCBV and a ‘microvascular’rCBV (i.e., weighted towards capillary-sized vessels),respectively. The steady-state approach has been ele-gantly validated using radiotracer measurements andintravital microscopy wherein it was demonstratedthat there was a good correlation between techniquesin terms of their ability to assess ‘angiogenic burden’



4.00(g)

3.50

3.00

2.50

2.00Rat

io

1.50

1.00

0.50

0.00Gray vs. White

MRI: Slope1/Slope2

(a) (b)

(c) (d) (e) (f)

*

Histology: fv1/fv2

Contra vs. Tumor

FIGURE 5 | (a) Post-Gd axial magnetic resonance imaging (MRI) ofthe rat brain illustrating the ROIs for the tumor (red) and contralateral(yellow) brain. (b) Post-monocrystalline iron oxide nanoparticle (MION)axial MRI of the rat illustrating the ROIs for the gray (gray) and white(white) matter, respectively. Binarized images of tissue sections ofmicrofilled vessels at 20× magnification of: (c) tumor, (d) contralateralbrain, (e) gray matter, and (f) white matter ROIs. The color of eachframe on the lower panel corresponds to the color of the ROI fromwhich the tissue sections were sampled for histology. (g) Comparison ofthe ratios of fractional volumes obtained from MRI and histology for thetumor versus contralateral, and gray versus white ROIs, respectively.The error bars represent the standard error of the mean for eachtechnique (N = 6 rats for the gray/white and N = 8 rats for thecontra/tumor). There was no significant (P = 0.525) differencebetween �R∗

2gray/�R∗2white and the ratio gray/white fractional

volume but a significant (P = 0.005) difference between�R∗

2tumor/�R∗2contra and the ratio tumor/contralateral fractional

volume. This result indicates that the GE calibration factor is the samefor gray and white matter but not the same for brain and tumor tissue.(Reprinted, with permission, from Ref. 39. Copyright 1994).

in a range of tumor models.52 This technique hasbeen employed to characterize tumor angiogenesis invarious experimental brain tumor models.28,42,48 Inconjunction with blood oxygenation level dependent(BOLD) MRI, another susceptibility contrast basedtechnique,53 it has been employed to simultaneouslyglean information about the tumor vascular architec-ture and its functional hemodynamic status.54 Morerecently the feasibility of using steady-state USPIO-

enhanced MRI for early quantitative monitoring ofantiangiogenic and vascular disrupting therapy wasdemonstrated in a murine fibrosarcoma model55 anda rat prolactinoma model,56 respectively. One studyemployed the USPIO NC100150 [Clariscan; Amer-sham Health] to assess the effects of overexpressingdimethylarginine dimethylaminohydrolase (DDAH),which metabolizes the endogenous inhibitors of NOsynthesis of asymmetric dimethylarginine and N-monomethyl-l-arginine, on tumor vascular morpho-genesis in a glioma model in vivo.57 Another studyutilized MION to assess the antiangiogenic effects ofVEGF receptor tyrosine kinase inhibitors in a drug-resistant colon carcinoma model.58 USPIO enhancedMRI can be especially useful in delineating the ‘co-opted’ (i.e., recruitment of preexisting vessels) vascularphenotype exhibited by low grade gliomas which doesnot enhance with conventional gadolinium based DCEMRI.59 Such measurements have also been used toassess tumor response to patupilone, a potent micro-tubule stabilizer and vascular disrupting agent60 andmore recently to interrogate the effects of tumor-derived platelet-derived growth factor (PDGF) ontumor angiogenesis.61 Several of these steady-stateapplications of USPIOs in animal models of diseaseare well summarized in Ref. 51.

Vessel-Size Magnetic Resonance ImagingAs mentioned above, the compartmentalization ofa superparamagnetic contrast agent within thevasculature induces magnetic field perturbations thatextend far into the tissue (Figure 1), increasingits relaxation rates (R2 and R∗

2). Assuming amonoexponential signal delay this enhancement inrelaxation rates (�R2 and �R∗

2) is given by Eq. 4.Also, as pointed out above no analytical solutionsexist for the intermediate exchange regime, whereinone must resort to Monte Carlo-type numericalsimulations for predicting the effect on �R2 and �R∗

2.These results are shown in Figure 2(b), which showsthat �R2 increases and peaks at a vessel radius of∼ 4–5 µm, while �R∗

2 increases and then plateaus toremain independent of vessel size for vessel radii>∼5–6 µm.21 Since Boxerman et al. had demonstratedcomputationally that both �R2 and �R∗

2 exhibit analmost linear response with the fractional vascularvolume and �χ , Dennie et al. proposed taking theratio �R∗

2/�R2 as a means of minimizing thisdependence of the relaxation rates on these usuallyunknown quantities.49 For a given SPIO dose, Dennieet al. demonstrated that this ratio �R∗

2/�R2 increasedlinearly with vessel size and could be employedas a metric of the average vessel size within a



35

100908070605040302010

00 10 20 30 40 50

3025201510

50

0 2500 5000 7500rCBV (au)

(f)

(b)(a)

(c)

(d)

(e)

(g)

Fractional volume (%)

No.

of o

bser

vatio

nsN

o. o

f obs

erva

tions

10000 12500 15000

Contra rCBVTumor rCBV

Contra FVTumor FV

FIGURE 6 | (a) Post-Gd MR image of a 9L gliosarcoma bearing rat brain illustrating the contralateral (C) and tumor (T) ROIs employed for MRrelative cerebral blood volume (rCBV) histogram analyses. (b) rCBV map computed from the first-pass of monocrystalline iron oxide nanoparticle(MION), overlaid on a high-resolution anatomical image: note elevated rCBV in the tumor, a finding consistent with a larger fractional blood volume(FV) relative to the contralateral brain. (c) Post-Gd MR image illustrating the contralateral (C) and tumor (T) grids that were sampled for stereologicalanalyses. H&E stained tissue section of vessels perfused with Microfil (a silicone injection compound) at 20× magnification of (d) tumor tissue and(e) normal brain. (f) A histogram of the GE rCBV data showing that the tumor rCBV is shifted to the right with respect to the contralateral rCBV, adifference that is also apparent from (g) the histogram of the stereologically calculated FV. (Reprinted, with permission, from Ref. 35. Copyright 2000Springer Verlag).

voxel, and also demonstrated that ‘vessel-size’ mapsobtained in a rat glioma model using an intravascularsuperparamagnetic iron-oxide nanoparticle (MION)contrast agent compared favorably to the predictedratio using histologically determined vessel sizes.49 Anexample of such a ratio map is shown in62 Figure 4.On the basis of the GE signal predicted by theYablonskiy and Haacke model63 in conjunction withthe SE signal predicted by the model proposed byKiselev and Posse,64 Tropres et al. formulated analternative expression for the vessel size index (R):28

R = 0.425(

Dγ�χB0

)1/2 (�R∗

2

�R2

)3/2

(15)

where absolute measurement of R requiresmeasurement of the diffusion coefficient (D) and ofthe increase in blood susceptibility (�χ ) after SPIOadministration. Finally, on the basis of the samesusceptibility models, Jensen and Chandra formulatedthe relaxation rate shift index (Q) that is independentof the contrast agent dose (above a certain thresholddose) and given by:65

Q ≡ �R2(�R∗

2)2/3 (16)

Wu et al. demonstrated the feasibility of imaging themouse brain microvasculature using this approach,and were able to show significant differencesbetween various cortical areas in agreement with thehistologically assessed MVD.66

While there have been several clinical studieswith SPIOs for a variety of targets and pathologies,24

clinical susceptibility based perfusion studies oftumors with SPIOs remain limited.67–69 With severaldifferent SPIOs in various stages of clinical trials, it isa matter of time before we see widespread applicationof susceptibility contrast enhanced MRI biomarkersin the clinical setting for a wide array of diseasesexhibiting anomalous vasculatures, including cancer.

CAVEATS AND ISSUESFinally, all DSC measurements are made assumingthat the calibration factor ‘k′ (Equation 4) is thesame for different tissue types and independent oftissue pathology. However, a recent study has shownthat k was the same for brain gray and whitematter but not the same for normal brain andtumor tissue for the GE pulse sequence42 (Figure 5).This difference was attributable to the grosslydifferent vascular morphologies of tumors because



of tumor angiogenesis, compared to normal brainand/or possibly differing blood rheological factorssuch as hematocrit. Consequently, the sensitivity toblood volume differences between tumor and normalbrain tissue may be underestimated when using GEsusceptibility contrast agent methods. Additionally,as summarized by Kiselev, susceptibility effects canbe considered to operate over various spatial scalesto induce relaxation,70 and range from the spin-spininteractions between the water protons and the con-trast agent ions, that is, the molecular or microscopicscale, to the scale at which protons are sensitive to thegeometry of the susceptibility induced magnetic fieldgradients, which in turn are defined by the microvascu-lar architecture and/or hematocrit of the vessels, thatis, the mesoscopic scale, and finally the scale at whichthere exist inhomogeneities in the magnetic field due toimperfect shimming and so on, that is the macroscopicscale. This dependence of contrast agent relaxivityon the microvascular architecture is especially rele-vant to the context of tumor angiogenesis whereinthe aberrant vessel morphology can drastically affectthe observed image contrast in DSC protocols42 (Fig-ure 6). This makes it imperative to develop simulationmethodologies and models that take the ‘de facto’microvascular architecture into account, such as thoseproposed in Refs 71,72. Finally, for the steady-state

imaging approaches, care must be taken to ensure thatcontrast agent is not in its saturated regime or thatthe relationship between the concentration of the con-trast agent and the relaxation rate is linear so that Eq.4 is still valid. For example, empiricial data49 havedemonstrated that this linear relation holds true forlow doses of MION (up to ∼ 5 mg Fe/kg) but becomessublinear for higher doses.28,51

CONCLUSION

As a result of its exquisite sensitivity to the underly-ing vasculature, in vivo susceptibility based contrastMRI is proving to be an important new tool for non-invasively elucidating the vascular remodeling thataccompanies tumor angiogenesis, vascular normaliza-tion and antiangiogenic therapy. Perfusion imagingwith SPIO nanoparticles has been an indispensabletool in the characterization of these phenomena. How-ever, it should be borne in mind that the mechanismsunderlying this contrast mechanism are complex andstill being elucidated. A better understanding of thecomplex interplay between the microvascular architec-ture and biophysics of the susceptibility induced MRsignal will result in the development of physiologicallymore relevant biomarkers.

NOTES

Supported by NIH P50CA103175—Career Development Award to APP.

REFERENCES

1. Atkinson AJ Jr., Colburn WA, De Gruttola VG, DeMets DL, Downing GJ, Hoth DF, Oates JA, Peck CC,Schooley RT, Spilker BA, Woodcock J, Zeger SL.Biomarkers and surrogate endpoints: preferred defini-tions and conceptual framework. Clin Pharmacol Ther2001, 69(3):89–95.

2. Folkman J. Tumor angiogenesis: therapeutic implica-tions. N Engl J Med 1971, 285(21):1182–1186.

3. Dvorak HF. Vascular permeability factor/vascularendothelial growth factor: a critical cytokine in tumorangiogenesis and a potential target for diagnosis andtherapy. J Clin Oncol 2002, 20(21):4368–4380.

4. Miller KD. Recent translational research: antiangio-genic therapy for breast cancer–where do we stand?Breast Cancer Res 2004, 6(3):128–132.

5. Siemann DW, Bibby MC, Dark GG, Dicker AP,Eskens FA, et al. Differentiation and definition of

vascular-targeted therapies. Clin Cancer Res 2005, 11(2Pt 1):416–420.

6. Banerjee S, Dowsett M, Ashworth A, Martin LA.Mechanisms of disease: angiogenesis and the manage-ment of breast cancer. Nat Clin Pract Oncol 2007,4(9):536–550.

7. Batchelor TT, Sorensen AG, di Tomaso E, Zhang WT,Duda DG, et al. AZD2171, a pan-VEGF receptor tyro-sine kinase inhibitor, normalizes tumor vasculature andalleviates edema in glioblastoma patients. Cancer Cell2007, 11(1):83–95.

8. Miller KD. E2100: A randomized phase III trial ofpaclitaxel versus paclitaxel plus bevacizumab as first-line therapy for locally recurrent or metastatic breastcancer. 41st Annual Meeting of the American Societyof Clinical Oncology. Orlando, FL; 2005.



9. Jain RK, Duda DG, Clark JW, Loeffler JS. Lessons fromphase iii clinical trials on anti-vegf therapy for cancer.Nat Clin Pract Oncol 2006, 3(1):24–40.

10. Le Serve AW, Hellmann K. Metastases and the nor-malization of tumour blood vessels by ICRF159: a newtype of drug action. Br Med J 1972, 1(5800):597–601.

11. Jain RK. Normalization of tumor vasculature: anemerging concept in antiangiogenic therapy. Science2005, 307(5706):58–62.

12. Gasparini G. Clinical significance of determination ofsurrogate markers of angiogenesis in breast cancer. CritRev Oncol Hematol 2001, 37(2):97–114.

13. Hlatky L, Hahnfeldt P, Folkman J. Clinical applicationof antiangiogenic therapy: microvessel density, whatit does and doesn’t tell us. J Natl Cancer Inst 2002,94(12):883–893.

14. Pathak AP, Gimi B, Glunde K, Ackerstaf E,Artemov D, et al. Molecular and functional imag-ing of cancer: advances in MRI and MRS. MethodsEnzymol 2004, 386:3–60.

15. Schenck JF. The role of magnetic susceptibility inmagnetic resonance imaging: MRI magnetic compat-ibility of the first and second kinds. Med Phys 1996,23(6):815–850.

16. Rosen BR. Susceptibility contrast imaging of cerebralblood volume: human experience. Magn Reson Med1991, 22:293–299.

17. Villringer A, Rosen BR, Belliveau JW, Ackerman JL,Lauffer RB, et al. Dynamic imaging with lanthanidechelates in normal brain: contrast due to magnetic sus-ceptibility effects. Magn Reson Med 1988, 6:164–174.

18. Zierler KL. Theoretical basis of indicator-dilutionmethods for measuring flow and volume. Circ Res1962, 10:393–407.

19. Fisel CR, Ackerman JL, Buxton RB, Garrido L,Belliveau JW, et al. MR contrast due to microscopi-cally heterogeneous magnetic susceptibility: numericalsimulations and applications to cerebral physiology.Magn Reson Med 1991, 17:336–347.

20. Kennan RP. Intravascular susceptibility contrast mech-anisms in tissues. Magn Reson Med 1994, 31:9–21.

21. Boxerman JL, Hamberg LM, Rosen BR, WeisskoffRM. MR contrast due to intravascular magnetic sus-ceptibility perturbations. Magn Reson Med 1995,34:555–566.

22. Yablonskiy DA. Quantitation of intrinsic magneticsusceptibility-related effects in a tissue matrix: phantomstudy. Magn Reson Med 1998, 39:417–428.

23. Kiselev VG. On the theoretical basis of perfusion mea-surements by dynamic susceptibility contrast MRI.Magn Reson Med 2001, 46:1113–1122.

24. Corot C, Robert P, Idee JM, Port M. Recent advancesin iron oxide nanocrystal technology for medical imag-ing. Adv Drug Deliv Rev 2006, 58(14):1471–1504.

25. Wang YX, Hussain SM, Krestin GP. Superparamag-netic iron oxide contrast agents: physicochemical char-acteristics and applications in MR imaging. Eur Radiol2001, 11(11):2319–2331.

26. Weissleder R, Elizondo G, Wittenberg J, Rabito CA,Bengele HH, et al. Ultrasmall superparamagnetic ironoxide: characterization of a new class of contrast agentsfor MR imaging. Radiology 1990, 175(2):489–493.

27. Weisskoff RM, Zuo CS, Boxerman JL, Rosen BR.Microscopic susceptibility variation and transverserelaxation: theory and experiment. Magn Reson Med1994, 31:601–610.

28. Tropres I, Grimault S, Vaeth A, Grillon E, Julien C,et al. Vessel size imaging. Magn Reson Med 2001,45(3):397–408.

29. Saini S, Frankel RB, Stark DD, Ferrucci JT. Mag-netism: a primer and review. Am J Roentgenol 1988,150:735–743.

30. Cullity BD. Definitions and units. Introduction to Mag-netic Materials: Reading, MA: Addison-Wesley; 1972,1–18.

31. Josephson L, Lewis J, Jacobs P, Hahn PF, Stark DD.The effects of iron oxides on proton relaxivity. MagnReson Imaging 1988, 6(6):647–653.

32. Pouliquen D, Perroud H, Calza F, Jallet P, Le Jeune JJ.Investigation of the magnetic properties of iron oxidenanoparticles used as contrast agent for MRI. MagnReson Med 1992, 24(1):75–84.

33. Tanimoto A, Oshio K, Suematsu M, Pouliquen D,Stark DD. Relaxation effects of clustered particles.J Magn Reson Imaging 2001, 14(1):72–77.

34. Virchow R: In: Hirschwald A: ed. Die krankhaftengeschwulste. Berlin: August Hirschwald; 1863.

35. Konerding M, van Ackern C, Fait E, Steinberg F,Streffer C. Morphological aspects of tumor angio-genesis and microcirculation. In: Molls M, Vaupel P,Brady LW, Heilmann HP: eds. Blood Perfusion andMicroenvironment of Human Tumors: Implicationsfor Clinical Radiooncology (Medical Radiology). NewYork: Springer Verlag; 2000: 5–17.

36. Jain RK. Determinants of tumor blood flow: a review.Cancer Res 1988, 48:2641–2658.

37. Maeda M, Itoh S, Kimura H, Iwasaki T, Hayashi N,et al. Tumor vascularity in the brain: evaluation withdynamic susceptibility-contrast MR imaging. Radiol-ogy 1993, 189:233–238.

38. Pathak AP, Schmainda KM, Ward BD, Linderman JR,Rebro KJ, et al. MR-derived cerebral blood volumemaps: Issues regarding histological validation andassessment of tumor angiogenesis. Magn Reson Med2001, 46(4):735–747.

39. Aronen HJ, Gazit IE, Louis DN, Buchbinder BR,Pardo FS, et al. Cerebral blood volume maps ofgliomas: comparison with tumor grade and histologicalfindings. Radiology 1994, 191:41–51.



40. Donahue KM, Krouwer HG, Rand SD, Pathak AP,Marszalkowski CS, et al. Utility of simultaneouslyacquired gradient-echo and spin-echo cerebral bloodvolume and morphology maps in brain tumor patients.Magn Reson Med 2000, 43(6):845–853.

41. Schmainda KM, Rand SD, Joseph AM, Lund R, WardBD, et al. Characterization of a first-pass gradient-echo spin-echo method to predict brain tumorgrade and angiogenesis. Am J Neuroradiol 2004,25(9):1524–1532.

42. Pathak AP, Rand SD, Schmainda KM. The effect ofbrain tumor angiogenesis on the in vivo relation-ship between the gradient-echo relaxation rate change(deltar2*) and contrast agent (MION) dose. J MagnReson Imaging 2003, 18(4):397–403.

43. Barbier EL, Lamalle L, Decorps M. Methodology ofbrain perfusion imaging. J Magn Reson Imaging 2001,13(4):496–520.

44. Ostergaard L. Principles of cerebral perfusion imag-ing by bolus tracking. J Magn Reson Imaging 2005,22(6):710–717.

45. Thompson HKJ, Starmer CF, Whalen RE, McIntoshHD. Indicator transit time considered as a gamma-variate. Circ Res 1964, 14:502–515.

46. Weisskoff RM. Pitfalls in MR measurement of tissueblood flow with intravascular tracers: which meantransit time? Magn Reson Med 1993, 29:553–559.

47. Quarles CC, Schmainda KM. Assessment of the mor-phological and functional effects of the anti-angiogenicagent su11657 on 9l gliosarcoma vasculature usingdynamic susceptibility contrast MRI. Magn Reson Med2007, 57(4):680–687.

48. Le Duc G, Peoc’h M, Remy C, Charpy O, Muller RN,et al. Use of T(2)-weighted susceptibility contrast MRIfor mapping the blood volume in the glioma-bearingrat brain. Magn Reson Med 1999, 42(4):754–761.

49. Dennie J, Mandeville JB, Boxerman JL, Packard SD,Rosen BR, et al. NMR imaging of changes in vascularmorphology due to tumor angiogenesis. Magn ResonMed 1998, 40:793–799.

50. Shen T, Weissleder R, Papisov M, Bogdanov A Jr,Brady TJ. Monocrystalline iron oxide nanoparticles(MION): physiochemical properties. Magn Reson Med1993, 29:599–604.

51. Wu EX, Tang H, Jensen JH. Applications of ultrasmallsuperparamagnetic iron oxide contrast agents in theMR study of animal models. NMR Biomed 2004,17(7):478–483.

52. Bremer C, Mustafa M, Bogdanov A Jr, Ntziachris-tos V, Petrovsky A, et al. Steady-state blood volumemeasurements in experimental tumors with differentangiogenic burdens a study in mice. Radiology 2003,226(1):214–220.

53. Ogawa S. Oxygenation-sensitive contrast in MR imageof rodent brain at high magnetic fields. Magn ResonMed 1990, 14:68–78.

54. Robinson SP, Rijken PF, Howe FA, McSheehy PM,van der Sanden BP, et al. Tumor vascular architectureand function evaluated by non-invasive susceptibil-ity MRI methods and immunohistochemistry. J MagnReson Imaging 2003, 17(4):445–454.

55. Persigehl T, Bieker R, Matuszewski L, Wall A,Kessler T, et al. Antiangiogenic tumor treatment: earlynoninvasive monitoring with USPIO-enhanced MRimaging in mice. Radiology 2007, 244(2):449–456.

56. Robinson SP, Howe FA, Griffiths JR, Ryan AJ, Water-ton JC. Susceptibility contrast magnetic resonanceimaging determination of fractional tumor blood vol-ume: a noninvasive imaging biomarker of responseto the vascular disrupting agent zd6126. Int J RadiatOncol Biol Phys 2007, 69(3):872–879.

57. Kostourou V, Robinson SP, Whitley GS, Griffiths JR.Effects of overexpression of dimethylarginine dimethy-laminohydrolase on tumor angiogenesis assessed bysusceptibility magnetic resonance imaging. Cancer Res2003, 63(16):4960–4966.

58. Reichardt W, Hu-Lowe D, Torres D, Weissleder R,Bogdanov A Jr. Imaging of vegf receptor kinaseinhibitor-induced antiangiogenic effects in drug-resistant human adenocarcinoma model. Neoplasia2005, 7(9):847–853.

59. Leenders WP, Kusters B, Verrijp K, Maass C,Wesseling P, et al. Antiangiogenic therapy of cere-bral melanoma metastases results in sustained tumorprogression via vessel co-option. Clin Cancer Res 2004,10(18 Pt 1):6222–6230.

60. Ferretti S, Allegrini PR, O’Reilly T, Schnell C,Stumm M, et al. Patupilone induced vascular disrup-tion in orthotopic rodent tumor models detected bymagnetic resonance imaging and interstitial fluid pres-sure. Clin Cancer Res 2005, 11(21):7773–7784.

61. Robinson SP, Ludwig C, Paulsson J, Ostman A. Theeffects of tumor-derived platelet-derived growth fac-tor on vascular morphology and function in vivorevealed by susceptibility MRI. Int J Cancer 2008,122(7):1548–1556.

62. Pathak AP, Schmainda KM, Ward BD, Rebro KJ,Rand SD: Assessing tumor angiogenesis with dynamicsusceptibility contrast fMRI: which morphologic corre-lates are relevant? Ninth Annual Meeting of the Inter-national Society for Magnetic Resonance in Medicine.Glasgow; 2001.

63. Yablonskiy DA, Haacke EM. Theory of NMR sig-nal behavior in magnetically inhomogeneous tissues:the static dephasing regime. Magn Reson Med 1994,32(6):749–763.

64. Kiselev VG, Posse S. Analytical model of susceptibility-induced MR signal dephasing: effect of diffusion in



a microvascular network. Magn Reson Med 1999,41(3):499–509.

65. Jensen JH, Chandra R. MR imaging of microvascula-ture. Magn Reson Med 2000, 44(2):224–230.

66. Wu EX, Tang H, Jensen JH. High-resolution MRimaging of mouse brain microvasculature using therelaxation rate shift index q. NMR Biomed 2004,17(7):507–512.

67. Hahn PF, Stark DD, Weissleder R, Elizondo G, Saini S,et al. Clinical application of superparamagnetic ironoxide to MR imaging of tissue perfusion in vascularliver tumors. Radiology 1990, 174(2):361–366.

68. Ichikawa T, Arbab AS, Araki T, Touyama K,Haradome H, et al. Perfusion MR imaging with asuperparamagnetic iron oxide using T2-weighted andsusceptibility-sensitive echoplanar sequences: evalua-tion of tumor vascularity in hepatocellular carcinoma.Am J Roentgenol 1999, 173(1):207–213.

69. Neuwelt EA, Varallyay CG, Manninger S, Solymosi D,Haluska M, et al. The potential of ferumoxytolnanoparticle magnetic resonance imaging, perfusion,and angiography in central nervous system malignancy:a pilot study. Neurosurgery 2007, 60(4):601–611, dis-cussion 611-2.

70. Kiselev VG. Transverse relaxation effect of mri contrastagents: a crucial issue for quantitative measurementsof cerebral perfusion. J Magn Reson Imaging 2005,22(6):693–696.

71. Marques JP, Bowtell RW: Simulations of the bold effectusing a realistic model of the vasculature. ISMRM, 12thAnnual Meeting. Kyoto; 2004.

72. Pathak AP, Ward BD, Schmainda KM. A novel tech-nique for modeling susceptibility-based contrastmechanisms for arbitrary microvascular geometries:the finite perturber method. Neuroimage 2008,40(3):1130–1143.


MR relaxation properties of superparamagnetic iron oxide particles.


Advanced Review

Optical nanoparticle sensors forquantitative intracellular imagingYong-Eun Koo Lee∗ and Raoul Kopelman∗

Real-time measurements of biological/chemical/physical processes, with nointerferences, are an ultimate goal for in vivo intracellular studies. To constructintracellular biosensors that meet such a goal, nanoparticle (NP) platforms seemto be most promising, because of their small size and excellent engineerability.This review describes the development of NP-based opical sensors and theirintracellular applications. The sensor designs are classified into two types, basedon the sensor structures regarding analyte receptor and signal transducer. Type1 sensors, with a single component for both receptor and transducer, work bymechanisms similar to those of ‘molecular probes’. Type 2 sensors, with a separatecomponent for receptor and transducer, work by different mechanisms that requirethe presence of specific NPs. A synergistic increase in optical signal or selectivityhas been reported for these second type of NP sensors. With ongoing rapidadvances in nanotechnology and instrumentation, these NP systems will soon becapable of sensing at the single-molecule level, at the point of interest within theliving cell, and capable of simultaneously detecting multiple analytes and physicalparameters. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 98–110

Intracellular imaging of the biochemistry andbiophysics of live cells has been of prime interest for

decades. Numerous approaches have been proposedto achieve real-time, noninvasive analysis of chemicaland physical properties at unperturbed cellularphysiological status. Significant progress in thevisualization of biological processes has been achievedby the dramatic advances in imaging processingtechnologies with high-performance computer systemsas well as by the continuous development of newanalyte-specific fluorescent molecular probes. Thesemolecular probes, however, have several drawbacksthat limit the indicator dyes available for reliableintracellular measurements. The indicator moleculeshave to be in a cell-permeable form, which oftenrequires properly derivatized indicator molecules.The measurement is often skewed by intracellularsequestration to specific organelles inside the cell,or by nonspecific binding to proteins and other cellcomponents. The cytotoxicity of the available dyesis sometimes a problem, as the mere presence of

∗Correspondence to: Yong-Eun Koo Lee, and Raoul Kopelman,University of Michigan, Ann Arbor, MI, USA.E-mail: [email protected]; [email protected]

DOI: 10.1002/wnan.002

these dye molecules may chemically perturb the cell.Furthermore, the dye is usually not ‘ratiometric’, i.e.,has only a single spectral peak, which then requirestechnologically more demanding techniques, suchas picosecond lifetime resolution or phase-sensitivedetection. We note that just loading into the cell aseparate reference dye, for ratiometric measurements,is not a solution, because of the aforementionedsequestration and nonspecific binding.

