Acta Biomaterialia 31 (2016) 301–311

Contents lists available at ScienceDirect

Acta Biomaterialia

journal homepage: www.elsevier .com/locate /actabiomat

Full length article

Recapitulating cranial osteogenesis with neural crest cells in 3-Dmicroenvironments

http://dx.doi.org/10.1016/j.actbio.2015.12.0041742-7061/� 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

⇑ Corresponding authors at: Department of Life Sciences, Imperial CollegeLondon, Silwood Park Campus Buckhurst Road, Ascot, Berkshire SL5 7PY, UnitedKingdom (A. Abzhanov). Demirci BAMM Labs, Canary Center at Stanford for EarlyCancer Detection, Department of Radiology, Department of Electrical Engineering(By courtesy), Stanford School of Medicine, Palo Alto, CA 94304, USA (U. Demirci).

E-mail addresses: a.abzhanov@imperial.ac.uk (A. Abzhanov), utkan@stanford.edu (U. Demirci).

1 These authors contributed equally to this work.2 Corresponding authors contributed equally.

Bumjin Namkoong a,b,1, Sinan Güven c,d,1, Shwathy Ramesan e, Volha Liaudanskaya c, Arhat Abzhanov a,f,g,⇑,2,Utkan Demirci c,e,⇑,2aDepartment of Organismic and Evolutionary Biology, Harvard University, Cambridge, MA 02138, USAbDepartment of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138, USAcDemirci BAMM Labs, Canary Center at Stanford for Early Cancer Detection, Department of Radiology, Department of Electrical Engineering (By courtesy),Stanford School of Medicine, Palo Alto, CA 94304, USAd Izmir Biomedicine and Genome Center, Dokuz Eylul University Health Campus, Balcova, 35350 Izmir, TurkeyeDemirci BAMM Labs, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USAfCurrent address: Department of Life Sciences, Imperial College London, Silwood Park Campus Buckhurst Road, Ascot, Berkshire SL5 7PY, United KingdomgCurrent address: Natural History Museum, Cromwell Road, London SW7 5BD, United Kingdom

a r t i c l e i n f o

Article history:Received 9 June 2015Received in revised form 11 November 2015Accepted 2 December 2015Available online 7 December 2015

Keywords:3-D cultureControlling 3-D tissue microenvironmentEctodermal to mesodermal differentiationBone

a b s t r a c t

The experimental systems that recapitulate the complexity of native tissues and enable precise controlover the microenvironment are becoming essential for the pre-clinical tests of therapeutics and tissueengineering. Here, we described a strategy to develop an in vitro platform to study the developmentalbiology of craniofacial osteogenesis. In this study, we directly osteo-differentiated cranial neural crestcells (CNCCs) in a 3-D in vitro bioengineered microenvironment. Cells were encapsulated in thegelatin-based photo-crosslinkable hydrogel and cultured up to three weeks. We demonstrated that thisplatform allows efficient differentiation of p75 positive CNCCs to cells expressing osteogenic markers cor-responding to the sequential developmental phases of intramembranous ossification. During the courseof culture, we observed a decrease in the expression of early osteogenic marker Runx2, while the othermature osteoblast and osteocyte markers such as Osterix, Osteocalcin, Osteopontin and Bone sialoproteinincreased. We analyzed the ossification of the secreted matrix with alkaline phosphatase and quantifiedthe newly secreted hydroxyapatite. The Field Emission Scanning Electron Microscope (FESEM) images ofthe bioengineered hydrogel constructs revealed the native-like osteocytes, mature osteoblasts, and cra-nial bone tissue morphologies with canaliculus-like intercellular connections. This platform provides abroadly applicable model system to potentially study diseases involving primarily embryonic craniofacialbone disorders, where direct diagnosis and adequate animal disease models are limited.

� 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction are regulated by the multitude of chemical and physical stimuli

Tissues and organs comprise cells and intercellular componentsthat are highly organized at the organismic level, which have adynamic, three-dimensional (3-D) nature especially highlightedduring embryonic development. Tissue functionality arises fromthe finely orchestrated interactions of all such elements, which

[1–4]. Tissue engineering and regenerative medicine hold a greatpromise to create models that mimic the natural biological tissuemicroenvironments and recapitulate tissue and organ formationandmaturation. Cell-encapsulating 3-D tissuemodels are emergingtools that bring inspiring in vitro venues to study the developmentof complex physiological systems. These models overcome the lim-itations of the two-dimensional (2-D) monolayer cultures in study-ing the significantly complex biological mechanisms involved inmulticellular behavior [5,6]. Microscale tissue models can effi-ciently utilize rare cells that are challenging for monolayer expan-sion [7]. Such microscale models are capable of recapitulatingregulatory dynamics of the germ layers and cellular condensationsforming at the early developmental stages. The design and fabrica-tion of 3-D microenvironments with well-defined architecture and

Fig. 1. Recapitulation of the osteogenic differentiation of CNCCs with direct differentiation in the 3-D hydrogels in vitro. Neural crest cells were isolated from E10.5 mouseembryo and initially expanded in monolayer with neural crest-specific media and later on encapsulated in hydrogels forming 3-D in vitro culture. Osteogenic differentiationphases with specific gene expressions in 3-D microenvironments were monitored for three weeks.

302 B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311

material compositions specific for progenitor cell types areessential for mimicking intricate events taking place in vivo.Cell-encapsulating hydrogel microenvironments are fabricated bygelation of extracellular matrix mimicry polymer solutions usingvarious techniques including thermal, enzymatic, chemical andUV photo-crosslinking methods. In particular, the flexibility tofine-tune the microscale tissue size and shape, together with thehigh throughput and reproducibility aspect of the UV photo-crosslinking platform, underscores its broad utility to createsuccessful tissue models [8,9].

