Science of the Total Environment 468–469 (2014) 968–976

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Particle size, surface charge and concentration dependent ecotoxicity ofthree organo-coated silver nanoparticles: Comparison between generallinear model-predicted and observed toxicity

Thilini Silva a,1, Lok R. Pokhrel a,1,2, Brajesh Dubey b,⁎, Thabet M. Tolaymat c, Kurt J. Maier a, Xuefeng Liu d

a Department of Environmental Health, College of Public Health, East Tennessee State University, Johnson City, TN 37614, United Statesb Environmental Engineering Program, School of Engineering, University of Guelph, 50 Stone Road East, Guelph, Ontario, Canadac USEPA, Office of Research and Development, National Risk Management Laboratory, 26 West Martin Luther King Drive, Cincinnati, OH 45224, United Statesd Department of Biostatistics and Epidemiology, College of Public Health, East Tennessee State University, Johnson City, TN 37614, United States

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Nanotoxicity predicted based on particleproperties for three organo-coated AgNPsagainst Escherichia coli and Daphniamagna.

• Particle size, surface charge, and concen-tration dependent AgNP toxicity wereobserved for both organisms.

• General linear model showed interac-tive effects of primary particle size andsurface charge which explained AgNPtoxicity.

⁎ Corresponding author. Tel.: +1 519 824 4120x52506E-mail address: bdubey@uoguelph.ca (B. Dubey).

1 These first co-authors contributed equally to this wor2 Current address: US Environmental Protection Agenc

0048-9697/$ – see front matter © 2013 Elsevier B.V. All rihttp://dx.doi.org/10.1016/j.scitotenv.2013.09.006

a b s t r a c t

a r t i c l e i n f o

Article history:Received 13 July 2013Received in revised form 2 September 2013Accepted 2 September 2013Available online xxxx

Editor: Damia Barcelo

Keywords:Organo-coated silver nanoparticlesGeneral linear modelNanotoxicologyNanoparticle characteristicsEscherichia coliDaphnia magna

Mechanismunderlying nanotoxicity has remained elusive. Hence, efforts to understandwhether nanoparticle prop-erties might explain its toxicity are ongoing. Considering three different types of organo-coated silver nanoparticles(AgNPs): citrate-coated AgNP, polyvinylpyrrolidone-coated AgNP, and branched polyethyleneimine-coated AgNP,with different surface charge scenarios and core particle sizes, herein we systematically evaluate the potential roleof particle size and surface charge on the toxicity of the three types of AgNPs against two model organisms,Escherichia coli and Daphnia magna. We find particle size, surface charge, and concentration dependent toxicity ofall the three types of AgNPs against both the test organisms. Notably, Ag+ (as added AgNO3) toxicity is greaterthan each type of AgNPs tested and the toxicity follows the trend: AgNO3 N BPEI-AgNP N Citrate-AgNP N PVP-AgNP. Modeling particle properties using the general linear model (GLM), a significant interaction effect of primaryparticle size and surface charge emerges that can explain empirically-derived acute toxicitywith great precision. Themodel explains 99.9% variation of toxicity in E. coli and 99.8% variation of toxicity in D. magna, revealing satisfactorypredictability of the regression models developed to predict the toxicity of the three organo-coated AgNPs. We an-ticipate that the use of GLM to satisfactorily predict the toxicity based on nanoparticle physico-chemical

.

k.y, National Health and Environmental Effects Research Laboratory, 200 SW 35th St., Corvallis, OR 97333, United States.

ghts reserved.

969T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

characteristics could contribute to our understanding of nanotoxicology and underscores the need to consider po-tential interactions among nanoparticle properties when explaining nanotoxicity.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Envisioned as a promising technology touted to offer greater benefitstomankind, nanotechnology has emerged as a cross-disciplinary scienceof the twenty first century encouraging collaboration among interdisci-plinary scientists in a way never witnessed before (Pokhrel and Dubey,2012). Progress in tuning engineered nanomaterial (ENM) functionalityfor desired applications has extended ENMs' fields of applications(Costanza et al., 2011), and therefore, a rapid commercialization ofnano-enabled products is taking place (www.nanotechproject.org;Tolaymat et al., 2010; National Academy of Sciences, 2012). As nano-enabled products can release chemical(s) in ‘nano’ and/or ionic forminto the environment (Impellitteri et al., 2009; Benn et al., 2010;Tolaymat et al., 2010), concerns over environmental contaminationand subsequent hazard to the receptor organisms have often been raised(Klaine et al., 2008; Vecitis et al., 2010; Yamashita et al., 2011; Pokhrelet al., 2012; Pokhrel and Dubey, 2012, 2013a; Rahaman et al., 2012). Be-cause aquatic systems can be regarded a major sink for environmentalcontaminants, systematic investigation of the potential aquatic toxicityof the most common nanomaterial types, such as silver nanoparticles(AgNP), may provide insights into if and how nanoparticles wouldcause toxicity should the exposure occur.

The mechanistic basis of AgNP toxicity to the biotic receptors—bothprokaryotic and eukaryotic organisms—has remained less clear as hasbeen the case for other (in)organic nanomaterials (Vecitis et al., 2010;El Badawy et al., 2011; Yamashita et al., 2011; Pokhrel et al., 2012;Barceló et al., 2013). Nonetheless, recent evidence of oxidative stressvia generation of reactive oxygen species (Choi and Hu, 2008), directphysical contact leading to membrane perturbation or cell-pitting(Choi and Hu, 2008; Fabrega et al., 2009; El Badawy et al., 2011), andDNA damage (Ahamed et al., 2008) have been documented in the liter-ature following exposure to AgNPs. Less well understood is whethernanoparticles or the associated ionic form is more toxic than the other(Xiu et al., 2012; Pokhrel and Dubey, 2013a), and whether there isany combined effects of the two forms (Pokhrel et al., 2012). Identifyingfactors enabling our understanding of nanoparticle toxicity has longremained challenging (Xiu et al., 2012; Pokhrel et al., 2013, in review;Pokhrel and Dubey, 2013b). Unclear is how nanoparticle characteristicswould interact with the biologic receptor characteristics at the nano-biointerface, as such interaction can potentially influence the toxicity(Vecitis et al., 2010; El Badawy et al., 2011).