In an attempt to solve these problems, whilemaintaining minimal physical interference, a new typeof sensor has been developed, utilizing nanoparticles(NPs) as platforms for immobilizing the sensorchemistry. The NP sensor is physically noninvasiveowing to its small size. An NP sensor of 20–600 nmin diameter takes up only 1 ppm to 1 ppb of amammalian cell’s volume.1 There are also standardmethods for delivery of NPs into cells, such as via agene gun, pico-injection, liposome incorporation, orendocytosis. This prevents unnecessary modification(derivatization) of the indicator dyes. The inertprotective matrix of the NPs eliminates interferencessuch as protein binding and/or membrane/organellesequestration.2 The NP matrix also obviates thetoxicity problem by protecting the cellular contentsfrom the indicator dyes and vice versa. The cell


WIREs Nanomedicine and Nanobiotechnology Optical nanoparticle sensors for quantitative intracellular imaging

viability after NP sensor delivery is about 99%,relative to control cells,3 indicating negligible physicaland chemical perturbation to the cell. Moreover, aratiomentric sensor can be easily constructed by co-loading of the indicator and reference componentswithin the NP matrix. We note that the NP sensorscan also be attached with specific molecular targetingmoieties, enabling the measurements of analytes atspecific cells or organelles of the cells.4

The NP has a high surface-to-volume ratio thatallows high accessibility of analytes to the indicatordyes/receptors as well as targeting factors towardsspecific cells or components of cells. Each NP canbe loaded with a high amount of components (singleor multiple) within the NP matrix as well as onthe surface. High loaded amounts of dyes in closeproximity to each other either within the restrictedNP volume or on the NP surface allow multipleinteractions with the sensing components, resultingin signal amplification.5 It is noteworthy that similaramplification effects have been reported for targetingefficiency by NPs with multiple targeting moieties onthe surface.6

Since the first NP sensor called PEBBLE (Pho-tonic Explorer for Biomedical use with BiologicallyLocalized Embedding) was reported by Kopelman andcolleagues,7,8 a number of possibilities have been pro-posed for immobilizing the sensor chemistry withinvarious kinds of NP matrixes, so as to constructNP sensors for specific intracellular applications. Itis also noted that optical detection has remains themost widely used method for sensing and imaging ofbiological systems.

This review focuses on the uses of syntheticNPs of 1–1000 nm in diameter for the design ofoptical sensors and their applications to intracellularmeasurements. It covers only untethered, that is free,NP sensors which are suitable for in situ measurementsin three dimensions; it does not cover mechanicallyfixed sensors like fiber-tip or film on glass slide, evenwhen they utilize NPs.

OPTICAL NP SENSOR DESIGN

The basic structure of a sensor requires twocomponents: an analyte recognizer that binds thetarget analyte, and a transducer that signals binding.

Optical Transduction ModalityFluorescence is a highly sensitive, specific means formonitoring cell activity, and a number of fluorescentreporters can be analyzed simultaneously. Fluores-cence has been and will be a major transduction

modality but has limiting factors such as photobleach-ing and interference due to autofluorescence fromcellular components.

Surface-enhanced Raman scattering (SERS) isa recently evolving optical modality for intracellularNP sensors and is complementary to fluorescence.9,10

It was found that the SERS effect is provided by avery small number of molecules located at specialsites in the gap between two nearly touching goldnanocrystals.11 Because of the high specificity ofa Raman spectrum, minute amounts of chemicalsinside living cells might be identified by their uniquefingerprint spectra.12 SERS requires the so-called‘SERS-active substrates’ such as nanometer-sized silveror gold structures, which target molecules thatget attached to them. Surface plasmon resonance(SPR) is another metal NP–based optical transducerthat draws much interest in biological detection,including immunoassay.13 SERS and SPR are freefrom photobleaching and self-quenching of the markermolecule. However, their sensitivity/reproducibilitystill needs validation.

NP MatrixNP matrices should exhibit excellent chemical stabilityand biocompatibility. A variety of NP matrices havebeen utilized for the design of optical nanosensors, aslisted in Table 1.

Polymeric NPs of various matrices and sizeswith surface-located reactive functional groups canbe prepared by various synthetic methods.53,54 Thesensor components can be loaded into the NPmatrix by various methods, including encapsulation,covalent linkage, physical adsorption, etc. The matrixfor polymeric NP-based sensors is selected by theaccessibility of an analyte to a recognition elementand the loading efficiency, within NP matrix, ofindicator/receptor and signal transducer.

Liposomes or micelles present limited utility forbiological sensing within the membrane-rich cellularenvironments, as they tend to mix with the native cellmembranes, degrading the sensor structure. However,polymer-capped stabilized liposomes or micelles havebeen utilized for designing sensors for intracellularmeasurements.40,41

Semiconductor quantum dots (QDs) are brighterand more stable against photobleaching than organicfluorophores, allowing real-time and continuousmonitoring.55 A study shows that the fluorescenceemission of QDs remains bright and stable inside cellsfor at least 14 days.56 The biosensing applicationsof QDs are usually based on fluorescence resonanceenergy transfer (FRET).55,57,58



TABLE 1 Matrices for Optical NP Sensors

NP type Optical transductionmodality

Matrix References

Polymeric NP Fluorescence Poly(acrylamide) 3,8,14–20

Poly(decylmethacrylate) 21–24

Poly(ethylene glycol) 25

Poly(methacrylate) 26

Poly(n-butylacrylate) 27

Polystyrene 28–30

Dendrimer 31

Latex 32

Organically modified silica 33,34

Silica 35–39

Polymerized liposome Fluorescence 1,2-Dioleoyl-sn-glycero-3-phosphocholine(DOPC) liposome with polymethacrylate shell

40

Plymerized micelle Silane-capped (polymerized) mixed micelle 41

Quantum dot Fluorescence CdS 42,43

ZnS-coated CdSe 44–46

Metal SERS, fluorescence Gold 47–49

SERS Silver 50

Metal/polymer hybrid SERS, SPR Gold nanoshell over silica 51,52

NP, nanoparticles; SERS, surface-enhanced Raman scattering; SPR, surface plasmon resonance

In a metallic or metal-coated polymer NP,incident light can couple to the plasmon excitation ofthe metal.59 This leads to enhanced optical detectionschemes utilizing SERS and plasmon resonance.These metal (gold, silver, or gold-coated silver)NPs have been utilized to detect a wide range ofbiological molecules through binding events involvinginteractions with surface-coated specific moleculesthat offer distinct SERS48,50,51,60 and SPR.52 GoldNPs have also been utilized to construct an opticalbiosensor for DNAs.47

Sensor ClassificationThe NP optical sensors that have been developed sofar for intracellular measurements can be classifiedinto two types (see Figure 1): (1) Type 1 sensorwhere the incorporated single component, usuallyfluorescent molecular probe, serves as an analyterecognizer as well as an optical signal transducer;(2) Type 2 sensor where the analyte recognizer andoptical transducer are distinct. Type 2 sensors enablea synergistic signal and selectivity enhancement aswell as sensitivity control that cannot be achievedwith free molecular probes.

In a Type 1 sensor, fluorescent or Raman-active dyes are either encapsulated in or covalently

linked to polymeric or metallic NPs. Upon bindingwith the analyte, the spectral change (fluorescencequenching/enhancement, fluorescence lifetime changeor fluorescence peak shift, SERS) of the indicator dyesoccurs. The sensitivity and selectivity of the sensormostly depend on the incorporated indicator dye butare also affected by the NP matrix.

In a Type 2 sensor, nonfluorescent selectiveanalyte-recognition elements or receptors (enzymes,antibodies, ligands, or aptamers) are either encap-sulated in or covalently linked to the polymeric ormetallic NPs. Binding of a specific analyte to thereceptors produces an effect on the optical reportersthat consist of co-loaded fluorescent dyes or the NPthemselves (as for QDs or metallic NPs).

Both Type 1 and 2 sensors have been developedto detect a variety of intracellular analytes, asexemplified in the following sections.

NP SENSORS FOR ION SENSING

Type 1 SensorsFluorescent SensorsType 1 fluorescent sensors, also called direct ionmeasurement PEBBLEs, have been developed forsensing H+, Ca2+, Mg2+, Zn2+, Cu+/2+, and



Nanoparticlematrix

Analyte

Receptor

Transducer

Incominglight

Outgoinglight

Nanoparticlematrix

Analyte

Receptor

Transducer

Incominglight Outgoing

light

(a)

(b)

FIGURE 1 | Schematic presentation of two kinds of nanoparticles(NP) sensors: (a) Type 1 where a single component serves as receptorand transducer; (b) Type 2 where receptor and transducer are separatedbut they communicate in order to produce optical signal change uponbinding.

Fe3+.3,8,14–18,61,62 The design includes a fluorescentindicator and a reference dye entrapped in orcovalently linked to an NP. The polyacrylamideNP has been used exclusively for this type of ionsensor because of its neutral and hydrophilic nature,which allows ions to readily permeate the NP matrixand interact with the indicator dye. The indicatordyes are mostly fluorescent molecular probes, butanalyte-sensitive biological molecules, such as a redfluorescent protein, have also been used.17 Type1 ion PEBBLEs have been applied successfully forintracellular measurements of pH, Mg2+, and Ca2+.As an example, ratiometric calcium nano-PEBBLEs,containing the ‘Calcium Green- 1′ (‘Molecular

Probes’) dye as sensing component and the sulforho-damine dye as reference, have been used to measurecalcium release from mitochondria upon introductionof toxins.63 Figure 264 shows a confocal microscopeimage of C6 glioma cells containing these PEBBLEs,after their selective delivery by liposomes (to thecytosol only). The sulforhodamine fluorescence is red(reference peak) in the image, while that of CalciumGreen is yellow/green. The ratio of the CalciumGreen/sulforhodamine intensity gives a good indica-tion of cellular (cytosolar) calcium levels, regardlessof dye or PEBBLE concentration, or fluctuations inlight source intensity. The toxin, m-dinitrobenzene(DNB), was introduced on the left side of the sample(microscope slide) and allowed to diffuse to the right.The effect of DNB is a severe disruption of the mito-chondrial function, followed by uncontrolled releaseof calcium (onset of a mitochondrial permeabilitytransition). This caused calcium PEBBLEs inside thecytosol of different cells to ‘light up’ from left to rightas a function of time. As a result, high resolution inboth the spatial and temporal domains was obtained.

Another interesting intracellular applicationof Type 1 PEBBLE sensors was made with theMg2+ PEBBLEs, to study the chemical changesinduced inside human macrophage cells by invadingsalmonella bacteria.65 The Mg2+ measurements bythe PEBBLE sensors showed conclusively that Mg2+

is not an important contributor in the controlof pathogens by macrophages, in contradiction toprevious reports.66

A different kind of direct ion NP sensor wasdesigned using additional layers of polyelectrolytes onthe surface of NPs for immobilizing indicator dyes.67

In this work, the potassium ion indicator, potassium-binding benzofuran isophthalate, was immobi-lized within poly(styrene sulfonate)/poly(allylaminehydrochloride) films assembled on the surface of flu-orescent europium NPs. The fluorescence from the(commercial) core nanoparicle serves as reference for aratiometric measurement. The indicator retains its sen-sitivity to potassium ions after immobilization within

FIGURE 2 | Confocal microscope image (time snapshot) ofthree human C6 glioma cells that contain CalciumGreen/sulfarhodamine Photonic Explorer for Biomedical usewith Biologically Localized Embeddings (PEBBLEs) (with m−dinitrobenzene (DNB) toxin dffusing from left to right).(Reprinted, with permission, from Ref. 64. Copyright 2003Taylor & Francis Group).



the films and exhibits sensitivity toward increases inpotassium concentration over a broad range.

SERS SensorsAn SERS pH sensor was developed with silverNPs (50–80 nm in diameter) functionalized withpara-mercaptobenzoic acid (4-MBA).50 The SERSspectrum from the functionalized silver NPs showsa characteristic response to the pH 6–8 of thesurrounding solution. There was a large variability inthe measured pH, as the SERS spectrum was observedonly from aggregated particle clusters. These sensorswere delivered into living Chinese Hamster Ovary(CHO) cells by passive uptake. The NP sensorsretained their robust signal and sensitivity to pHwithin a cell. The spectrum indicates that the pH sur-rounding the NP is below 6, which is consistent withthe particles being located inside a lysosome (pH 5).

A similar SERS pH sensor was designed on thebasis of a gold nanoshell/silica core NP coated with alayer of para-MBA.52 The nanosensor was capable ofmeasuring pH in its local vicinity continuously overthe range of 5.80–7.60 pH units.

Type 2 SensorsType 2 NP sensors for cobalt, copper, hydrogen,nickel, potassium, silver, sodium, zinc, and chlorideions have been developed.

NP sensors called ion-correlation PEBBLEs havebeen developed for Na+, K+, and Cl− ions.21–23 Thesensor is made of poly(decylmethacrylate) (PDMA)NPs embedded with three components: a nonfluores-cent ionophore that binds selectively to the ion ofinterest, a fluorescent hydrogen ion–selective dye thatplays the role of a reporter, and a lipophilic additivethat maintains ionic strength. The operation of theentire system is based on having a thermodynamicequilibrium that controls ion exchange (for sensingcations) or ion co-extraction (for sensing anions), i.e.an equilibrium-based correlation between differention species. The degree of protonation measured fromthe fluorescence change of the hydrogen ion–selectivedye is related to the concentration of the analyte ionby the theory developed for optical absorption–basedion- correlation sensors.68,69 The hydrophobic PDMAmatrix is selected to ensure a local chemical equilib-rium among embedded components within NPs in theaqueous phase. The composition of the matrix, i.e. thecross-linker-to-monomer ratio, was found to affect thedynamic response range. Intracellular measurementsof K+ and Cl− ions were made by this type of PEBBLEsensors. The PDMA K+ PEBBLE sensors, as an exam-ple, were introduced into rat C6 glioma cells using

a BioRad (Hercules, CA) Biolistic PDS-1000/He genegun system.21 The confocal images confirmed that thesensors were localized in the cytoplasm of the cells.The response of the PEBBLE sensors inside the cellsto the addition of kainic acid, a K+-channel-openingagonist, indicated an increase in K+ concentration, theexpected trend. Another K+ NP sensor based on thesame mechanism was designed with a different matrix,i.e. a poly(n-butyl acrylate) (PnBA) nano sphere ofless than 200 nm in diameter.27 This study showsthat the composition of the three sensor components(ionophore, hydrogen-sensitive dye, and lipophilicadditive) affects the characteristics of the sensors, suchas its dynamic range, selectivity, and response time. Itshould be noted that the selectivity of these two Type2 K+ NP sensors21,27 is higher than that of Type 1 K+NP sensor67 by a factor of 1000–10,000.

Several FRET-based Type 2 ion NP sensors havebeen designed.

A Type 2 pH nanosensor was developed bycoating a pH-insensitive fluorescent polystyrene bead(200 nm in diameter) with a layer of polyaniline(PANI) of only a few nanometers thick.28 PlainPANI films display no fluorescence in the visibleand near-IR range, but they do display characteristicpH-dependent absorption spectra that are due toprotonation and deprotonation, respectively, of theemeraldine form of the PANI. Because of the fluores-cence spectra of the beads being overlapped with theabsorption spectra of PANI, the fluorescence intensitychanges in accordance with the changes in pH.

Silica NPs have been utilized for two differentdesigns of Type 2 fluorescence sensors for copper ions.In one design, the surface of silica NPs was covalentlylinked with a picolinamide subunit (selective Cu2+ligand) and fluorescent dansylamide.36–38 The graftingof the ligand and the dye subunits to the NP’s surfacenot only ensures the intercomponent communicationin the sensor but also induces cooperative processesin the binding of the substrate. The sensitivity ofthe sensors was tuned by changing the ligand-to-dyeratio. In another silica-based design, silica NPs wereprepared from a monomer containing chemosensor-like unit (similar to a molecular probe) made bycoupling polyamine chains (receptor) and dansylunits (fluorophore).5 These sensors may be classifiedas Type 1.5 sensors. The sensors were selectivefor copper, cobalt, and nickel ions and showed agreatly improved sensitivity from the occurrence ofmulticomponent cooperative photophysical processes.

Another Type 2 copper ion sensor was developedutilizing latex NPs. The hydrophobic fluorophore(BODIPY) is entrapped within the particle core, andthe copper-chelating receptor (cyclam) is covalently



linked to the polymer backbone. The fluorescence ofthe dye is quenched upon binding Cu2+ to cyclambecause of FRET between the dye and copper cyclamcomplexes. The response of the sensors is fast, with90% quenching within 1 s.32,70

QDs have also been used for designing ionnanosensors. The ligands coated on the surface ofQDs were found to have a profound effect on theluminescence response of QDs to physiologicallyimportant metal cations. l-Cysteine- and thioglycerol-capped CdS QDs were used to detect zinc and copperions in physiological buffer samples, respectively. Thedetection limits were 0.8 µM for zinc (II) and 0.1 µMfor copper (II) ions.42 Pentapeptide Gly-His-Leu-Leu-Cys-coated CdS QDs (2.4 ± 1.5 nm by transmissionelectron microscopy (TEM)) were designed to detectCu2+ and Ag+ selectively, with high sensitivity, below0.5 µ m.43

It is noted that the copper ion nanosensorhas been the most studied among Type 2 NPsensors. Table 2 compares Type 1 and 2 NP sensorsfor copper ion that have been developed so far. Itshould be noted that, none of them has been appliedfor intracellular studies because the dynamic ranges ofthe developed sensors are above the normal unboundcopper ion level, which is only femtomolar.71,72

These sensors may be applied for cells under stressedconditions that could increase the free copper ion con-centration to micromolar levels.73 In order to studythe copper ion homeostasis under normal conditions,a sensor with higher sensitivity needs to be developed.

NP SENSORS FOR SMALL MOLECULES

Dissolved Oxygen SensorAll the NP sensors that have been developed fordetecting dissolved oxygen belong to Type 1. The

first NP sensor for dissolved oxygen was developedusing hydrophilic silica NPs paired with rutheniumindicator dyes, and reference dyes.35 The sensor wasused successfully for the reliable oxygen imagingdone inside live cells. NP sensors with enhancedsensitivity and dynamic range were developed usingthe more sensitive platinum-based oxygen-sensitivedyes and reference dyes, embedded in a hydrophobicmatrix, organically modified silica (ormosil),33 orPDMA.24 The hydrophobic matrix is usually bettersuited for oxygen sensing than the hydrophilic onebecause of its higher oxygen solubility. The embeddedplatinum (II) octaethylporphine ketone, an oxygen-sensitive dye, has infrared (IR) fluorescence andmakes the sensors work in human plasma samples,24

unaffected by light scattering and autofluorescence.These PEBBLE nanosensors exhibit a perfectly linearStern–Volmer calibration curve over the entire rangeof dissolved oxygen concentration, an ideal butpreviously unachieved goal for any fluorescent oxygensensors. The sensitivity was very high with QDO of97–97.5% is the quenching response to dissolvedoxygen, defined by

QDO = (IN2 − IO2)/IN2 × 100 (1)

where IN2 is the fluorescence intensity of the indicatordye or the indicator/reference intensity ratio infully deoxygenated water, and IO2 is that in fullyoxygenated water.

These oxygen nanosensors were also successfullyapplied for real-time imaging of oxygen inside livecells, monitoring metabolic changes inside live C6Glioma cells.33

Oxygen sensors with additional layers of poly-electrolytes on the NP surface have been developed.The polyelectrolyte layers are used either to con-trol the dye loading or to systematically assemble

TABLE 2 Copper Iona NP Sensors

Sensor type Matrix NP size (nm) Recognitioncomponent

Signal pro-ducer

Optical signal Detectablerange

References

Type 1 Poly (acrylamide) 85 DsRed DsRed Fluorescence 200–5000 nM 17

Type 2 Silica 18–75 Picolinamide Dansylamide Fluorescence 4.7–200µM 36–38

Silica 30 Polyamine Dansyl unit Fluorescence 50–1000µM 5

Latex 16 Cyclam BODIPYderivative

Fluorescence 1 nM–5µM 32,70

CdS QD 3.5 Thioglycerol QD Fluorescence 0.1–1600µM 42

CdS QD 2.4 Peptide(Gly-His-Leu-Leu-Cys

QD Fluorescence 100 nM–2µM 43

aType 1 NP sensor based on DsRed is designed to sense both Cu+ and Cu2+, while all Type 2 NP sensors are made for Cu2+ only.NP, nanoparticles; QD, quantum dots



the sensors on the cell membranes. In one realiza-tion, commercial fluorospheres (100 nm) are coatedwith a multilayer of polyelectrolytes via layer-by-layer self-assembly, and then a ruthenium-basedoxygen-sensitive fluorophore, (tris(4,7-diphenyl-1,10-phenanthroline)ruthenium(II), is post-loaded withinthe deposited polyelectrolyte multilayers.74 The fluo-rescent NPs act as physical scaffolds and provide areference peak for a ratiometric measurement. Thesensitivity was medium level, with QDO of 60%.The sensors were successfully delivered to the inte-rior of human dermal fibroblasts via endocytosis withno apparent loss in cell viability. In another design,the same ruthenium-based dye is entrapped in com-mercial polystyrene beads of 100 nm in diameter,and poly(ethyleneimine) (PEI) is covalently linked tothe NP surface via glutaraldehyde chemistry.29 Thesenanosensors were assembled on individual Saccha-romyces cerevisiae cells via electrostatic interactionsbetween the positively charged PEI and negativelycharged cell surfaces. This work demonstrates a proofof concept for self-assembly of nanosensors ontoindividual cell surfaces in a controlled manner fornoninvasive examination of the oxygen concentrationin the proximity of individual yeast cells.

Oxygen nanosensors were also developedon the basis of a nanometer-sized, polymerizedphospholipid vesicle (liposome).40 The liposomesof 150 nm diameter were prepared from 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) orDOPC doped with small (< 1%) mole percentages of1,2-dioleoyl-sn-glycero-3-phosphoethanol amine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl) (NBD-PE). Thesevesicles were then stabilized via a cross-linking poly-merization of hydrophobic methacrylate monomers,partitioned into the hydrophobic interior of theDOPC bilayer. For oxygen detection, a ruthenium-based dye was encapsulated into the aqueous interiorof the polymerized liposome. NBD-PE was used asa reference dye for ratiometric measurements. TheStern–Vomer plot provides a straight line over theentire dissolved oxygen range and the QDO is 76%.

The oxygen NP sensors described abovehave all utilized fluorescence intensity for measure-ments. Lifetime measurement based oxygen sensorswere also constructed by encapsulating Pt(II)-tetra-pentafluorophenyl-porphyrin (PtPFPP) in polystyrenebeads of 0.3–1µm in diameter.30 The sensors wereinjected into plant cells using glass microcapillar-ies, and an optical multifrequency phase-modulationtechnique was used to discriminate the sensor signalfrom the strong autofluorescence of the plant tis-sue. The same sensors were injected into the salivary

glands of the blowfly to quantify the changes in oxy-gen content within individual gland tubules duringhormone-induced secretory activity.75

The NP sensors for dissolved oxygen aresummarized in Table 3.

NP Sensors for Reactive Oxygen SpeciesNP sensors have been developed for two molecularreactive oxygen species (ROS) (singlet oxygen andhydrogen peroxide) and one radical ROS (hydroxylradical). These sensors were designed to showirreversible responses towards ROS, due to highreactivities and short lifetimes of the ROS.

Singlet Oxygen SensorRatiometric NP sensors for singlet oxygen have beendeveloped using ormosil NPs.34 These sensors incor-porate the singlet oxygen–sensitive 9,10-dimethylanthracene as an indicator dye and a singlet oxy-gen–insensitive dye, octaethylporphine, as a referencedye for ratiometric fluorescence-based analysis. Theencapsulation of these dyes into the hydrophobicormosil matrix results in a higher specificity towardsinglet oxygen, as the matrix blocks the entry of short-lived polar ROS, such as OH and superoxide radicals.These nanoprobes have been used to monitor thesinglet oxygen produced by ‘dynamic nanoplatforms’that were developed for photodynamic therapy.76

OH Radical SensorsThe hydroxyl radical is the most reactive ROS,presenting two problems for the construction ofsensors: (1) inability to penetrate significantly intoany matrix without being destroyed; (2) ability tooxidize (and photobleach) most potential referencedyes. A sensor was designed to get around theseproblems by attaching the hydroxyl indicator dyecoumarin-3-carboxylic acid (CCA) onto the NP sur-face, while encapsulating the reference dye deep insideit.20 The detection of this probe was based on theirreversible hydroxylation of a nonfluorescent form ofCCA, resulting in a fluorescent product (7-hydroxy-coumarin-3-carboxylic acid). This nanoprobe demon-strates a proof of principle of a ratiometric hydroxylradical probe, with good sensitivity and reversibility.

Hydrogen Peroxide SensorA poly(ethylene glycol) (PEG) hydrogel nanosphere(250–350 nm) with the encapsulated enzymehorseradish peroxide (HRP) was prepared andutilized as a sensor for hydrogen peroxide, based onthe Amplex Red assay.25 In the presence of HRP,Amplex Red (10-acetyl-3,7-dihydroxyphenoxazine)



TABLE 3 Dissolved Oxygen NP Sensors

Sensor type Matrix NP size Oxygen indicator Optical signal QDO (%) References

Type 1 Silica 20–300 nm Ru(II)- tris(4,7-diphenyl-1,10-phenanthroline)dichloride

Fluorescenceintensity

80 35

Ormosil 120 nm Pt(II) octaethylporphineketone


97 33

Ormosil 120 nm Pt(II) octaethylporphine Fluorescenceintensity

97 33

Poly(decylmethacrylate)

150–250 nm Pt(II) octaethylporphineketone


97.5 24

Commercialfluorophorewith multilayerof polyelec-trolytes

100 nm Ru(II)- tris(4,7-diphenyl-1,10-phenanthroline)dichloride


60 74

Polystyrene withcovalentlylinkedpolyethylen-imine



Not available 29

Polymerizedliposome



76 40

Polystyrene 300 nm–1µm Pt(II)-tetra-pentafluorophenyl-porphyrin

Fluorescencelifetime

Not available 30,75

reacts with H2O2, in a 1:1 stoichiometry, to producethe red fluorescent oxidation product, resorufin. Theresponse of the HRP-loaded PEG NPs changed as afunction of H2O2 concentration in the presence ofexternally introduced Amplex Red, indicating that theenzyme activity of HRP was still maintained withinthe NPs. The HRP-loaded NPs were introduced viaphagocytosis inside macrophages and were foundto respond to exogenous H2O2 (100µm) as well asendogenous peroxide induced by lipopolysaccharide(1 µg/mL).

Glucose SensorA poly(acrylamide) NP-based fluorescent glucosesensor was developed by incorporating glucoseoxidase (GOx), an oxygen-sensitive ruthenium-baseddye, and a reference dye.19 This is a Type 2 sensor inwhich the enzymatic oxidation of glucose to gluconicacid results in the local depletion of oxygen, whichis measured by the oxygen-sensitive dye. It should benoted that the traditional ‘naked’ molecular probescannot be used to achieve this kind of synergistic task.The dynamic range was found to be ∼ 0.3–8 mM,with a linear range between 0.3 and 5 mM.

Maltose SensorThree different designs of QD-based maltose sensorshave been reported with maltose-binding proteins(MBPs) as maltose receptors. Two of them utilize theβ-cyclodextrin-acceptor dye conjugates that are capa-ble of binding within the saccharide-binding pocketof MBP and thus compete effectively with maltose,the MBP’s preferred substrate.44 In one configuration,a β-cyclodextrin–QSY9 conjugate is bound to anMBP located on the QD surface, resulting in FRETquenching of the QD photoluminescence. Added mal-tose displaces the β-cyclodextrin–QSY9, and the QDphotoluminescence increases in a systematic manner.In another configuration, QDs were coupled withCy3-labeled MBPs bound to β-cyclodextrin-Cy3.5.In this case, the QD donor drives the sensor functionthrough a two-step FRET mechanism that overcomesinherent QD donor–acceptor distance limitations. Aratiometric measurement was made on the basis ofthe emission peaks of Cy3 and Cy 3.5. In these twodesigns, the loss of displaceable quenchers may causeerrors. A QD–MBP-based maltose sensor was devel-oped without quencher molecules.45,46 In this design,



a ruthenium complex ([(tetraamine)(5-maleimido-phenanthroline)ruthenium]-[PF6]2) is covalentlylinked to MBP. The interaction (distance) betweenthe Ru complex and QD changes in accordance withthe conformational change of MBP upon bindingwith maltose, resulting in a concentration-dependentincrease in QD fluorescence.