Often referred to as the ‘‘fourth embryonic germ layer” [10,11],neural crest is a transient, multipotent embryonic population ofcells, which migrate from their site of origin at the dorsal neuraltube and differentiate into a diverse array of specialized cell typesduring vertebrate development [12]. Cranial neural crest cells(CNCCs) can differentiate into chondrogenic and osteogenic cellsthat, in turn, generate most of the cranial skeletal tissues (bothbones and cartilages) in the vertebrate head [12,13]. Cranial der-mal bones are unique in the way they are formed – dermal bonesform directly from the neural crest of the cranial dermis withoutforming a cartilage precursor, through a process known asintramembranous ossification [12,14–20]. Despite these uniquedevelopmental features of the cranial bones, most of our currentunderstanding of bone development, in general, is from studieson the limb and trunk bones of mesodermal origin. The skeletoge-nesis using CNCCs, and the morphogenesis of cranial bone forma-tion, are not very well understood, particularly at the tissue level,at least in part due to the lack of 3-D microscale tools to modeland manipulate cranial skeletal tissues in vitro.

Recent studies investigating cranial neural crest cell biologyutilize 2-D monolayer culture models, which do not replicate thenative, three-dimensionally defined microenvironments [21–23].In this study, we created a broadly applicable platform that allowsthe direct differentiation of multipotent progenitor cells, includingthose of neural crest origin. We produced in vitro 3-D tissue modelconstructs with relatively high throughput efficiency to mimic theosteogenic differentiation conditions of the mammalian post-migratory CNCCs (Fig. 1). The objective was to recapitulate thein vivo development of cranial dermal bone tissue from CNCCs inwell-defined 3-D bio-mimicking hydrogel units. Such a platformprovides in vitro study models for embryonic craniofacial disorders,where our understanding of etiology is limited, and few adequateanimal disease models exist. To this purpose, we have developedan array of microenvironments allowing successful 3-D direct

differentiation of neural crest cells, forming cranial skeletal tissuesand mimicking the embryonic phases of organogenesis.

2. Materials and methods

2.1. Mouse cranial neural crest cell isolation and culture

Mouse cranial neural crest cells were isolated according toEtchevers [24] under Harvard IACUC, Animal Experimentation Pro-tocol No 26-04. Briefly, frontonasal, maxillary and mandibularprominences of E10.5 mouse embryos were dissected into smallpieces in a petri dish containing DMEM/F12 media (Life Technolo-gies) with 1% penicilin–streptomycin (Life Technologies). These tis-sue blocks were digested for 5 min in a mixture of warm 0.1%collagenase (Sigma) and 0.025% trypsin (Thermo Scientific), whilebeing pipetted gently up and down to generate single cell suspen-sion of mouse CNCCs. These cells were washed with fresh mediumto stop the digestion, counted and seeded in laminin-coated plates.Cells were maintained and expanded in a monolayer culture inDMEM/F12 media with N2 supplement (Life Technologies),10 ng/ml EGF and 10 ng/ml bFGF recombinant proteins (R&DSystems).

2.2. Synthesis of methacrylated gelatin hydrogel

Gelatin (Sigma) was dissolved in 50 �C phosphate buffered sal-ine (PBS) with 10% (w/v) of final concentration. Methacrylic anhy-dride (Sigma) was added (0.8 ml/g) drop wise to the dissolvedgelatin and stirred for two and a half hours. The unreactedmethacrylic anhydride was dialyzed, and the final product was lyo-philized. Methacrylation was monitored and controlled with NMRanalysis (Oxford Instruments, NMR AS600) in deuterium oxide(Supplementary Fig. 1).

2.3. Cell encapsulation

Cell-encapsulated 3-D hydrogels were generated usingmethacrylated gelatin (GelMA) as a prepolymer solution. CNCCsbetween passage 3 and 5 were resuspended in 5% (w/v)GelMA prepolymer solution containing 0.3% (w/v) 2-hydroxy-1-(4-(hydroxyethoxy)phenyl)-2-methyl-1-propanone photoinitiator(Irgacure 2959; CIBA Chemicals) with the cell densities of1 � 106, 5 � 106, 1 � 107 or 4 � 107 cells/ml. The mixture of the

B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311 303

cell-prepolymer solution was placed between 150 lm-thickspacers and covered with a glass coverslip (Fig. 1). The geometryof the hydrogel units was shaped by a photomask with circularmotifs of 500 lm in diameter, placed on top of the glass coverslip.The prepolymer solution with cells was crosslinked with UV light(1 mW/cm2) for 20 s, and the excess uncrosslinked polymersolution was washed out with PBS. The resulting cell-encapsulated 3-D hydrogels were cultured in osteogenic media(OM) composed of alpha-MEM (Life Technologies), 10% fetalbovine serum (Life Technologies), 100 U/mL penicillin (LifeTechnologies), 100 mg/mL streptomycin (Life Technologies),100 nM dexamethasone (Sigma), 10 mM b-glycerophosphate(Sigma), 200 lM ascorbic acid (Sigma) and 50 ng/ml BMP2 (R&DSystems) for direct differentiation toward osteogenic lineage.These cell-encapsulated 3-D hydrogels were harvested weekly forthree weeks for further analysis. 3-D hydrogel samples culturedin the neural crest growth media (control media, CM) were usedas controls.

2.4. Characterization of cranial neural crest cells

Prior to 3-D hydrogel encapsulation, the isolated andmonolayer-expanded post-migratory CNCCs were characterizedwith ectoderm- and mesenchyme-specific cell surface markers.Cells were stained with p75 antibody conjugated with fluoresceinisothiocyanate (FITC) (Abcam) and CD73 antibody conjugated withphycoerythrin (PE) (BD Biosciences) in 1% BSA solution for 40 minat 4 �C and analyzed using flow cytometry (BD FACS Calibur) andFlowJo software (Fig. 2A). Cells stained with IgG FITC and IgG PEwere used as negative controls. The viability and proliferation ofthe encapsulated CNCCs in 3-D in vitro culture were assessed withLive–Dead staining (Life Technologies).