Amongst the factors identified to mediate nanoparticle toxicity, pri-mary particle size has generally remained central (Choi and Hu, 2008;Jiang et al., 2008; Park et al., 2011); however, it is beginning to be under-stood that particle dissolution into toxic ions, particle state of aggregation,other transformations that could potentially occur once nanoparticlesenter different environments, and surface charge/coating agent couldalso influence the toxicity and therefore studies suggest consideringthem during toxicity assessment (Liu and Hurt, 2010; El Badawy et al.,2011; Lowry et al., 2012; Pokhrel et al., 2012, in review; Tejamaya et al.,2012; Xiu et al., 2012).

Citrate, polyvinylpyrrolidone (PVP), and branched polyethyleneimine(BPEI) represent, in part, the most commonly employed coating/stabilizing agents (Tolaymat et al., 2010) as they enable effective nano-particle dispersion (El Badawy et al., 2012); whilst citrate has beenwidely used as a reductant for nanoparticle synthesis (El Badawyet al., 2012; Pokhrel et al., 2012). In this study, not only that these coat-ing materials differentially charge AgNPs surface, they also confer sta-bility via distinctly different mechanisms, namely, electrostatic (forcitrate-coated AgNPs), steric (for PVP-coated AgNPs), and electrosteric(for BPEI-coated AgNPs). Different primary particle sizes are obtained

employing different methods of synthesis (El Badawy et al., 2012). Wesystematically evaluate AgNP toxicity using the three different typesof organo-coated AgNPs: Citrate-AgNP, PVP-AgNP, and BPEI-AgNP,and characterize for hydrodynamic diameters (HDD), particlemorphol-ogy (i.e., diameter and shape using transmission electron microscopy(TEM)), state of aggregation, ion release rate, solution pH, and zeta (ζ)potential including that of the biologic receptor surfaces. Our resultsdemonstrate particle size, surface charge, and concentration dependenttoxicity of the three different organo-coated AgNPs against both theprokaryotic (Escherichia coli) and eukaryotic (Daphnia magna) organ-isms. Using the regressionmethod of general linearmodel (GLM), here-in we demonstrate for the first time a significant interactive effect ofprimary particle size (TEM diameter) and surface charge to satisfactori-ly explain acute toxicity of the three different organo-coated AgNPsagainst both the test organisms. Notably, our GLM shows an associationbetween a minimum set of AgNP properties and the biologic responsesin E. coli and D. magna.

2. Materials and methods

2.1. Organo-coated silver nanoparticle synthesis and characterization

This study considers three different organo-coated silver nano-particles (AgNPs): citrate-coated AgNP (Citrate-AgNP), polyvinylpyrroli-done-coated AgNP (PVP-AgNP), and branched polyethyleneimine-coated AgNP (BPEI-AgNP), presenting different surface charge scenariosand particle sizes which enabled us to study the potential main effectsof the particle size and surface charge, including their interactions, onAgNP toxicity. These AgNPs were synthesized as described previouslyby El Badawy et al. (2010) (detailed in Supplementary Information),and purified using a tangential flow filtration (TFF) system equippedwith 10 kD hollow fiber polysulfone membranes (www.spectrumlabs.com). The purification protocol was previously described in detail(Pokhrel et al., 2012; Pokhrel and Dubey, 2012). The NPs were well-characterized as follows: the hydrodynamic diameter (HDD) and zeta(ζ) potential were measured using the dynamic light scattering (DLS)and phase analysis light scattering (PALS), respectively, by a NICOMPpar-ticle sizer/zeta potential unit (PSS NICOMP Particle Sizing Systems, CA);the surface plasmon resonances were recorded as absorbance spectrausing an UV/vis spectrophotometer (HACH DR 5000, HACH Company,CO); the particles were imaged using a Transmission electronmicroscope(TEM, Philips EM 420) and analyzed for particle size and shape; and thesuspension pHwas determined using a pHmeter. Sampleswere digestedusing ultrapure HNO3 (Method 3050B) (USEPA 1996), and total Ag con-centration was quantified using a Graphite furnace-atomic absorptionspectrometry (GF-AAS, Varian Spectra 220Z) following the USEPAmethod 7010 (USEPA, 2007). For comparison, ion-specific toxicity wasconducted using AgNO3 (Fisher Scientific, Cat. # S486-100).

For nano-bio interactions to commence, the NPs are expected tohave some degree of physical contact with the biologic surface whichoccurs due to attractive or repulsive forces (Vecitis et al., 2010; ElBadawy et al., 2011). Estimating organisms' surface charge (as ζ poten-tial) in tandem with particles' surface charge in the test matrix shouldoffer insight into the strength of interaction at the nano-bio interface(Vecitis et al., 2010; El Badawy et al., 2011; Pasquini et al., 2012). To de-termine ζ potential of D. magna body surface, ten b 24 h old neonateswere cleaned several times with moderately hard water (MHW), care-fully removed the gastrointestinal tract under a dissecting microscope,washed three times with MHW, then the integument (skin) wasresuspended in MHW and disintegrated using a tissue homogenizer

970 T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

(Cole-Parmer Ultrasonic Homogenizer, 125 w/cm2, 25 kHz, Cat. # R-04717-00) before measuring the ζ potential using PALS. For E. coli, thecells (as recommended by the supplier) were reconstituted in MHWfor 15 min, followingwhich its surface chargewas recorded using PALS.

2.2. Toxicity bioassays

Two different toxicity bioassays very well representing aquatic tox-icitywere performed, namely, E. coli bioassay andD. magna 48 h surviv-al assay.

2.2.1. Escherichia coli bioassayThe potential of three different organo-coated AgNPs to inhibit

β-galactosidase enzyme activity was evaluated in E. coli. Chlorophenol-red β-galactopyranoside (CPRG) used as a chromogenic substrateis catalyzed at the glycosidic bond by β-galactosidase, forminggalactopyranose and chlorophenol red as the reaction products (Bittonet al., 1994; Pokhrel et al., 2012). The details of this bioassay have beende-scribed in our previous publication (Pokhrel et al., 2012). Briefly, 100 μL ofreconstituted bacteria was added to 900 μL of the sample and incubatedat 35 °C for 90 min in a glass test tube. Then 200 μL of this suspension(sample with bacteria) was transferred to a 96 well plate and added100 μL of CPRG to each well, following which the well plate was incubat-ed at 35 °C for another 90 min; thus the total exposure durationwas 3 h.The bacteria were exposed to a wide range of concentrations of organo-coated AgNPs (0.01–41.2 mg/L as total Ag), following which enzyme ac-tivitywas recorded at 570 nmusing amicroplate reader. TheMHW(con-ductance 560 μS cm−1, hardness 280 mg/L as CaCO3) and 1 mg Cu2+/L(as CuSO4) represented negative and positive controls, respectively(Pokhrel et al., 2012). At least triplicate samples were analyzed for eachsample dilution, including the controls. All test solutionsweremaintainedin a narrow pH range of 7.0–7.2. AgNO3 was also used to test ion-specifictoxicity. To eliminate the potential effect of coating material on AgNPtoxicity, tests were performed for each coating material separately (i.e.,sodium citrate dihydrate, PVP, and BPEI) at the highest theoretical con-centration present in the nanosuspension. The bacterial reagent andCPRG were obtained from M2B Research & Innovative Technologies,LLC, Gainesville, FL.