Metronidazole SensorA Type 1 nanosensor for detecting metronidazole,a drug for the treatment of anaerobic protozoanand bacterium infections, was developed by cova-lent immobilization of indicator dye, 3-amino-9-ethylcarbazole (AEC), in poly(methacrylate) NP ofthe size less than 100 nm in diameter.26 The obtainedsensors have higher photostability and lower toxicityin comparison with free AEC. The results revealed thatthe probe showed good selectivity and had a linearresponse to the analyte in the range from 2.0 × 10−5

to 1.0 × 10−3 mol L−1 with a detection limit of9.0 × 10−6 mol L−1

NP SENSORS FOR LARGE BIOLOGICALMOLECULES

NP sensors have been used for detecting large biolog-ical molecules such as DNAs and proteins. The basicdesign is composed of NPs functionalized with recep-tors (antibodies,52 DNAs,48 or aptamers77) for target-specific detection. The analysis was done with variousapproaches including optical methods such as fluores-cence, SERS, and SPR. These sensors have been devel-oped for diagnostic assay, i.e. laboratory measure-ments of the analytes in biological samples like blood,which is not really relevant for this review articlefocused on direct intracellular measurements. Readersinterested in these sensors are referred to the literaturefor reviews on NP-based diagnostic assay.49,78–80

NP SENSORS FOR CELLULARACTIVITY

Apoptosis SensorA nanosensor for detecting apoptosis of cells wasdeveloped by conjugating a caspase-specific FRET-based apoptosis reagent (PhiPhiLux G1D2) to the G5poly(amidoamine) (PAMAM) dendrimer for apopto-sis detection and folic acid for specific targeting.31 Thenanosensors were applied for apoptosis measurementsin two different cell lines: KB cell (folate receptor pos-itive) and UMSCC-38 cell (folate receptor negative).The cells were first incubated with either 0.45µM NP

sensors or phosphate-buffered saline (PBS) (untreatedcells) for 30 min, added with either the apoptosis-inducing agent staurosporine at a concentration of0.5µM or PBS (control), and incubated again for anadditional 3 h. The apoptosis was observed on thebasis of the fluorescence of the detached cells using aflow cytometer. The cell death by apoptosis was notmonitored. The apoptotic KB cells increased the fluo-rescence intensity to a much greater degree, while theapoptotic UMSCC-38 cells did not show any increasein fluorescence intensity over the background fluores-cence of stained control cells. These results suggestthat the sensor can measure the intracellular activi-ties or analytes in the specific location selected by thetargeting moieties linked to the NP surface.

NP Sensor for Lipid PeroxidationA nanosensor for detecting lipid peroxidation bychemiluminescence was designed by conjugatingCoumarin C343 (C16H15NO4) to silica NPs (15 nm)and then entrapping these dye-linked silica in asol–gel silica NP (∼ 100 nm).39 Coumarine C343is known to enhance the weak chemiluminescenceassociated with lipid peroxidation. The producednanosensor enhanced low-level chemiluminescence byapproximately 100%.

NP SENSORS FOR INTRACELLULARPHYSICAL PROPERTIES

NP Sensors for Electric FieldIntracellular electric fields have been measured byvoltage dyes or patch/voltage clamps. These tech-niques frequently require lengthy calibration steps foreach cell or cell type measured, and the measurementsare confined to cellular membranes. A nanodevice todetermine electric field inside any live cell or cellularcompartment, called E-PEBBLE, was developed usingpolymerized micelles.41 The E-PEBBLE is preparedby encasing the fast-response, voltage-sensitive dyedi-4-ANEPPS inside the hydrophobic core of asilane-capped (polymerized) mixed micelle, whichprovides a uniform environment for the moleculesand therefore allows for universal calibration.

The E-PEBBLEs are calibrated externally andapplied for in vitro E-field determinations, withno further calibration steps. The PEBBLEs wereintroduced into immortalized rat astrocytes, DITNCcells, by endocytosis and enabled, for the first time,complete three-dimensional electric field profilingthroughout the entire volume of living cells (notjust inside membranes). This new ability is expectedto greatly enhance the understanding of the role



of cellular E-fields in influencing and/or regulating bio-logical processes, with wider implications for cellularbiology, biophysics, and biochemistry.

NP Sensor for Local Viscosity MeasurementsA new type of sensors for local viscosity measure-ments has been developed with so called ‘MOONs’(MOdulated Optical Nanoprobes). The MOONs arehalf-metal-capped fluorescent NPs whose fluorescencesignals can be modulated according to their orienta-tions, as the metal-coated side reflects the excitationlight. The Brownian rotation or the rotational behav-iors of the MOONs under an external magnetic fieldhave been utilized to measure the local viscosity, whichaffects the rotation rate of the MOONs.81,82 We notethat the same rotational behavior of MOONs alsoallows the sensor’s signal-to-noise (background) ratios(SNR) to be enhanced by up to 4000 times.83 So far,MOON-based sensors have been developed using amicron size particle owing to the size-related diffi-culties for efficient magnetization or high fluorescentintensity of individual sensor particle. With recentprogress on nanotechnology and coating technol-ogy, such as molecular beam epitaxy, nanometer-sizedMOONs are being developed. This sensor design isquite attractive, as adding a metal coating on onehemisphere of any NP sensor containing a fluorescentindicator allows the simultaneous measurement of thelocal viscosity as well as the concentration of a chemi-cal analyte. It also increases tremendously the SNR ofthe chemical sensing part.

CONCLUSIONA variety of NP-based opical sensors have been devel-oped in concurrence with advances in nanomaterials.These sensors provide minimal physical as well aschemical interferences owing to the combination oftheir small size and their protective NP matrix, orsurface coatings. Some, but not all of these NP sensorshave been successfully utilized for real-time measure-ments of important intracellular analytes. It has beenreported that single-cell analysis has the potential fordiagnosing diseases at an early stage, at which changeson a tissue level are not yet evident but chemicalchanges within cells are observable.84 Getting chem-ical or physical information from a single cell or a

specific location within a single cell would be oneof the important future applications of NP sensors.The following issues must be considered in order toimprove the performance of the NP sensors for widerand more effective future intracellular applications:

Sensitivity and Signal-to-NoiseThe goal will eventually be to enable single analytemolecule (ion) detection, in a single cell, in vitro orin vivo, despite the large background. Owing to thelimited numbers of analyte molecules (ions) within asmall volume single cell, instrumentation and sensingtechnology must meet stringent detection limits. Oneof the promising future sensor designs for enhancedsensitivity may be based on ‘MOONs’ that providea background-free detection. This technique can beuseful for samples with highly scattering and/or fluo-rescent backgrounds, or for experiments with severalfluorescent probes.

SelectivityThe selectivity of the sensors toward the analytes ofinterest is mainly determined by that of the molecu-lar probes or receptors. A higher level of selectivitycan be obtained by locating the NP sensors at a spe-cific location in a live cell, either through moleculartargeting groups conjugated to the NP surface, orthrough remote steering means such as magnetic orlaser tweezers. A recent study demonstrates the poten-tial use of magnetic tweezers for remote control of theorientation and position of the NP sensors.85

Multiplexing CapabilityDetection of multiple analytes can be made possi-ble by a properly designed single NP sensor. An NPsensor containing multiple molecular probes or recep-tors that are specific to different analytes could be,for example, one in which the various optical sig-nals are well resolved. The MOON-based fluorescentNP sensors provide another example of multitask-ing sensors that can measure the chemical propertysimultaneously with a physical property, such as localtemperature or viscosity.73,74,81,82 A third examplecould be given in which confocal microscopy resolvesthe individual signals from a number of cell-embeddednanosensors.

NOTESThe authors would like to acknowledge funding from NSF grant DMR0455330 and NCI ContractN01-CO-37123. We would also like to thank Ron Smith for the schematic diagram for Figure 1.



REFERENCES

1. Buck SM, Xu H, Brasuel M, Philbert MA, Kopel-man R. Nanoscale probes encapsulated by biologicallylocalized embedding (PEBBLEs) for ion sensing andimaging in live cells. Talanta 2004, 63(1):41–59.

2. Graber ML, Dilillo DC, Friedman BL, Pastoriza-munoz E. Characteristics of fluoroprobes for measuringintracellular pH. Anal Biochem 1986, 156(1):202–212.

3. Clark HA, Hoyer M, Parus S, Philbert MA, Kopel-man R. Optochemical nanosensors and subcellularapplications in living cells. Mikrochim Acta 1999,131(1–2):121–128.

4. Harrell JA, Kopelman R. Biocompatible probes mea-sure intracellular activity. Biophotonics Int 2000,7:22–24.

5. Montalti M, Prodi L, Zaccheroni N. Fluorescencequenching amplication in silica nanosensors for metalions. J Mater Chem 2005, 15(27–28):2810–2814.

6. Montet X, Funovics M, Montet-Abou K, Weissleder R,Josephson L. Multivalent effects of RGD peptidesobtained by nanoparticle display. J Med Chem 2006,49(20):6087–6093.

7. Sasaki K, Shi ZY, Kopelman R, Masuhara H. Three-dimensional pH microprobing with an optically-manipulated fluorescent particle. Chem Lett 1996,(2):141–142.

8. Clark HA, Barker SLR, Brasuel M, Miller MT, Mon-son E, et al. Subcellular optochemical nanobiosen-sors: probes encapsulated by biologically localisedembedding (PEBBLEs). Sens Actuators, B 1998,51(1–3):12–16.

9. Kneipp K, Kneipp H, Itzkan I, Dasari RR, Feld MS.Surface-enhanced Raman scattering and biophysics.J Phys Condens Matter 2002, 14(18):R597–R624.

10. Kneipp J, Kneipp H, McLaughlin M, Brown D,Kneipp K. In vivo molecular probing of cellular com-partments with gold nanoparticles and nanoaggregates.Nano Lett 2006, 6(10):2225–2231.

11. Bosnick KA, Jiang J, Brus LE. Fluctuations and localsymmetry in single-molecule rhodamine 6G Ramanscattering on silver nanocrystal aggregates. J Phys ChemB 2002, 106(33):8096–8099.

12. Nabiev IR, Morjani H, Manfait M. Selective analysisof antitumor drug-interaction with living cancer cellsas probed by surface-enhanced Raman-spectroscopy.Eur Biophys J 1991, 19(6):311–316.

13. Ramanavieius A, Herberg FW, Hutschenreiter S, Zim-mermann B, Lapenaite I, et al. Biomedical applicationof surface plasmon resonance biosensors (review). ActaMed Lit 2005, 12(3):1–9.

14. Clark HA, Hoyer M, Philbert MA, Kopelman R. Opti-cal nanosensors for chemical analysis inside single livingcells. 1. Fabrication, characterization, and methods forintracellular delivery of PEBBLE sensors. Anal Chem1999, 71(21):4831–4836.

15. Park EJ, Brasuel M, Behrend C, Philbert MA, Kopel-man R. Ratiometric optical PEBBLE nanosensors forreal-time magnesium ion concentrations inside viablecells. Anal Chem 2003, 75(15):3784–3791.

16. Sumner JP, Aylott JW, Monson E, Kopelman R. A flu-orescent PEBBLE nanosensor for intracellular free zinc.Analyst 2002, 127(1):11–16.

17. Sumner JP, Westerberg N, Stoddard AK, Fierke CA,Kopelman R. Cu+ and Cu2+ sensitive PEBBLE fluo-rescent nanosensors using Ds red as the recognitionelement. Sens Actuators, B 2005, 113(2):760–767.

18. Sumner JP, Kopelman R. Alexa Fluor 488 as aniron sensing molecule and its application in PEBBLEnanosensors. Analyst 2005, 130(4):528–533.

19. Xu H, Aylott JW, Kopelman R. Fluorescent nano-PEBBLE sensors designed for intracellular glucoseimaging. Analyst 2002, 127(11):1471–1477.

20. King M, Kopelman R. Development of a hydroxyl rad-ical ratiometric nanoprobe. Sens Actuators, B 2003,90(1–3):76–81.

21. Brasuel M, Kopelman R, Miller TJ, Tjalkens R,Philbert MA. Fluorescent nanosensors for intracellu-lar chemical analysis: decyl methacrylate liquid polymermatrix and ion-exchange-based potassium PEBBLE sen-sors with real-time application to viable Rat C6 gliomacells. Anal Chem 2001, 73(10):2221–2228.

22. Brasuel M, Kopelman R, Kasman I, Miller TJ,Philbert MA. Ion concentrations in live cells fromhighly selective ion correlations fluorescent nano-sensors for sodium. Proc IEEE Sens 2002, 1:288–292.

23. Brasuel MG, Miller TJ, Kopelman R, Philbert MA.Liquid polymer nano-PEBBLES for Cl− analy-sis and biological applications. Analyst 2003,128(10):1262–1267.

24. Cao YF, Koo YE, Kopelman R. Poly (decylmethacrylate)-based fluorescent PEBBLE swarmnanosensors for measuring dissolved oxygen in biosam-ples. Analyst 2004, 129(8):745–750.

25. Kim SH, Kim B, Yadavalli VK, Pishko MV. Encapsula-tion of enzyme within polymer spheres to create opticalnanosensors for oxidative stress. Anal Chem 2005,77(21):6828–6833.

26. Tan SZ, Jiang JH, Yan BN, Shen GL, Yu RQ. Prepara-tion of a novel fluorescence probe based on covalentimmobilization by emulsion polymerization and itsapplication to the determination of metronidazole. AnalChim Acta 2006, 560(1–2):191–196.

27. Ruedas-Rama MJ, Hall EAH. K+-selective nano-spheres: maximizing response range and minimizingresponse time. Analyst 2007, 131(12):1282–1291.

28. Pringsheim E, Zimin D, Wolfbeis OS. Fluorescentbeads coated with polyaniline: a novel nanomaterial for optical sensing of pH. Adv Mater 2001,13(11):819–822.



29. Kuang YN, Walt DR. Detecting oxygen consumptionin the proximity of saccharomyces cerevisiae cells usingself-assembled fluorescent nanosensors. Biotechnol Bio-eng 2007, 96(2):318–325.

30. Schmalzlin E, van Dongen JT, Klimant I, Marmodee B,Steup M, et al. An optical multifrequency phase-modulation method using microbeads for measuringintracellular oxygen concentrations in plants. BiophysJ 2005, 89(2):1339–1345.

31. Myc A, Majoros IJ, Thomas TP, Baker JR. Dendrimer-based targeted delivery of an apoptotic sensor in cancercells. Biomacromolecules 2007, 8(1):13–18.

32. Meallet-Renault R, Pansu R, Amigoni-Gerbier S, Lar-pent C. Metal-chelating nanoparticles as selective flu-orescent sensor for Cu2+. Chem Commun 2004,20:2344–2345.

33. Koo YE, Cao Y, Kopelman R, Koo SM, Brasuel M,et al. Real-time measurements of dissolved oxygeninside live cells by ormosil (organically modified sil-icate) fluorescent PEBBLE nanosensors. Anal Chem2004, 76(9):2498–2505.

34. Cao Y, Koo YEL, Koo S, Kopelman R. Ratiometricsinglet oxygen nano-optodes and their use for monitor-ing photodynamic therapy nanoplatforms. PhotochemPhotobiol 2005, 81(6):1489–1498.

35. Xu H, Aylott JW, Kopelman R, Miller T, Philbert M. AReal-time ratiometric method for the determination ofmolecular oxygen inside living cells using sol gel-basedspherical optical nanosensors with applications to ratC6 glioma. Anal Chem 2001, 73(17):4124–4133.

36. Brasola E, Mancin F, Rampazzo E, Tecilla P, Tonel-lato U. A fluorescence nanosensor for Cu2+ on silicaparticles. Chem Commun 2003, 24:3026–3027.

37. Rampazzo E, Brasola E, Marcuz S, Mancin F,Tecilla P, et al. Surface modification of silica nanopar-ticles: a new strategy for the realization of self-organized fluorescent chemosensors. J Mater Chem2005, 15(27–28):2687–2696.

38. Arduini M, Marcuz S, Montolli M, Rampazzo E,Mancin F, et al. Turning fluorescent dyes into Cu(II)nanosensors. Langmuir 2005, 21(20):9314–9321.

39. Baker N, Greenway GM, Wheatley RA, Wiles C. Achemiluminescence nanosensor to monitor lipid per-oxidation. Analyst 2007, 132(2):104–106.

40. Cheng ZL, Aspinwall CA. Nanometre-sized molecularoxygen sensors prepared from polymer stabilized phos-pholipid vesicles. Analyst 2006, 131(2):236–243.

41. Tyner KM, Kopelman R, Philbert MA. ‘‘Nano-sizedvoltmeter’’ enables cellular-wide electric field mapping.Biophys J 2007, 93:1163–1174.

42. Chen YF, Rosenzweig Z. Luminescent CdS quantumdots as selective ion probes. Anal Chem 2002,74(19):5132–5138.

43. Gattas-Asfura KM, Leblane RM. Peptide-coated CdSquantum dots for the optical detection of copper(II)and silver(I). Chem Commun 2003, 21:2684–2685.

44. Medintz IL, Clapp AR, Mattoussi H, Goldman ER,Fisher B, et al. Self-assembled nanoscale biosensorsbased on quantum dot FRET donors. Nat Mater 2003,2(9):630–638.

45. Sandros MG, Gao D, Benson DE. A modularnanoparticle-based system for reagentless smallmolecule biosensing. J Am Chem Soc 2005,127(35):12198–12199.

46. Sandros MG, Shete V, Benson DE. Selective, reversible,regentless maltose biosensing with core-shell semicon-ducting nanoparticles. Analyst 2006, 131(2):229–235.

47. Maxwell DJ, Taylor JR, Nie S. Self-assemblednanoparticle probes for recognition and detection ofbiomolecules. J Am Chem Soc 2002, 124:9606–9612.

48. Elghanian R, Storhoff JJ, Mucic RC, Letsinger RL,Mirkin CA. Selective colorimetric detection of polynu-cleotides based on the distance-dependent opti-cal properties of gold nanoparticles. Science 1997,277(5329):1078–1081.

49. Thaxton CS, Georganopoulou DG, Mirkin CA. Goldnanoparticle probes for the detection of nucleic acidtargets. Clin Chim Acta 2006, 363(1–2):120–126.

50. Talley CE, Jusinski L, Hollars CW, Lane SM, Huser T.Intracellular pH sensors based on surface-enhanced Raman scattering. Anal Chem 2004,76(23):7064–7068.

51. Bishnoi SW, Rozell CJ, Levin CS, Gheith MK, John-son BR, et al. All-optical nanoscale pH meter. NanoLett 2006, 6(8):1687–1692.

52. Hirsch LR, Jackson JB, Lee A, Halas NJ, West J. Awhole blood immunoassay using gold nanoshells. AnalChem 2003, 75(10):2377–2381.

53. Kumar MNVR. Nano and microparticles as controlleddrug delivery devices. J Pharm Pharm Sci 2000,3(2):234–258.

54. Koo YEL, Fan W, Hah HJ, Kopelman R, Xu H, et al.Photonic explorers based on multifunctional nano-platforms for biosensing and photodynamic therapy.Appl Opt 2007, 46(10):1924–1930.

55. Bailey RE, Smith AM, Nie S. Quantum dots in biologyand medicine. Physica E 2004, 25(1):1–12.

56. Sun YH, Liu YS, Vernier PT, Liang CH, Chong SY,et al. Photostability and pH sensitivity ofCdSe/ZnSe/ZnS quantum dots in living cells. Nanotech-nology 2006, 17(17):4469–4476.

57. Clapp AR, Medintz IL, Mauro JM, Fisher BR,Bawendi MG, et al. Fluorescence resonance energytransfer between quantum dot donors and dye-labeled protein acceptors. J Am Chem Soc 2004,126(1):301–310.

58. Willard DM, Van Orden A. Quantum dots: resonantenergy-transfer sensor. Nat Mater 2003, 2(9):575–576.



59. Alivisatos P. The use of nanocrystals in biological detec-tion. Nat Biotechnol 2004, 22(1):47–52.

60. Kneipp K, Haka AS, Kneipp H, Badizadegan K,Yoshizawa N, et al. Surface-enhanced Raman spec-troscopy in single living cells using gold nanoparticles.Appl Spectrosc 2002, 56(2):150–154.

61. Buck SM, Koo YEL, Park E, Xu H, Philbert MA, et al.Optochemical nanosensor PEBBLEs: photonic explor-ers for bioanalysis with biologically localized embed-ding. Curr Opin Chem Biol 2004, 8(5):540–546.

62. Sun H, Scharff-Poulsen AM, Gu H, Almdal K. Synthe-sis and characterization of ratiometric, pH sensingnanoparticles with covalently attached fluorescent dyes.Chem Mater 2006, 18(15):3381–3384.

63. Clark HA, Kopelman R, Tjalkens R, Philbert MA.Optical nanosensors for chemical analysis inside sin-gle living cells. 2. Sensors for pH and calcium andthe intracellular application of PEBBLE sensors. AnalChem 1999, 71(21):4837–4843.

64. Monson E, Brasuel M, Philbert M, Kopelman R. PEB-BLE nanosensors for in vitro bioanalysis. In: Vo-Dinh T: ed. Biomedical Photonics Handbook. BocaRaton, FL: CRC Press LLC; 2003:, 59–1–59–14.

65. Martin-Orozco N, Touret N, Zaharik ML, Park E,Kopelman R, et al. Visualization of vacuolaracidification-induced transcription of genes ofpathogens inside macrophages. Mol Biol Cell 2006,17(1):498–510.

66. Vescovi EG, Soncini FC, Groisman EA. Mg2+ asan extracellular signal: environmental regulation ofSalmonella virulence. Cell 1996, 84(1):165–174.

67. Brown JQ, McShane MJ. Core-referenced ratiometricfluorescent potassium ion sensors using self-assembledultrathin films on europium nanoparticles. IEEE Sens J2005, 5(6):1197–1205.

68. Buhlmann P, Pretsch E, Bakker E. Carrier-based ion-selective electrodes and bulk optodes, 2. Ionophoresfor potentiometric and optical sensors. Chem Rev 1998,98(4):1593–1687.

69. Bakker E, Simon W. Selectivity of Ion-sensitive bulkoptodes. Anal Chem 1992, 64(17):1805–1812.

70. Meallet-Renault R, Herault A, Vachon JJ, Pansu RB,Amigoni-Gerbier S, et al. Fluorescent nanoparticles asselective Cu(II) sensors. Photochem Photobiol Sci 2006,5(3):300–310.

71. Changela A, Chen K, Xue Y, Holschen J, Outten CE,et al. Molecular basis of metal-ion selectivity andzeptomolar sensitivity by CueR. Science 2003,301(5638):1383–1387.

72. Finney LA, O’Halloran TV. Transition metal specia-tion in the cell: insights from the chemistry of metal ionreceptors. Science 2003, 300(5621):931–936.

73. Bush AI. Metals and neuroscience. Curr Opin ChemBiol 2000, 4(2):184–191.

74. Guice KB, Caldorera ME, McShane MJ. Nanoscaleinternally referenced oxygen sensors produced fromself-assembled nanofilms on fluorescent nanoparticles.J Biomed Opt 2005, 10(6):1–10.

75. Schmalzlin E, Walz B, Klimant I, Schewe B,Lohmannsroben HG. Monitoring hormone-inducedoxygen consumption in the salivary glands of theblowfly, Calliphora vicina, by use of luminescentmicrobeads. Sens Actuators, B 2006, 119(1):251–254.

76. Reddy GR, Bhojani MS, McConville P, Moody J, Mof-fat BA, et al. Vascular targeted nanoparticles for imag-ing and treatment of brain tumors. Clin Cancer Res2006, 12(22):6677–6686.

77. Levy M, Cater SF, Ellington AD. Quantum-dotaptamer beacons for the detection of proteins. Chem-biochem 2005, 6(12):2163–2166.

78. Penn SG, Hey L, Natan MJ. Nanoparticles for bioanal-ysis. Curr Opin Chem Biol 2003, 7(5):609–615.

79. Kubik TK, Bogunia-Kubik K, Sugisaka M. Nanotech-nology on duty in medical applications. Curr PharmBiotechnol 2005, 6(1):17–33.

80. Fischer NO, Tarasow TM, Tok JBH. Heightened sensefor sensing: recent advances in pathogen immunoassaysensing platforms. Analyst 2007, 132(3):187–191.

81. Behrend CJ, Anker JN, McNaughton BH, Kopel-man R. Microrheology with modulated opticalnanoprobes (MOONs). J Magn Magn Mater 2005,293(1):663–670.

82. McNaughton BH, Agayan RR, Wang JX, Kopel-man R. Physiochemical microparticle sensors basedon nonlinear magnetic oscillations. Sens Actuators, B2007, 121(1):330–340.

83. Anker JN, Kopelman R. Magnetically modulatedoptical nanoprobes. Appl Phys Lett 2003,82(7):1102–1104.

84. Soper SA, Brown K, Ellington A, Frazier B, Garcia-Manero G, et al. Point-of-care biosensor systems forcancer diagnostics/prognostics. Biosens Bioelectron2006, 21(10):1932–1942.

85. Anker JN, Koo YE, Kopelman R. Magnetically con-trolled sensor swarms. Sens Actuators, B 2007,121(1):83–92.


Imaging nanoscale events in live cells.Magnetic nanoparticle biosensors.


Advanced Review

Synthesis of poly(alkylcyanoacrylate)-based colloidalnanomedicinesJulien Nicolas∗ and Patrick Couvreur1

Nanoparticles developed from poly(alkyl cyanoacrylate) (PACA) biodegradablepolymers have opened new and exciting perspectives in the field of drug deliverydue to their nearly ideal characteristics as drug carriers in connection withbiomedical applications. Thanks to the direct implication of organic chemistry,polymer science and physicochemistry, multiple PACA nanoparticles withdifferent features can be obtained: nanospheres and nanocapsules (either oil- orwater-containing) as well as long-circulating and ligand-decorated nanoparticles.This review aims at emphasizing the synthetic standpoint of all these nanoparticlesby describing the important aspects of alkyl cyanoacrylate chemistry as well as theexperimental procedures and the different techniques involved for the preparationof the corresponding colloidal devices. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 111–127

Nanotechnology has emerged as a promisingarea of research in which scientists from both

academia and industry put a lot of effort, hopingfor the best with regard to life in future. It isa highly multidisciplinary field which consists ofengineering functional systems at the molecular scaleand covers applied physics, materials science, interfaceand colloid science, supramolecular chemistry as wellas chemical, mechanical, and electrical engineering.One of the direct applications of nanotechnology isdevoted to the medical and pharmacology areas, alsocalled nanomedicine, the most famous example beingnanoparticle drug delivery.

Indeed, a crucial impulse was given tonanomedicine with the development of varioustypes of drug-carrier nanodevices, made possible bymeans of multidisciplinary approaches–organic andpolymer chemistry, physicochemistry, pharmacology,etc. Among suitable nanodevices for drug delivery,nanoparticles on the basis of biodegradable poly(alkylcyanoacrylate) (PACA) polymers have appeared asan established technology for colloidal nanomedicine.

∗Correspondence to: Julien Nicolas, Laboratoire de Physico-Chimie, Pharmacotechnie et Biopharmacie, UMR CNRS8612, Univ Paris-Sud, 92296 Chatenay Malabry, France.E-mail: [email protected] de Physico-Chimie, Pharmacotechnie et Biopharmacie,UMR CNRS 8612, Univ Paris-Sud, 92296 Chatenay Malabry,France.

DOI: 10.1002/wnan.015

Introduced more than 25 years ago in the field ofpharmacology, PACA drug carriers have demon-strated significant results in numerous pathologiessuch as cancer, severe infections (viral, bacteriologic,parasite) as well as several metabolic and autoimmunediseases, well-reviewed in the recent literature.1–6 Asa complementary work, the objective of the presentreview is to emphasize the synthetic aspect of thesecolloidal carriers by describing, as precisely as pos-sible, the chemistry of the cyanoacrylate monomers,their polymerization as well as the different structuresand morphologies of the corresponding nanoparticles.In particular, description of this PACA-based nan-otechnology will start from the simplest nanocarriersto more sophisticated and ‘smart’ drug deliverydevices. The reader who would like a more exhaustivepoint of view about the biologic and pharmaceuticalaspects of PACA nanoparticles as well as the drugssuccessfully incorporated in such colloidal devices isreferred to the above-mentioned references.

ALKYL CYANOACRYLATEMONOMERS AND RELATEDPOLYMERS

General Features of Alkyl CyanoacrylateMonomersAlkyl cyanoacrylates are widely known monomers,extremely appreciated for their very high reactivity



and the excellent adhesive properties of the resultingpolymers. On one hand, the famous Superglue(manufactured by Henkel), which contains shortalkyl chain cyanoacrylates, is commonly employedby the general public for repairing and do-it-yourselfactivities, whereas longer alkyl chain cyanoacrylateshave been developed for biomedical purposes such assurgical glue for the closure of skin wounds7–13 andembolitic material for endovascular surgery.10,11,14

Indeed, several commercial products have emergedfrom the use of cyanoacrylates in the biomedicalarea, mainly devoted to tissue adhesion. For instance,methyl cyanoacrylate (MCA, Figure 1) is the maincomponent of the Biobond tissue adhesive and longeralkyl ester chain cyanoacrylates, such as n-butylcyanoacrylate (nBCA, Figure 1) or octyl cyanoacrylate(OCA, Figure 1), were commercialized under thetrademarks of Indermil, Liquidband, and Dermabond,respectively.

The synthesis of alkyl cyanoacrylate monomershas been described in the patent literature since1949.15–18 Basically, the main strategy to achieveα-cyanoacrylates comprises two steps. First, thecorresponding alkyl cyanoacetate is reacted withformaldehyde in the presence of a basic catalyst, toform PACA oligomers (by the so-called Knoevenagelcondensation reaction). The catalyst is a base, eitherinorganic (e.g., sodium or potassium hydroxide,ammonia) or organic (e.g., quilonine, piperidine,dialkyl amines). Then, pure alkyl cyanoacrylatemonomer is recovered by a thermal depolymerizationreaction of the previously obtained oligomers, usingsuitable stabilizers such as protonic or Lewis acids

CN

OO

CN

OO

CN

OO

14

CN

OO

CN

OO

CN

OO

13

CN

OO

MCA ECA nBCA IBCA

IHCA ISCA HDCA

CN

OO

OCA

4

CN

OO

On

MePEGCA

FIGURE 1 | Structure of alkyl cyanoacrylates described in theliterature: methyl cyanoacrylate (MCA), ethyl cyanoacrylate (ECA),n-butyl cyanoacrylate (nBCA), isobutyl cyanoacrylate (IBCA), isohexylcyanoacrylate (IHCA), octyl cyanoacrylate (OCA), isostearylcyanoacrylate (ISCA), hexadecyl cyanoacrylate (HDCA), andmethoxypoly(ethylene glycol) cyanoacrylate (MePEGCA).

