in vitro 3d corneal tissue model with epithelium, stroma ... · in vitro 3d corneal tissue model...

lable at ScienceDirect

Biomaterials 112 (2017) 1e9

Contents lists avai

Biomaterials

journal homepage: www.elsevier .com/locate/biomater ia ls

In vitro 3D corneal tissue model with epithelium, stroma, andinnervation

Siran Wang a, b, Chiara E. Ghezzi b, Rachel Gomes b, c, d, Rachel E. Pollard b,James L. Funderburgh e, David L. Kaplan a, b, *

a National Engineering Laboratory for Modern Silk, College of Textile and Clothing Engineering, Soochow University, Suzhou 215021, People's Republic ofChinab Department of Biomedical Engineering, Tufts University, Medford, MA 02155, USAc New England Eye Center, Tufts Medical Center, Boston, MA, USAd Federal University of S~ao Paulo, School of Medicine, S~ao Paulo, Brazile Department of Ophthalmology, University of Pittsburgh, Pittsburgh, PA 15213, USA

a r t i c l e i n f o

Article history:Received 7 August 2016Received in revised form23 September 2016Accepted 26 September 2016Available online 4 October 2016

Keywords:SilkCorneaInnervationTissue modelStroma

* Corresponding author.E-mail address: [email protected] (D.L. Kapla

http://dx.doi.org/10.1016/j.biomaterials.2016.09.0300142-9612/© 2016 Elsevier Ltd. All rights reserved.

a b s t r a c t

The interactions between corneal nerve, epithelium, and stroma are essential for maintaining a healthycornea. Thus, corneal tissue models that more fully mimic the anatomy, mechanical properties andcellular components of corneal tissue would provide useful systems to study cellular interactions, cornealdiseases and provide options for improved drug screening. Here a corneal tissue model was constructedto include the stroma, epithelium, and innervation. Thin silk protein film stacks served as the scaffoldingto support the corneal epithelial and stromal layers, while a surrounding silk porous sponge supportedneuronal growth. The neurons innervated the stromal and epithelial layers and improved function andviability of the tissues. An air-liquid interface environment of the corneal tissue was also mimickedin vitro, resulting in a positive impact on epithelial maturity. The inclusion of three cell types in co-culture at an air-liquid interface provides an important advance for the field of in vitro corneal tissueengineering, to permit improvements in the study of innervation and corneal tissue development,corneal disease, and tissue responses to environmental factors.

© 2016 Elsevier Ltd. All rights reserved.

1. Introduction

The cornea is the outermost layer of the human eye and is animportant part of the ocular light path. The cornea has threedistinct layers: the epithelium, stroma, and endothelium. Theexternal epithelial layer has 3e5 layers of epithelial cells and pro-tects the inner structures [1,2]. The middle stromal layers arecomposed of aligned corneal stromal cells guided by parallelcollagen lamellae [2]. The innermost layer of the cornea is theendothelial layer. Current tissue models fail to incorporate each ofthese distinct anatomical layers and are not able to realize diseaseconditions of corneal tissue. Corneal opacity is one of the principalcauses of bilateral blindness, affecting 7 million people around theworld [3]. Among them, 2.85 million people have diminished or anabsence of sensation due to corneal nerve dysfunction or degen-eration [4]. Every year 46,000 patients in the USA receive corneal

n).

transplantation surgery.The cornea is the most densely innervated surface in the human

body [5]. Neuronal innervation is closely related to the health ordisease state of the corneal epithelium and stroma. Innervation isdistributed throughout the epithelium and stroma but is absent inthe endothelial layer [6]. Stromal nerve trunks with a density of33e71/mm2 [7] arise from the limbal plexus and enter the pe-ripheral corneal stroma radially. In the stroma, nerves are orga-nized parallel to the collagen lamellae and branch into smallerfascicles as they proceed toward the superficial stroma [8]. Thenerve fibers then penetrate the epithelial layer with a density ofapproximately 600 termini/mm2 [6,8]. The nerves interact physi-cally and chemically with corneal tissue, providing sensing andreleasing trophic factors including neurotransmitters and neuro-peptides tomaintain homeostasis [6]. During corneal development,nerve growth is modulated bymany growth factors, including brainderived nerve growth factor (BDNF), nerve growth factor (NGF),glial cell derived neurotrophic factor (GDNF), and neurotrophin-3(NT-3) [9e12]. Among these growth factors, NGF is critical for

S. Wang et al. / Biomaterials 112 (2017) 1e92

corneal nerve survival, axonal branching, elongation, sprouting andregeneration [10]. The lack of these trophic factors can lead toneurotrophic keratopathy, a disorder that make cornea more sus-ceptible to injury [13].