2.2.2. Daphnia magna mortality bioassayA static, non-renewable D. magna 48 h bioassay was performed fol-

lowing the standard USEPA guidelines, including the culture and main-tenance of the daphnids (USEPA, 1987). Briefly, ten b 24 h old neonateswere introduced randomly into each 50 mL test beaker containingAgNP treatment (0.01–40.3 μg/L as total Ag) or the control. As the neg-ative and positive controls, MHW and CuSO4 were used, respectively.Animals were maintained at 20 ± 1 °C with a 16 h photoperiod cycleand were unfed during the test. Triplicate test runs were performedfor each sample concentration. Total dead daphnids in each test beakerwere recorded at the end of the test.

2.3. Analysis of dissolved silver in AgNP samples

The dissolved Ag fraction in three different organo-coated AgNPsamples was determined by incubating a 50 mL sample at 20 ± 1 °C ina centrifuge tube (Fisher Scientific, Cat. # 06-443-18) and maintaininga 16 h photoperiod cycle for 48 h—the condition similar to D. magnatest described above. Soon the samples were centrifuged for 30 min at~3150 g (4000 rpm, Thermo Electron, IEC Centra CL3 Series Centrifuge),then 5 mL supernatant was removed for digestionwith HNO3 (ultrapuregrade) following the method 3050B and analyzed in duplicates forthe Ag amount released from each AgNP sample using a GF-AAS(Pokhrel andDubey, 2012). Typical sample analysis consisted of the sam-ple blanks, internal standards, spiked samples, and sample duplicates;the recovery of Ag was in the range 96–101%.

2.4. Statistical analysis

As data satisfied normal probability distribution (Kolmogorov–Smirnov test, p N 0.1 in all cases), they were used untransformed. TheEC50 (i.e., effective concentration for 50% enzyme activity inhibition inE. coli) or LC50 (i.e., lethal concentration that kills 50% of D. magna neo-nates) values were estimated using the linear regression analysis. Com-parison of the means of EC50 or LC50 among different treatments wasperformed using the ANOVA followed by the Dunnett t test (2-tailedpost-hoc) for multiple comparisons.

The only association that has been quantitatively described previ-ously in “nano” environmental health and safety (EHS) studies hasbeen the use of univariate correlation method (Vecitis et al., 2010;Diedrich et al., 2012; Ma et al., 2012), explaining whether (or not) agiven physico-chemical property (e.g., particle size, electron structure)of the model ENM correlates with the observed toxicity. Not pursuedpreviously, but the important questions addressed in this work arewhether interactions exist between the physico-chemical characteris-tics of the three organo-coated AgNPs and how such interactions(should those occur) could explain AgNP toxicity against both the pro-karyotic and eukaryotic organisms. Using the general linear model(GLM), herein we address these important questions by investigatingthe potential main effects of TEM diameter (TEMdia) and surface charge(measured as ζ potential), and the potential interactive effects of TEMdia

and ζ potential on the toxicity of the three different organo-coatedAgNPs. TEMdia representing the primary particle size and ζ potentialmeasured as a function of surface charge were used as the covariatesin the linear model. The GLM-predicted toxicity (as LC50/EC50) of thethree types of AgNPs, whichwas quantified using the parameters, TEM-dia and ζ potential, is presented and comparedwith their experimentallyderived LC50 or EC50 values. Any unexplained variance of each model isdescribed by its respective error term (ε), and the precision of the pre-dictive power of eachmodel was determined by the coefficient of deter-mination (R2). Statistical analyses were performed using the PASW(a.k.a. SPSS) Statistics 18.0 (PASW, 2009).

3. Results and discussion

The physico-chemical characteristics of the TFF-purified three differ-ent organo-coated AgNP samples used in this study are summarized inTable 1. These NPs had primary particle size (TEMdia) within the100 nm size range, with variable average TEM diameter enabling usfor size-dependent toxicity to be evaluated (Table 1). Not only thatthe three different, yet commonly used, organo-coatings imparted sta-bility to AgNPs by three distinct stabilization mechanisms as previouslystated, they also offered variable surface charge scenarios (Table 1),hence allowing us for assessing the potential interactive roles of surfacecharge and primary particle size on AgNP toxicity against both the pro-karyotic (E. coli) and eukaryotic (D.magna) organisms. All three types oforgano-coated AgNPs had similar pH and hydrodynamic diameter(HDD) (Table 1), hence excluding their potential roles on AgNP toxicityin this study. The representative TEM images, particle size distributions,and the characteristic surface Plasmon resonance spectra of the AgNPsamples are presented in the Supplementary Information (Fig. S1).

To evaluate the potential toxicity in E. coli upon exposure to AgNPs,we measured inhibition of β-galactosidase (β-Gal), an extensively con-served metabolic enzyme found in plants, fungi, and animals, includingthe humans (Taron et al., 1995). In animals, β-Gal catalyzes the conver-sion of complex sugars (e.g., lactose) to simple sugars (e.g., glucose andgalactose) that are utilized as substrates in cellular respiration (Hansenand Gitzelmann, 1975). While in plants, β-Gal activity has been widelyassociated with ontogenic changes in the tissues such as seeds,seedlings, pollens, and fruits (Figueredo et al., 2011). A significantconcentration-dependent β-Gal activity inhibition was observed in E.coli for all the three types of organo-coated AgNPs, including for thefree Ag+ (as AgNO3) (Fig. 1). At comparable concentrations, BPEI-

Table 1Characteristics of the evaluated organo-coated AgNPs for the toxicity studies.