2.5. Quantitative real-time PCR

Total RNA (3 replicates for each group) was extracted by TRIzol�

(Life Technologies) and RNeasy MINI kit with QIAshedder columns(Qiagen) according to the manufacturer’s instructions. Afterextraction, total RNA was resuspended in the final volume of30 ll with RNAse-free water and quantified using spectropho-tometer (Nanodrop ND-1000, Delaware, USA). For the best resultsin the PCR array, all RNA samples were selected for purity using anA260/A280 UV spectrophotometry ratio greater the 1.99. 1 lg oftotal RNA was used for the reverse transcription reaction usingHigh-Capacity cDNA Reverse Transcription Kit (Applied Biosys-tems). Amplification and detection were carried out using SYBRGreen (Kapa Biosystems) with Eppendorf Mastercycler using 2-step cycle for 40 cycles. Primer sequences: (i) GAPDH Forward(50-CTA CAC TGA GGA CC AGG TTG TCT-30) Reverse (50-TTG TCATAC CAG GAA ATG AGC TT-30), (ii) p75 Forward (50-GCA TTG TGGTAG GCC AGA CC-30) Reverse (50-CCT GAA AGT CAC TCC ATCCC-30) (iii) Sox9 Forward (50-AGC TCA CCA GAC CCT GAG AA-30)Reverse (50-TCC CAG CAA TCG TTA CCT TC-30), (iv) Osterix Forward(50-CTG CTT GAG GAA GAA GCT C-30) Reverse (50-TTC TTT GTG CCTCCT TTC C-30), (v) Osteocalcin Forward (50-CTG ACA AAG CCT TCATGT CCA A-30) Reverse (50-GCG CCG GAG TCT GTT CAC TA-30),(vi) Bone Sialoprotein Forward (50-CAG GGA GGC AGT GAC TCTTC-30) Reverse (50-AGT GTG GAA AGT GTG GCG TT-30), (vii) Alkalinephosphatase Forward (50-AGT TAC TGG CGA CAG CAA GC-30) Rev-erse (50-GAG TGG TGT TGC ATC GCG-30), (viii) Aggrecan Forward(50-TGG CTT CTG GAG ACA GGA CT-30) Reverse (50-TTC TGC TGTCTG GGT CTC CT-30), (ix) Collagen Type 2 Forward (50-CAA CACAAT CCA TTG CGA AC-30) Reverse (50-TCT GCC CAG TTC AGG TCTCT-30). mRNA expressions were assayed in triplicate for eachsample and normalized to GAPDH housekeeping gene expressionsto calculate DCT. The undifferentiated and unencapsulated neural

crest cells were used as control samples to calculate the relativegene expression levels in fold change (2DDCT value).

2.6. Immunohistochemistry

For each time point, samples were fixed in 4% paraformalde-hyde solution for 20 min, permeabilized with 0.1% TritonX-100(Sigma) and blocked with 1% bovine serum albumin for 2 h at roomtemperature. Primary antibodies for Osterix (OSX) (Abcam),RUNX2 (Novus-Biologicals), Osteopontin (OPN) (GenWay Biotech),Osteocalcin (OCN) (EMD Millipore), bone sialoprotein (BSP) (Bioss)and Ki67 (Abcam) were used at the concentrations recommendedby the manufacturers, and the hydrogel samples were incubatedwith the antibodies overnight at 4 �C. AlexaFluor 488 and Alexa-Fluor 568 (Life Technologies) were used as secondary antibodies.Alexa Fluor 647 Phalloidin (Life Technologies) was used to visual-ize the actin cytoskeleton. The images were taken with Zeiss LSM780 confocal microscope.

2.7. Mineralization assay

The mineralization matrixes produced by the osteo-differentiated CNCCs were detected and quantified with alkalinephosphatase staining kit (Sigma) and hydroxyapatite quantifyingOsteoimageTM (Lonza) assay, according to manufacturers’ instruc-tions. The mineralization levels in the cell-encapsulated hydrogelswere assessed weekly for three weeks. The secreted hydroxyap-atite was quantified by the relative fluorescence intensities of thecell-encapsulated hydrogels using ImageJ software (NIH, Bethesda,Maryland, USA).

2.8. Image quantification

The unprocessed raw confocal z-stack files were analyzed withFiji software (ImageJ, NIH, Bethesda, Maryland, USA) with 3-Dobjects counter plugins. The total number of cells were countedusing DAPI staining. For OSX, cells with a positive signal in thenucleus were counted while cells with the positive signal in thecytoplasm were counted for OCN and OPN. Signals above athreshold set by the software were counted as positive and wereconfirmed manually.

2.9. Freeze-fracture cryo-scanning electron microscopy

Freeze-fracture cryo-SEM imaging was performed with a ZeissSupra55VP FESEM. A fixed single hydrogel was placed on a samplestub and submerged into a slurry bath of liquid nitrogen. Thefrozen sample was transferred to a cryo-transfer chamber andwas fractured in a vacuum with a sharp blade to expose the crosssections of the sample. The surfaces of the fractured samples weresputter coated with platinum and imaged using the SE2 detector ata voltage of 4 kV while maintained at �160 �C.

2.10. Statistical analysis

Statistical analyses were performed with GraphPad Prism, ver-sion 5 (GraphPad Software, Inc., San Diego). Results were analyzedusing analysis of variance with Tukey’s post hoc test for multiplecomparisons and Student’s two-tailed t test for single comparisons,with statistical significance set at p < 0.05. Unless otherwise stated,the mean values represent three experiments with two or threechannels per experiment, and the error bars represent the standarderror of mean.