Despite the importance of corneal innervation, the interactionsbetween nerve and corneal tissue in healthy and diseased corneaare not fully understood [5]. This is partially due to the limitationswith rabbit, mice, pig and human in vivo models, including thecomplexity of the in vivo environments, differences between hu-man and animal corneal tissues, and the challenges with studyinghuman embryo cornea. These drawbacks underscore the need fornew research tools.

In vitro corneal tissue models have unique advantages forstudying cellular interactions including the ability to simplify thecomplex in vivo environment, utilize human cells, be cost effectivewhen compared with animal and human studies, and be designedfor high throughput analysis. An in vitro tissue model of humancorneal innervation can also support studies of corneal nervefunctions. Current corneal tissue in vitro models mainly focus oncorneal epithelial and stromal cells and use collagen as substrates[14,15]. Among the few co-culture studies that used corneal cellsand neurons [15,16], layers of collagen hydrogel were used toresemble the lamellar structure of cornea but failed to recapitulatethe alignment of the stromal cells and the multi-layer features ofthe epithelial cells. Further, the native density of nerve endings andbranches has also not been achieved through in vitro cultures.Collagen as the substrate also poses significant limitations due tolow mechanical stiffness, leading to mismatched mechanicalproperties and contraction in long-term culture [17,18].

Silk is a biodegradable protein material with highly tunablemechanical properties that can be cast into optically clear films[18,19]. Silk films, physically crosslinked through water vaporannealing, provided elastic moduli of 67.7 kPa [20] that matchedthe stiffness of the cornea, which is 40e60 kPa [21,22]. Silk filmswith surface patterns and functionalized with RGD supported thealignment, growth, andmatrix synthesis by human corneal stromalcells [23]. These films also did not contract and slowly degradedin vitro, providing support for sustained in vitro tissue models. Thesilk can also be formed into sponges which can support neurongrowth and the formation of neuronal connections [24e26]. In ourprevious study [16], 2D human corneal stromal stem cells (hCSSCs)and dorsal root ganglion neurons (DRG) in a co-culture systemwasgenerated and we observed that axon length increased when theDRGs were co-cultured with the hCSSCs.

The goal of the present study was to generate a 3D silk proteinbased co-culture system including the corneal stromal layer,epithelial layer, and DRG neurons, to further understand the in-teractions between corneal innervation and corneal tissues. Thescaffold designwas to closelymimic corneal anatomy, with silk filmstacks for corneal epithelial and stromal cell growth surrounded bya silk sponge seededwith the DRG simulating the limbus tissue. Theguidance for neuronal extensions was generated by the addition ofNGF in the epithelial layer scaffold. An air-liquid interface was alsodesigned for a bioreactor support system to house the corneal tis-sues and to better mimic the native corneal environment. This newcorneal tissue construct supported dense innervation in theepithelial and stromal regions as well as sustained cultivationin vitro, critical outcomes for the system utility toward the furtherstudy of corneal development, function and dysfunction.

2. Materials and methods

2.1. Preparation of silk solution

Silk solution was prepared from cocoons of Bombyx mori

silkworm based on the procedure developed in our previousstudies [20,27]. Silk cocoons were purchased from Tajima Shoji Co.(Yokohama, Japan) and boiled for 30 min in 0.02 M Na2CO3 solution(Sigma-Aldrich. St Louis, MO). The boiled silk was rinsed withdeionizedwater 6 times and dried overnight. The extracted silk wasthen dissolved in a 9.3 M LiBr solution and dialyzed against distilledwater for 2 days to obtain a silk aqueous solution (5e7% w/v).