Material pH Particle size distribution (nm) Particle circularity Average zeta potential (mV) Mechanism of stabilization

Hydrodynamic diametera

Mean ± S.D.TEM diameterMean ± S.D.

BPEI-AgNP 7.1 Before:10.9 ± 0.8After: 10.9 ± 0.8

10.0 ± 4.6(n = 63)

0.87 +28.8 Electrosteric

Citrate-AgNP 7.2 Before:10.9 ± 0.8After: 11.0 ± 0.7

56.0 ± 14.0(n = 98)

0.88 −20.08 Electrostatic

PVP-AgNP 7.0 Before:11.0 ± 0.7After: 11.0 ± 0.9

72.0 ± 24.0(n = 57)

0.87 −7.49 Steric

a Volume-weighted hydrodynamic diameter measured in the test matrix (moderately hard water) using the DLS method before and after the toxicity tests were conducted; particlecircularity of 1 indicates that the particle is a perfect circle in a 2D TEM imagery; PVP-AgNP, polyvinylpyrrolidone-coated AgNPs; Citrate-AgNP, citrate-coated AgNP; BPEI-AgNP, branchedpolyethyleneimine-coated AgNPs. ImageJ 1.44 program was used to analyze particle size distributions and circularity of the AgNPs from the representative TEM imageries.

971T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

AgNPs exhibited significantly greater enzyme activity inhibition amongthe tested AgNP types, while free Ag+ (as AgNO3) showed significantlygreater toxicity than all types of AgNPs evaluated (Fig. 1). In fact, Ag+

(as AgNO3) was, on average, 2.5, 6.4, and 16.4 times acutely inhibitorythan BPEI-AgNPs, Citrate-AgNPs, and PVP-AgNPs, respectively; the β-Galinhibition followed the trend: AgNO3 N BPEI-AgNP N Citrate-AgNP N

PVP-AgNP.Similar to E. coli bioassay, a concentration-dependent mortality of

D. magna neonates was observed in the 48 h bioassay for all types ofAgNPs evaluated, including for the free Ag+ (as AgNO3). The neonatalmortality also followed the same toxicity trend as observed for E. coliassay presented above. Among the three organo-coated AgNP types,BPEI-AgNPs caused significantly higher mortality of the daphnids,while PVP-AgNPs resulted in least toxicity at the comparable concentra-tions (Fig. 2). It was the free Ag+ (as added AgNO3) that led to thehighest mortality of daphnids compared to the AgNPs tested; on aver-age, Ag+ was 1.1, 8.0, and 13.3 times more lethal than BPEI-AgNPs,

Fig. 1. Concentration-dependent inhibition of β-galactosidase activity, compared to negative coAg+ ions (as AgNO3). Error bars represent ±1 standard deviation of the triplicate runs. PVP-Abranched polyethyleneimine-coated AgNPs.

Citrate-AgNPs, and PVP-AgNPs, respectively. The respective EC50 orLC50 values of the organo-coated AgNPs for both the test organismsare presented in Table 2.

Because the TEMdia, not the HDD (Pearson r = −0.27, p = 0.483),correlated significantly with ζ potential (Pearson r = −0.877, p =0.002), and to avoid any redundancy in the GLM, TEMdia and ζ potentialwere used as the covariates in the linear models. Ma et al. (2012) haverecently shown TEMdia, not the HDD, as a good predictor of organo-coated AgNP dissolution into Ag ions, a premise consistent to our anal-ysis. In the current work, we documented for the first time a significantmain effect of primary particle size (TEMdia) and surface charge, and asignificant interaction effect of primary particle size and surface charge,explaining empirically-derived acute toxicity correctly using the GLM(Tables 2 and 3). As the intercept term is not included in the model bythe model selection procedure, this means that if the particle size andsurface charge are zero, the toxicity in E. coli and D. magnawould be se-curely anchored at zero and other factorswould not significantly impact

ntrol, in Escherichia coli exposed to the three types of organo-coated AgNPs, including thegNP, polyvinylpyrrolidone-coated AgNPs; Citrate-AgNP, citrate-coated AgNPs; BPEI-AgNP,

Fig. 2. Concentration-dependentmortality ofDaphniamagna exposed to the three types of organo-coated AgNPs, including the Ag+ ions (as AgNO3), as shown by 48 h test bioassay. Errorbars represent ±1 standard deviation of the triplicate samples. PVP-AgNP, polyvinylpyrrolidone-coated AgNPs; Citrate-AgNP, citrate-coated AgNPs; BPEI-AgNP, branchedpolyethyleneimine-coated AgNPs.

972 T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

the toxicity. Our models explaining the toxicity of AgNPs in E. coli andD. magna are presented in Eqs. (1) and (2), respectively.

Escherichia coli EC50 ¼ 37:599� TEMdia–17:222� ζþ 1:476 TEMdia � ζð Þ þ εi ð1Þ

Daphnia magna LC50 ¼ 0:077� TEMdia–0:031� ζþ 0:002 TEMdia � ζð Þ þ εii ð2Þ

where TEMdia denotes TEM diameter, ζ represents zeta potential, and εiand εii are the respective error terms of the models representing any

Table 2General linearmodel and parameter estimates showing themain and interactive effects ofTEM diameter (TEMdia) and zeta potential on the toxicity of AgNPs (used as EC50 values, adependent variable in the model) against Escherichia colia.

Dependent variable:EC50

Type III sumof squares

Meansquares

F p

Source

Model 1.465E7 4.883E6 2373.454 b0.0001TEMdia 6.202E6 6.202E6 3014.463 b0.0001Zeta potential 2.896E5 2.896E5 140.771 b0.0001TEMdia × Zeta potential 9.515E5 9.515E5 462.510 b0.0001Error 1.234E4 2.057E3

Predictor Coefficient B Std. error t p

TEMdia 37.599 0.685 54.904 b0.0001Zeta potential −17.222 1.452 −11.865 b0.0001TEMdia x Zeta potential 1.476 0.069 21.506 b0.0001

a The coefficient of determination (R2) of the model was 0.999, suggestive of goodmodel fit with sufficient predictive power.

variance unaccounted for by each model. The parameter estimatesof the models are presented in Tables 2 and 3. The coefficients of deter-minations (R2) of the linear models explaining the toxicity in E. coli andD. magna are 0.999 and 0.998, respectively, suggestive of goodmodel fitwith sufficient predictability of the models. A schematic summarizesthese results in Fig. 3.