(a)

(b)

OOR

CN

n

− H2O+ n CH2O

CN

COORn

OOR

CN

n

CN

OO

R

n∆

FIGURE 2 | Synthesis of alkyl cyanoacrylate monomer viaKnoevenagel condensation reaction (a) and subsequent thermaldepolymerization (b).

with small amounts of a free-radical inhibitors toprevent repolymerization (Figure 2).

From that moment on, the synthetic protocolremained almost unchanged. It was only slightly mod-ified and improved essentially by playing with thenature of the solvent mixture,19,20 by applying atransesterification approach for making cyanoacry-lates bearing longer alkyl ester chains,21 or by usinga more efficient catalyst (namely pyrrolidine) for thecondensation step.22

Polymerization of Alkyl Cyanoacrylatesin Homogeneous MediaOn the fringe of typical vinyl monomers [styrenics,(meth)acrylates, etc.] is the alkyl cyanoacrylate family,which seems to be an exotic class of polymerizablecompounds. Indeed, due to the presence of twopowerful electro-withdrawing groups in the α-carbonof the double bond, namely ester (COOR) andcyano (CN), alkyl cyanoacrylate monomers exhibita remarkable reactivity toward nucleophiles suchas anions (hydroxide, iodide, alcoholate, etc.) orweak bases (alcohol, amine, etc.), resulting in a veryhigh polymerization rate. Even traces of one of theabove-mentioned compounds in the reaction mediumare sufficient to initiate such a fast polymerization.This explains why alkyl cyanoacrylates are extremelydifficult to handle under their pure form and thatbatches of these monomers are usually maintain stablewith a small amount of acidic stabilizers (e.g., SO2,sulfonic acid, etc.).

PACA can be synthesized according tothree distinct types of polymerization: (1) anionic,(2) zwitterionic, and (3) radical (Figure 3). In prac-tice, because of the exceptional reactivity of alkylcyanoacrylate derivatives, anionic and zwitterionicpolymerization mechanisms are by far predominantunder conventional experimental conditions with


WIREs Nanomedicine and Nanobiotechnology Synthesis of PACA-based colloidal nanomedicines

CN

OO

R

B–

CN

OO

R

B

CN

OO

R

OOR

CNB

n

CN

OO

R

NuCN

OO

R

Nu

CN

OO

R

OOR

CNNu

n

CN

OO

R

P

CN

OO

R

P

CN

OO

R

OOR

CNP

n

(a)

(b)

(c)

–

–

FIGURE 3 | Initiation and propagation steps involved during anionic (a), zwitterionic (b), and radical (c) polymerizations of alkyl cyanoacrylatemonomer initiated by a base (B−), a nucleophile (Nu), and a radical (P•), respectively.

respect to a pure radical process. This explains whystudies on alkyl cyanoacrylates polymerization in bothhomogeneous (i.e., bulk or solution) and hetero-geneous (i.e., emulsion, microemulsion) media weremainly devoted to anionic and zwitterionic processes.

Synthesis of HomopolymersIn this field, an extensive work has been accom-plished by Pepper and coworkers to get a betterunderstanding of the involved polymerization mecha-nisms depending on the experimental conditions.23–26

Indeed, the homopolymerization in solution of ethylcyanoacrylate (ECA, Figure 1) and nBCA were initi-ated either by simple anions (CH3COO−, CN−, I−,etc.) or by covalent organic bases (Et3N, pyridine,etc.), leading to anionic or zwitterionic polymer-izations, respectively.23 For zwitterionic polymeriza-tion of nBCA, the influence of the nature of theinitiator as well as other experimental conditions(inhibiting species, presence of water, etc.) on boththe main characteristics of the obtained polymer(number-average molecular weight, molecular weightdistribution) and polymerization kinetics (monomerconversion, polymerization rate, etc.) were investi-gated through a small library of covalent organic basessuch as phosphine,26–29 pyridine24,27 and amine25,27,30

derivatives. Considering anionic polymerization, thesame research group used tetrabutyl ammoniumsalts (hydroxide, bromide, acetate, and substituted

acetates) as the initiating species for the polymeriza-tion of nBCA at 20–40◦C in tetrahydrofuran (THF)and reported a nearly ideal living polymerization inthe case of the hydroxide-based initiator.31–33

Even though anionic and zwitterionic mecha-nisms are more likely to occur for the polymerizationof alkyl cyanoacrylates, free-radical polymerizationwas believed to be the main chain-extensionprocess during homopolymerization34–37 andcopolymerization37,38 carried out in bulk when a suit-able inhibitor is introduced in the reaction medium.However, even under these specific inhibition condi-tions, anionic polymerization is not totally suppressedbut is made negligible regarding the timescale of thepolymerization reaction. In particular, Canale et al.34

used in 1960, boron trifluoride–acetic acid complexwhile conducting free-radical bulk polymerizationof MCA at 60◦C initiated by azobisisobutyronitrile(AIBN), whereas Bevington et al.36 used propane-1,3-sultone as an efficient inhibitor against anionicpolymerization for the free-radical polymerization ofMCA in bulk or in 1,4-dioxane at 60◦C, initiated byAIBN or benzoylperoxide (BPO). In 1983, Yamadaet al.37 polymerized ECA in bulk at 30◦C with asmall amount of acetic acid or propane-1,3-sultoneand from their results, they extracted very highpropagation rate constants: kp = 1622 l mol−1 s−1

in the presence of acetic acid and kp = 1610 l mol−1

s−1 in the presence of propane-1,3-sultone. As a



comparison, methyl methacrylate (MMA) whichis considered as a highly reactive monomer gavekp = 450 l mol−1 s−1 at 30◦C.39

Synthesis of CopolymersAlkyl cyanoacrylates were also copolymerized withmore ‘common’ vinyl monomers through a free-radical process (using trifluoride–acetic acid complexas an efficient inhibitor against anionic polymeriza-tion) to give different kinds of copolymers, dependingon the nature of the comonomer.38 Random copoly-mers with MMA were achieved in bulk, whereasalternating copolymers with styrene were reported inbenzene solution at 60◦C under AIBN initiation. Con-sidering bulk properties, random copolymers with10% MMA exhibit physical properties similar tothe PMCA homopolymer, whereas alternating copoly-mers with styrene had an enhanced thermal stabilitycompared with random copolymers. Hall et al., whopreviously investigated the reactions of electron-richolefins with electron-poor olefins,40–42 confirmed thealternating copolymer starting from a 1:1 styrene :MCA mixture, either initiated by AIBN under UVlight at 40◦C in benzene solution or produced spon-taneously at room temperature.43 However, whenusing other comonomers such as isobutyl vinyl ether,p-methoxystyrene or β-bromostyrene, copolymeriza-tions with MCA led to mixtures of (co)polymersand/or small adducts.43

In 1978, a comprehensive synthetic approach ofbis(alkyl cyanoacrylate)s was proposed by Buck start-ing from anthracene adducts.44 These difunctionalalkyl monomers derived from cyanoacrylates werecopolymerized with monofunctional alkyl cyanoacry-lates such as MCA and isobutyl cyanoacrylate(IBCA, Figure 1), resulting in crosslinked macro-molecular adhesive compositions exhibiting supe-rior mechanical properties under both dry and wetenvironments than the noncrosslinked counterparts,which could be advantageously employed as pit andfissure sealant in dentistry.

More sophisticated macromolecular architec-tures such as diblock and triblock copolymers com-prising poly(ethylene glycol) (PEG) and PACA blockswere also synthesized in homogeneous media via zwit-terionic polymerization.45 The synthesis involved thepreparation of triphenylphosphine end-capped mono-hydroxyl and dihydroxyl PEGs, giving the correspond-ing monofunctional and difunctional macrozwitteri-onic initiator. The polymerization of IBCA was theninitiated with each one of the macroinitiators inTHF at ambient temperature to afford PIBCA-b-PEG

CN

OO

14

+Me2NH

CN

OO

On

EtOH, 20°C

CN CN

OO OO

O

x y

n14

+ CH2O

FIGURE 4 | Synthesis of random, comb-like poly[(hexadecylcyanoacrylate)-co-methoxypoly(ethylene glycol) cyanoacrylate][P(HDCA-co-MePEGCA)] copolymer via Knoevenagel condensationreaction.

diblock and PIBCA-b-PEG-b-PIBCA triblock copoly-mers with tuneable compositions in good match withthe initial stoichiometry.

Synthesis of poly[(hexadecyl cyanoacrylate)-co-methoxypoly(ethylene glycol) cyanoacrylate][P(HDCA-co-MePEGCA)] comb-like copolymersexhibiting amphiphilic properties was reported byPeracchia et al.46 This original approach derived fromKnoevenagel condensation reaction where corre-sponding cyanoacetates, namely hexadecyl cyanoac-etate and PEG monomethyl ether cyanoacetate, werereacted with formaldehyde in the presence of dimethy-lamine as the catalyst (Figure 4). Thanks to the slow,in situ formation of the cyanoacrylate monomers, itallowed the polymerization process to be better con-trolled compared with a direct anionic polymerization.Besides, the composition of the copolymer (and thusits hydrophilicity/hydrophobicity) can be adjustedsimply by varying the initial cyanoacetates feed ratio.

POLY(ALKYLCYANOACRYLATE)-BASEDNANOPARTICLES

General Consideration on the Synthesisof Poly(alkyl cyanoacrylate) NanoparticlesNanoparticle is a collective name for two differenttypes of colloidal objects, namely nanospheres (NS)and nanocapsules (NC), which can be separatelyobtained depending on the preparation process.Basically, nanospheres are matrix systems constitutedby the polymer in which the drug is physicallyand uniformly dispersed, whereas nanocapsules arevesicular systems in which the drug is solubilized ina liquid core, either water (w-NC) or oil (o-NC),surrounded by a thin polymer layer (Figure 5).

During the last 25 years, an important break-through in this field has been witnessed with thedevelopment of PACA nanoparticles as colloidaldrug carriers. Polymerizations in heterogeneousmedia (i.e., emulsion, dispersion, miniemulsion,



NS w-NC o-NC

FIGURE 5 | Schematic representation of nanospheres (NS),water-containing nanocapsules (w-NC), and oil-containingnanocapsules (o-NC).

microemulsion)47,48 and spontaneous emulsificationtechniques49–51 are two well-known approaches forthe preparation of polymeric particles, which havealso been intensively used for the confection of PACAnanoparticles as colloidal drug carriers for in vivoadministration.

Synthesis of NanospheresIn 1979, Couvreur et al. first developed a simpleprocess to directly generate stable MCA or ECAnanospheres, consisting of a dropwise additionof the monomer into a vortexed HCl solution(2 < pH < 3) containing a nonionic or a macro-molecular surfactant.52 Since then, numerous studiesaiming at establishing relevant parameters governingthe polymerization kinetics as well as the characteris-tics of the macromolecules and the nanospheres havebeen reported. It has been shown that the nature andthe concentration of the surfactant played a significantrole on the particle size,53–61 whereas the type of boththe monomer and the surfactant strongly influencedthe molar mass of the obtained polymer.55–58 Besidesthe monomer concentration,53,59,60,62 the pH of thereaction medium53,55–58,60–64 and the concentration ofsulfur dioxide (acting as a polymerization inhibitor)57

were also crucial parameters which strongly affectedthe macromolecular and/or colloidal properties ofthe nanospheres. The size of the colloidal objectswhich can be obtained usually ranged from 50 to300 nm,54,59,60 which is a well-adapted windowfor colloidal drug delivery devices, especially byintravenous administration.

For a more fundamental standpoint, severaltentative mechanisms have been postulated.65,66

It has been reported that the emulsion/dispersionpolymerization in acidic medium is not that trivial andproceeds via a stepwise, anionic mechanism compris-ing reversible propagation and reversible terminationsteps63,64 (Figure 6). Basically, PACA oligomers areformed in the monomer droplets and are reversibly

CN

OO

R

CN

OO

R

HO(a)

(b)

(c)

OOR

CNHO

n

CN

OO

CN

OO

R

HO CN

OO

R

+ n

OOR

CNHO

n

CN

OO

H

OOR

CNHO

n

CN

OOH

HO

R

RR

HO–

–

– –

–

–

FIGURE 6 | Schematic representation of poly(alkyl cyanoacrylate)formation via the stepwise anionic polymerization mechanism inemulsion/dispersion. Initiation step (a), reversible propagation step (b),and reversible termination step (c).

terminated by the acid-inhibiting agents present inthe monomer. This step is followed by a re-initiationreaction of terminated species by still living chains,leading to further polymerization until a molecularweight balancing is reached, similar to depolymeriza-tion/repolymerization events.66 One should be awarethat in all these mechanisms, the polymerization ispostulated to be initiated by the hydroxyl ions fromthe aqueous phase independently of other reactantsexisting in the polymerization medium.

On the basis on an interfacial polymerizationmechanism,67,68 Limouzin et al. polymerized nBCAin emulsion and miniemulsion in the presenceof dodecylbenzenesulfonic acid (DBSA) acting asboth surfactant and terminating agent (also termedtersurf).69 By releasing protons at the water/oilinterface, DBSA allowed the interfacial, anionicpolymerization to be drastically slowed down througha (reversible) termination reaction and to proceedunder a fairly controlled fashion leading to stablehigh solids content (∼20%) PnBCA nanospheres.The miniemulsion technique was also used by Weisset al. for the preparation of PnBCA nanospheres. Byvarying the concentration of the surfactant (SDS),and by adding sodium hydroxide as the initiatingspecies, high solids content dispersions up to 10%with average diameters ranging from 110 to 360 nmwere obtained.70

Synthesis of NanocapsulesNanocapsules are reservoir-type nanoparticles inwhich drugs can be encapsulated according to their



intrinsic solubility. In other words, oil-containingnanocapsules will be able to encapsulate hydropho-bic drugs, whereas hydrophilic ones will be efficientlyencapsulated into water-containing nanocapsules.71

The nature of the nanocapsules (i.e., water-containingor oil-containing) is determined by the nature of thedispersed phase involved in a heterogeneous poly-merization process, usually emulsion or microemul-sion. Basically, the macromolecular shell is formedby the spontaneous anionic polymerization of alkylcyanoacrylate occurring at the interface between thedispersed and the continuous phase. Historically,oil-containing nanocapsules were first developed byFallouh et al.49 through a simple protocol: a solu-tion of monomer and oil in a water-miscible solvent(usually ethanol) is poured into an aqueous solutionof surfactant (usually Poloxamer 188) under vigor-ous stirring, leading to small oil/monomer dropletsat the interface of which the polymerization is ini-tiated by hydroxide ions present in water. Gallardoet al.65 reported that the crucial parameters for achiev-ing nanocapsules lies: (1) in the diffusion behavior ofthe organic solvent (acting as a monomer support)within the aqueous phase, which ultimately governsthe reservoir nature of the nanoparticles, and (2) inthe simultaneous precipitation of the polymer at thewater/oil interface (i.e., the polymer should be insol-uble in both the aqueous and the organic phase).Usually, nanocapsules exhibit average diameter rang-ing from 200 to 350 nm, the latter being governed byseveral physicochemical parameters such as the natureand the concentration of the monomer and encapsu-lated drug, the amount of surfactant and oil as well asthe speed of diffusion of the organic phase within theaqueous phase. However, Altinbas et al. have demon-strated that when a miniemulsion is applied insteadof an emulsion, nanocapsules of an average diameterbelow 100 nm can be obtained.72

The main drawback often encountered in thisapproach is the contamination of the nanocapsulepopulation by a substantial amount of nanospheres,resulting from a partial polymerization in the organicphase.65 However, it has been shown that anoptimized ethanol/oil ratio,65,73 the acidification ofthe organic phase,74 and the inhibition of thepolymerization in the organic phase by aproticsolvents75 (acetonitrile, acetone) each avoided theformation of matrix-type nanoparticles.

Water-containing nanocapsules have been devel-oped more recently than were the oil-containingcounterparts. They are usually prepared by waterin oil (w/o) (micro)emulsion, also called an inverse(micro)emulsion, using polysorbate, sorbit monoleateor poly(ethylene oxide) lauryl ester (Brij 35)

as surfactants. Basically, the alkyl cyanoacrylatemonomer is added to the preformed (micro)emulsionand, in a similar way to that of oil-containing nanocap-sules, spontaneous anionic polymerization occurredat the water/oil interface to form a thin PACAlayer surrounding an aqueous core. Depending onthe nature of the surfactant and the starting system(emulsion or microemulsion), which are parametersgoverning the surface properties of these colloidalobjects, this process led to 50–350 nm diameter, stablenanocapsules.76–79

However, because the inverse (micro)emulsionprocesses conduct to water-containing nanocapsulesdispersed in oil (which are suitable for oral routeadministration), intravenous injection cannot bedirectly performed with a nonaqueous dispersingmedium. To circumvent this limitation, a recentmethod aiming at transferring the nanocapsulesfrom an oil-dispersing medium to a water-dispersingmedium was recently suggested by Couvreur andcoworkers and consisted in a centrifugation step ofthe nanocapsules onto an aqueous layer.77,79

To synthesize nanocapsules with preformedpolymers, homopolymer of alkyl cyanoacrylate arerequired and synthesized separately, for instance bydripping the monomer in pure water, the polymerbeing subsequently recovered by lyophilization. Thenanocapsules preparation method, also called interfa-cial deposition, consists of the addition of a solutionof the homopolymer and a small amount of oil,for instance Miglyol (which will constitute the oilycore of the nanocapsules), into an aqueous phase.The oil-containing nanocapsules form instantaneouslyby deposition of the homopolymer at the oil/waterinterface, which precipitate as a macromolecularshell.71,80,81 In general, a surfactant is added in theaqueous phase to ensure colloidal stability of thesenanocapsules.

Synthesis of Poly(alkyl cyanoacrylate)Nanoparticles with Controlled SurfacePropertiesIn this topic, the major breakthrough is undoubt-edly the grafting of PEG, a nonionic, flexible, andhydrophilic polymer, onto nanoparticles (which alsoapplies for other colloidal drug carriers such asliposomes). This approach, termed ‘PEGylation’, rep-resented a milestone in the drug delivery area.82,83

Indeed, non-‘PEGylated’ nanoparticles are quicklyeliminated from the bloodstream because of theadsorption of blood proteins (opsonins) onto theirsurface, which triggers the recognition of the mononu-clear phagocyte system (MPS) by the macrophages.



As a consequence, these nanoparticles are ineluctablyaccumulated in MPS organs such as the liver andthe spleen, restricting the therapeutic activity of theentrapped compounds to liver diseases (i.e., hepaticprimary hepatocarcinoma or metastasis as well as liverintracellular infections). In contrast, when covered byPEG chains, the obtained nanoparticles are able toefficiently escape this recognition system, resulting inlong-circulating, colloidal devices, also called ‘stealth’nanoparticles.82,83

After it has been demonstrated that PACAnanoparticles can be seen as very promisingbiodegradable drug carriers (the BioAlliance Pharmaspin-off company is now producing doxorubicin-loaded PACA nanoparticles for clinical uses inphase II/III trials with resistant liver hepatocarcinomaas main indication), their complexity was furtherincreased by performing appropriate tuning of theirsurface properties in order to control their in vivo fate.

Surface Modification of NanospheresFirst attempts concerning surface modification ofPACA nanospheres logically concerned the ‘PEGy-lation’ concept, either via a simple adsorption of PEGchains onto the nanoparticles or by a covalent linkageof PEG chains with PACA polymers. However, theadsorption approach does not fit the covalent linkagecriteria and is not really suitable as long as it has beendemonstrated that these kinds of assemblies (PACAnanoparticles on which poloxamer 388 or poloxam-ine 908 was adsorbed) are not stable during in vivoadministration, resulting in a loss of coating and nosignificant influence on the biodistribution pattern.84

Thus the covalent bond of the PEG chains at the sur-face of the nanoparticles is a prerequisite for this kindof application.

Basically, different types of hydrophilicmolecules have been anchored, on purpose, to thesurface of PACA nanoparticles (Figure 7). Efficientsurface modification of nanospheres can be achievedeither in situ during the polymerization in aqueousdispersed media or from preformed amphiphiliccopolymers during emulsification processes.

Concerning previous studies about anionic/zwitterionic emulsion polymerization of alkyl cyanoacry-late, the hydrophilic molecules introduced in therecipes (SDS, dextran, poloxamer, Tweens, cyclodex-trins, etc.) were solely used as stabilizing agents forinvestigating their effect on the stability, the averagediameter, and the particle size distribution. However,it was not fully understood at this time that some ofthem, especially those containing nucleophilic func-tional groups, might take part in the initiation ofthe polymerization, leading to a partial formation

of surface-active macromolecules. This point is ofgreat importance since nanoparticles with covalentlyanchored stabilizing moieties at their surface wouldbehave differently in a biologic medium than thosewith adsorbed surfactants. As a consequence, thisis only later on that researches have been strictlydevoted to surface engineering of PACA nanoparticlesin order to investigate any subtle change of the surfaceproperties of the nanoparticles on their in vivo fate.

However, almost unmarked, early worksby Douglas et al. postulated that dextran orβ-cyclodextrin may also initiate the polymerization ofbutyl cyanoacrylate (BCA) resulting in the formationof amphiphilic copolymers, helping to stabilize thenascent nanoparticles.54 This approach was revisitedby Peracchia et al. using different linear PEGs actingas stabilizers and initiators for the emulsion poly-merization of IBCA85,86 (Figure 7(a) and (b)). It wasdemonstrated that PEG chains exhibited different con-formations at the surface of the nanospheres: (1) hairynanospheres with PEG monomethyl ether due to asingle initiation site (Figure 7(a)) or (2) long loopsusing PEG due to the divergent chain growth (twoinitiating sites) during the polymerization of IBCA87

(Figure 7(b)). In the same spirit, the use of polysac-charides, such as dextran, dextran sulfate, chitosan,and thiolated chitosan, as stabilizing/initiating agentsunder similar experimental conditions also led to sta-ble nanospheres in the 100–500 nm range, exhibitingdifferent surface properties; for instance, positivelycharged with chitosan61,88,89 and from rather neutralto negatively charged with dextran derivatives.88,90,91

So far, anionic (mini)emulsion polymerizationwas the most widespread and straightforward tech-nique to synthesize PACA nanospheres. Even though,in that case, the mechanism is on the basis ofanionic propagating species,63,64 Chauvierre et al.recently adapted Couvreur’s original protocol to afree-radical emulsion polymerization process, thanksto the polysaccharide/cerium IV (Ce4+) ions redoxcouple as the initiator92 (Figure 8). Because of the fastradical initiation rate, anionic polymerization is negli-gible regarding the timescale of the experiment whichmakes way for a free-radical chain growth process.This technique was also employed for the emulsionpolymerization of alkyl cyanoacrylate using differentkinds of polysaccharides,89,90,93,94 allowing a directcomparison with nanospheres obtained from anionicemulsion polymerization. The first difference is theconformation of polysaccharide chains at the surfaceof the nanospheres in direct relation with the structureof the copolymer. Indeed, anionic emulsion polymer-ization led to grafted copolymers, whereas linear block



HO

OH

∗

nHO

OCH3

∗

n

NCO

∗

OCH3

O

n

NH2

OH

O

∗

COOH

COOH

COOH

COOH

COOH

O

O

HO∗

O

OHO∗

O

O

HO

O

OHO

∗

(a) (b) (c)

(d) (e) (f)

FIGURE 7 | Schematic representation of poly(alkyl cyanoacrylate)-based nanospheres with controlled surface properties using poly(ethyleneglycol) monomethyl ether (a), poly(ethylene glycol) (b), poly[(hexadecyl cyanoacrylate)-co-methoxypoly(ethylene glycol) cyanoacrylate] copolymer (c),polysaccharide chains under anionic initiation (d), polysaccharide chains under redox initiation (e), and amino acids (f). The moiety anchored at thesurface of the nanoparticles is marked by single asterisk.

copolymers were achieved under redox radical ini-tiation (Figure 8), leading respectively to compactloops (Figure 7(d)) and hairy polysaccharide chains(Figure 7(e)) at the surface of the nanospheres.88,93

The size of the polysaccharide-decoratednanospheres was in the 80–800 nm range anddepended on: (1) the molecular weight of the polysac-charide, where a minimum value of about 6000 gmol−1 was required for ensuring an efficient colloidalstability88,89 and (2) on the nature of the polysac-charide: dextran-decorated nanospheres exhibited anaverage diameter below 300 nm, dextran sulfate andchitosan led to a larger average diameter of about350–600 nm,88 whereas the use of heparin conductedto 90-nm nanospheres.95,96

Another crucial difference resulting from thesurface conformation of the hydrophilic chains, foreither PEG derivatives or polysaccharides, concernsthe measure of the complement activation,87,88 which

is known to play a significant role in the non-specific recognition events of the immune system.Indeed, according to Peracchia et al., nanospheresbearing big loops because of α,ω-dihydroxyl PEG(Figure 7(b)) were shown to better prevent comple-ment consumption than do the hairy nanoparticlesobtained from PEG monomethyl ether87 (Figure 7(a)).Besides, Bertholon et al. demonstrated that, for bothdextran and chitosan, an increase of the length of thecompact loops (Figure 7(d)) resulted in an increaseof complement activation, whereas the opposite effectwas obtained by increasing the length of the hairypolysaccharide chains88 (Figure 7(e)), which clearlydemonstrated that complement activation is highlysensitive to any change of the surface chain conforma-tion. In a recent work, it was also suggested that theconformation of the coating material also affects thecytotoxicity profile of PACA nanoparticles.97



FIGURE 8 | Anionic emulsion polymerization of alkyl cyanoacrylates initiated by hydroxyl groups of dextran (a) and redox radical emulsionpolymerization (RREP) of alkyl cyanoacrylates initiated by dextran/cerium IV (Ce4+) ions redox couple (b).

Recently, an interesting synthetic pathwayto functionalize PACA nanospheres using aminoacids was proposed by Weiss et al.70 The authorsused a miniemulsion process to prepare a stablepH 1 dispersion of nBCA nanoparticles stabilizedby SDS as the surfactant. Polymerization was thentriggered by the addition of nucleophilic compoundssuch as amino acids (for instance, glycine), leadingto functionalized, stable nanospheres (as alreadydiscussed earlier, the similar miniemulsion process hasbeen applied to nonfunctionalized nanospheres whensodium hydroxide was added as the initiator). Thismethod allowed: (1) the solids content to be increasedup to 10 wt% with average diameter ranging from 80to 350 nm, depending on the amount of surfactantas well as the nature of the amino acid and (2) aconvenient surface functionalization by amino acidmoieties (Figure 7(f)).

The preparation of ‘PEGylated’ nanoparti-cles from preformed polymers is a well-established

technique which first requires the synthesis ofamphiphilic copolymers with PEG segments. PIBCA-b-PEG diblock and PIBCA-b-PEG-b-PIBCA tri-block copolymers were synthesized from phosphineend-capped PEG macroinitiators.45 With diblockcopolymers, unimodal size distribution and sta-ble nanoparticles in the range of 100–700 nmwere obtained by nanoprecipitation or emulsifica-tion/solvent evaporation, the average diameter beingcontrolled mainly by the amount of organic solventand by the composition of the polymers. However, thepresence of phosphine groups within the synthesizedpolymers may be a toxicological issue.

The amphiphilic, biodegradable copolymerscomprising poly(hexadecyl cyanoacrylate) hydropho-bic units and methoxypoly(ethylene glycol) cyanoacrylate hydrophilic units (Figure 4) were used toprepare the corresponding P(HDCA-co-MePEGCA)nanospheres exhibiting a biodegradable PACAcore and a shell of excretable PEG chains46,98,99



PHDCA-co-PEGCAPHDCAPHDCA-P80PHDCA-Polox 908

ab

c

0.25

0.20

0.15

0.10

0.05

0.00

% d

ose

/ g ti

ssue

FIGURE 9 | Concentration of radioactivity in right hemisphere (a),left hemisphere (b), and cerebellum (c), after intravenous administrationof 60 mg kg−1 of [14C]-P(HDCA-co-MePEGCA) nanoparticles,poloxamine 908-coated [14C]-PHDCA nanoparticles, polysorbate80-coated [14C]-PHDCA nanoparticles, and uncoated [14C]-PHDCAnanoparticles (mice at 1 h postinjection).