Fig. 2. The characterizations of the CNCCs prior and after hydrogel encapsulation. (A) FACS analysis of the CNCCs prior to the encapsulation in 3-D hydrogels. 63.21% ± 20.28of cells were p75 positive indicating the neural crest cells phenotype (left), and 66.22% ± 11.64 were CD73 positive showing mesenchymal lineage characteristic (right).Corresponding IgG staining was used as a negative control (blue contour). (B) The Live–Dead staining of neural crest cells encapsulated in the 3-D hydrogel. Live cells werestained for green and dead cell are shown in red. (C) The proliferation capacity of neural crest cells was evaluated with Ki67 staining after 1 week of encapsulation in the 3-Dhydrogels. Proliferating cells express Ki67 (red) in the nuclei, co-localizing with DAPI (blue). The cell cytoskeletons were shown with phalloidin staining (green). (D–I) Effectsof cell encapsulation density on osteogenic differentiation of primary CNCCs. (D–F) Bright field images of the cell-encapsulated hydrogels generated using different celldensities. (G–I) The immunocytochemistry results for OSX (green) and cytoskeleton (phalloidin, magenta) in 3-week osteo-differentiated hydrogels showed that high celldensity is more favorable for osteogenic differentiation. Scale bars 100 lm. (For interpretation of the references to color in this figure legend, the reader is referred to the webversion of this article.)

304 B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311

B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311 305

3. Results

3.1. Characterization of hydrogel and neural crest cells

3-D cellular microenvironments were designed from highlybiocompatible gelatin-based hydrogels, supporting cell viabilityand growth [25]. Gelatin was methacrylated and lyophilized forthe generation of UV photo-crosslinkable hydrogels (Supplemen-tary Fig. 1). Mouse CNCCs were isolated from the frontonasal, max-illary and mandibular prominences of E10.5 mouse embryos andexpanded in the monolayer culture prior to 3-D differentiation(Fig. 1). The CNCCs before encapsulation showed 63.21% ± 20.28positivity for ectodermal marker p75 and 66.22% ± 11.64 positivityfor mesodermal marker CD73 confirming the mesenchymal-neuralphenotype as previously shown (Fig. 2A) [26,27]. The effect of theencapsulation process on cell viability were assessed withLive–Dead staining (Fig. 2B). More than 90% of encapsulated cellswere alive after the formation of 3-D constructs. The encapsulationprocess or the hydrogel itself do not restrict or alter the prolifera-tion capacity of neural crest cells as shown with Ki67 immunos-taining (Fig. 2C). The encapsulated cells were then differentiatedtoward the osteogenic lineage in the 3-D microenvironment toevaluate their potential for ectomesenchymal transformation andthe natural state osteogenesis. We investigated the effect of the cellencapsulation density on cranial osteogenesis under four differentcell concentrations (1 � 106 to 4 � 107 cells/ml). Cell densitiesabove 107 cells/ml showed significant expressions of the osteo-genic markers (Fig. 2D–I), whereas the lower cell concentrationsresulted with poor osteogenic morphology. All followingexperiments were performed at a cell density of 4 � 107 cell/ml.

3.2. Monitoring the osteogenic differentiation of cranial neural crestcells in 3-D platform

3-D differentiated CNCCs were analyzed for chondrogenic andosteogenic gene expressions with qRT-PRC (Fig. 3). The neural crestmarker p75 only maintained a high level of expression in the cellscultured in the neural crest cell-specific media and its expressiondecreased by more than 70% after the first week of culture in theosteogenic media (Fig. 3A). Since CNCCs have the potential to dif-ferentiate into chondrocytes, we also checked the expression levelof Sox9, collagen type II, and aggrecan. Sox9 is required for chondro-genesis, as it induces the expressions of cartilage-specific extracel-lular matrix molecules, such as collagen type II, IX and XI andaggrecan [28–31]. The Sox9 expression was slightly increased upto the second week but eventually decreased below the originallevel, confirming the cells were not in the course of chondrogenicdifferentiation (Fig. 3B) [32,33]. The expression of collagen type IIand aggrecan did not show differences in expression between thecontrol and the osteogenic samples (Fig. 3C–D). To evaluate theosteogenic differentiation, expression levels of four major bonespecific genes, Osterix (Osx, also known as Sp7), Alkaline phos-phatase (Alp), Osteocalcin (Ocn) and Bone sialoprotein (Bsp) wereanalyzed. All four genes showed dramatic increase in their expres-sion when cultured in the osteogenic differentiation media, espe-cially by the third week of incubation, where markers indicatedthe presence of mixed population of osteoblasts and osteocytes(Fig. 3E–H).

The expressions of bone-specific markers of 3-D encapsulatedCNCC were also evaluated at the protein level for three weeks withimmunocytochemistry (Fig. 4). The differentiation of CNCCs to bonecells was demonstrated by the accumulation of OSX (Fig. 4A and D),OPN (Fig. 4B and E) and OCN proteins (Fig. 4C and F). The expressionof the early osteogenic marker RUNX2 was higher in cells culturedin the osteogenic media compared to the control cells at the initial

differentiation phase, and RUNX2 expression later graduallydecreased as the bone tissue started to form (Fig. 4G and Supple-mentary Fig. 2). On the other hand, more than 30% of encapsulatedcells cultured in the osteogenic media showed OSX expression inthe nucleus after three weeks of incubation (Fig. 4D). The OPNand OCN protein depositions could be detected in the cytoplasmand the extracellular matrix of osteogenic cells cultured for 3 weeksin osteogenic media (Fig. 4E and F), as in the late developmentphase of the in vivo bone tissue, confirming the successful osteo-genic differentiation of the CNCCs using the 3-D hydrogel platform.We detected BSP depositions in the peripheral parts of the 3-Dhydrogels starting from week 1, which gradually expanded itsexpression toward the core parts of the hydrogel during the 3-week osteogenic differentiation period (Supplementary Fig. 2). Byweek 3, BSP accumulation could be observed around the lacunaeof the osteogenic cells.