We calculated the percentage precision of our GLM to predictnanotoxicity (as LC50 or EC50) using the covariates, TEMdia and zetapotential, for all the three types of AgNPs (Table 4). Comparison ofGLM-predicted toxicity to that of experimentally-derived toxicity forboth the model organisms is presented in Fig. 3 and Table 4. Resultsshowed that the precision of the linear model developed to predict

Table 3General linearmodel and parameter estimates showing themain and interactive effects ofTEM diameter (TEMdia) and zeta potential on the toxicity of AgNPs (used as LC50 values, adependent variable in the model) against Daphnia magna.a

Dependent variable:LC50

Type III sum ofsquares

Mean squares F p

Source

Model 94.373 31.458 1207.334 b0.0001TEMdia 26.014 26.014 998.396 b0.0001Zeta potential 0.920 0.920 35.296 0.001TEMdia × Zeta potential 1.445 1.445 55.474 b0.001Error 0.156 0.026

Predictor Coefficient B Std. error t p

TEMdia 0.077 0.002 31.597 b0.0001Zeta potential −0.031 0.005 −5.941 0.001TEMdia × Zeta potential 0.002 0 7.448 b0.001

a The coefficient of determination (R2) of the model was 0.998, suggestive of goodmodel fit with sufficient predictive power.

Fig. 3. Schematic showing the significant main effects and the significant interactive effects of primary particle size (TEM diameter) and surface charge explaining empirically-derivedacute toxicity in E. coli and D. magna using the general linear model (GLM). Bar graph to the right shows concentration-, particle size-, and surface charge-dependent toxicity of thethree different organo-coated AgNPs. PVP-AgNP, polyvinylpyrrolidone-coated AgNPs; Citrate-AgNP, citrate-coated AgNPs; BPEI-AgNP, branched polyethyleneimine-coated AgNPs.

973T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

nanotoxicity of the three different organo-coated AgNPs was in therange 99.8–100% for E. coli. For D. magna survival assay, some degreeof deviation recorded in % precision of predicting toxicity by the GLMcan be attributed to the higher sensitivity of zooplanktons to AgNPs,resulting in narrow differences in the experimentally-derived LC50values (Table 4), which is in good agreement with the smaller coeffi-cient values of the predictors (i.e., TEMdia, zeta potential) including ofthe interaction term (i.e., TEMdia × zeta potential) as noted in Table 3.

Potential energy barrier, which has been hypothesized to act as alimiting factor (Vecitis et al., 2010; El Badawy et al., 2011), betweenthe biologic surface and AgNP needs to be removed before the particlescould chemically interact with the receptor molecules within the cellenvelope (cell wall and/or cell membrane) and/or with the cellular con-tent following permeation of nanoparticle into the cell (El Badawy et al.,2011), which may likely be governed by its nano-size or material-specific characteristics (Pokhrel et al., 2012). Applying the single-chainmean field (SCMF) theory, studies have modeled the potential interac-tion of rod shaped biomolecules, including the carbon nanotubes, withthe phospholipid bilayer and suggested that the change in surface pat-terning, and therefore the energy associated with the surface, of therod shaped objects may influence, or rather enhance, object-cell mem-brane association, and perhaps cell entry (Pogodin et al., 2011, 2012),and subsequently the toxicity. Fig. 4 strongly supports primary particlesize and surface charge dependent toxicity as predicted by our modelsfor the three types of AgNPs against both the test organisms. Ouranalysis of the magnitude of charge difference between the nano-

Table 4Impact of surface charge on the toxicity of organo-coated AgNPs to Escherichia coli and Daphni

Nanoparticle Zeta potentiala (mV) Magnitude of charge

AgNP (A) E. coli (B) D. magnac (C) E. coli |A–B| D. m

BPEI-AgNP +28.8 39.8 31.3Citrate-AgNP −20.08 −11.0 −2.55 9.08 17.5PVP-AgNP −7.49 3.51 4.9

Note greater difference in charge indicates higher attraction (or lower repulsion) whereaspolyvinylpyrrolidone-coated AgNPs; Citrate-AgNP, citrate-coated AgNP; BPEI-AgNP, branched

a Zeta potential measured in the test matrix.b Magnitude of charge difference is taken as an absolute value.c Daphnia integument suspended in MHW (detailed in Materials and methods section, see S

bio interfaces, revealing higher attraction forces between (each of)the biologic receptors’ surface and the BPEI–AgNP surface, mighthave enabled potential physical contact between BPEI–AgNP andthe receptor organism, leading to higher toxicity (Tables 5). Like-wise, the lowest charge difference as observed for PVP-AgNPs sug-gests that the dominant repulsive forces (lower attraction) mighthave played a key role in keeping PVP-AgNPs away from the biologicsurfaces and this may explain the lowest observed toxicity (Table 5).A previous study has shown surface charge-dependent toxicity oforgano-coated AgNPs on the physiological activity and mortality inGram positive Bacillus species, demonstrating BPEI-AgNPs as themost toxic among the types of AgNPs evaluated (El Badawy et al.,2011). The authors attributed higher toxicity of BPEI-AgNPs to thegreater charge difference between the surface of Bacillus cell walland the AgNPs, which was large enough to overcome the energy(electrostatic) barrier at the nano-bio interface, therefore leadingto higher toxicity. This finding is consistent with our results(Fig. 4A,B). However, the toxicity trend for other two types ofAgNPs (i.e., Citrate-AgNP and PVP-AgNP) observed in the previousstudy (El Badawy et al., 2011) was modified in the present study;this suggests that the disparity in toxicity observed could be largelydue to the differences in the sensitivity and/or surface charge of thetypes of test organisms used including that of the Citrate-AgNPs inthese two studies (El Badawy et al., 2011). In good agreement withour findings for Citrate-AgNP and PVP-AgNP, Kennedy et al. (2010)had observed similar toxicity trend for these two organo-coated

a magna.

differenceb Strength of interaction Expected toxicity Observed toxicity

agna |A–C|

5 Higher attraction Higher Higher3 Lower attraction Medium Medium4 Higher repulsion Lower Lower

lower difference in charge indicates higher repulsion (or lower attraction). PVP-AgNP,polyethyleneimine-coated AgNPs.

upporting Information).