(Figure 7(c)). Nanoprecipitation or emulsifica-tion/solvent evaporation techniques employingP(HDCA-co-MePEGCA) polymers led to very stable‘PEGylated’ nanospheres with average diametersin the 100–200 nm range and monomodal sizedistributions.98 These materials showed a reducedcytotoxicity toward mouse peritoneal macrophages,and the presence of the PEG segments was foundto increase the degradability of the polymer in thepresence of calf serum.98 Besides, as a result of thePEG coating, an extended circulation time in thebloodstream was demonstrated.100

The impressive result deriving from the useof these stealth nanoparticles is their ability tosignificantly cross the blood–brain barrier (BBB)compared with non-PEGylated counterparts and thosewith preadsorbed surfactants such as polysorbate 80or poloxamine 9082,101–104 (Figure 9).

This unique feature suggested that P(HDCA-co-MePEGCA) nanospheres exhibited appropriate prop-erties for entering the central nervous system (CNS)via the BBB. Even though a passive diffusion becauseof an increased permeability of the BBB (when locallydisrupted at the tumor site) may not be ruled out,the mechanism by which those nanoparticles prefer-entially crossed the healthy BBB was assigned to aspecific adsorption of apolipoprotein E and B-100(Apo E and B-100) onto P(HDCA-co-MePEGCA)nanospheres leading to their translocation mediatedby low-density lipoprotein receptors (LDLR).105–107

The involvement of Apo E on the translocationthrough the BBB of polysorbate 80-covered PACAnanoparticles was also reported by Kreuter’s group

who hypothesized the formation of lipoprotein parti-cle mimics recognized by the LDLR gene family in thebrain endothelial cells of the BBB.108

Synthesis of ‘PEGylated’ NanocapsulesTo the best of our knowledge, the only exam-ples of ‘PEGylated’ PACA nanocapsules werereported by Brigger et al.81 and Li et al.,109,110

both using P(HDCA-co-MePEGCA) copolymers.46

Although Brigger et al.81 prepared the correspondingstealth, oil-containing nanocapsules by the interfacialdeposition technique, Li et al. used a water-in-oil-in-water (w/o/w) double emulsion process to achieve’PEGylated’, water-containing nanocapsules as tumornecrosis factor-α carriers.109,110 This two-step emul-sification protocol started by the emulsification of theaqueous phase containing the drug into the organicphase in which the P(HDCA-co-MePEGCA) copoly-mer was dissolved (w/o), followed by its addition intoan aqueous PVA solution (w/o/w). Stable nanocap-sules of about 140–150 nm in diameter were thencollected by centrifugation.

Addressed Poly(alkyl cyanoacrylate)Biodegradable NanoparticlesFor the forthcoming years, the most exciting chal-lenge in drug delivery, irrespective of the nature of thedrug carriers (i.e., liposome, nanoparticles), willbe undoubtedly the synthesis of efficient ligands-decorated colloidal devices for achieving specificcells targeting, on the basis of molecular recog-nition processes. Indeed, the main drawback ofprevious generation of drug carriers is their non-specific drug release behavior. Nanoparticles areindeed unable to be efficiently addressed to thedesired cells and the therapeutic activity of theencapsulated drug may be partly hampered. Evenfor the remarkable case of brain-targeted P(HDCA-co-MePEGCA) nanospheres,2,101–104 the linkage of ajudicious ligand at their surface would certainly resultin a strongly higher extravasation yield across theBBB.

Thus, if a great deal of effort has been alreadydevoted to this area, a lot of works remaindue to be done. The only example of the so-calledthird-generation PACA nanoparticles involves folate-decorated P(HDCA-co-MePEGCA) nanospheresto target the folate receptor, which is overex-pressed at the surface of many tumor cells. Forthis purpose, the synthetic route for P(HDCA-co-MePEGCA) copolymers46 was adapted to thesynthesis of a poly[(hexadecyl cyanoacrylate)-co-aminopoly(ethylene glycol) cyanoacrylate]



[P(HDCA-co-H2NPEGCA)] copolymer, startingfrom a protected aminopoly(ethylene glycol)cyanoacetate.111

Then, the corresponding nanospheres wereobtained by nanoprecipitation showing a narrowsize distribution for an average diameter of 80 nm.The conjugation with N-hydroxysuccinimide–folate(NHS–folate) occurred via an amidation pathwaydirectly at the surface of the nanospheres bearing avail-able amino groups (Figure 10). The specific interactionoccurring between the folate-conjugated nanospheresand the folate-binding protein was demonstrated bysurface plasmon resonance. The apparent affinity ofthe folate bound to the nanospheres appeared 10-foldhigher than the free folate in solution, because of themultivalency of the folate-decorated nanoparticles.

Biocompatibility and Biodegradationof Poly(alkyl cyanoacrylate) PolymersThe degradation and toxicity of PACA nanoparti-cles are a crucial point, especially for biomedicalapplications. Indeed, a drug carrier device is suitablefor in vivo applications only if it is made of biocom-patible, possibly biodegradable, or at least excretable(e.g., by the kidneys) materials. In fact, PACAs arebioerodible polymers for which different degradationpathways have been reported so far (Figure 11).

The predominant mechanism occurs via thehydrolysis of their side chain ester functions,55,112,113

producing the corresponding alkyl alcohol andpoly(cyanoacrylic acid) as the degradation products,the latter being fully water-soluble and readilyeliminated by kidney filtration (Figure 11(a)). Thishydrolysis, which is believed to be the main degrada-tion mechanism in vivo, proceeds typically in a coupleof hours for PACA nanoparticles and is strongly

(a)

(b)

(c)

(d)

OOR

CN

n

OHO

CN

n+ n ROH

OOR

CN

n + n CH2OCN

COORn

OOR

CN

nOO

R

CN

m +CN

OO

Rm << n

(n – m)

OOR

CNHO

n + CH2O

OOR

CN

n –1H

CN

OR

O

FIGURE 11 | Possible degradation pathways for poly(alkylcyanoacrylate) (PACA) polymers: hydrolysis of ester functions (a),’unzipping’ depolymerization reaction (b), the inverse Knoevenagelcondensation reaction (c), and release of formaldehyde from hydrolysisof the α-hydroxyl functions (d).

affected by: (1) the length of the alkyl side chains;the longer the alkyl side chains, the lower the toxicitybut the slower the hydrolysis55,114,115 and (2) the sur-rounding environment as it can be strongly catalyzedby esterases from serum, lysosomes, and pancreaticjuice.116,117 However, a complete excretion of these

NCO

∗NH2

O

n

N O

OO

O

NCO

∗NH

On

O

Folate

Folate

Folate

Folate

Folate

Folate

FIGURE 10 | Synthesis of folate-decorated poly[(hexadecyl cyanoacrylate)-co-aminopoly(ethylene glycol) cyanoacrylate] [P(HDCA-co-H2

NPEGCA)] nanospheres. The moiety anchored at the surface of the nanoparticles is marked by single asterisk.



materials would occur only for low-molecular-weightPACA polymers, typically below 10,000 g mol−1.

It has been postulated that the ‘unzipping’depolymerization reaction, initiated by a base, couldalso take part in the biodegradation pathway ofPACA,66 especially in biologic media where it can betheoretically induced by amino acids of proteins (Fig-ure 11(b)). Following the depolymerization of parentpolymers, instant repolymerization to form lower-molecular-weight polymers would occur, even if noclear description of this mechanism has been shownyet, possibly because of its too fast occurrence to beunambiguously observed.

Finally, another suggested mechanism for thedegradation of PACA polymers is on the basis ofthe well-known inverse Knoevenagel condensationreaction, which produces the corresponding alkylcyanoacetate and formaldehyde (Figure 11(c)), eventhough the release of formaldehyde might also resultfrom hydrolysis of the α-hydroxyl functions of thepolymer chains, provided the hydroxyl ions have beeninitially used as an initiator69 (Figure 11(d)). How-ever, the inverse Knoevenagel condensation reactionhas been reported to a lesser extent in aqueous solu-tion at physiological pH and too slow to competewith the above-mentioned enzyme-catalyzed hydroly-sis mechanism.118–120

CONCLUSION

Even though the chemistry of alkyl cyanoacrylates isnot as straightforward as for other ‘common’ vinylmonomers, the convergent involvements of organicchemistry, polymer science, and physicochemistrymade possible the development of more and moresophisticated, biodegradable PACA-based nanopar-ticles, successfully used as drug delivery devices.Indeed, from the pharmacology standpoint, PACAnanoparticles fulfill important requirements of idealdrug delivery systems: ease and reproducibility ofpreparation, ease of storage and administration ina sterile form, satisfying drug-loading capacity, lowtoxicity, excellent biodegradability, and feasibility forscale-up production. By playing with experimentalconditions (nature and amount of reactants, pro-cess of preparation, etc.), various types of PACAnanoparticles can be obtained, each of them exhibit-ing specific features regarding the nature of thedrug and/or the way the drug is encapsulated:nanospheres (matrix-type nanoparticles; oil-solubledrug) or nanocapsules (reservoir-type nanoparticles;either oil-soluble or water-soluble drug), nonsurface-modified (mainly devoted to MPS organs) or ‘PEGy-lated’ nanoparticles (long-circulating drug carriers) aswell as ligand-decorated nanospheres (addressed drugdelivery devices). As a result of constant efforts inthis field, PACA nanotechnologies have thus openedexciting perspectives for the discovery of novel andmore efficient nanomedicines.

NOTES

The authors thank Christine Vauthier, Francois Ganachaud, and Didier Desmaele for their fruitful discussions.The CNRS and the University Paris-Sud are also acknowledged for financial support.

REFERENCES

1. Vauthier C, Dubernet C, Fattal E, Pinto-AlphandaryH, Couvreur P. Poly(alkylcyanoacrylates) as biode-gradable materials for biomedical applications. AdvDrug Deliv Rev 2003, 55:519–548.

2. Garcia-Garcia E, Andrieux K, Gil S, Couvreur P. Col-loidal carriers and blood-brain barrier (BBB) translo-cation: a way to deliver drugs to the brain? Int JPharm 2005, 298:274–292.

3. Couvreur P, Vauthier C. Nanotechnology: intelligentdesign to treat complex disease. Pharm Res 2006,23:1417–1450.

4. Couvreur P, Gref R, Andrieux K, Malvy C. Nan-otechnologies for drug delivery: application to cancerand autoimmune diseases. Prog Solid State Chem2006, 34:231–235.

5. Vauthier C, Labarre D, Ponchel G. Design aspects ofpoly(alkylcyanoacrylate) nanoparticles for targeteddrug delivery. J Drug Target 2007, 15:641–663.

6. Fattal E, Andrieux K, Barratt G, Couvreur P,Labarre D, et al. Recent advances inpoly(alkylcyanoacrylate) nanoparticles for drugdelivery. Nanoparticles for Pharmaceutical



Applications. Stevenson Ranch, CA: American Scien-tific Publishers; 2007, 213–232.

7. Skeist I, Miron J. In: Bikales N. ed. Encyclopedia ofPolymer Science and Engineering, Vol. 2. New York:Wiley-Interscience; 1977, 1–19.

8. Coover HW, Dreifus DW, O’Connor JT. In: Skeist I.ed. Handbook of Adhesives. New York: Van Nos-trand Reinhold; 1990, 463–477.

9. King ME, Kinney AY. Tissue adhesives: a new methodof wound repair. Nurse Pract 1999, 24:66–73.

10. Oowaki H, Matsuda S, Sakai N, Ohta T, Iwata H,et al. Non-adhesive cyanoacrylate as an embolicmaterial for endovascular neurosurgery. Biomaterials2000, 21:1039–1046.

11. Reece TB, Maxey TS, Kron IL. A prospectus on tissueadhesives. Am J Surg 2001, 182:40S–44S.

12. Hollock GG. Expanded applications for octyl-2-cyanoacrylate as a tissue adhesive. Ann Plast Surg2001, 46:185–189.

13. Marcovich R, Williams AL, Rubin MA, Wolf JS.Comparison of 2-octyl cyanoacrylate adhesive, fibringlue, and suturing for wound closure in the porcineurinary tract. Urology 2001, 57:806–810.

14. Pollak JS, White RI Jr: The use of cyanoacrylate adhe-sives in peripheral embolization. J Vasc Interv Radiol2001, 12:907–913.

15. Joyner FB, Hawkins GF. α-cyanoacrylates. US PatentNo. 2721858. 1955.

16. Joyner FB, Shearer NH. Monomeric α-cyanoacrylates. US Patent No. 2756251. 1956.

17. Jeremias CG. Monomeric α-cyanoacrylates. US PatentNo. 2763677. 1956.

18. Ardis AE. Monomeric alkyl α-cyanoacrylates. USPatent No. 2467926. 1949.

19. Tseng YC, Hyon SH, Ikada Y. Modification ofsynthesis and investigation of properties for2-cyanoacrylates. Biomaterials 1990, 11:73–79.

20. Sahadev CC, Krishna CCG, Ravindranath K,Reddy NJ, Sahadev N. A process for preparationof high purity octyl and isoamylcyanoacrylates. USPatent No. 2467926. 2003.

21. Malofsky B, Badejo IT. Transesterification methodand catalysts for making cyanoacrylates fromcyanoacetates. US Patent No. 2001036380. 2001.

22. Yadav JS, Reddy BVS, Basak AK, Visali B, Narsa-iah AV, et al. Phosphane-catalyzed knoevenagel con-densation: a facile synthesis of alpha-cyanoacrylatesand alpha-cyanoacrylonitriles. Eur J Org Chem 2004,3:546–551.

23. Donnelly EF, Johnston DS, Pepper DC, Dunn DJ.Ionic and zwitterionic polymerization of n-alkyl 2-cyanoacrylates. J Polym Sci Polym Lett Ed 1977,15:399–405.

24. Johnston DS, Pepper DC. Polymerization viamacrozwitterions. 2. Ethyl and butyl cyanoacrylatespolymerized by pyridine and poly(vinylpyridine).Macromol Chem Phys 1981, 182:407–420.

25. Johnston DS, Pepper DC. Polymerization viamacrozwitterions. 3. Ethyl and butyl cyanoacrylatespolymerized by benzyldimethyl, triethyl, and tribenzy-lamines. Macromol Chem Phys 1981, 182:421–435.

26. Johnston DS, Pepper DC. Polymerization viamacrozwitterions. 1. Ethyl and butyl cyanoacrylatespolymerized by triethyl and triphenylphosphines.Macromol Chem Phys 1981, 182:393–406.

27. Pepper DC. Kinetics and mechanism of zwitterionicpolymerizations of alkyl cyanoacrylates. Polym J1980, 12:629–637.

28. Pepper DC, Ryan B. Initiation processes in polymer-izations of alkyl cyanoacrylates by tertiary amines:inhibition by strong acids. Makromol Chem 1983,184:383–394.

29. Cronin JP, Pepper DC. Zwitterionic polymerizationof butyl cyanoacrylate by triphenylphosphine andpyridine. Makromol Chem 1988, 189:85–102.

30. Eromosele IC, Pepper DC, Ryan B. Water effects onthe zwitterionic polymerization of cyanoacrylates.Makromol Chem 1989, 190:1613–1622.

31. Eromosele IC, Pepper DC. Free- and paired-ion prop-agation in the polymerization of butyl cyanoacrylateby tetrabutylammonium salts. Makromol Chem,Rapid Commun 1986, 7:639–643.

32. Eromosele IC, Pepper DC. Anionic polymerization ofbutyl cyanoacrylate by tetrabutylammonium salts. 2.Propagation rate constants. Makromol Chem 1989,190:3095–3103.

33. Eromosele IC, Pepper DC. Anionic polymerization ofbutyl cyanoacrylate by tetrabutylammonium salts.1. Initiation processes. Makromol Chem 1989,190:3085–3094.

34. Canale AJ, Goode WE, Kinsinger JB, Panchak JR,Kelso RL, et al. Methyl a-cyanoacrylate. I. Free-radical homopolymerization. J Appl Polym Sci 1960,4:231–236.

35. Otsu T, Yamada B. Determination of Q,e parametersof methyl α-cyanoacrylate. Makromol Chem 1967,110:297–299.

36. Bevington JC, Jemmett JAL, Onyon PF. Polymeriza-tion of methyl α-cyanoacrylate. II. Conditions for rad-ical polymerization. Eur Polym J 1976, 12:255–257.

37. Yamada B, Yoshioka M, Otsu T. Determination ofabsolute rate constants for radical polymerizationand copolymerization of ethyl α-cyanoacrylate in thepresence of effective inhibitors against anionic poly-merization. Makromol Chem 1983, 184:1025–1033.

38. Kinsinger JB, Panchak JR, Kelso RL, Bartlett JS,Graham RK. Methyl α-cyanoacrylate. II. Copolymer-ization studies. J Appl Polym Sci 1965, 9:429–437.



39. Kamachi M, Liaw DJ, Nozakura S. Solvent effect onradical polymerization of methyl methacrylate. PolymJ 1981, 13:41–50.

40. Hall HK, Ykman P. Reactions of zwitterionsfrom trisubstituted electron-deficient ethylenes andelectron-rich olefins. J Am Chem Soc 1975,97:800–807.

41. Hall HK, Dunn LC, Padias AB. Zwitterion cyclo-additions and free-radical polymerizations of para-(dimethylamino)styrene with tris-(carbomethoxy)ethylenes and tetrakis(carbomethoxy)ethylenes. J OrgChem 1980, 45:835–838.

42. Hall HK, Daly RC. Electron-deficient trisubstitutedolefins - new class of reactive comonomers. Macro-molecules 1975, 8:22–31.

43. Rasoul HAA, Hall HK Jr: Cycloaddition and poly-merization reactions of methyl α-cyanoacrylatewith electron-rich olefins. J Org Chem 1982,47:2080–2083.

44. Buck CJ. Unequivocal synthesis of bis(2-cyanoacrylate) monomers. I. Via anthracene adducts.J Polym Sci, Polym Chem Ed 1978, 16:2475–2507.

45. Choi YK, Bae YH, Kim SW. Block copoly-mer nanoparticles of ethylene oxide andisobutyl cyanoacrylate. Macromolecules 1995,28:8419–8421.

46. Peracchia MT, Desmaele D, Couvreur P, d’Angelo J.Synthesis of a novel poly(MePEG cyanoacrylate-co-alkyl cyanoacrylate) amphiphilic copolymer fornanoparticle technology. Macromolecules 1997,30:846–851.

47. Lovell P, El-Aasser MS. eds. Emulsion Polymeriza-tion and Emulsion Polymers. Chichester: John Wiley& Sons; 1997.

48. Gilbert RG. Emulsion Polymerization: A MechanisticApproach. London: Academic Press; 1995.

49. Fallouh NA, Roblottreupel L, Fessi H, Devis-saguet JP, Puisieux F. Development of a new processfor the manufacture of polyisobutylcyanoacrylatenanocapsules. Int J Pharm 1986, 28:125–132.

50. Thioune O, Fessi H, Devissaguet JP, Puisieux F.Preparation of pseudolatex by nanoprecipitation:Influence of the solvent nature on intrinsic viscos-ity and interaction constant. Int J Pharm 1997,146:233–238.

51. Ganachaud F, Katz JL. Nanoparticles and nanocap-sules created using the Ouzo effect: spontaneousemulsification as an alternative to ultrasonic and high-shear devices. ChemPhysChem 2005, 6:209–216.

52. Couvreur P, Kante B, Roland M, Guiot P, Bauduin P,et al. Polycyanoacrylate nanocapsules as potentiallysosomotropic carriers - preparation, morphologicaland sorptive properties. J Pharm Pharmacol 1979,31:331–332.

53. Douglas SJ, Illum L, Davis SS, Kreuter J. Particle-sizeand size distribution of poly(butyl-2-cyanoacrylate)nanoparticles. 1. Influence of physicochemical factors.J Colloid Interface Sci 1984, 101:149–158.

54. Douglas SJ, Illum L, Davis SS. Particle-size and sizedistribution of poly(butyl 2-cyanoacrylate) nanopar-ticles. 2. Influence of stabilizers. J Colloid InterfaceSci 1985, 103:154–163.

55. Vansnick L, Couvreur P, Christiaensleyh D,Roland M. Molecular-weights of free and drug-loaded nanoparticles. Pharm Res 1985, 1:36–41.

56. El-Egakey MA, Bentele V, Kreuter J. Molecular-weights of polycyanoacrylate nanoparticles. Int JPharm 1983, 13:349–352.

57. Lescure F, Zimmer C, Roy D, Couvreur P. Opti-mization of polyalkylcyanoacrylate nanoparticlepreparation - influence of sulfur-dioxide and Phon nanoparticle characteristics. J Colloid Interface Sci1992, 154:77–86.

58. Douglas SJ, Davis SS, Holding SR. Molecular-weights of poly(butyl 2-cyanoacrylate) producedduring nanoparticle formation. Br Polym J 1985,17:339–342.

59. Alonso MJ, Sanchez A, Torres D, Seijo B, Vila-Jato JL. Joint effects of monomer and stabilizerconcentrations on physicochemical characteristics ofpoly(butyl 2-cyanoacrylate) nanoparticles. J Microen-capsul 1990, 7:517–526.

60. Seijo B, Fattal E, Roblottreupel L, Couvreur P. Designof nanoparticles of less than 50 Nm diameter - prepa-ration, characterization and drug loading. Int J Pharm1990, 62:1–7.

61. Yang SC, Ge HX, Hu Y, Jiang XQ, Yang CZ. Forma-tion of positively charged poly(butyl cyanoacrylate)nanoparticles stabilized with chitosan. Colloid PolymSci 2000, 278:285–292.

62. Tuncel A, Cicek H, Hayran M, Piskin E. Monosizepoly(ethylcyanoacrylate) microspheres - preparationand degradation properties. J Biomed Mater Res1995, 29:721–728.

63. Behan N, Birkinshaw C. The mechanism of polymeri-sation of butyl cyanoacrylate in aqueous dispersions.Macromol Rapid Commun 2000, 21:884–886.

64. Behan N, Birkinshaw C, Clarke N. Poly n-butylcyanoacrylate nanoparticles: a mechanistic study ofpolymerisation and particle formation. Biomaterials2001, 22:1335–1344.

65. Gallardo M, Couarraze G, Denizot B, Treupel L,Couvreur P, et al. Study of the mechanisms offormation of nanoparticles and nanocapsules ofpolyisobutyl-2-cyanoacrylate. Int J Pharm 1993,100:55–64.

66. Ryan B, McCann G. Novel sub-ceiling temperaturerapid depolymerization-repolymerization reactions of



cyanoacrylate polymers. Macromol Rapid Commun1996, 17:217–227.

67. Maitre C, Ganachaud F, Ferreira O, Lutz JF,Paintoux Y, et al. Anionic polymerization of phenylglycidyl ether in miniemulsion. Macromolecules 2000,33:7730–7736.

68. Cauvin S, Sadoun A, Dos Santos R, Belleney J,Ganachaud F, et al. Cationic polymerization of p-methoxystyrene in miniemulsion. Macromolecules2002, 35:7919–7927.

69. Limouzin C, Caviggia A, Ganachaud F, Hemery P.Anionic polymerization of n-butyl cyanoacrylate inemulsion and miniemulsion. Macromolecules 2003,36:667–674.

70. Weiss CK, Ziener U, Landfester K. A route tononfunctionalized and functionalized poly(n-butylcyanoacrylate) nanoparticles: preparation inminiemulsion. Macromolecules 2007, 40:928–938.

71. Mayer C. Nanocapsules as drug delivery systems. IntJ Artif Organs 2005, 28:1163–1171.

72. Altinbas N, Fehmer C, Terheiden A, Shukla A,Rehage H, et al. Alkyl cyanoacrylate nanocapsulesprepared from mini-emulsions: a comparison withthe conventional approach. J Microencapsul 2006,23:567–581.

73. Aboubakar M, Puisieux F, Couvreur P, Deyme M,Vauthier C. Study of the mechanism of insulin encap-sulation in poly(isobutylcyanoacrylate) nanocapsulesobtained by interfacial polymerization. J BiomedMater Res 1999, 47:568–576.

74. Wohlgemuth M, Mayer C. Pulsed field gradient NMRon polybutylcyanoacrylate nanocapsules. J ColloidInterface Sci 2003, 260:324–331.

75. Puglisi G, Fresta M, Giammona G, Ventura CA.Influence of the preparation conditions onpoly(ethylcyanoacrylate) nanocapsule formation. IntJ Pharm 1995, 125:283–287.

76. Vranckx H, Demoustier M, Deleers M. A newnanocapsule formulation with hydrophilic core: appli-cation to the oral administration of salmon calcitoninin rats. Eur J Pharm Biopharm 1996, 42:345–347.

77. Lambert G, Fattal E, Pinto-Alphandary H, Gulik A,Couvreur P. Polyisobutylcyanoacrylate nanocapsulescontaining an aqueous core as a novel colloidal carrierfor the delivery of oligonucleotides. Pharm Res 2000,17:707–714.

78. Krauel K, Davies NM, Hook S, Rades T. Using dif-ferent structure types of microemulsions for thepreparation of poly(alkylcyanoacrylate) nanoparticlesby interfacial polymerization. J Controlled Release2005, 106:76–87.

79. Hillaireau H, Le Doan T, Chacun H, Janin J,Couvreur P. Encapsulation of mono- and oligo-nucleotides into aqueous-core nanocapsules in

presence of various water-soluble polymers. Int JPharm 2007, 331:148–152.

80. Barratt G, Puisieux F, Yu WP, Foucher C, Fessi H,et al. Anti-metastatic activity of MDP-L-alanyl-cholesterol incorporated into various types ofnanocapsules. Int J Immunopharmacol 1994,16:457–461.

81. Brigger I, Armand-Lefevre L, Chaminade P,Besnard M, Rigaldie Y, et al. The stenlying effect ofhigh hydrostatic pressure on thermally and hydrolyt-ically labile nanosized carriers. Pharm Res 2003,20:674–683.

82. Stolnik S, Illum L, Davis SS. Long circulatingmicroparticulate drug carriers. Adv Drug Deliv Rev1995, 16:195–214.

83. Storm G, Belliot SO, Daemen T, Lasic DD. Surfacemodification of nanoparticles to oppose uptake bythe mononuclear phagocyte system. Adv Drug DelivRev 1995, 17:31–48.

84. Douglas SJ, Davis SS, Illum L. Biodistribution ofpoly(butyl 2-cyanoacrylate) nanoparticles in rabbits.Int J Pharm 1986, 34:145–152.

85. Peracchia MT, Vauthier C, Puisieux F, Couvreur P.Development of sterically stabilized poly(isobutyl2-cyanoacrylate) nanoparticles by chemical couplingof polyethylene glycol. J Biomed Mater Res 1997,34:317–326.

86. Peracchia MT, Vauthier C, Popa M, Puisieux F, Cou-vreur P. Investigation of the formation of stericallystabilized poly(ethylene glycol/isobutyl cyanoacrylate)nanoparticles by chemical grafting of polyethy-lene glycol during the polymerization of isobutylcyanoacrylate. STP Pharma Sci 1997, 7:513–520.

87. Peracchia MT, Vauthier C, Passirani C, Couvreur P,Labarre D. Complement consumption by polyethy-lene glycol in different conformations chemicallycoupled to poly(isobutyl 2-cyanoacrylate) nanoparti-cles. Life Sci 1997, 61:749–761.

88. Bertholon I, Vauthier C, Labarre D. Complementactivation by core-shell poly(isobutylcyanoacrylate)-polysaccharide nanoparticles: influences of surfacemorphology, length, and type of polysaccharide.Pharm Res 2006, 23:1313–1323.

89. Bravo-Osuna I, Ponchel G, Vauthier C. Tuning ofshell and core characteristics of chitosan-decoratedacrylic nanoparticles. Eur J Pharm Sci 2007,30:143–154.

90. Labarre D, Vauthier C, Chauvierre U, Petri B,Muller R, et al. Interactions of blood proteins withpoly(isobutylcyanoacrylate) nanoparticles decoratedwith a polysaccharidic brush. Biomaterials 2005,26:5075–5084.

91. Bertholon-Rajot I, Labarre D, Vauthier C. Influenceof the initiator system, cerium-polysaccharide, onthe surface properties of poly(isobutylcyanoacrylate)nanoparticles. Polymer 2005, 46:1407–1415.



92. Chauvierre C, Labarre D, Couvreur P, Vauthier C.Radical emulsion polymerization of alkylcyanoacry-lates initiated by the redox system dextran-cerium(IV)under acidic aqueous conditions. Macromolecules2003, 36:6018–6027.

93. Bertholon I, Lesieur S, Labarre D, Besnard M,Vauthier C. Characterization of dextran-poly(isobutylcyanoacrylate) copolymers obtained byredox radical and anionic emulsion polymerization.Macromolecules 2006, 39:3559–3567.

94. Bertholon I, Hommel H, Labarre D, Vauthier C.Properties of polysaccharides grafted on nanoparticlesinvestigated by EPR. Langmuir 2006, 22:5485–5490.

95. Chauvierre C, Labarre D, Couvreur P, Vauthier C.Novel polysaccharide-decorated poly(isobutylcyanoacrylate) nanoparticles. Pharm Res 2003,20:1786–1793.