3.3. Osteogenic mineralization and the morphological analysis of theneo-tissues

ALP expression starts from the early osteogenic differentiationsteps and continues through the maturation of the newly formedbone tissues. The engineered 3-D cell-encapsulated hydrogels wereanalyzed for mineralization by examining the ALP activity in thecells differentiated over the 3-week culture period (Fig. 5A). Thecells cultured in the osteogenic condition inside the hydrogelsexhibited high level of ALP activity after two weeks, whichincreased 2.56 folds by week 3. The undifferentiated cells in thecontrol hydrogels did not show any ALP activity (Fig. 5A). We eval-uated the functionality of osteo-differentiated CNCCs also withmineralization of hydrogels and the secretion of hydroxyapatite.The newly secreted hydroxyapatite in the 3-D cell-encapsulatedhydrogels was quantified with Osteoimage� Mineralization assay(Fig. 5B). The secretion of hydroxyapatite crystals was detected inthe osteogenic hydrogels starting from week 1, which graduallyincreased with further osteogenic differentiation and maturationby week 3. The undifferentiated control hydrogels did not showany fluorescent signals indicative of hydroxyapatite secretion.

Importantly, osteo-differentiated CNCC encapsulated hydrogelsshowed formation of in vivo-grade osteo lacuna structures, charac-teristic of the living bone tissue with embedded mature osteocytes(Fig. 6, Supplementary Fig. 3). Within the 3-D constructs osteo-differentiated cells also build intercellular connections providingcell–cell interactions (Fig. 6C, Supplementary Video 1 and Video2). Morphological analysis of the osteo-differentiated hydrogelswith FESEM revealed round, osteocyte-shaped cells located in thelacunae, and also mature osteoblasts throughout the hydrogel(Fig. 7A–C). Moreover, we observed differentiated bone cells witha large number of long projections called filopodia extending fromthe cell body, which is another key morphological characteristic ofthe mature osteocytes (Fig. 7A, red colored) [34].

4. Discussion

Along with many other important roles in the embryonic devel-opment, neural crest cells are essential for the proper developmentof the vertebrate head. The human skull is composed of more thantwenty intricately shaped and interconnected bones, which themajority is originated from the neural crest cells. Thus, deeperknowledge of the cranial neural crest cell behavior, and eventuallyhow they properly differentiate to bone, is essential for under-standing both the normal head development and the myriad ofcraniofacial abnormal conditions present in up to one-third of allcongenital birth defects in humans [35]. The vertebrate craniumis also an important and unique structure from the evolutionary

Fig. 3. qRT-PCR analysis of the neural crest-specific gene p75 and bone and cartilage-specific genes in cells cultured in control or osteogenic media for three weeks (n = 3 foreach sample. Error bar indicates standard error). (A) p75 was highly expressed in encapsulated cells cultured in control media. (B) The expression of cartilage-specific markerSox9 did not change in both control and osteogenic media. (C) Col II expression slightly decreased after week 2. (D) Expression of chondrogenic marker aggrecan did not altersignificantly during the differentiation. (E–H) Bone specific marker Osx, Alp, Ocn, and Bspwere highly expressed in the cells cultured in osteogenic media especially at the thirdweek of incubation. (One asterisks denote significance at P < 0.05, two asterisks denote significance at P < 0.01, Student t-test).

306 B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311

perspective as it represents one of the innovations that defines thevertebrate lineage, and vertebrate heads display a variety of adap-tive and species-specific characteristics [36]. 3-D differentiationculture systems support efficient tissue formation and allow themanipulation of the cells in their native-like microenvironments[23]. Notably, our work demonstrates the generation of 3-D cellu-lar niches for the efficient differentiation of CNCCs while maintain-ing their viability and proliferation capacity. To achieve a 3-D

in vitro differentiation platform that recapitulates the in vivoosteogenic differentiation of CNCCs, we monitored and analyzedthe developmental phases of cranial bone in our system. In thecourse of embryogenesis, the neural crest cells have a uniquestatus as these cells of neuroectodermal origin can give rise tothe specialized cell types that are typically derived from the meso-dermal germ layer, such as connective tissue, cartilage and bones[26]. In this study, after monolayer amplification we encapsulated

Fig. 4. The cell-encapsulated hydrogels cultured either in control media (CM) (A–C) or osteogenic media (OM) (D–F) were stained for OSX, OPN and OCN after three weeks. (A,D) The nuclear localization of osteoblastic marker OSX (green) could be seen in the differentiated cells cultured in the osteogenic media. (B, C, E, F) Late osteogenic markersOPN (red) and OCN (green) were expressed in the cytoplasm and the extracellular matrix of osteo-differentiated cells (E and F). (G) The expression of early osteogenic markerRUNX2 decreased during the course of osteogenic differentiation of CNCCs, whereas the expression of OPN and OSX increased (n = 3 for each sample. Error bar indicatesstandard error). All scale bars 50 lm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311 307

post-migratory CNCCs expressing both neural and mesenchymalmarkers, p75 and CD73 (Fig. 2A). Clonal study performed on neuralcrest cells demonstrated that those cells are mostly formed asheterogeneous populations consisting of multiple progenitor celltypes (e.g., melanocytes, glia, neuron, and myofibroblast) [37]. Toreach the high CNCC differentiation yield, we investigated theeffects of hydrogel properties and differential cell density inmicro-tissue formation. Substrate stiffness and focal polarizationof progenitor cells in conventional 2-D culture conditions dramat-ically affect the fate and physiology of cells [22,38]. The 3-D plat-form that we generated allows tuning of both hydrogel stiffnessand cell encapsulation density. According to our findings, highcell-to-cell contact dramatically assists the osteogenic differentia-tion in the microscale hydrogels. The highest level of cell-to-cellcontacts and the native-like osteogenic differentiation in both mor-phological and molecular aspects were achieved in our 3-D culturesystem with the cell density of 4 � 107 cell/ml.