Fig. 4. Primary particle size (TEM diameter)- and surface charge (as ζ potential)-dependent toxicity of organo-coated AgNPs to (A) E. coli and (B) D. magna. Two barswith different lettersindicate significant difference between the two treatments at the p = 0.01 level. Comparison of the means for E. coli EC50 and D. magna LC50 among different treatments was performedusing the one-way ANOVA followed by Dunnett t test (2-tailed posthoc) for multiple comparisons. E. coli ζ potential was−11 mV and that of D. magnawas−2.5 mV; details of surfacecharge are presented in Table 4. PVP-AgNP, polyvinylpyrrolidone-coated AgNPs; Citrate-AgNP, citrate-coated AgNPs; BPEI-AgNP, branched polyethyleneimine-coated AgNPs. NRWQC de-notes USEPA national recommended water quality criterion for fresh water system (USEPA, 2009). The lines are meant to guide the eyes.

974 T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

AgNPs against D. magna and Pimephales promelas. Results of theother studies also indicate a potential for direct physical interactionbetween the nanoparticles (e.g., various fullerene derivatives) andbiologic receptors (but not the effect of reactive oxygen speciesbeing formed), which could likely be an important mechanism ofnanotoxicity (Ali et al., 2004; Tang et al., 2007).

Potential oxidation of AgNPs under aerobic experimental conditions,as in this study, has been shown to promote AgNP dissolution into Agions (Xiu et al., 2012). Our analysis showed more dissolved Ag ions re-leased from PVP-AgNP suspension (3.74 μg/L), rather than from BPEI-AgNP (3.17 μg/L), or Citrate-AgNP (2.39 μg/L) suspension. Because theconcentrations of dissolved Ag were in the range 33–52 times lowerthan the EC50 value for Ag+ (as AgNO3) in E. coli, the observed toxicityof the types of AgNPs may not be associated with the aerobically

Table 5Comparison of the general linearmodel (GLM)-predicted toxicity (LC50/EC50) versus experimenDaphnia magna.

Nanoparticle Daphnia magna

Experimental LC50 ± S.D. (μg/L) GLM-predicted

LC50 (μg/L) % Preci

BPEI-AgNP 0.41 ± 0.1 0.45 109.7Citrate-AgNP 2.88 ± 0.08 2.69 93.4PVP-AgNP 4.79 ± 0.25 4.71 98.3

GLM, general linear model; % Precision = (GLM predicted LC50 or EC50 / Experimental LC50 orAgNP; BPEI-AgNP, branched polyethyleneimine-coated AgNPs.

released dissolved Ag ion emanating from the AgNP suspensions; thissuggests negligible contribution of dissolved Ag ion to the observed tox-icity of AgNPs. Our analysis of HDD of the three types of AgNPs evaluat-ed revealed higher stability of the particles in the test matrix (i.e.,moderately hardwater) during the test periods as no change inHDD oc-curred (Table 1), confirming an absence of aggregation or settling ofparticles in both the test media; thus particle aggregation was not a fac-tor accounting for the observed differences in toxicity among the typesof AgNPs evaluated against both the test organisms. Employing chemi-cally different, yet commonly used, coating agents during synthesis ofthe three types of AgNPs contributed to the colloidal particle stabilityvia different stabilization mechanisms (see Table 1) by hindering parti-cle aggregation, and enabled us to acquire desired particle propertiessuch as size, surface charge, and shape of the synthesized AgNPs (El

tally-derived toxicity of the three types of organo-coated AgNPs against Escherichia coli and

Escherichia coli

Experimental EC50 ± S.D. (μg/L) GLM-predicted

sion EC50 (μg/L) % Precision

305 ± 33 305.08 100793 ± 71 791.63 99.8

2041 ± 5 2040.14 99.9

EC50) × 100. PVP-AgNP, polyvinylpyrrolidone-coated AgNPs; Citrate-AgNP, citrate-coated

975T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

Badawyet al., 2010). Because all the three types of organo-coatedAgNPsevaluated in this study did not appear morphologically distinct fromeach other in shape, as theywere roughly oval or spherical with narrowaverage circularity (Table 1; Supplementary Fig. S1), particle shapemaynot explain the observed differences in acute toxicity among the typesof organo-coated AgNPs. Because the potential toxicity of each coatingmaterial used in this study was negligible at the highest theoreticallyrelevant concentration (tri sodium citrate = 10 mM, PVP = 0.25%,and BPEI = 0.5 mM) against both the test organisms, it negates thepossibility that the coating material alone could have any contributionto the observed AgNP toxicity. This is in good agreement with the pub-lished literature using the same coating materials and similar modelspecies (El Badawy et al., 2011; Pokhrel et al., 2012).

Under section 304 of the Clean Water Act (CWA), water qualitycriteria are establishedwith an aim to protect human health and aquaticlife from the contaminants of concerns. Recently revised USEPA nationalrecommended water quality criterion (NRWQC) for fresh water systemis 3.2 μg Ag/L (USEPA, 2009). As our experimentally obtained or theGLM-predicted LC50 values against D. magna for BPEI-AgNP and Cit-rate-AgNP fall below the NRWQC threshold, this suggests that thedaphnid population might be vulnerable to AgNPs in the fresh watersystems. Recently, a study has shown that a low, environmentally rele-vant concentration of Citrate-AgNPs (2 μg Ag/L) could influencedaphnid migratory behaviors (i.e., horizontal and vertical migrationpatterns), survival, and reproductive potential in the presence of itscommon Odonate predator, Anax junius nymphs (Pokhrel and Dubey,2012). Growing demand for AgNPs as an antimicrobial agent, includingof other various kinds of nanomaterials (e.g., ZnO NP, TiO2 NP, CdSequantum dots, C60/C70 fullerenes), in hundreds of consumer productswill inevitably contaminate the aquatic systems, potentially affectingthe biota, and perhaps the ecosystems, therein (Lovern et al., 2007;Brausch et al., 2011; El Badawy et al., 2011; Pokhrel et al., 2012;Pokhrel and Dubey, 2012).

Overall, among the three types of organo-coated AgNPs evaluated inthis study, BPEI-AgNP was the most biocidal due to its smaller primaryparticle size, and greater charge differences between the NP surfaceand the biologic surface leading to higher attractive forces, potentiallypromoting the toxicity. Notably, however, Ag+ (as added AgNO3) wasthemost toxic of all the chemicals tested against E. coli, while the toxic-ity of free Ag+ (as AgNO3) and BPEI-AgNP did not differ significantlyagainst D. magna (p N 0.5; Fig. 3). Potential release of dissolved Agions from the evaluated AgNPs in our experimental scenarios was inad-equate to explain the observed toxicity. Likewise, particle state of aggre-gation, suspension pH, coatingmaterial alone, and particle shape did notcontribute to the toxicity of the AgNPs evaluated. This study, therefore,highlights the significance of considering various measureable physico-chemical characteristics of nanoparticles including the particle size, sur-face charge of nanoparticles including that of the receptors' surface, anda proper evaluation of the potential main and interaction effects of thefundamental characteristics of nanoparticles using the quantitativemodeling, such as the GLM applied in this work, could significantly con-tribute to our understanding of nanotoxicology.