96. Chauvierre C, Marden MC, Vauthier C,Labarre D, Couvreur P, et al. Heparin coatedpoly(alkylcyanoacrylate) nanoparticles coupled tohemoglobin: a new oxygen carrier. Biomaterials2004, 25:3081–3086.

97. Chauvierre C, Leclerc L, Labarre D, Appel M,Marden MC, et al. Enhancing the tolerance ofpoly(isobutylcyanoacrylate) nanoparticles with amodular surface design. Int J Pharm 2007,338:327–332.

98. Peracchia MT, Vauthier C, Desmaele D, Gulik A,Dedieu J-C, et al. PEGylated nanoparticles from anovel methoxypolyethylene glycol cyanoacrylate-hexadecyl cyanoacrylate amphiphilic copolymer.Pharm Res 1998, 15:550–556.

99. Peracchia MT, Fattal E, Desmaele D, Besnard M,Noel JP, et al. Stealth (R) PEGylated polycyanoacry-late nanoparticles for intravenous administrationand splenic targeting. J Controlled Release 1999,60:121–128.

100. Peracchia MT, Harnisch S, Pinto-Alphandary H,Gulik A, Dedieu JC, et al. Visualization of in vitroprotein-rejecting properties of PEGylated stealth(R) polycyanoacrylate nanoparticles. Biomaterials1999, 20:1269–1275.

101. Calvo P, Gouritin B, Chacun H, Desmaele D,D’Angelo J, et al. Long-circulating PEGylated poly-cyanoacrylate nanoparticles as new drug carrier forbrain delivery. Pharm Res 2001, 18:1157–1166.

102. Calvo P, Gouritin B, Villarroya H, Eclancher F, Gian-navola C, et al. Quantification and localizationof PEGylated polycyanoacrylate nanoparticles inbrain and spinal cord during experimental allergicencephalomyelitis in the rat. Eur J Neurosci 2002,15:1317–1326.

103. Brigger I, Morizet J, Aubert G, Chacun H, Terrier-Lacombe MJ, et al. Poly(ethylene glycol)-coated

hexadecylcyanoacrylate nanospheres display a com-bined effect for brain tumor targeting. J PharmacolExp Ther 2002, 303:928–936.

104. Garcia-Garcia E, Gil S, Andrieux K, Desmaele D,Nicolas V, et al. A relevant in vitro rat model forthe evaluation of blood-brain barrier translocation ofnanoparticles. Cell Mol Life Sci 2005, 62:1400–1408.

105. Kim HR, Andrieux K, Delomenie C, Chacun H,Appel M, et al. Analysis of plasma protein adsorp-tion onto PEGylated nanoparticles by complementarymethods: 2-DE, CE and protein Lab-on-chip((R))system. Electrophoresis 2007, 28:2252–2261.

106. Kim HR, Andrieux K, Gil S, Taverna M, Chacun H,et al. Translocation of poly(ethylene glycol-co-hexadecyl)cyanoacrylate nanoparticles into rat brainendothelial cells: role of apolipoproteins in receptor-mediated endocytosis. Biomacromolecules 2007,8:793–799.

107. Kim HR, Gil S, Andrieux K, Nicolas V, Appel M,et al. Low-density lipoprotein receptor-mediatedendocytosis of PEGylated nanoparticles in rat brainendothelial cells. Cell Mol Life Sci 2007, 64:356–364.

108. Kreuter J, Shamenkov D, Petrov V, Ramge P,Cychutek K, et al. Apolipoprotein-mediated transportof nanoparticle-bound drugs across the blood-brainbarrier. J Drug Target 2002, 10:317–325.

109. Li YP, Pei YY, Zhou ZH, Zhang XY, Gu ZH, et al.PEGylated polycyanoacrylate nanoparticles as tumornecrosis factor-a carriers. J Controlled Release 2001,71:287–296.

110. Li Y-P, Pei Y-Y, Zhou Z-H, Zhang X-Y, Gu Z-H,et al. Stealth polycyanoacrylate nanoparticles astumor necrosis factor-a carriers: pharmacokinet-ics and anti-tumor effects. Biol Pharm Bull 2001,24:662–665.

111. Stella B, Arpicco S, Peracchia MT, Desmaele D,Hoebeke J, et al. Design of folic acid-conjugatednanoparticles for drug targeting. J Pharm Sci 2000,89:1452–1464.

112. Lenaerts V, Couvreur P, Christiaensleyh D, Joiris E,Roland M, et al. Degradation of poly(isobutylcyanoacrylate) nanoparticles. Biomaterials 1984,5:65–68.

113. Langer K, Seegmuller E, Zimmer A, Kreuter J. Char-acterization of polybutylcyanoacrylate nanoparticles:quantification of PBCA polymer and dextran. IntJ Pharm 1994, 110:21–27.

114. Leonard F, Kulkarni RK, Brandes G, Nelson J,Cameron JT. Synthesis and degradation of poly(alkyla-cyanoacrylates). J Appl Polym Sci 1966,10:259–272.

115. Lherm C, Muller RH, Puisieux F, Couvreur P.Alkylcyanoacrylatedrug carriers. 2. cytotoxicity ofcyanboacrylate nanoparticles with different alkylchain-length. Int J Pharm 1992, 84:13–22.



116. Muller RH, Lherm C, Herbort J, Couvreur P. Invitromodel for the degradation of alkylcyanoacrylatenanoparticles. Biomaterials 1990, 11:590–595.

117. Scherer D, Robinson JR, Kreuter J. Influence of theenzymes on the stability of polybutylcyanoacrylatenanoparticles. Int J Pharm 1994, 101:165–168.

118. Vezin WR, Florence AT. In vitro heterogeneousdegradation of poly(n-alkyl-cyanoacrylates). J BiomedMater Res 1980, 14:93–106.

119. Ley D, Couvreur P, Lenaerts V, Roland M, Speiser P.Etude du mecanisme de degradation des nanopar-ticules de polycyanoacrylate d’alkyle. Labo Pharma-Probl Tech 1984, 32:100–104.

120. Cooper AW, Harris PJ, Kumar GK, Tebby JC.Hydrolysis of alkyl cyanoacrylate and ethoxycar-bonylpropenoylphosphonate polymers. J Polym Sci,Part A: Polym Chem 1989, 27:1967–1974.


Polyalkylcyanoacrylate nanoparticles for delivery of drugs across the blood–brain barrier.


Advanced Review

Hydrogel mediated delivery oftrophic factors for neural repairJoshua S. Katz1 and Jason A. Burdick∗

Neurotrophins have been implicated in a variety of diseases and their delivery tosites of disease and injury has therapeutic potential in applications including spinalcord injury, Alzheimer’s disease, and Parkinson’s disease. Biodegradable polymers,and specifically, biodegradable water-swollen hydrogels, may be advantageous asdelivery vehicles for neurotrophins because of tissue-like properties, tailorabilitywith respect to degradation and release behavior, and a history of biocompatibility.These materials may be designed to degrade via hydrolytic or enzymaticmechanisms and can be used for the sustained delivery of trophic factors invivo. Hydrogels investigated to date include purely synthetic to purely natural,depending on the application and intended release profiles. Also, flexibility inmaterial processing has allowed for the investigation of injectable materials, thedevelopment of scaffolding and porous conduits, and the use of composites fortailored molecule delivery profiles. It is the objective of this review to describewhat has been accomplished in this area thus far and to remark on potential futuredirections in this field. Ultimately, the goal is to engineer optimal biomaterialsto deliver molecules in a controlled and dictated manner that can promoteregeneration and healing for numerous neural applications. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 128–139

Disruption of central nervous system (CNS) orperipheral nervous system (PNS) tissues such

as the spinal cord, optic nerve, and motor neuronscan severely affect a patient’s motor, sensory,and autonomic functions, and depending on theseverity of the injury, the patient’s quality of lifecan decline dramatically.1–3 Unfortunately, currentclinical treatment options are severely limited formany of these injuries and diseases and are unableto restore complete function to these patients. Forinstance, in the spinal cord, one significant barrierto regeneration is the extremely complex cascade ofevents (e.g., inflammation, glial scarring, release ofinhibitory molecules) that occurs after injury thatmust be addressed to restore functional recovery tothe patient.4,5 However, one promising therapy is thedelivery of neurotrophins that can influence the localfunction of cells within and surrounding the injury site.

∗Correspondence to: Jason A. Burdick, Department of Bioengineer-ing, University of Pennsylvania, 240 Skirkanich Hall, 210 S. 33rdStreet, Philadelphia, PA 19104, USA.E-mail: [email protected] of Bioengineering, University of Pennsylvania, 240Skirkanich Hall, 210 S. 33rd Street, Philadelphia, PA 19104, USA

DOI: 10.1002/wnan.010

Neurotrophins have been widely investigated fortheir influence on cell mortality, differentiation, andfunction in both the CNS and the PNS.6 These neu-rotrophins include factors such as nerve growth factor(NGF), brain-derived neurotrophic factor (BDNF),neurotrophin-3 (NT-3), neurotrophin-4/5 (NT-4/5),and glial derived neurotrophic factor (GDNF). Neu-rotrophins bind to tropomyosin-related kinase (Trk)receptors (TrkA for NGF, TrkB for BDNF and NT-4/5, and TrkC for NT-3) and the pan-neurotrophinreceptor p75.7 The functions of neurotrophins invivo are many and include controlling neural cellgrowth and survival, influencing glial development,and functions in non-neural tissues such as in thecardiovascular and immune systems.8–12 Addition-ally, neurotrophins can mediate axon signals or acton myelinating glia to influence the remyelinationof axons.13,14 For example, neurotrophins can pro-mote axonal growth, neuronal survival, and plasticityafter injury to the spinal cord.15 Lu and coworkers16

recently illustrated the ability of NT-3 in combinationwith cyclic adenosine monophosphate to induce regen-eration of sensory axons past a spinal cord lesion.Additionally, the overexpression of neurotrophinsafter injury induced sprouting of corticospinal tract


WIREs Nanomedicine and Nanobiotechnology Hydrogel mediated delivery of trophic factors

axons past the injury site.17 Techniques such as genetherapy, delivery via stem cells, and polymeric deliveryvehicles are being investigated for the supplementationof neurotrophins to injured neural tissues.6

There are several methods which have beenexplored for the delivery of neurotrophins and drugsto the nervous system, including mini-pumps, genet-ically modified cells, and polymer formulations.18–20

Hydrogels are water-swollen insoluble polymer net-works that have a wide range of chemical compo-sitions and properties.21–23 Hydrogels can be formedthrough a variety of mechanisms, including both phys-ical (e.g., ionic or hydrogen bonding) and chemicalgelation (e.g., covalent bonding). Through alterationsin the chemical structure, important properties such asswelling, degradation (e.g., hydrolytic or enzymatic),and mechanics can be controlled. The delivery oflarge molecules, such as neurotrophins, is typicallyaccomplished by encapsulating the molecules duringgelation, which are subsequently released via diffu-sion and degradation mechanisms. This process isrelatively complex and dynamic as the hydrogel meshsize changes as the material degrades and swells. Withrecent advances in polymer synthesis and our under-standing of biological polymers, our ability to controlhydrogels, and consequently, molecule delivery is con-stantly improving.24

There are numerous factors that make hydrogelsideal delivery vehicles for neurotrophic factors andrepair of neural tissue. First, this approach does notintroduce either live tissue (e.g., grafts) or viral vectors,eliminating potential issues with graft rejection andadverse responses. Next, hydrogel delivery eliminatesthe need for devices like pumps and catheters that canmalfunction. Finally, hydrogels can provide constantand tailorable delivery of either one or numerousmolecules to a desired in vivo location. Because ofthe short in vivo half life of neurotrophins, sustaineddelivery to the injury site results in significantly betterrecovery compared to a single injection.25,26 Becauseof the complexity of injuries, the appropriate deliveryprofile depends on the injured tissue and timing oftherapies. Several hydrogels have been investigatedfor the controlled delivery of neurotrophins andit is the objective of this review to outline pastwork in this area and look forward to futuredirections. The hydrogels investigated have ranged incomposition from purely synthetic (e.g., poly (ethyleneglycol) (PEG)) to purely natural (e.g., collagen), andin physical structure from uniform gels to porousscaffolds and composite materials.

INJECTABLE HYDROGELSHydrogels are made injectable through numerousmeans including free-radical polymerizations (i.e.,thermal, photo, or redox initiation), self-assemblyof materials, or ionic crosslinking.27 One of thebiggest advantages to using injectable materials forthese applications is the non-invasiveness of hydrogeldelivery, which can limit further tissue damage. Forinstance, disruption of the dura cover to many tissues(including the brain and spinal cord) results in theloss of many potentially stimulatory molecules, whichcould be avoided with injection through the dura.Additionally, many imaging techniques could be usedin combination with the injection procedures topotentially deliver these hydrogels in a closed surgery.Several clinically used biomaterials are alreadyinjected in vivo, such as poly(methyl methacrylate)bone cements28,29 and photocurable resins for fillingdental caries,30 and neural applications could benefitfrom similar procedures. This section focuses on thevarious injectable hydrogels that have been exploredfor delivery of growth factors for neurologicalapplications.

Agarose is a polysaccharide derived fromseaweed and comprised of repeating galactopyranoseunits. It has been used for a variety of biomedicalapplications and can be thermally induced to form ahydrogel through intermolecular hydrogen bondinginteractions.31,32 At elevated temperatures whenhydrogen bonds cannot form, agarose solutions donot gel. However, as the solution is cooled, hydrogenbonds begin to form, leading to gelation. Jain andcoworkers33 used cooled nitrogen gas to gel solutionsof agarose in situ. Following injury to the spinalcord, a solution of agarose containing BDNF-loadedmicrotubules was pipetted into the injury site. Thesolution was then cooled by nitrogen gas whichwas passed over a bath of dry ice to produce agel. A schematic of this cooling system is shown inFigure 1(a). The presence of BDNF greatly enhancedthe regeneration of axons and their ability to penetrateinto and through the scaffold.

Another natural polymer, collagen, crosslinksat physiological conditions through ionic interac-tions with salts present in solution. Hamann andcoworkers35,36 injected aqueous solutions of collagencontaining growth factors (epidermal growth factor(EGF) and/or FGF-1) into the intrathecal space sur-rounding the spinal cord as a drug delivery system.Rather than acting as a mechanical support for axonalregeneration as seen in many other systems, this sys-tem chemically supports the regenerative response tospinal cord injury (SCI) by the delivery of therapeu-tic agents directly and locally to the site of injury.



N2 Tank

Dry ice box

Aluminum rodsurrounded by dry ice

(a)

(b)

Liquid

Gel

Dura materintrathecal space

Spinal cord

FIGURE 1 | Examples of techniques for in situ gelation of hydrogelsolutions. (a) For agarose, nitrogen is passed from the tank through abath of dry ice and acetone (box) through an aluminum rod surroundedby dry ice to cool the solution and induce gelation. (b) Shear forces,induced through syringe injection, allow methylcellulose andacetate-modified hyaluronan (HAMC) gels to flow and then re-gel atphysiological temperatures in vivo, potentially in the intrathecal space.

Significantly higher levels of cavitation were observedin animals that did not receive the growth factors atthe site of injury. However, in a functional recoverytest with this specific delivery system, there was nosignificant difference between the treated animals andthe controls.

To address some of the issues associatedwith the injectable collagen system, such as slowgelation and cell infiltration within the dura, Guptaand coworkers34 created mechanically-reversible gelsconsisting of a blend of methylcellulose and acetate-modified hyaluronan (HAMC). Methylcellulose (MC)is a modified natural material that gels with increasingtemperature. As hydrogen bonds break, hydrophobicinteractions force the MC to gel. Hyaluronan(HA) is a naturally occurring polysaccharide thatmay spontaneously gel in water though hydrophilicassociations with the water and has been exploredwidely for biological applications.37–39 However,when sheared, the gel breaks as the molecules alignwith the direction of the shear. One limitationof using unmodified HA alone is that it quicklydisperses when injected in vivo because of its highwater solubility. HAMC blends gel at room andphysiologic temperatures, but flow when subjected

to shear stresses (Figure 1(b)). A volume of 10 µLof HAMC was injected into the intrathecal spacefollowing a clip compression injury to rats. Whilefunctional recovery of the animals was better withthe injection of the HAMC than with an injection ofartificial cerebrospinal fluid (aCSF) into the intrathecalspace, the difference was not significant after 1 week.However, no growth factors were injected with theHAMC in this study, and the presence of such factorscould potentially cause significant improvement in theanimal’s recovery.

PEG has been widely investigated as a bio-material for many years, primarily because it is arelatively inert material.40,41 As protein adsorption toPEG hydrogels is minimal, non-specific protein bind-ing and cellular interactions can be avoided. Hubbelland coworkers42,43 modified PEG hydrogels with bothdegradable units and then reactive groups to formbiodegradable hydrogels based on PEG. The chemi-cal structure of this material is shown in Figure 2(a).These synthetic macromers form a hydrogel via a rad-ical polymerization, which typically is initiated usinga photoinitiation process. For the encapsulation andrelease of a growth factor (e.g., neurotrophin), thePEG macromer is dissolved in a buffer solution con-taining the photoinitiator and the growth factor. Thissolution is exposed to light to form a hydrogel andthe growth factor is released through a combina-tion of diffusion and degradation.43–45 This hydrogelhas been explored extensively for both tissue engi-neering and drug delivery applications,46,47 primarilybecause of the control that is afforded over the tempo-ral material properties. For instance, degradation canbe altered through parameters such as the molecularweight of the PEG, the type (e.g., lactic vs caproicacid) of degradable groups, the number of degradablegroups, and the concentration of the macromer insolution.

For the delivery of neurotrophins, Burdickand coworkers48 monitored the release kinetics ofseveral factors from PEG hydrogels. Example releaseprofiles are shown in Figure 2(b) for a range ofneurotrophins. They found that the release kineticsof these factors were controlled by changes inthe network crosslinking density, which influencesneurotrophin diffusion and subsequent release fromthe gels, with total release times ranging fromweeks to several months. The release and activity ofone neurotrophic factor, ciliary-neurotrophic factor(CNTF), was assessed with a cell based proliferationassay and an assay for neurite outgrowth from retinalexplants. CNTF released from a degradable hydrogelabove an explanted retina was able to stimulateoutgrowth of a significantly higher number of



100

80

60

40

20

00 5 10 15 20 25

Time (days)

Cum

ulat

ive

% r

elea

seO O

O

O

O

O

O

O

R R

n

5mm

orR =

(a)

(b)

FIGURE 2 | (a) Structure of PEG modified with degradable groupsand reactive acrylates to allow for degradation andphotopolymerization, respectively. (b) Cumulative release ofneurotrophins ciliary-neurotrophic factor (CNTF) (•), BDNF (�), andNT-3 (�) from 10 wt% PEG hydrogels. (Reprinted, with permission,from Ref. 48. Copyright 2006 Elsevier).

neurites than controls without CNTF. Finally, uniquemicrosphere/hydrogel composites were developed tosimultaneously deliver multiple neurotrophins withindividual release rates.48

This system was also exploited for application tothe injured spinal cord.49 The failure of injured axonsto regenerate in the mature CNS can have significantimplications in spinal cord injury. However, thedelivery of neurotrophins can promote axon growthin injured CNS and thus, the delivery of NT-3 wasused in an attempt to alter anatomical and behavioraloutcomes. NT-3 was delivered to a dorsal hemisectionlesion in adult rats using PEG hydrogels by injectingthe macromer and initiator solution directly on thelesion and exposing to visible light (using a dentalcuring lamp) for 60 s for gelation. The treated animalsshowed improved recovery in both an open-field BBBtest and in a horizontal ladder walk test comparedto untreated rats. Also, the treated animals showedmuch greater axon growth in both the corticospinaland raphespinal tracts over control animals. Thus, thisprovides great scope to the use of injectable hydrogelsfor trophic factor delivery to influence outcomes ininjured patients.

POROUS SCAFFOLDS AND GUIDANCECHANNELSWhile the primary goal of injectable hydrogels isto chemically support axonal regeneration throughneurotrophin delivery locally at the site of injury,it may be advantageous to also physically supportcellular function and growing axons. To date, themajority of this support has come through thedevelopment of nerve guidance channels, which aimto template regenerating axons through the injurysite.50 While originally designed from non-degradable,non-hydrogel materials, more recent scaffolds havebegun to incorporate many features common amongtissue engineering scaffolds, such as porosity andswelling.

Poly(hydroxyethyl methacrylate) (poly(HEMA))hydrogels have been widely explored as a biocompat-ible, non-degradable material for drug delivery andtissue engineering applications.51,52 Gels form throughthe free-radical polymerization of the methacrylateunits in the presence of a crosslinker (i.e., dimethacry-late). These materials have been used recently for thedelivery of neurotrophins to cells for axonal regen-eration. Piotrowicz and Shoichet53 synthesized nerveguidance channels from copolymers of HEMA andmethyl methacrylate (MMA). They incorporated NGFinto the channels either by adding poly (lactic-co-glycolic acid) (PLGA) microspheres to the formulationor by adding a second layer of HEMA containingNGF to the interior wall of the channel. Sustainedrelease of NGF was observed for both systems overa 30-day period, but the release was much higherfor the HEMA/NGF coated nerve guidance channels.Shoichet and coworkers54,55 also induced concentra-tion gradients of NGF and NT-3 into macroporousHEMA nerve guidance channels using a gradientmixer. Neurite outgrowth within the channels wasobserved to follow the gradient of NGF. Additionally,they found that NGF and NT-3 work synergistically,with less NGF required to successfully guide neuritegrowth when NT-3 was present.

Belkas and coworkers56 filled HEMA-co-MMAnerve tubes with collagen and implanted them intorats that had 10 mm sections of the sciatic nerveremoved. A bimodal response was observed inwhich approximately 60% of the rats improvedin comparison to rats which received autografts inplace of the nerve tube, while approximately 40%saw no significant healing, potentially as a result ofchannel collapse. In an expansion of this work, Tsaiand coworkers57 filled HEMA-co-MMA hydrogelguidance channels with MatrigelTM, MC, fibrin, ortype I collagen. Some channels also contained eitherFGF-1 or NT-3. Channels were implanted into rats



(a) (b)

(c) (d)

FIGURE 3 | Insertion of collagen-filledHEMA-co-MMA nerve tubes into a complete transectionspinal cord injury. (a) Injury pre-insertion. (b) Filling oftubes with collagen. (c) Half-implanted tube.(d) Fully-inserted nerve tube at site of injury. (Reprinted,with permission, from Ref. 57. Copyright 2006 Elsevier).

that had undergone a complete spinal cord transection(Figure 3). The presence of the growth factors alteredthe density and improved the orientation of thegrowing axons. Two channel formulations (fibrinand multiple channels within a larger channel) ledto consistent improvement in recovery as measured byBBB score.

To better mimic the mechanical properties ofthe spinal cord, Bakshi and coworkers58 designedHEMA microporous gels containing 85% water.These gels had compressive moduli of 3–4 kPa,similar to that of the spinal cord. The flexibilityof the gels allows them to easily fit into a defectcaused by spinal cord injury, and the presence ofthe micropores allows for cells to migrate and growthrough the gels. Gels were implanted into injuredrats and after 1, 2, and 4 weeks, the rats weresacrificed and examined for response to the gels. Allgels prompted a moderate inflammatory response,though there was little scarring. In gels soakedin BDNF prior to implantation, axon regenerationwas observed, though only transiently (2 weeks). Inorder to make the nerve conduits bioactive, Yu andShoichet59 copolymerized HEMA with 2-aminoethylmethacrylate, which can be easily modified withpeptides. In this work, two peptides derived fromlaminin (i.e., YIGSR and IKVAV) were used. Peptide-modified conduits templated on polycaprolactone(PCL) fibers (which were subsequently removed,yielding hollow channels) were much more conduciveto cell growth and neurite extension compared tounmodified channels.

In a recent study, Bryant and coworkers60 useda photomask and ultraviolet light to selectively gelcertain regions of a precursor solution, creating

channels within poly(HEMA) hydrogels. The gelswere created by pouring a HEMA solution contain-ing a biodegradable crosslinker and photoinitiatorover templated poly(methyl methacrylate) (PMMA)spheres, which is a technique to create highly-orderedporous hydrogels.61 Selective blocking of the UV lightand exploitation of reaction behavior was used tocreate open channels of several hundred microns indiameter, while the rest of the gel contained open,highly interconnected pores of approximately fiftymicrons in diameter. While the authors did not addressthe potential application of this system to neuralregeneration, it is easy to postulate that within thesystem, the large pores could facilitate axonal growthand guidance, while the smaller surrounding poresenabled the rapid transport of nutrients to and fromthe growing cells.

Stokols and Tuszynski62,63 created agarosescaffolds with linearly-oriented channels by growingice crystals along a temperature gradient through asolution of agarose followed by freeze drying. Arepresentative image of the scaffolds is shown inFigure 4, showing longitudinal pores with a honey-comb structure cross-section. BDNF was incorporatedinto the gels by swelling the freeze-dried gels inthe presence of the protein. Prior to implantationin rats, the channels were also filled with collagen.The presence of a growth factor allowed for 2–3 timesas many axons to penetrate through the channelscompared to negative controls (no BDNF), and theimmune response was minimal. However, the authorsdo not report on the functional recovery of theanimals. Stokols, Tuszynski, and coworkers64 alsotemplated agarose channels on polystyrene fibers,yielding a scaffold with uniform channels following



FIGURE 4 | Aligned agarose nerve tubes synthesizedby ice crystallization. (a, b) SEM images of freeze-driedtubes a| longitudinally and (b) cross-sectionally.(c) Axonal penetration in vivo through the agarose tubes.Scale bars = 100 µm. (Reprinted, with permission, fromRef. 63. Copyright 2006 Elsevier).

(a) (b)

(c)

polystyrene dissolution. Prior to implantation, theauthors filled the channels with bone marrow stromalcells (MSCs) or MSCs engineered to produce BDNF.BDNF production by the MSCs in vivo greatlyenhanced the ability of regenerating axons to penetrateinto the scaffold and span the length of the scaffold.However, the functional recovery of the animals wasnot reported.

PATTERNED AND COMPOSITEHYDROGELS

To provide spatial chemical cues and to enhancematerial properties, numerous techniques such aspatterning and the use of composite materials havebeen investigated. As demonstrated by Nomura andcoworkers,65 both the mechanical and chemical prop-erties of implants are important for the developmentof a successful implant. Loading one material withinanother material allows for further tuning of bothtypes of properties in tandem with each other.

The 2-nitrobenzyl moiety is well-known as aphoto-protecting group, and recently has found aniche in the development of functional biomaterialswhich can be patterned in multiple dimensions.66,67 Ina pioneering work, Luo and Shoichet68,69 conjugatedagarose polymers with 2-nitrobenzyl cysteine. Follow-ing gelation, selected portions of the gels were exposedto ultraviolet light, liberating the thiol sidechain of theblocked cysteine. Further conjugation of the liberatedcysteine residues allowed for biofunctionalization ofthe gels with small peptides or proteins. Rat dorsalroot ganglia were then seeded on the gels and seento grow through the columns of gel containing the

peptide, as patterned via the light activation. Thiswork reports the spatial control of only two dimen-sions or at best a gradient in the z direction. However,it is easy to see how such a system could easilybe expanded to three dimensional patterning usingadvanced microscopy and laser techniques and couldincorporate the delivery of neurotrophic factors.

To study how cells move along a moleculargradient, Dodla and Bellamkonda70 immobilizedlaminin-1 (LN-1) in an agarose gel. Gradients ofdifferent degrees were created by diffusion of LN-1 through the gel followed by photoimmobilization.Interestingly, the lower concentration gradients ofLN-1 did a better job of promoting directionalityin neurite extension and growth. These results arepotentially useful for the development of regenerativematerials that could be modified with gradients ofneurotrophins and used in vivo for axonal growthand guidance.

Fibrin is another natural polymer that hasbeen widely used in the biomaterials field.71,72

Sakiyama-Elbert and coworkers73–77 have shownprogress in the controlled release of neurotrophinsfrom fibrin gels through interactions with heparinor peptides. Heparin interacts non-covalently withvarious neurotrophins, such as NGF, BDNF and NT-3. Heparin was attached to fibrin gels that containedimmobilized heparin binding peptides within thematrix to influence the release of neurotrophins fromthe fibrin gels. A schematic of this process is illustratedin Figure 5. Decreased release rates from the gels wereobserved and reported for both NGF74 and NT-3,75

and were observed to significantly improve neuriteextension in vitro.74 In in vivo studies, regeneration of



GG

G

G

G

GG

G

GG

G

G

Fibrin

Peptide

Heparin

Growth factor

FIGURE 5 | Schematic of growth factor immobilization to a fibrinmatrix through interactions with heparin. (Reprinted, with permission,from Ref. 76. Copyright 2006 Elsevier).

axons was significantly improved by the inclusion ofheparin into the fibrin matrices to control the releaseof NGF73 or NT-3.76 As another method to slow therelease of NGF from fibrin matrices, phage displaywas used to positively select an NGF binding peptide,which was then immobilized in the fibrin matrix.77 Animproved response in neurite extension was observedrelative to free NGF in a fibrin matrix and the responsewas both dose and pH dependent.