We systematically investigated the intramembranousosteogenic differentiation progress of CNCCs in our 3-D platformby analyzing the expression patterns of multiple neural crest andskeletogenic markers, including p75, Sox9, Col II, aggrecan, Runx2,Osx, Alp, Ocn, Opn and Bsp (Fig. 3 and 4). Cells encapsulated in

hydrogels cultured in osteogenic media lost the expression of p75and even in early phase (week 1) are committed to osteogenic lin-eage. On the other hand, the ectodermal phenotype of CNCCs incontrol condition resumed up to week 2 with elevaded expressionof p75. Intramembranous ossification differs from the endochon-dral ossification, as it lacks the formation of hypertrophic cartilagetemplate prior to generating mineralized calcified bone [39]. Sox9is a well-known marker for chondrogenic cells in the process ofendochondral ossification, but it also plays an important role inthe differentiation of CNCCs to early bone progenitor cells duringintramembranous ossification [19]. Accordingly, we detected theelevated levels of Sox9 expression during the early phase of theosteogenic differentiation, which then decreased below the initiallevel at the later stages of the culture process (Fig. 3B). We alsoevaluated the chondrogenesis for potential endrochondral ossifica-tion with regulation of Col II and aggrecan expression. After week 2,Col II expression slightly decreased in osteogenic conditions andaggrecan expression did not alter significantly for both studygroups. These findings indicate there is no chondrogenic tissue for-mation and the osteogenic differentiation is through theintramembranous course. Early phases of the osteogenesis processare directly regulated by Runx2, an early marker and regulator of

Fig. 5. The osteogenic mineralization in the control and osteo-differentiated cell-encapsulated 3-D hydrogels were determined and quantified at weeks 1, 2 and 3 withalkaline phosphatase (A), and hydroxyapatite (B) staining. The osteogenic differentiation of CNCCs showed hydroxyapatite secretion and mineralized the hydrogels graduallyafter week 1 (C). Scale bars 100 lm.

308 B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311

many other osteoblast markers, such as Bsp, Ocn and Alp [19]. Inour 3-D differentiation system, RUNX2 was expressed during thefirst week of osteogenesis and its expression decreased signifi-cantly at the later stages of bone tissue culture (Fig. 4G). Theselater stage osteogenically committed cells no longer requireRUNX2 for further ossification [40]. All of the later stage osteogenicmarkers (e.g., Osx, Ocn, Opn, BSP) suggest that skeletal development

process was successfully recapitulated and led to the formationof mature osteoblasts and osteocytes. During the second andthird week of 3-D osteo-differentiatiation CNCCs demonstrate theexpression of (i) OSX, a bone-specific transcription factor that isspecifically expressed in osteoblasts of all skeletal elements [41],(ii) OCN, an osteoblast-derived hormone expressed only in fully dif-ferentiated osteoblasts and osteocytes, (iii) OPN, a late osteogenic

Fig. 6. The hydrogel-encapsulated CNCCs change their morphology in the 3-D osteogenic differentiation platform. (A) Hydrogel-encapsulated CNCCs incubated in controlmedia exhibited elongated morphology typical for mesenchymal cells. (B) Spherical morphology of osteo-differentiated cells in lacunae-like niches. Cytoskeletons werestained with phalloidin (magenta) and cell nuclei with DAPI (blue). (C) The confocal image of the hydrogel with encapsulated cells differentiated for two weeks in osteogenicmedia. The immunostaining results showed early osteogenic cells positive for RUNX2 (red) and osteocytes positive for OSX (green). The cytoskeletons stained with phalloidin(white) revealed filopodia projections and cell-to-cell connections (red arrowheads). Scale bars: 100 lm in A and B, 10 lm in C. (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of this article.)

B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311 309

protein that regulates the biomineralization and boneremodelling[42,43] and (iv) BSP, a non-collagenous protein that is a keycomponent of the bone extracellular matrix [44,45] (Fig. 3E, F,Fig. 4B, C, E, F, G and Supplementary Fig. 2). The formation ofmature bone tissue is further characterized by the mineralizationin terms of hydroxyapatite deposition in the newly formed extra-cellular matrix [46]. We demonstrate the ability of newly differen-tiated osteogenic cells to mineralize the hydrogels withhydroxyapatite from second week. Similarly, high expression ofalkaline phosphatase (ALP), which is highly involved in the miner-alization of neo-tissues [47] supports the osteogenesis and in vitrotissue maturation (Fig. 5). Another decisive method to evaluate theformation of bone tissue and the emergence of the differentiated

osteoblasts and osteocytes is to analyze their morphologies com-pared to the native mature bone tissues with osteoblasts andosteocytes embedded. As early as one week after the initiation ofthe 3-D osteogenic differentiation process, the cell morphology ofthe neural crest cells began to change to reflect their differentiationinto osteoblasts (Figs. 6, 7, Supplementary Video 1 and Video 2).The appearance of the lacunae cavities containing matureosteoblasts and osteocytes within the hydrogel is strikingly similarto those seen in the natural bone tissue [48]. Such cell morphologychanges during the course of osteogenesis can be observed throughthe freeze-fractured FESEM imaging. While the newly formed, lar-ger osteoblast cells were tightly embedded in the hydrogel matrix(Fig. 7B), their morphology changed to the small and round

Fig. 7. The representative freeze-fractured hydrogel FESEM images of the bioengineered neo-tissues, after three weeks of osteogenic differentiation. Cell-encapsulatedhydrogels revealed cells in different phases of osteogenic differentiations, recapitulating native-like microenvironment. (A) An osteocyte (red) in a lacuna with extendedfilopodia that could also be observed in the bone canaliculus-like structure (white arrowheads), connecting to the neighboring cells. (B) A focused image of a single osteoblaststarting to remodel the surrounding hydrogel into a lacuna-like structure. (C) Representative image of the osteogenic niche in 3-D hydrogel microenvironment. Scale bars;3 lm in A and B, 10 lm in C. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

310 B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311

osteocyte-looking cells that reside in the spacious lacunae at laterstages, and send multiple filopodia extensions through the sur-rounding matrix and the neighboring bone cells (Fig. 7A). Filopodiapenetrate the bone tissue through canal-like structures called bonecanaliculi and are believed to sense biomechanical stress and com-municate with other bone cells within the same tissue [34]. Thesecellular projections also observed in our mature 3-D bone tissuesextended beyond the immediate lacunae boundaries and contin-ued through the surrounding matrix toward other neighboringbone cells (Fig. 7A, white arrowheads). These cell–cell interactionswere clearly detected in the cell-encapsulated 3-D hydrogels, start-ing from the second week of osteogenic differentiation.