Our choice of different non-toxic surface coatings, which rendereddifferent charge scenarios, on AgNPs surface also enabled the assess-ment of surface charge-dependent toxicity of AgNPs against both theprokaryotic and eukaryotic model organisms. The toxicity profiles de-veloped here for the three organo-coated AgNPs demonstrate the im-portance of selecting coating materials in tandem with particle sizeand surface charge while considering safer AgNPs for environmentalpurposes, or alternately antimicrobial AgNPs for biomedical applica-tions or for delaying membrane biofouling in water purification sys-tems. Current paradigm that the particle size and surface chargematter still prevails. Notably, however, the results of this study suggestprobing for the potential interaction effects amongst the differentphysico-chemical properties of the nanomaterials by deploying appro-priate quantitative modeling, thus offering a new perspective on the

way that the nanomaterial toxicity can be better explained or predictedwhen empirical toxicity data are lacking (Pokhrel et al. in review). Totest an assertion that a minimum set of properties of the well-characterized nanomaterials, as was previously proposed (Pasquiniet al., 2012) and clearly demonstrated in this study, could adequatelyexplain the potential toxicity of other types of nanomaterials andagainst other test organisms will require additional systematic and fo-cused studies.

Acknowledgments

This studywas supported in part by the East Tennessee State Univer-sity (ETSU) Research Development Council Grant# 82064 and the ETSUOffice of Research and Sponsored Programs Grant# 83003. The authorsthank TEMAnalysis Services Lab, TX for support with TEM characteriza-tion of NPs. This study has not been subjected to the US EPA internal re-view, and the opinions expressed are those of the authors and do notreflect that of the associated institutions. Any mention of the tradenames does not imply their endorsements or recommendations for use.

Appendix A. Supplementary data

Methods for organo-coated AgNP synthesis, TEM characterization,particle size distribution, and UV–vis spectra; plot showing particle size-dependent toxicity; and correlation between parameters of AgNPs.Supplementary data associated with this article can be found onlineat doi: http://dx.doi.org/10.1016/j.scitotenv.2013.09.006.

References

Ahamed M, Karns M, Goodson M, Rowe J, Hussain SM, Schlager JJ, et al. DNA damage re-sponse to different surface chemistry of silver nanoparticles in mammalian cells.Toxicol Appl Pharmacol 2008;233:404–10.

Ali SS, Hardt JI, Quick KL, Kim-Han JS, Erlanger BF, Huang TT, et al. A biologically effectivefullerene (C60) derivative with superoxide dismutase mimetic properties. Free RadicBiol Med 2004;37(8):1191–202.

Barceló D, Farré M, Bennett J, Hanson M. Nanomaterials in the environment. Sci Total En-viron 2013. [http://www.journals.elsevier.com/science-of-the-total-environment/virtual-special-issues/nanomaterials-in-the-environment/].

Benn T, Cavanagh B, Hristovski K, Posner JD, Westerhoff P. The release of nanosilver fromconsumer products used in the home. J Environ Qual 2010;39:1875–82.

Bitton G, Jung K, Koopman B. Evaluation of a microplate assay specific for heavy metaltoxicity. Arch Environ Contam Toxicol 1994;27:25–8.

Brausch K, Anderson TA, Smith PN, Maul JD. The effect of fullerenes and functionalizedfullerenes on Daphnia magna phototaxis and swimming behavior. Environ ToxicolChem 2011;30:878–84.

Choi O, Hu ZQ. Size dependent and reactive oxygen species related nanosilver toxicity tonitrifying bacteria. Environ Sci Technol 2008;42:4583–8.

Costanza J, El Badawy AM, Tolaymat TM. Comment on “120 years of nanosilver history:implications for policy makers”. Environ Sci Technol 2011;45:7591–2.

Diedrich T, Dybowska A, Schoot J, Valsami-Jones E, Oelkers EH. The dissolution rates ofSiO2 nanoparticles as a function of particle size. Environ Sci Technol 2012;46(9):4909–15.

El Badawy AM, Luxton TP, Silva RG, Scheckel KG, Suidan MT, Tolaymat TM. Impact of en-vironmental conditions (pH, ionic strength, and electrolyte type) on the surfacecharge and aggregation of silver NPs suspensions. Environ Sci Technol 2010;44(4):1260–6.

El Badawy AM, Silva RG, Morris B, Scheckel KG, Suidan MT, Tolaymat TM. Surfacecharge-dependent toxicity of silver nanoparticles. Environ Sci Technol 2011;45(1):283–7.

El Badawy AM, Scheckel KG, Suidan MT, Tolaymat TM. The impact of stabilization mech-anism on the aggregation kinetics of silver nanoparticles. Sci Total Environ 2012;429:325–31.

Fabrega J, Renshaw JC, Lead JR. Interactions of silver nanoparticles with Pseudomonasputida biofilms. Environ Sci Technol 2009;43:9004–9.

Figueredo SA, Lashermes P, Aragao FJL. Molecular characterization and functional analysisof the β-galactosidase gene during Coffea arabica (L.) fruit development. J Exp Bot2011;62(8):2691–703.

Hansen RG, Gitzelmann R. The metabolism of lactose and galactose. In: Jeanes A, Hodge J,editors. Physiological effects of food carbohydratesACS Symposium Series; 1975.p. 100–22.

Impellitteri CA, Tolaymat TM, Scheckel KG. The speciation of silver nanoparticles in antimi-crobial fabric before and after exposure to a hypochlorite/detergent solution. J EnvironQual 2009;38:1528–30.

Jiang W, Kim BYS, Rutka JT, Chan WCW. Nanoparticle-mediated cellular response issize-dependent. Nat Nanotechnol 2008;3:145–50.