2.2. Preparation of growth factor stamped flat silk films

Flat, optically clear, porous silk films were prepared by casting120 mL of 1% w/v silk solution with 0.05% w/v of polyethylene oxide(PEO, MW ¼ 900,000, SigmaeAldrich) on a 12 mm diameter glasscoverslip (Electron Microscopy Science, Hatfield, PA) [28]. The filmswere then dried overnight. High and low concentration NGF inkswere used for stamping the silk films. The inks were composed of50 ml (4 mg/mL) acetic acid-type I collagen solution (rat-tail tendon,BD, Franklin Lake, NJ) containing 100 ng/mL keratinocyte growthfactor (KGF) (Sigma), 100 ng/mL hepatic growth factor (HGF)(Sigma), 200 ng/mL epithelium growth factor (EGF) (Thermo Fisher,Waltham MA), and either a high concentration of NGF (400 ng/mL)or a low concentration of NGF (200 ng/mL) (R&D Systems, Min-neapolis, MN). Multi-circular, radial and uniform stamp patternswere employed (Supplementary Fig. 1). The multi-circular stampswere formed by dipping a 12 mm outside diameter and a 6 mminside diameter donut shape polydimethylsiloxane (PDMS) (FisherScientific Co. Fair Lawn, NJ) stamp in the low NGF ink and pressingonto the dried silk film. The center was stampedwith a 6mmPDMScylinder carrying the high concentration NGF ink. The radialpattern was stamped with the high NGF ink with its shape indi-cated in Supplementary Fig. 1. The whole surface of the uniformlystamped silk film was covered with high NGF ink. The silk filmswere annealed inwater filled desiccators at�25mmHg for 2.5 h forphysical cross-linking. Before use, the silk films were exposed to UVlight for 30 min on each side and the soaked in DI water for 48 h toextract any residual PEO to form the pores.

2.3. Preparation of patterned silk films

Patterned silk films were also prepared based on the proceduresdeveloped in our previous studies [29,30]. Briefly, 1% w/v silk so-lution with 0.05%w/v of PEO (MW ¼ 900,000; SigmaeAldrich) wascast onto patterned PDMS molds with 600 lines/mm grating. ThePDMSmolds were prepared using our previously reportedmethods[30]. The silk films were dried at room temperature overnight andwater annealed with the same methods used with the stamped silkfilms. The patterned silk films were then peeled off from the moldsusing established methods [30]. The silk films were immersed in DIwater for 2 days to extract the PEO to generate the pores(0.5e5 mm) [28], and then immersed in 70% ethanol for 8 min andwashed with PBS (5X) before cell seeding.

2.4. Preparation of silk sponges

Salt leached silk scaffolds with 500e600 mm pores were pre-pared using our previously reported procedure [31]. The scaffoldswere mounted in a custom designed well fabricated to be 1 mmdepth depressed into a Delrin sheet (McMaster-Carr, Robbinsville,NJ). The scaffold was sliced into 1 mm thick layers using microtomeblade and cut into donut shapes (15 mm outer diameter, 12 mminner diameter) with a biopsy punch (McMaster-Carr, Robbinsville,NJ). The silk sponge donuts were sterilized by autoclave before cellseeding.

S. Wang et al. / Biomaterials 112 (2017) 1e9 3

2.5. RGD and PDL surface modification

Arginine-Glycine-Aspartic acid-Serine (RGD) peptide (Bachem,Torrance, CA) functionalized patterned silk films were preparedusing methods from our previous work [29]. Stamped flat silk filmsand salt leached silk scaffolds were soaked in 1mL of 10 mg/ml poly-D-lysine (PDL) solution overnight at 4 �C.

2.6. Preparation of collagen hydrogels

Collagen gels were prepared by adding 100 ml of 10x DMEM(Sigma) to 900 ml (4 mg/mL) acetic acid-type I collagen solution(rat-tail tendon, Corning, Corning NY), followed by neutralizationwith 20 ml 1 M NaOH (Sigma).