In summary, we demonstrate a 3-D differentiation platformthat enables to recapitulate an important developmental processin vitro. The platform we generated successfully supports theintramembranous ossification of cranial bone tissue using primaryCNCCs. Thus, it is an essential tool added for studying the normaland abnormal osteogenic development in the model and non-model vertebrates. This platform has the potential to be used ina broader context, especially when combined with other bioengi-neering technologies [6,8,49,50]. This platform can be easily scaledup by assembly technologies, and be utilized in bottom-up tissueengineering, enabling to generate larger, multi-hydrogel constructsneeded to understand the morphogenesis of specific skeletal struc-tures. Neural crest cells differentiated from induced pluripotentstem cells (iPSCs) that are derived from human patients with cran-iofacial diseases could be used to monitor disease-specific pheno-types during osteogenic differentiation and eventually bone andother skeletal tissue formations for medical and pharmaceutical

research. The adaptive and flexible nature of the platform will alsoallow it to be used to investigate the evolution of the diverse cran-iofacial bone shapes by including morphogenic factors in the speci-fic regions of the assembled multi-hydrogels to recapitulate andmanipulate the morphogenic processes in vitro.

Acknowledgements

UD acknowledges that this material is based in part upon worksupported by the NSF CAREER Award Number 1150733, NIHR01EB015776, NIH R15HL115556 and NIH R21HL112114. AAacknowledges NSF 1257122 and Templeton Foundation RFP-12-01. Dr. U Demirci is a founder of, and has an equity interest in:(i) DxNow Inc., a company that is developing microfluidic andimaging technologies for point-of-care diagnostic solutions, and(ii) Koek Biotech, a company that is developing microfluidic IVFtechnologies for clinical solutions. U.D.’s interests were viewedand managed in accordance with the conflict of interest policies.The other authors indicated no potential conflicts of interest.Authors thank to Yesim Bagatur (M.S.) and Zeynep Esencan (M.S.)for their technical assistance.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.actbio.2015.12.004.

B. Namkoong et al. / Acta Biomaterialia 31 (2016) 301–311 311

References

[1] R.J. Davenport,What controls organ regeneration, Science 309 (5731) (2005) 84.[2] R. Perez-Castillejos, Replication of the 3D architecture of tissues, Mater. Today

13 (1–2) (2010) 32–41.[3] Y.S. Song et al., Engineered 3D tissue models for cell-laden microfluidic

channels, Anal. Bioanal. Chem. 395 (1) (2009) 185–193.[4] F. Xu et al., A three-dimensional in vitro ovarian cancer coculture model using a

high-throughput cell patterning platform, Biotechnol. J. 6 (2) (2011) 204–212.[5] S. Guven et al., Engineering of large osteogenic grafts with rapid engraftment

capacity using mesenchymal and endothelial progenitors from human adiposetissue, Biomaterials 32 (25) (2011) 5801–5809.

[6] S. Guven et al., Multiscale assembly for tissue engineering and regenerativemedicine, Trends Biotechnol. 33 (5) (2015) 269–279.

[7] C.A. Bichsel et al., Diagnostic microchip to assay 3D colony-growth potential ofcaptured circulating tumor cells, Lab Chip 12 (13) (2012) 2313–2316.

[8] S. Tasoglu et al., Untethered micro-robotic coding of three-dimensionalmaterial composition, Nat. Commun. 5 (2014) 3124.

[9] S. Tasoglu et al., Guided and magnetic self-assembly of tunablemagnetoceptive gels, Nat. Commun. 5 (2014) 4702.

[10] B.K. Hall, The neural crest as a fourth germ layer and vertebrates asquadroblastic not triploblastic, Evol. Dev. 2 (1) (2000) 3–5.

[11] B.K. Hall, The neural crest and neural crest cells: discovery and significance fortheories of embryonic organization, J. Biosci. 33 (5) (2008) 781–793.

[12] N. Le Douarin, C. Kalcheim, The Neural Crest, second ed., Developmental andCell Biology Series, vol. 36, Cambridge University Press, Cambridge, UK, NewYork, NY, USA, 1999. Xxiii, 445 p.

[13] G.F. Couly, P.M. Coltey, N.M. Le Douarin, The triple origin of skull in highervertebrates: a study in quail-chick chimeras, Development 117 (2) (1993)409–429.

[14] D.M. Noden, The role of the neural crest in patterning of avian cranial skeletal,connective, and muscle tissues, Dev. Biol. 96 (1) (1983) 144–165.

[15] D.M. Noden, Cell movements and control of patterned tissue assembly duringcraniofacial development, J. Craniofac. Genet. Dev. Biol. 11 (4) (1991) 192–213.

[16] B.K. Hall, T. Miyake, The membranous skeleton: the role of cell condensationsin vertebrate skeletogenesis, Anat. Embryol. (Berl) 186 (2) (1992) 107–124.

[17] X. Jiang et al., Tissue origins and interactions in the mammalian skull vault,Dev. Biol. 241 (1) (2002) 106–116.

[18] J.A. Helms, D. Cordero, M.D. Tapadia, New insights into craniofacialmorphogenesis, Development 132 (5) (2005) 851–861.