976 T. Silva et al. / Science of the Total Environment 468–469 (2014) 968–976

Kennedy AJ, Hull MS, Bednar AJ, Goss JD, Gunter JC, Bouldin JL, et al. Fractionatingnanosilver: importance for determining toxicity to aquatic test organisms. EnvironSci Technol 2010;44:9571–7.

Klaine SJ, Alvarez PJJ, Batley GE, Fernandes TF, Handy RD, Lyon DY, et al. Nanomaterials inthe environment: behavior, fate, bioavailability, and effects. Environ Toxicol Chem2008;27(9):1825–51.

Liu J, Hurt RH. Ion release kinetics and particle persistence in aqueous nano-silver colloids.Environ Sci Technol 2010;44:2169–75.

Lovern SB, Strickler JR, Klaper R. Behavioral and physiological changes in Daphnia magnawhen exposed to nanoparticle suspensions (titanium dioxide, nano-C60, andC60HxC70Hx). Environ Sci Technol 2007;41:4465–70.

Lowry GV, Gregory KB, Apte SC, Lead JR. Transformation of nanomaterials in the environ-ment. Environ Sci Technol 2012;46:6893–9.

Ma R, Levard C,Marinakos S, Chen Y, Liu J, Brown Jr GE, et al. Size-controlled dissolution oforganic-coated silver nanoparticles. Environ Sci Technol 2012;46(2):752–9.

Method 3050B. Acid digestion of sediments, sludges, and soils. United States Environ-mental Protection Agency; 1996 [December].

Method 7010. Graphite furnace atomic absorption spectrometry. United States Environmen-tal Protection Agency; 2007 [February]. http://www.epa.gov/osw/hazard/testmethods/sw846/pdfs/7010.pdf.

National Academy of Sciences. A research strategy for environmental, health, and safety as-pects of engineered nanomaterials. http://www.nap.edu/catalog.php?record_id=13347, 2012.

Park MV, Neigh AM, Vermeulen JP, de la Fonteyne LJ, Verharen HW, Briedé JJ, et al. Theeffect of particle size on the cytotoxicity, inflammation, developmental toxicity andgenotoxicity of silver nanoparticles. Biomaterials 2011;32(36):9810–7.

Pasquini LM, Hashmi SM, Sommer TJ, Elimelech M, Zimmerman JB. Impact of surfacefunctionalization on bacterial cytotoxicity of single-walled carbon nanotubes. Envi-ron Sci Technol 2012;46:6297–305.

PASW (SPSS). Base 18.0 user's guide. Chicago, IL: SPSS Inc.; 2009.Pogodin S, Slater NKH, Baulin VA. Surface patterning of carbon nanotubes can enhance

their penetration through a phospholipid bilayer. ACS Nano 2011;5(2):1141–6.Pogodin S, Slater NKH, Baulin VA. Biomolecule surface patterning may enhance mem-

brane association. ACS Nano 2012;6(2):1308–13.Pokhrel LR, Dubey B. Potential impact of low-concentration silver nanoparticles on

predator–prey interactions between predatory dragonfly nymphs and Daphniamagna as a prey. Environ Sci Technol 2012;46(14):755–62.

Pokhrel LR, Dubey B. Evaluation of developmental responses of two crop plants exposedto silver and zinc oxide nanoparticles. Sci Total Environ 2013a;452–453:321–32.

Pokhrel LR, Dubey B. Untangling species sensitivity paradox in environmental research.Expert Opin Environ Biol 2013b;2(1):1–2.

Pokhrel LR, Dubey B, Silva T, El Badawy AM, Tolaymat TM, Scheuerman PR. Rapid screen-ing of aquatic toxicity of several metal-based nanoparticles using the MetPLATE™bioassay. Sci Total Environ 2012;426:414–22.

Pokhrel LR, Dubey B, Scheuerman PR. Evaluation of experimental design options in envi-ronmental nano-science research. Expert Opin Environ Biol 2013;2(1):1–6.

Pokhrel LR, Dubey B, Scheuerman PR. Natural water chemistry (dissolved organic carbon,pH, and hardness) modulates colloidal stability, dissolution, and antibacterial activityof citrate functionalized silver nanoparticles; 2013w [in review].

Project on Emerging Nanotechnologies Inventory. www.nanotechproject.org.Rahaman MdS, Vecitis CD, Elimelech M. Electrochemical carbon nanotube filter perfor-

mance towards virus removal and inactivation in the presence of natural organicmatter. Environ Sci Technol 2012;46:1556–64.

Tang YJ, Ashcroft JM, Chen D, Min G, Kim C-H, Murkhejee B, et al. Charge-associated ef-fects of fullerene derivatives onmicrobial structural integrity and central metabolism.Nano Lett 2007;7(3):754–60.

Taron CH, Benner JS, Hornstra LJ, Guthrie EP. A novel beta-galactosidase gene isolatedfrom the bacterium Xanthomonas manihotis exhibits strong homology to several eu-karyotic beta-galactosidases. Glycobiology 1995;5(6):603–10.

Tejamaya M, Römer I, Merrifield RC, Lead JR. Stability of citrate, PVP, and PEG coatedsilver nanoparticles in ecotoxicology media. Environ Sci Technol 2012;46(13):7011–7.

Tolaymat TM, El Badawy AM, Genaidy A, Schekel KG, Luxton TP, Suidan M. Anevidence-based environmental perspective ofmanufactured silver nanoparticle in syn-theses and applications: a systematic review andcritical appraisal of peer-reviewed sci-entific papers. Sci Total Environ 2010;408(5):999–1006.

United States Environmental Protection Agency. Procedures for conducting Daphniamagna toxicity bioassay; 1987 [EPA/600/8-87/011, March].

United States Environmental Protection Agency. National recommended water qualitycriteria. http://water.epa.gov/scitech/swguidance/standards/current/upload/nrwqc-2009.pdf, 2009.

Vecitis CD, Zodrow KR, Kang S, Elimelech M. Electronic structure-dependent bacterial cy-totoxicity of single-walled carbon nanotubes. ACS Nano 2010;4(9):5471–9.

Xiu Z, Zhang Q, Puppala HL, Colvin VL, Alvarez PJJ. Negligible particle-specific antibacterialactivity of silver nanoparticles. Nano Lett 2012;12(8):4271–5.

Yamashita K, Yoshioka Y, Higashisaka K, Mimura K, Morishita Y, Nozaki M, et al. Silica andtitanium dioxide nanoparticles cause pregnancy complications in mice. NatNanotechnol 2011;6:321–8.