Towards another composite structure, Nomuraand coworkers65 carried out a study similar to thatcarried out by Belkas and coworkers56 (discussedabove) with HEMA or HEMA-co-MMA channelsloaded with fibrin, acidic fibroblast growth factor(aFGF), and heparin and reinforced with a PCLcoil. They employed the coil to alleviate theproblems of tube collapse that was observed inthe previous study.56 The coil was successful inpreventing collapse of the nerve tubes; however, noaxonal regeneration was observed as a result of thedevelopment of syringomyelia (a cyst which causesnerve degeneration) and migration of the rostralstump. This work emphasizes the importance ofmechanical properties of the hydrogel towards thesuccessful regeneration of the tissues.

Combining synthetic polymers as drug releasingdevices within a natural polymer matrix, Goraltchoukand coworkers78 incorporated PLGA 50/50 micro-spheres containing bovine serum albumin (BSA) orepidermal growth factor (EGF) within chitosan/chitinnerve guidance channels. PLGA is an example of apoly(α-hydroxy ester) that has been widely exploredas a degradable material for tissue engineering anddrug delivery. They observed controlled release ofprotein from the microspheres into solution over an84-day period compared to only 70 days for the free

microspheres. The released EGF from the nerve guid-ance channels was able to successfully promote theformation of neurospheres in culture for the first 14days of release. However, by day 21, the released EGFwas no longer biologically active. The lack of avail-ability of bioactive EGF is a very serious potentiallimitation that must be addressed for the devel-opment of sustained release implants. Others haveentrapped numerous neurotrophins in microspheresfor delivery in vivo.79,80 Yu and Bellamkonda81 loadedpoly(sulfone) guidance channels with LN-1-modifiedagarose and lipid microtubules loaded with NGF forslow release. The channels were implanted in a 10-mm defect in the sciatic nerve of rats. Recovery andregeneration were statistically the same as autografts,though neither group returned to ‘normal’ function.As an additional example, Yang and coworkers82

formed porous nerve conduits of PLGA 75/25 contain-ing NGF (illustrated in Figure 6). Successful release ofNGF was observed and release could be sustained upto a 40-day period by changes in porosity, mechanismof NGF incorporation, and polymer molecular weight.Others have used non-porous scaffolding for neuralregeneration applications,83,84 but this is beyond thescope of this review.

FUTURE DIRECTIONS

As reviewed above, a substantial amount of workhas been performed in the area of hydrogel deliveryof trophic factors. This work has led to significantadvances in the development of potential therapeuticsfor individuals with damage to their neural system.However, recent advances in the engineering ofmaterials, polymer synthesis, and neural biology openup further avenues for research in this area. Severalimportant and developing technologies are outlinedbelow.

Stimuli responsive hydrogels are being developedfor numerous applications, where some external stim-uli (pH, temperature, light) are used to alter hydro-gel properties, and consequently, molecule release.For instance, poly(N-isopropylacrylamide) hydrogelsrespond and deswell with increased temperatures,85

which can lead to the release of growth factors. Dasand coworkers86 recently incorporated light-sensitivegold nanoshells into these hydrogels, where the goldnanoshells absorb light at a predefined wavelength,heat up, and initiate the hydrogel thermal response.In this system, light can be transmitted transdermallyto trigger release, leading to a well defined releaseprofile. Stayton and coworkers87,88 have developedseveral novel copolymer systems that are responsive



Side view

(b)Top view

Screens

Screens

Fixers

Fixers

PINSPINS

Side view

Center rod

(a)

Center rod

Top view

(c) (d)

FIGURE 6 | PLGA porous nerve conduits. (a, b) Schematics for fabrication of single and multiple lumen conduits, respectively. (c, d) Images ofsingle and multiple lumen porous conduits. Scale bars: (c) 700 µm, (d) 500 µm. (Reprinted, with permission, from Ref. 82. Copyright 2005 Elsevier).

to pH. These polymers are used as gene delivery vehi-cles which swell at low pH and facilitate release fromthe endosome, allowing more specific, local release ofthe gene. Such a polymer could be potentially use-ful for intracellular delivery or delivery around theinjury where the pH has been shown to be slightlylower than surrounding tissues.89 Thus, advances inpolymer synthesis and processing are leading to novelapproaches for growth factor, and potentially, neu-rotrophin delivery.

In a different area, Mi and coworkers identifiedpleiotrophin (PTN, also known as heparin bindingneurotrophic factor) as a neurotrophic factor thataids in the recovery of damaged motor neurons in thespinal cord.90 Following nerve injury, a significantupregulation of PTN was observed in recoveringnerves near the site of injury for up to 1 month.In vitro studies showed that PTN could induce axonalgrowth in a culture of spinal cord explants in thedirection of the PTN source. Following a sciaticnerve transection, HEK−293PTN cells were deliveredin a silicone tube to the site of injury. After 8weeks, there was a ten-fold increase in axon densitycompared to control, and some functional recoverywas observed. Additionally, when HEK−293PTN cellswere transplanted in gelfoams to the site of facialnerve injury in mouse pups, 63% of the facial motorneurons recovered compared to 12% of the controls.The emergence of PTN as another neurotrophicfactor that plays a role in the recovery of injuredneurons should further aid in the search for theideal delivery system/delivery agent(s) combination.

Finally, recent trends are towards combinatorialapproaches for the delivery of multiple stimulatoryfactors that together can have an additive effect onregeneration. One example by Lu and coworkers16

illustrated that the delivery of both cAMP andNT-3 provided elevated recovery in animals withspinal cord injury over the delivery of only oneof the molecules. Similar results were also seenin the work done by Moore and coworkers withNT-3 and NGF.55 Biomaterials, and specifically,hydrogels, are ideal candidates for the deliveryof multiple factors, each with individual releaseprofiles, as a result of our excellent control overpolymer behavior. Additionally, various stem cells andnovel scaffolding may be combined in combinationwith neurotrophin delivery to accelerate healing andrecovery.

CONCLUSION

The application of synthetic and natural materials forneurotrophin delivery has only recently been used andis finding widespread success. While no single systemexplored has yet to provide what many would considerto be complete recovery, the results from many ofthese studies are promising and for many of theseapplications, incremental improvements correlate tosignificant enhancement in quality of life. Continuedresearch in this field will shed new light on the roleeach neurotrophin plays in the healing of neuraltissues, as well as on the role that the physical material



can play in the delivery of the neurotrophin andsupport of growing axons at the injury site. Recenttrends and expected directions in this field promiseeven more advanced systems of responsive materials,

novel trophic factors, and combinatorial approachesthat will allow for promising treatment methods forpatients with debilitating neurological conditions.

REFERENCES

1. Johnson EO, Zoubos AB, Soucacos PN: Regenerationand repair of peripheral nerves. Injury 2005, 36(suppl4):S24–S29.

2. Ma W, Fitzgerald W, Liu QY, O’Shaughnessy TJ,Maric D, et al.: CNS stem and progenitor celldifferentiation into functional neuronal circuits inthree-dimensional collagen gels. Exp Neurol 2004,190(2):276–288.

3. Weber RV, Mackinnon SE: Bridging the neural gap.Clin Plast Surg 2005, 32(4):605–616.

4. Gautier SE, Oudega M, Fragoso M, Chapon P,Plant GW, et al.: Poly(alpha-hydroxyacids) for appli-cation in the spinal cord: Resorbability and biocom-patibility with adult rat Schwann cells and spinal cord.J Biomed Mater Res 1998, 42:642–654.

5. Fitch MT, Doller C, Combs CK, Landreth GE, Silver J:Cellular and molecular mechanisms of glial scarringand progressive cavitation: In Vivo and In Vitro analy-sis of inflammation-induced secondary injury after CNSTrauma. J Neurosci 1999, 19:8182–8198.

6. Blesch A, Lu P, Tuszynski MH: Neurotrophic factors,gene therapy, and neural stem cells for spinal cordrepair. Brain Res Bull 2002, 57:833–838.

7. Arevalo JC, Wu SH: Neurotrophin signaling:many exciting surprises!. Cell Mol Life Sci 2006,63(13):1523–1537.

8. Aguayo AJ, Clarke DB, Jelsma TN, Kittlerova P, Fried-man HC, et al.: Effects of neurotrophins on the survivaland regrowth of injured retinal neurons. Ciba FoundSymp 1996, 196(:):135–144 discussion 144–138.

9. Barde YA: Neurotrophins: a family of proteins sup-porting the survival of neurons. Prog Clin Biol Res1994, 390:45–56.

10. Casaccia-Bonnefil P, Kong H, Chao MV: Neu-rotrophins: the biological paradox of survival fac-tors eliciting apoptosis. Cell Death Differ 1998,5(5):357–364.

11. Saito A, Tominaga T, Chan PH: Neuroprotective roleof neurotrophins: relationship between nerve growthfactor and apoptotic cell survival pathway after cerebralischemia. Curr Atheroscler Rep 2005, 7(4):268–273.

12. Tessarollo L: Pleiotropic functions of neurotrophinsin development. Cytokine Growth Factor Rev 1998,9(2):125–137.

13. Jean I, Lavialle C, Barthelaix-Pouplard A, Fressin-aud C: Neurotrophin-3 specifically increases matureoligodendrocyte population and enhances remyelina-tion after chemical demyelination of adult rat CNS.Brain Res 2003, 972(1–2):110–118.

14. Rosenberg SS, Ng BK, Chan JR: The quest for remyeli-nation: a new role for neurotrophins and their recep-tors. Brain Pathol 2006, 16(4):288–294.

15. Zhou L, Baumgartner BJ, Hill-Felberg SJ, McGowenLR, Shine HD: Neurotrophin-3 expressed In Situinduces axonal plasticity in the adult injured spinalcord. J Neurosci 2003, 23:1424–1431.

16. Lu P, Yang H, Jones LL, Filbin MT, Tuszynski MH:Combinatorial therapy with neurotrophins and cAMPpromotes axonal regeneration beyond sites of spinalcord injury. J Neurosci 2004, 24(28):6402–6409.

17. Zhou L, Shine HD: Neurotrophic factors expressed inboth cortex and spinal cord induce axonal plastic-ity after spinal cord injury. J Neurosci Res 2003,74:221–226.

18. Harbaugh RE, Saunders RL, Reeder RF: Use ofimplantable pumps for central nervous-system druginfusions to treat neurological disease. Neurosurgery1988, 23(6):693–698.

19. Nomura H, Tator CH, Shoichet MS: Bioengineeredstrategies for spinal cord repair. J Neurotrauma 2006,23(3–4):496–507.

20. Whittlesey KJ, Shea LD: Delivery systems for smallmolecule drugs, proteins, and DNA: the neu-roscience/biomaterial interface. Exp Neurol 2004,190(1):1–16.

21. Anseth KS, Burdick JA: New directions in photopoly-merizable biomaterials. MRS Bull 2002, 27:130–138.

22. Khare AR, Peppas NA: Release behavior of bioactiveagents from pH-sensitive hydrogels. J Biomater SciPolym Ed 1993, 4(3):275–289.

23. McNair AM: Using hydrogel polymers for drug deliv-ery. Med Device Technol 1996, 7(10):16–22.

24. Anderson DG, Burdick JA, Langer R: Mate-rials science. Smart biomaterials. Science 2004,305(5692):1923–1924.

25. Krewson CE, Klarman ML, Saltzman WM: Distribu-tion of nerve growth-factor following direct delivery tobrain interstitium. Brain Res 1995, 680(1–2):196–206.



26. Sayer FT, Oudega M, Hagg T: Neurotrophins reducedegeneration of injured ascending sensory and corti-cospinal motor axons in adult rat spinal cord. ExpNeurol 2002, 175(1):282–296.

27. Lee KY, Mooney DJ: Hydrogels for tissue engineering.Chem Eng Prog 2001, 101(7):1869–1879.

28. Carrodeguas RG, Lasa BV, Del Barrio JS: Injectableacrylic bone cements for vertebroplasty with improvedproperties. J Biomed Mater Res B Appl Biomater 2004,68(1):94–104.

29. Lewis G: Injectable bone cements for use in ver-tebroplasty and kyphoplasty: state-of-the-art review.J Biomed Mater Res B Appl Biomater 2006,76(2):456–468.

30. Anseth KS, Newman SM, Bowman CN: Polymeric den-tal composites: properties and reaction behavior ofmultimethacrylate dental restorations. Adv Polym Sci1995, 122:177–217.

31. Balgude AP, Yu X, Szymanski A, Bellamkonda RV:Agarose gel stiffness determines rate of DRG neu-rite extension in 3D cultures. Biomaterials 2001,22(10):1077–1084.

32. Mauck RL, Soltz MA, Wang CC, Wong DD, Chao PH,et al.: Functional tissue engineering of articular carti-lage through dynamic loading of chondrocyte-seededagarose gels. J Biomech Eng 2000, 122(3):252–260.

33. Jain A, Kim YT, McKeon RJ, Bellamkonda RV: In situgelling hydrogels for conformal repair of spinal corddefects, and local delivery of BDNF after spinal cordinjury. Biomaterials 2006, 27(3):497–504.

34. Gupta D, Tator CH, Shoichet MS: Fast-gellinginjectable blend of hyaluronan and methylcellulose forintrathecal, localized delivery to the injured spinal cord.Biomaterials 2006, 27(11):2370–2379.

35. Hamann MCJ, Tator CH, Shoichet MS: Injectableintrathecal delivery system for localized administra-tion of EGF and FGF-2 to the injured rat spinal cord.Exp Neurol 2005, 194(1):106–119.

36. Hamann MCJ, Tsai EC, Tator CH, Shoichet MS:Novel intrathecal delivery system for treatment ofspinal cord injury. Exp Neurol 2003, 182(2):300–309.

37. Chung C, Mesa J, Miller GJ, Randolph MA, Gill TJ,et al.: Effects of auricular chondrocyte expansion onneocartilage formation in photocrosslinked hyaluronicacid networks. Tissue Eng 2006, 12(9):2665–2673.

38. Chung C, Mesa J, Randolph MA, Yaremchuk M, Bur-dick JA: Influence of gel properties on neocartilageformation by auricular chondrocytes photoencapsu-lated in hyaluronic acid networks. J Biomed Mater ResA 2006, 77A(3):518–525.

39. Peattie RA, Rieke ER, Hewett EM, Fisher RJ, Shu XZ,et al.: Dual growth factor-induced angiogenesis invivo using hyaluronan hydrogel implants. Biomaterials2006, 27(9):1868–1875.

40. Elisseeff J, McIntosh W, Anseth K, Riley S, Ragan P,et al.: Photoencapsulation of chondrocytes inpoly(ethylene oxide)-based semi-interpenetrating net-works. J Biomed Mater Res 2000, 51(2):164–171.

41. Metters AT, Anseth KS, Bowman CN: Fundamentalstudies of a novel, biodegradable PEG-b-PLA hydrogel.Polymer 2000, 41(11):3993–4004.

42. Sawhney AS, Pathak CP, Hubbell JA: Bioerodiblehydrogels based on photopolymerized poly(EthyleneGlycol)-Co-Poly(Alpha-Hydroxy Acid) diacrylatemacromers. Macromolecules 1993, 26(4):581–587.

43. West JL, Hubbell JA: Photopolymerized hydrogelmaterials for drug-delivery applications. React Polym1995, 25(2–3):139–147.

44. Lu SX, Anseth KS: Release behavior of high molec-ular weight solutes from poly(ethylene glycol)-based degradable networks. Macromolecules 2000,33(7):2509–2515.

45. Mason MN, Metters AT, Bowman CN, Anseth KS:Predicting controlled-release behavior of degradablePLA-b-PEG-b-PLA hydrogels. Macromolecules 2001,34(13):4630–4635.

46. Anseth KS, Metters AT, Bryant SJ, Martens PJ, Elisse-eff JH, et al.: In situ forming degradable networks andtheir application in tissue engineering and drug delivery.J Control Release 2002, 78(1–3):199–209.

47. Burdick JA, Mason MN, Hinman AD, Thorne K,Anseth KS: Delivery of osteoinductive growth factorsfrom degradable PEG hydrogels influences osteoblastdifferentiation and mineralization. J Controlled Release2002, 83(1):53–63.

48. Burdick JA, Ward M, Liang E, Young MJ, Langer R:Stimulation of neurite outgrowth by neurotrophinsdelivered from degradable hydrogels. Biomaterials2006, 27(3):452–459.

49. Piantino J, Burdick JA, Goldberg D, Langer R,Benowitz LI: An injectable, biodegradable hydrogel fortrophic factor delivery enhances axonal rewiring andimproves performance after spinal cord injury. ExpNeurol 2006, 201(2):359–367.

50. Bellamkonda RV: Peripheral nerve regeneration: anopinion on channels, scaffolds and anisotropy. Bio-materials 2006, 27(19):3515–3518.

51. Goda T, Ishihara K: Soft contact lens biomaterials frombioinspired phospholipid polymers. Expert Rev MedDevices 2006, 3(2):167–174.

52. Oral E, Peppas NA: Hydrophilic molecularly imprintedpoly(hydroxyethyl-methacrylate) polymers. J BiomedMater Res A 2006, 78(1):205–210.

53. Piotrowicz A, Shoichet MS: Nerve guidance chan-nels as drug delivery vehicles. Biomaterials 2006,27(9):2018–2027.

54. Kapur TA, Shoichet MS: Immobilized concentrationgradients of nerve growth factor guide neurite out-growth. J Biomed Mater Res A 2004, 68A(2):235–243.



55. Moore K, Macsween M, Shoichet M: Immobilized con-centration gradients of neurotrophic factors guide neu-rite outgrowth of primary neurons in macroporousscaffolds. Tissue Eng 2006, 12(2):267–278.

56. Belkas JS, Munro CA, Shoichet MS, Midha R: Periph-eral nerve regeneration through a synthetic hydro-gel nerve tube. Restor Neurol Neurosci 2005,23(1):19–29.

57. Tsai EC, Dalton PD, Shoichet MS, Tator CH: Matrixinclusion within synthetic hydrogel guidance chan-nels improves specific supraspinal and local axonalregeneration after complete spinal cord transection.Biomaterials 2006, 27(3):519–533.

58. Bakshi A, Fisher O, Daggi T, Himes BT, Fischer I,et al.: Mechanically engineered hydrogel scaffolds foraxonal growth and angiogenesis after transplanta-tion in spinal cord injury. J Neurosurg Spine 2004,1(3):322–329.

59. Yu TT, Shoichet MS: Guided cell adhesion and out-growth in peptide-modified channels for neural tissueengineering. Biomaterials 2005, 26(13):1507–1514.

60. Bryant SJ, Cuy JL, Hauch KD, Ratner BD: Photo-patterning of porous hydrogels for tissue engineering.Biomaterials 2007, 28(19):2978–2986.

61. Stachowiak AN, Bershteyn A, Tzatzalos E, Irvine DJ:Bioactive hydrogels with an ordered cellular struc-ture combine interconnected macroporosity and robustmechanical properties. Adv Mater 2005, 17(4):399.

62. Stokols S, Tuszynski MH: The fabrication and char-acterization of linearly oriented nerve guidance scaf-folds for spinal cord injury. Biomaterials 2004,25(27):5839–5846.

63. Stokols S, Tuszynski MH: Freeze-dried agarose scaf-folds with uniaxial channels stimulate and guide linearaxonal growth following spinal cord injury. Biomateri-als 2006, 27(3):443–451.

64. Stokols S, Sakamoto J, Breckon C, Holt T, Weiss J,et al.: Templated agarose scaffolds support lin-ear axonal regeneration. Tissue Eng 2006,12(10):2777–2787.

65. Nomura H, Katayama Y, Shoichet MS, Tator CH:Complete spinal cord transection treated by implanta-tion of a reinforced synthetic hydrogel channel resultsin syringomyelia and caudal migration of the rostralstump. Neurosurgery 2006, 59(1):183–192.

66. Doh J, Irvine DJ: Photogenerated polyelectrolyte bilay-ers from an aqueous-processible photoresist for multi-component protein patterning. J Am Chem Soc 2004,126(30):9170–9171.

67. Katz JS, Doh J, Irvine DJ: Composition-tunable prop-erties of amphiphilic comb copolymers containingprotected methacrylic acid groups for multicomponentprotein patterning. Langmuir 2006, 22(1):353–359.

68. Luo Y, Shoichet MS: Light-activated immobilizationof biomolecules to agarose hydrogels for con-trolled cellular response. Biomacromolecules 2004,5(6):2315–2323.

69. Luo Y, Shoichet MS: A photolabile hydrogel for guidedthree-dimensional cell growth and migration. NatMater 2004, 3(4):249–253.

70. Dodla MC, Bellamkonda RV: Anisotropic scaffoldsfacilitate enhanced neurite extension in vitro. J BiomedMater Res A 2006, 78A(2):213–221.

71. Pittier R, Sauthier F, Hubbell JA, Hall H: Neuriteextension and in vitro myelination within three-dimensional modified fibrin matrices. J Neurobiol2005, 63(1):1–14.

72. Wong C, Inman E, Spaethe R, Helgerson S: Fibrin-based biomaterials to deliver human growth factors.Thromb Haemost 2003, 89(3):573–582.

73. Lee AC, Yu VM, Lowe JB, Brenner MJ, Hunter DA,et al.: Controlled release of nerve growth factorenhances sciatic nerve regeneration. Exp Neurol 2003,184(1):295–303.

74. Sakiyama-Elbert SE, Hubbell JA: Controlled release ofnerve growth factor from a heparin-containing fibrin-based cell ingrowth matrix. J Control Release 2000,69(1):149–158.

75. Taylor SJ, McDonald JW, Sakiyama-Elbert SE: Con-trolled release of neurotrophin-3 from fibrin gelsfor spinal cord injury. J Controlled Release 2004,98(2):281–294.

76. Taylor SJ, Rosenzweig ES, McDonald JW, Sakiyama-Elbert SE: Delivery of neurotrophin-3 from fibrinenhances neuronal fiber sprouting after spinal cordinjury. J Control Release 2006, 113(3):226–235.

77. Willerth SM, Johnson PJ, Maxwell DJ, Parsons SR,Doukas ME, et al.: Rationally designed peptides forcontrolled release of nerve growth factor from fibrinmatrices. J Biomed Mater Res A 2007, 80A(1):13–23.

78. Goraltchouk A, Scanga V, Morshead CM,Shoichet MS: Incorporation of protein-eluting micro-spheres into biodegradable nerve guidance channelsfor controlled release. J Controlled Release 2006,110(2):400–407.

79. Mahoney MJ, Saltzman WM: Transplantation ofbrain cells assembled around a programmablesynthetic microenvironment. Nat Biotechnol 2001,19(10):934–939.

80. Ward MS, Khoobehi A, Lavik EB, Langer R,Young MJ: Neuroprotection of retinal ganglion cells inDBA/2J mice with GDNF-loaded biodegradable micro-spheres. J Pharm Sci Technol 2007, 96(3):558–568.

81. Yu XJ, Bellamkonda RV: Tissue-engineered scaffoldsare effective alternatives to autografts for bridgingperipheral nerve gaps. Tissue Eng 2003, 9(3):421–430.

82. Yang Y, De Laporte L, Rives CB, Jang JH, Lin WC,et al.: Neurotrophin releasing single and multiple



lumen nerve conduits. J Control Release 2005,104(3):433–446.

83. Lavik EB, Klassen H, Warfvinge K, Langer R,Young MJ: Fabrication of degradable polymer scaffoldsto direct the integration and differentiation of retinalprogenitors. Biomaterials 2005, 26(16):3187–3196.

84. Levenberg S, Burdick JA, Kraehenbuehl T, Langer R:Neurotrophin-induced differentiation of human embry-onic stem cells on three-dimensional polymeric scaf-folds. Tissue Eng 2005, 11(3–4):506–512.

85. Chilkoti A, Dreher MR, Meyer DE, Raucher D: Tar-geted drug delivery by thermally responsive polymers.Adv Drug Deliv Rev 2002, 54(5):613–630.

86. Das M, Sanson N, Fava D, Kumacheva E: Microgelsloaded with gold nanorods: photothermally triggeredvolume transitions under physiological conditions.Langmuir 2007, 23(1):196–201.

87. Henry SM, El-Sayed MEH, Pirie CM, Hoff-man AS, Stayton PS: pH-responsive poly(styrene-alt-maleic anhydride) alkylamide copolymers forintracellular drug delivery. Biomacromolecules 2006,7(8):2407–2414.

88. Yin X, Hoffman AS, Stayton PS: Poly(N-isopropylacrylamide-co-propylacrylic acid) copolymersthat respond sharply to temperature and pH. Biomacro-molecules 2006, 7(5):1381–1385.

89. Akino M, Odonnell M, Robitaille PML, Stokes BT:Phosphorus-31 magnetic resonance spectroscopy stud-ies of pig spinal cord injury - Myelin changes, intra-cellular pH, and bioenergetics. Invest Radiol 1997,32(7):382–388.

90. Mi RF, Chen WR, Hoke A: Pleiotrophin is a neu-rotrophic factor for spinal motor neurons. Proc NatlAcad Sci USA 2007, 104(11):4664–4669.


Neuroregenerative scaffolds for CNS repair.Delivery across the blood-brain barrier using Pluronics.


Focus Article

Mesoporous silica-basednanomaterials for drug delivery:evaluation of structural propertiesassociated with release rateMaria Strømme,∗ Ulrika Brohede,1 Rambabu Atluri1 and Alfonso E.Garcia-Bennett1

We present here a study of the controlled release of amino acid-derived amphiphilicmolecules from the internal pore structure of mesoporous nanoparticle drugdelivery systems with different structural properties, namely cubic and hexagonalstructures of various degrees of complexity. The internal pore surface of thenanomaterials presented has been functionalised with amine moieties througha one-pot method. Release profiles obtained by conductivity measurementsare interpreted in terms of specific structural and textural parameters of theporous nanoparticles, such as pore geometry and connectivity. Results indicatethat diffusion coefficients are lower by as much 4 orders of magnitude in two-dimensional structures in comparison to three-dimensional mesoporous solids.A fast release in turn is observed from mesocaged materials AMS-9 and AMS-8,where the presence of structural defects is thought to lead to a slightly lowerdiffusion coefficient in the latter. We conclude that the use of single or mixedphases of these porous systems can be utilized to provide sustained release overlong time periods and expect their use in a variety of formulations. 2008 John Wiley & Sons, Inc. Wiley Interdiscipl. Rev. Nanomed. Nanobiotechnol. 2009 1 140–148

INTRODUCTION

Drug delivery systems (DDS) are used to facilitatethe delivery of pharmaceuticals or other thera-

peutics (below called drugs). They are devices createdby combining knowledge of formulation technologyand material science with knowledge of chemicalinteractions and biological processes. Traditional DDSusually administer drugs with an initial burst of drugrelease. As the amount of drug falls below the mini-mum effective serum concentration, the therapeuticeffect is lost. In order to keep the concentrationwithin the therapeutic range, several administrations

Additional Supporting Information may be found in the onlineversion of this article.∗Correspondence to: Maria Strømme, Department ofEngineering Sciences, The Angstrom Laboratory, Upp-sala University, Box 534, SE-751 21 Uppsala, Sweden.E-mail: [email protected] of Engineering Sciences, The Angstrom Laboratory,Uppsala University, Box 534, SE-751 21 Uppsala, Sweden.

DOI: 10.1002/wnan.013

are usually required. Therefore, to avoid frequentadministration, controlled DDS are needed. Such sys-tems can either speed up or slow down the in vivodrug uptake. Systems providing sustained, extendedor immediate release, as well as depot delivery, fastdissolving and chewable tablets are all variations ofrate-controlling technologies that might be used tohelp deliver drugs. There is no ‘one-type-fits-all’ drugdelivery technology that fulfils every need or desire todeliver a drug in a particular way.

In the sustained-release category drug deliveryvehicles based on polymer and polysaccharidematrixes, coated reservoirs and other traditionaldissolution or diffusion controlled systems1 are totallydominating the market. However, as stated by oneof the editors of Modern Drug Discovery ‘Innovativepharmaceutical treatments require innovative meth-ods of administration’,2 and therefore innovativeDDS.

In general, the smaller the size of the drug-carrying vehicle, the better is the absorption in thebody. With new therapeutic agents such as proteins,


WIREs Nanomedicine and Nanobiotechnology Mesoporous silica-based nanomaterials for drug delivery

nucleic acids and poorly soluble drugs, new nanotypecarriers may offer not-yet-realized possibilities to drugadministration.3

One of the most promising categories ofnanoparticles and microparticles for which an internalordered pore structure may be tailored to hostdifferently sized drug molecules as well as to createpredetermined release profiles are the mesoporous,amorphous silica–based materials. Such materialshave been shown to be able to incorporate highdosages of drugs in the internal pore system4 owingto their large internal surface areas and pore volumesas well as their tuneable pore sizes. It has alreadybeen shown that a sustained drug release over severaldays can be obtained using mesoporous particles.5 Inthat seminal study, a mesoporous material, MCM-41,with cylindrical pore geometry was used to host andrelease ibuprofen, with the drug release over 3 days.These materials offer further interesting strategies formore precise control of the chemical surface propertiesdetermining both the drug release and transportproperties of the carrier. These may include surfacefunctionalisation and geometrical design of the poresfor specific drug candidates.