[19] A. Abzhanov et al., Regulation of skeletogenic differentiation in cranial dermalbone, Development 134 (17) (2007) 3133–3144.

[20] D.M. Noden, R.A. Schneider, Neural crest cells and the community of plan forcraniofacial development: historical debates and current perspectives, Adv.Exp. Med. Biol. 589 (2006) 1–23.

[21] E.T. Pineda, R.M. Nerem, T. Ahsan, Differentiation patterns of embryonic stemcells in two- versus three-dimensional culture, Cells Tissues Organs 197 (5)(2013) 399–410.

[22] M.P. Lutolf, P.M. Gilbert, H.M. Blau, Designing materials to direct stem-cellfate, Nature 462 (7272) (2009) 433–441.

[23] M. Ravi et al., 3D cell culture systems: advantages and applications, J. Cell.Physiol. 230 (1) (2015) 16–26.

[24] H. Etchevers, Primary culture of chick, mouse or human neural crest cells, Nat.Protoc. 6 (10) (2011) 1568–1577.

[25] U.A. Gurkan et al., Simple precision creation of digitally specified, spatiallyheterogeneous, engineered tissue architectures, Adv. Mater. 25 (8) (2013)1192–1198.

[26] N.M. Le Douarin, G.W. Calloni, E. Dupin, The stem cells of the neural crest, CellCycle 7 (8) (2008) 1013–1019.

[27] C. von Levetzow et al., Modeling initiation of Ewing sarcoma in human neuralcrest cells, PLoS ONE 6 (4) (2011) e19305.

[28] V. Lefebvre et al., SOX9 is a potent activator of the chondrocyte-specificenhancer of the pro alpha1(II) collagen gene, Mol. Cell Biol. 17 (4) (1997)2336–2346.

[29] W. Bi et al., Sox9 is required for cartilage formation, Nat. Genet. 22 (1) (1999)85–89.

[30] C. Healy, D. Uwanogho, P.T. Sharpe, Regulation and role of Sox9 in cartilageformation, Dev. Dyn. 215 (1) (1999) 69–78.

[31] Y. Mori-Akiyama et al., Sox9 is required for determination of the chondrogeniccell lineage in the cranial neural crest, Proc. Natl. Acad. Sci. USA 100 (16)(2003) 9360–9365.

[32] L.J. Ng et al., SOX9 binds DNA, activates transcription, and coexpresses withtype II collagen during chondrogenesis in the mouse, Dev. Biol. 183 (1) (1997)108–121.

[33] Q. Zhao et al., Parallel expression of Sox9 and Col2a1 in cells undergoingchondrogenesis, Dev. Dyn. 209 (4) (1997) 377–386.

[34] E.M. Aarden, E.H. Burger, P.J. Nijweide, Function of osteocytes in bone, J. Cell.Biochem. 55 (3) (1994) 287–299.

[35] N.C. Jones, P.A. Trainor, The therapeutic potential of stem cells in the treatmentof craniofacial abnormalities, Expert Opin. Biol. Ther. 4 (5) (2004) 645–657.

[36] C. Gans, R.G. Northcutt, Neural crest and the origin of vertebrates: a new head,Science 220 (4594) (1983) 268–273.

[37] G.W. Calloni, N.M. Le Douarin, E. Dupin, High frequency of cephalic neuralcrest cells shows coexistence of neurogenic, melanogenic, and osteogenicdifferentiation capacities, Proc. Natl. Acad. Sci. USA 106 (22) (2009) 8947–8952.

[38] M.A. Bucaro et al., Fine-tuning the degree of stem cell polarization andalignment on ordered arrays of high-aspect-ratio nanopillars, ACS Nano 6 (7)(2012) 6222–6230.

[39] C. Scotti et al., Recapitulation of endochondral bone formation using humanadult mesenchymal stem cells as a paradigm for developmental engineering,Proc. Natl. Acad. Sci. USA 107 (16) (2010) 7251–7256.

[40] T. Takarada et al., An analysis of skeletal development in osteoblast-specificand chondrocyte-specific runt-related transcription factor-2 (Runx2) knockoutmice, J. Bone Miner. Res. 28 (10) (2013) 2064–2069.

[41] K. Nakashima et al., The novel zinc finger-containing transcription factorosterix is required for osteoblast differentiation and bone formation, Cell 108(1) (2002) 17–29.

[42] G.K. Hunter, Role of osteopontin in modulation of hydroxyapatite formation,Calcif. Tissue Int. 93 (4) (2013) 348–354.

[43] S.T. Choi et al., Osteopontin might be involved in bone remodelling rather thanin inflammation in ankylosing spondylitis, Rheumatology (Oxford) 47 (12)(2008) 1775–1779.

[44] P.V. Hauschka et al., Osteocalcin and matrix Gla protein: vitamin K-dependentproteins in bone, Physiol. Rev. 69 (3) (1989) 990–1047.

[45] L.W. Fisher et al., Matrix sialoprotein of developing bone, J. Biol. Chem. 258(20) (1983) 12723–12727.

[46] K. Gkioni et al., Mineralization of hydrogels for bone regeneration, Tissue Eng.Part B: Rev. 16 (6) (2010) 577–585.

[47] Y. Mikuni-Takagaki et al., Matrix mineralization and the differentiation ofosteocyte-like cells in culture, J. Bone Miner. Res. 10 (2) (1995) 231–242.

[48] U.E. Pazzaglia et al., Osteoblast-osteocyte transformation. A SEM densitometricanalysis of endosteal apposition in rabbit femur, J. Anat. 224 (2) (2014) 132–141.

[49] P. Chen et al., Biotunable acoustic node assembly of organoids, Adv. Healthc.Mater. 4 (13) (2015) 1937–1943.

[50] S. Tasoglu et al., Magnetic levitational assembly for living material fabrication,Adv. Healthc. Mater. 4 (10) (2015) 1469–1476. 1422.