Recently, mesoporous silicas with controllednanoparticle size have been prepared using anionicamino acid–derived amphiphiles and alkoxysilane co-structure-directing agents (CSDAs), denoted AMS-nmesoporous silicas.6–8 Briefly, the synthesis mecha-nism of AMS-n mesoporous materials is thought tobe governed by the interaction of the CSDA withthe self-assembling (micellar) amphiphiles. Using well-established sol–gel techniques, a silicate wall is thencondensed surrounding the micellar–CSDA complex,and a mesoporous solid is eventually achieved throughthe preferential calcination (or solvent extraction) ofthe organic amphiphile from the condensed silica par-ticle. The micellar ‘liquid crystal’ phase of the anionicamphiphile is hence reproduced or templated into sil-ica form. The use of CSDAs has the double advantage,as it not only acts as a structural directing agent but,if the amphiphile is removed through solvent extrac-tion, it also leads to a homogeneous coating of organicfunctional groups on the internal surface of the pores.A schematic diagram of the synthesis mechanism andthe final material composition is shown in Figure 1.

In our investigations of the silicate particles asDDS, we have utilised the incorporated amphiphilicmolecules–which are anchored via electrostatic inter-actions to the internal surface of the mesopores–asthe model drugs. Considering the synthesis mech-anism of AMS-n materials, we can obtain a verywell defined and consistent (reproducible) interac-tion between model drug and particle surface for

CO2H

NCH3−(CH2)10

HCO2H

3NH SiO

OO

SiO

OOO

N-Lauroylglutamic acid

co-structuredirecting agent,

APES

Amphiphile/Drugrelease

silicate-amphiphilemesoporous material

micellar complex

TEOS silica source

O

+

FIGURE 1 | Schematic representation of the self-assemblymechanism of AMS-n mesoporous materials and the use of the aminoacid–derived amphiphile, N-lauroyl-glutamic acid, as a model drug. Forcomparison purpose, the use of the amphiphile as a model drug ensurescomplete drug loading and complete electrostatic interaction of themodel drug with the internal porous surface.

all the structures in this report, as well as certaintyof the position of the amphiphilic molecule withinthe porous matrix. Furthermore, the loading amountof the molecule is predetermined from the synthesismixture and can be confirmed through conventionalthermogravimetric analysis. It has already been shownthat the incorporation of model drugs molecules canbe achieved by direct synthesis through three dis-tinct routes: (1) by utilizing the hydrophobic core ofthe amphiphile micelle in order to solubilise non-polar molecules;9 (2) through electrostatic interactionwith the hydrophilic headgroup; and (3) by a simpleaddition to the synthesis gel leading to inhomoge-neous incorporation and loss of porous structure.10

Post-synthetic methods in which functionalisation ofinternal pore surfaces has been utilized are also a usefuland facile alternative for drug loading. In such a case,the ideal confirmation is a homogeneous function-alisation of the entire internal pore surface togetherwith a stoichiometric concentration of drug moleculesover such a surface.3 The use of amino acid–derivedamphiphilic molecules in this study allows us to probethe release from such ideal confirmations, enabling usto for the first time study release from a variety ofperiodic porous structures.

In this study mesoporous nanoparticles andmicroparticles forming mesocaged cubic AMS-8(Fd3m), tetragonal AMS-9 (P42/mnm), cylindrical,bicontinuous cubic AMS-6 (Ia3d) and hexagonalAMS-3 (p6mm) structures have been prepared andcharacterised extensively using N-lauroyl glutamic


Focus Article www.wiley.com/wires/nanomed

acid (C12-Glut) and N-lauroyl alanine (C12-Ala)anionic amphiphiles in combination with amino-propyl triethoxysilane (APES) as CSDA as previ-ously described in the literature.11 The mesoporousstructures AMS-6 and AMS-3 are analogous tothe well-known MCM-48 and MCM-41 materi-als. A recent study of the immunological responseof AMS-6 and AMS-8 particles showed low tox-icity profiles on monocyte-derived dendritic cells,and the particles provided a promising approachworth further development into drug/vaccine deliverysystems.12

The aim of the present work is to evaluate howthe control over structural properties such as poregeometry, connectivity (two- or three-dimensional)and pore wall chemistry, within the context ofmesoporous nanoparticles and microparticles, can beused to achieve well-defined, controlled release overextended time periods. An extended version of thepresent work will be published elsewhere.13

In molecular terms, the release process may beconsidered as a random walk of molecules inside thepore system of the carrier. The first 60% of releasecurves has been shown to be adequately described bya semi-empirical power-law expression:14–18

Q = a + b · tk (1)

for various carrier symmetries such as planar,cylindrical and spherical. Here Q is the amount ofmolecules released per unit exposed area of the carrier,t denotes time and a, b and k are constants.

The power-law function, Eq. (1), is related to theWeibull function that has been suggested as a universaltool for describing release from both Euclidian andfractal systems, and may be considered as a short-time approximation of the latter.19 The constant atakes initial delay and burst effects into account, andb is a kinetic constant.20 The power-law exponent, k,also called the transport coefficient, characterises thediffusion process, and equals 0.5 for ordinary CaseI–or carrier controlled–diffusion in systems for whichno swelling of the carrier material occurs.21

Diffusion-controlled release from a planarsystem, in which the carrier structure is inert, may bedescribed by the Higuchi square-root-of-time law:14

Q =√

DeffCs(2Cm − εCs) · t (2)

Here Deff is the effective diffusion coefficient of thedrug inside the carrier, and Cm is the total amountof drug present in the carrier per unit volume, ε

is the total porosity of the carrier, defined as thevolume fraction of pores when the drug material

has been removed (here referred to as the porosityof the calcined samples), and Cs is solubility of thedrug in the release medium. For model drugs that areamphiphiles, the critical micelle concentration (CMC)can be treated as the solubility limit for the mono-molecularly dispersed species, especially when theaggregation number of the micelles is high19 as isthe case for the amphiphiles used as model drugs inthe present work.22

For detailed expressions for Deff, ε, Cm, andthe total outer particle area Atot exposed to theliquid in the release measurements see SupplementaryInformation 1 (SI1).

EXPERIMENTAL

Synthesis of mesoporous nanoparticlesThe general synthesis of AMS-n mesoporous materialshas been reported previously.7 For details on thesynthesis conditions used in the present work, seeSupplementary Information 2 (SI2).

Structural, textural and compositionalcharacterisationN2 adsorption–desorption isotherms were measuredat −196 ◦C on an ASAP2020 Micromeritics Instru-ment. The calcined mesoporous materials were out-gassed for a period of 6 h at 200 ◦C and 0.3 kPapressure. The BET specific surface area23 was evalu-ated from the adsorption data in the relative pressurerange from 0.05 to 0.3. The total volume of intra-particle pores Vtot was extracted from the adsorptiondata as the amount adsorbed at a relative N2 gas pres-sure of 0.9. This upper pressure limit was chosen inorder to exclude the inter-particle porosity contribu-tion to the pore volume. Pore size distribution curveswere derived from desorption curves using the non-local density functional theory (NLDFT) method24

assuming a spherical model of cages.Scanning electron microscopy (SEM) images

were recorded using a LEO 1550 SEM, equipped witha Schottky field-emisson gun. The SEM was operatedat 3 kV and at magnifications of between 20,000 and50,000×.

Thermogravimetric (TG) analyses were per-formed on a Mettler TGA instrument by heatingthe samples from 25 to 900

oC at a heating rate of

10oC/min on an alumina holder under the flow of air

at 20 mL/min. From these data, the particle composi-tion as well as the amount of amphiphilic moleculesreleased could be assessed.



Release studied by conductivitymeasurementsRelease measurements were performed on three to fivesamples each from the four types of as-synthesised(un-calcined) mesoporous particle types under studyin 20 mL of distilled water as dissolution medium byusing ac conductivity measurements as implementedby the alternating ionic current (AIC) technique.25

The water was kept at a temperature of 65 ± 3 ◦Cduring the entire release measurement by keeping themeasuring cell in an incubator (incucell IC 55, BMTa.s., Brno, Czechnia). This elevated temperature wasused because of the rather poor solubility of alanine atroom temperature. The experimental set-up used wasidentical to that described elsewhere,23 except that thecubic measuring cell was somewhat larger (4 × 4 × 4cm3) and equipped with stainless steel electrodescovering two opposite surfaces of the cell. The cellwas covered by a lid to prevent water evaporationduring the measurements. A 12-mm magnet, rotatingat approximately 60 rpm, stirred the liquid during theentire measurement. A function generator (HP 3325A)applied an alternating voltage (1VRMS, 10 kHz) to theelectrodes of the cell, and the conductance of thecell was calculated from measurements of the voltageacross the cell and the current that passed throughit using two digital multimeters (Agilent 34401A).The number of molecules released was thus assessedfrom the conductance measured across the dissolutionmedium. This experimental set-up minimises delayscaused by transport times to the measurement site,and therefore enhances the temporal resolution ofthe measurement. In order to assess the CMC ofthe two types of amphiphilic molecules used (C12Glutand C12Ala), conductivity versus concentration curveswere recorded at the same conditions as for the releaseexperiments. The CMC value was then extracted inthe usual manner: as the concentration for which adiscontinuity in the slope of the curve appeared.

RESULTS AND DISCUSSION

Analysis of the synthesised materialThe N2 adsorption–desorption isotherms are dis-played in Figure S3.1 of Supplementary Information3 (SI3), together with insets showing the pore sizedistribution as obtained using the NLDFT approachassuming cylindrical and slit-type pore geometries (forcage structures). The textural properties of the samplesare shown in Table 1. As evident from the adsorp-tion isotherm curves, the relative pressure of 0.9 usedfor the total mesopore volume calculation is locatedwell below the region where capillary condensation in

inter-particle voids occurs for all particle types understudy. The total pore volume for sample AMS-9 isconsiderably smaller than that reported previously inthe literature and possesses a smaller pore size. Thismay be due to pore-blocking of cavities within thetetragonal structure. This effect has been observedpreviously and may be due to prolonged hydrother-mal treatment.11 Smaller cage/cavity windows mayinfluence the release rate considerably, providing ageometrical inhibition for the release of the moleculesunder study, and hence this sample was deemed ade-quate within the scope of this project.

Analysis of SEM images show that all syn-thesized particles are approximately spherical andfairly mono-dispersed (Figure 2). The AMS-3 sampleprepared in this project, however, contains some rod-shaped particles of approximately the same diameter(but with a somewhat larger length) as the sphericalones (Figure 2(a)). The radii used for the assessmentof the drug-releasing particle area, Atot, needed toobtain Q, were obtained by extracting the parti-cle sizes from more than 10 SEM images of eachsample. The radii are presented in Table 1, and,as can be seen, AMS-8 and AMS-9 samples con-sist of microparticles, while for AMS-3 and AMS-6the particle radii are ∼ 200 and 160 nm, respec-tively.

Analysis of the release processThe amphiphile CMC values (Cs in Eqs (2) and (S1.1))obtained from conductivity measurements at 65 ◦C areshown in Table S1.1 in Supplementary Information1 (SI1). As expected, the measured CMC values65 ◦C are somewhat higher than corresponding valuespreviously obtained at 40 ◦C.26,27

TG data normalized at 100 ◦C for an AMS-8 sample before and after a release experimentare displayed in Figure S3.2 in SupplementaryInformation 3 (SI3). All samples show TG curvescharacterised by an initial weight loss below ∼ 100 ◦Cowing to the loss of adsorbed moisture and solvent.Three further distinctive weight losses are seen ataround 200 (step I), 400 (step II) and 600 ◦C (stepIII). The latter step has been previously associatedwith dehydration of framework silanol groups andis quite prominent in all materials synthesised inthis study, with an average loss of 10 wt%. Theloss in step II takes place over a broad temperaturerange (250–450 ◦C) and can be associated with thedecomposition of both the anionic amphiphile andamino-functional groups bound to the pore wall. Thedecomposition of the surface-bound functional groupsat higher temperatures than the ‘free’ functionalgroup (measured at 200 ◦C for APES) is consistent



TAB

LE1

Text

ural

and

Surf

acta

ntRe

leas

eD

ata

ofth

eA

MS-

nM

esop

orou

sM

ater

ials

Und

erSt

udy

AMS-

nPo

reSu

rface

Tota

lPor

eAv

erag

eU

nrea

cted

Reac

ted

Amph

iphi

leb

Out

erPa

rticl

eε

Cm

Def

f

Diam

eter

aAr

eaVo

lum

ePa

rtic

leAP

ESb

APES

b(T

G,%

)Su

rface

Area

(m2/g

)(a

mph

iphi

les/

(cm

2/s

)

(DFT

,A)

(BET

,V

tot

Radi

us(T

G,%

)(T

G,%

)A

c tot

m3)

m2/g

)(c

m3/g

)(S

EM,n

m)

Cylin

dric

alhe

xago

nalA

MS-

341

.055

9.3

0.56

120

0±

153.

665.

7326

.15

10.6

7±

1.5

0.60

28.

45×

1026

1.37

×10

−17

±0.0

9×

10−1

7

Cylin

dric

alcu

bic

AMS-

652

.052

0.0

0.56

916

0±

132.

856.

5519

.27

12.7

8±

2.7

0.58

26.

459

×10

262.

22×

10−1

4

±0.1

2×

10−1

4

Cage

dcu

bic

AMS-

833

.154

7.6

0.32

111

35±

110

3.60

8.98

13.7

61.

73±

0.32

0.43

13.

985

×10

264.

05×

10−1

2

±0.3

2×

10−1

2

Cage

dte

trag

onal

AMS-

916

.641

9.0

0.19

630

45±

175

5.47

7.1

25.0

70.

68±

0.07

0.34

97.

024

×10

267.

71×

10−1

1

±0.5

1×

10−1

1

a Bas

edon

acy

lindr

ical

pore

geom

etry

for

AM

S-3

and

AM

S-6

and

asp

heri

calp

ore

geom

etry

for

AM

S-8

and

AM

S-9.

bO

btai

ned

from

TG

mea

sure

men

tsof

as-s

ynth

esis

edpa

rtic

les.

c Obt

aine

dfr

omT

Gan

dSE

Mda

taus

ing

Eq.

(S1.

2)



(a) (b)

(c) (d)

FIGURE 2 | SEM images of calcined samples, with scaling bar lengthwithin parenthesis, of (a) AMS-3 (1 µm), (b) AMS-6 (1 µm), (c) AMS-8(2 µm) and (d) AMS-9 (2 µm), showing the variation in particlemorphology and particle size.

with other published works on amino and otherfunctional groups.28 The weight loss in step Iis therefore considered here to be due to thedecomposition of the aminopropyl groups nottaking part in the co-assembly process with theamphiphilic molecule. A plot of absolute values ofthe first derivative of the weight loss with respectto temperature is also included in Figure S3.2 ofSupplementary Information 3. The amounts of APESand amphiphile present in the as-synthesised and thereleased particles have been estimated by subtractingthe weight loss in step II from the total weight of APESadded to the synthesis gel. In Table 1, the amount ofAPES and amphiphiles present in the as-synthesisedsamples as well as the weight loss estimated from step Iare summarized, together with the Atot and Cm valuescalculated from Eq. (S1.2) and (S1.4), respectively(Supplementary Information 1).

Figure 3 shows the initial part of the typicalrelease curves for the different mesoporous structuresunder study. For clarity, log-lin plots of the full releasecurves have been included in the figure inset. The Qvalues on the vertical axis have been obtained fromTG analyses of each released sample and by usingthe Atot values (divided by 10 since 100 mg, not 1g, was used in the release experiments) in Table 1.From release curves similar to those in Figure 3, aneffective diffusion coefficient for the release processwas extracted according to Eq. (S1.1). The results aresummarized in Table 1.

None of the mesoporous particles released thetotal amount of amphiphilic molecules hosted in thestructure during the time span of a release experiment.

The cage-type AMS-8 and AMS-9 released 53 ± 5%and 51 ± 2%, respectively, of the total amphiphilecontent during the first 8 h of the release experiment. Itshould be noted that the significantly different valuesof Q reached at the end of the release experimentfor the AMS-8 and AMS-9 samples shown in Figure3 does not imply that very different amounts ofamphiphiles have been released from these structures;Q is the released amount of molecules normalizedby Atot, and this area differs by a factor of ∼ 2.5between the structures (Table 1). There appearsto be very little steric constraint imposed on thediffusion of molecules by the presence of a cageconnecting window for both of these structures.Although there are considerable differences in thetextural properties, these two structures are verysimilar crystallographically; both are assumed to formfrom cubic and tetragonal close packing of sphereswith intricate 3D connectivities. The AMS-8 structureis known to be composed of two networks of cages ofdifferent sizes, while the AMS-9 structure is thoughtto be composed of three networks of cages withdifferent sizes and has lower symmetry than AMS-8.8,9 Adsorption isotherm data measured for AMS-9shows low textural properties, which might haveresulted from a prolonged hydrothermal treatment.Furthermore, most samples of AMS-8 contain acertain amount of structural defects, the sampleprepared here being no exception as determined byTEM.13 All these factors appear to have little effect onthe amount released after 8 h; however, the diffusioncoefficient for AMS-9 is approximately one order ofmagnitude higher than that of AMS-8, which couldbe attributed to presence of stacking fault defectsin AMS-8 and the resulting more intricate system ofcages (higher tortuosity). Unexpectedly, AMS-9 showsthe fastest diffusion coefficient of all mesoporousstructures studied here.

For the cylindrical type AMS-3 and AMS-6 structures, the corresponding amount releasedwas 6.3 ± 1.5% and 39 ± 2% after about 57 and18 h, respectively. When comparing the diffusioncoefficients for C12-Ala in the AMS-3 and AMS-6structures, we find that Deff in AMS-3 is extremelylow and more than 3 orders of magnitude smaller thanin the AMS-6 structure. This significant differencebetween AMS-3 and AMS-6 can be explained interms of the connectivity of the two cylindricalpore structures, since textural properties, morphologyand amount of functionalisation are all in thesame range (Table 1). The 2D hexagonal structure(AMS-313) shows the slowest diffusion coefficient ofall AMS-n materials presented here, while all 3Dcubic structures (AMS-6 and AMS-813) show larger



1020

1019

1018

AMS-3

AMS-6

AMS-8

AMS-9

00

2000 4000 6000 8000 10,000

Time(s)

3 1020

2 1020

1 1020

0 5 1041 1051.5105 2 105

Q (

surf

acta

nts/

m2 )

FIGURE 3 | Selected release curves for the various types of AMS-nstructures under study. The inset shows a log-lin plot of the samecurves.

diffusion coefficients and larger amounts released perunit time.

Clearly, the diffusion coefficients of C12-Glu inthe AMS-8 and AMS-9 structures are several ordersof magnitude larger than those for C12-Ala in theAMS-3 and AMS-6 structures. For comparison, thepore wall surface structures of mesoporous materialsAMS-3, AMS-6 and AMS-8 are shown in Figure 4.The intertwined bicontinous surface of AMS-6 andthe 3D connectivity in the structure of AMS-8 areworth noting. From these models, it is possible toappreciate how a larger amount may be releasefrom cage structures. There are numerous connectionsbetween cages and more connections to the externalsurface, which, although increase the tortuosity of thecage structure, may contribute largely to the higherdiffusion coefficients calculated for these structures.In the case of the cylindrical pore systems, thereis no connection between cylindrical channels, anddiffusion is limited by the number of openings tothe external surface. The nature of the interactionbetween two molecules C12-Glu and C12-Ala and therespective silica surface is clearly different, since C12-Glu contains two acid moieties in the head group,while C12-Ala contains only one. Functional groups

on the surface of the mesopores can be consideredas anchors for the drug molecules in a potential drugdelivery system. TG analysis suggests that the molarratio of surfactant to APES in the final material isclose to 1:0.3, 1:0.4, 1:1 and 1:0.4 for the AMS-3,AMS-6, AMS-8 and AMS-9 structures, respectively.These values appear to support our findings for thecage containing structures, inasmuch as a higherconcentration of surface functional groups wouldresult in a slower release rate or Deff, in particularif the model drug molecule contains two anchoringsites. However, this argument does not explain theapparent differences in Deff between the cage andcylindrical structures or, indeed, between the 2D and3D cylindrical systems studied here.

CONCLUSION

In conclusion, a detailed study of the controlled releaseachieved from mesoporous materials containingcylindrical and cage pore systems of different texturaland morphological properties has been conducted.Amino acid–derived amphiphilic molecules, used hereas model drug molecules to probe the release, were

FIGURE 4 | Representation of the pore wall surfacestructures of mesoporous materials with 2D hexagonalcylindrical pores (AMS-3 left), 3D bicontinous cubiccylindrical pores (AMS-6, middle) and 3D cubic cage-typepores (AMS-8 right). Pores sizes and unit cells are not toscale and shown for comparison only.



loaded directly into the mesoporous structures via theco-assembly of micelles with amino silanes and a sil-ica source, ensuring that for all structural systems theamphiphile is electrostatically bound to the internalsurface of the pores. This provides an ideal modellingtool for the investigations of the structure–functionrelation with respect to the potential use of meso-porous particles as drug carriers and in DDS. It isclear from this work that the connectivity (2D or 3D)of the pores is the key to controlling the release ratewith an increase of several orders of magnitude indiffusion coefficient from 3D porous networks withsimilar textural properties. From this study, it canalso be concluded that pore geometry can play animportant role in controlling release rates, and wehave shown release curves from cage-type orderedmesoporous systems determining that the degree ofstructural defects may contribute towards a slowerdiffusion coefficient, and that the presence of cageconnecting windows may play an unexpected part inspeeding up the release of molecules from the internalcavities as exemplified by the higher diffusion coeffi-cients in the cage structures AMS-8 and AMS-9. Ourresults do not indicate any clear correlation between

electrostatic binding and release rate and suggest thatthe control of structural properties such as connectiv-ity or pore geometry may be a much more efficientstrategy in order to control the release rate. The 2Dhexagonal structure of AMS-3 showed the slowestdiffusion coefficient of all mesoporous materials stud-ied here as well as the smallest amount of aminefunctionalisation.

Likewise, samples with particle sizes below 500nm show a lower release rate than those with largerparticle size, although it is not possible to draw signifi-cant conclusions at this stage with respect to the effectof particle size on release rates from ordered meso-porous materials. This will form part of our futurework.

Overall, we have shown the large, potential useof mesoporous materials with cylindrical and cage-type structures as controlled drug release systemsusing amino acid–derived amphiphilic molecules asmodel drugs. It is clear that the use of single phase ormixed phases of these porous systems can be utilizedto provide sustained release over long time periodsand we expect their use in a variety of formulations.

NOTES

We are grateful to Prof. Osamu Terasaki (Stockholm University) for useful discussions and use of the electronmicroscope. We thank Prof. Andreas Fischer (Kungliga Tekniska H&ogskolan) for access to XRD facilities. TheSwedish Research Council, the Knut and Alice Wallenberg Foundation, the G&oran Gustafsson Foundationand the Swedish Foundation for Strategic Research are acknowledged for their financial support of this project.One of the authors (MS) is a Swedish Royal Academy of Sciences Research Fellow and would like to thank theAcademy for their support.

REFERENCES

1. Jantzen GM, Robinsson JR: In: Banker GS, RhodesCT, eds. Modern Pharmaceutics. 4th ed, Chap 15.New York: Marcel Dekker; 2002.

2. Vogelson CT. Advances in drug delivery systems. ModDrug Discov 2001, 4:49.

3. Vallet-Regı M. Ordered mesoporous materials in thecontext of drug delivery systems and bone tissue engi-neering. Chem Eur J 2006, 12:5934.

4. Kresge CT, Leonowicz ME, Roth WJ, Vartuli CJ,Beck JS. Ordered mesoporous molecular sieves synthe-sized by a liquid–crystal template mechanism. Nature1992, 359:710.

5. Vallet-Regı M, Ramila A, del Real RP, Perez-Pariente J.A new property of MCM-41: Drug delivery system.Chem Mater 2001, 13:308.

6. Che S, Garcia-Bennett AE, Yokoi T, Sakamoto K,Kunieda H, et al. A novel anionic surfactant templatingroute for synthesizing mesoporous silica with uniquestructure. Nat Mater 2003, 2:801.

7. Garcia-Bennett AE, Che S, Terasaki O, Tatsumi T.Structural investigations of AMS-n mesoporous mate-rials by transmission electron microscopy. Chem Mater2004, 16:813.

8. Garcia-Bennett AE, Kupferschmidt N, Sakamoto Y,Che S, Terasaki O. Synthesis of mesocage structures bykinetic control of self-assembly in anionic surfactants.Angew Chem Int Ed 2005, 44:5317.

9. Kataokaa K, Harada A, Nagasaki Y. Block copolymermicelles for drug delivery: design, characterization andbiological significance. Adv Drug Deliv Rev 2001,47:113.



10. Coutinho D, Madhugiri S, Balkus KJ Jr. Synthesis andcharacterization of organosilane functionalized DAM-1mesoporous silica. J Porous Mater 2004, 11:239.

11. Garcia-Bennett AE, Che S, Miyasaka K, Terasaki O.Structural solution of mesocaged material AMS-8.Chem Mater 2004, 16:3597.

12. Vallhov H, Gabrielsson S, Strømme M, Scheynius A,Garcia Bennett AE. Mesoporous silica particles inducesize dependent effects on human dendritic cells. NanoLett 2007, 7:3576.

13. Brohede U, Atluri R, Garcia-Bennett AE, Strømme M.Sustained release from mesoporous nanoparticles: Eval-uation of structural properties associated with releaserate. Curr Drug Deliv 2008, 5:177–185.

14. Higuchi T. Mechanism of sustained-action medication:theoretical analysis of rate of release of solid drugsdispersed in solid matrices. J Pharm Sci 1963, 52:1145.

15. Bonny JD, Leuenberger H. Matrix type controlledrelease systems: I. Effect of percolation on drug dis-solution kinetics. Pharm Acta Helv 1991, 66:160.

16. Korsmeyer RW, Gurny R, Doelker E, Buri P, Pep-pas NA. Mechanisms of solute release from poroushydrophilic polymers. Int J Pharm 1983, 15:25.

17. Kim H, Fassihi R. Application of a binary polymer sys-tem in drug release rate modulation .1. Characterizationof release mechanism. J Pharm Sci 1997, 86:316.

18. Rinaki E, Valsami G, Macheras P. The power law candescribe the ’entire’ drug release curve from HPMC-based matrix tablets: A hypothesis. Int J Pharm 2003,255:199.

19. Kosmidis K, Argyrakis P, Macheras P. Fractal kineticsin drug release from finite fractal matrices. J Chem Phys2003, 119:6373.

20. Jamzad S, Tutunji L, Fassihi R. Analysis of macro-molecular changes and drug release from hydrophilicmatrix systems. Int J Pharm 2005, 292:75.

21. Ritger PL, Peppas NA. A simple equation for descrip-tion of solute release. I. Fickian and non-fickian releasefrom non-swellable devices in the form slabs, spheres,cylinder or discs. J Controlled Release 1987, 5:23.

22. Tanford C. Micelle shape and size. J Phys Chem 1972,76:3020.

23. Brunauer S, Emett PH, Teller E. Absorption of gasesand multimolecular layers. J Am Chem Soc 1938,60:309.

24. Olivier JP, Conklin WB, Szombathely MV. Determina-tion of pore size distribution from density functionaltheory: A comparison of nitrogen and argon results. In:Ronquerd J, Rodregaez-Keiso F, Sing KSW, Unger KK,eds. Characterization of porous solids III. Proceed-ings of the IUPAC Symposium, Vol. 87, Amsterdam:Elsevier, 1994, 81.

25. Frenning G, Ek R, Strømme M. A new method for char-acterizing the release of drugs from tablets in low liquidsurroundings. J Pharm Sci 2002, 91:776.

26. Yoshida R, Takehara M, Sakamoto K. Surfactantsderived from amino acids. I. Critical micelle concen-trations of the salts of optically active and racemicN-acylamino acids. J Jpn Oil Chem Soc 1975, 24:538.

27. Sakamoto K. In Surfactant Science Series, Vol. 101,Protein-Based Surfactants (eds Xia, J. & Nnanna, I.A.)Chap. 10, New York: Dekker; 2001, 261–280.

28. Dıaz I, Perez-Pariente J. Synthesis of spongelike func-tionalized MCM-41 materials from gels containingamino acids. Chem Mater 2002, 14:4641.


Multifunctional polymeric micelles as cancer-targeted, MRI-ultrasensitive drug delivery systems.Therapeutic nanoreactors: combining chemistry and biology in a novel triblock copolymer drug delivery system.Lyotropic systems for drug delivery.


reviews - nano medicine and nano biotechnology, vol.1, issue 1 (2009)

Documents