2.7. Human corneal stromal stem cell (hCSSCs) culture

Human corneal stromal stem cells (HCSSCs) were isolated fromcollagenase digests of limbal stromal tissue dissected from de-identified human corneas, obtained from the Center for OrganRecovery and Education, (CORE; Pittsburgh, PA), as described pre-viously [32]. Ethical aspects of the research protocols wereapproved by the Committee for Oversight of Research Involving theDead (CORID) Protocol #161. HCSSCs were passaged 4 times beforeseeding. Cells were detached with 0.25% trypsin (GIBCO) solutionand seeded on the surface of the sterilized patterned porous silkfilms at a concentration of 15,000 cells/cm2. Cell seeding wasaccomplished by adding the cell suspension dropwise on top of thefilms. The films were incubated for 30 min to allow time for cellattachment. Seeded silk films were cultured in proliferation me-dium containing DMEM/MCDB-201 in the ratio of 3 to 2 (v/v) with2% fetal bovine serum, 10 ng/mL platelet-derived growth factor,1 mg/mL lipid-rich bovine serum albumin (Albumax, Life Tech-nologies, Grand Island, NY), 10 ng/mL epidermal growth factor,5 mg/mL transferrin, 5 ng/mL selenous acid, 0.1 mM ascorbic acid-2-phosphate, 10�8 M dexamethasone, 100 IU/mL penicillin, 100 mg/mL streptomycin, 50 mg/mL gentamicin, and 100 ng/mL choleratoxin until confluent (2 days). After cells reached confluency,hCSSCs were differentiated on the silk films into keratocytes withdifferentiation medium composed of advanced DMEM (Life Tech-nologies), containing 1.0 mM L-ascorbic acid-2-phosphate (Sigma),50 mg/mL gentamicin (Life Technologies), 2 mM L-alanyl-L-gluta-mine (Life Technologies), 100 mg/mL penicillin, 100 mg/mL strepto-mycin (Mediatech, Manassas, VA), 0.1 ng/mL transforming growthfactor-beta3 (TGFb-3, Sigma), and 10 ng/mL basic fibroblast growthfactor (FGF-2, Sigma) [33].

2.8. Human corneal epithelial cell (hCECs) culture

Primary hCECs (C0185C, Thermo Fisher) were passaged 5 timesbefore seeding. Cells were detachedwith 0.25% trypsin (GIBCO) andseeded on top of sterilized stamped silk films at a density of150,000 cells/cm2. The films were then incubated for 4 h to allowtime for cell attachment and then cultured in keratinocyte SFMmedium (Thermo Fisher) for 2 days to reach confluency.

2.9. Chicken dorsal root ganglion (DRG) cell culture

DRG explants were dissected from day 8 chicken embryosfollowing protocols developed in our prior study [34]. The explantswere then carefully placed on the surface of the salt leached silkscaffolds with forceps and incubated for 2 h to allow time for cellattachment. The scaffolds were then flipped over and cultured inDMEM containing 20% FBS and 50 ng/mL NGF.

2.10. Co-culture of hCSSCs hCECs and DRG neurons

The scaffolds for co-culture were designed to mimic cornealanatomy (Fig. 1). For convenience, E, S and D will be used torepresent hCECs, hCSSCs and DRG neurons, respectively. To preparethe co-culture scaffolds, 3 layers of patterned silk films seeded withS were stacked with their patterns in a crisscross pattern. Then, thesilk film stacks were cut to 12 mm diameter with a biopsy punch(McMaster-Carr) and transferred to the center of the silk spongedonuts. The flat silk films seeded with E were then transferred withforceps and gently placed on top of the film stacks. To achieveintegrity of the scaffold, 500 ml type I rat tail collagenwas casted ontop and absorbed into the scaffold. In order to guide axons towardthe top of the scaffolds, 50 ml of collagen hydrogel containing400 ng/mL of NGF was cast on top of the film stack. The scaffoldswere then incubated at 37 �C for 30 min to complete the cross-linking. After this, the whole scaffold was immersed in hCSSCsdifferentiation medium and cultivated for 2 days. A customizeddesigned waffle-shaped PDMS floating shelf (5 mm thick, 5 cmdiameter, with 16 � 1 mm2 holes) was prepared by casting PDMSon top of Delrin® molds (McMaster-Carr). This PDMS shelf allowedthe top of the scaffold to remain at the air-liquid interface (ALIC)while the bottomwas immersed in hCSSCs differentiation medium.The cultivation in liquid (LC) and air-liquid interface (ALIC) lasted28 days. The co-cultures of E and D (ED-LC and ED-ALIC), S and D(SD-LC and SD-ALIC), and single cultures (E-LC, E-ALIC, S-LC, S-ALIC,D-LC, D-ALIC), were also processed as comparisons for the three celltypes in tri-cultures (ESD-LC, ESD-ALIC). The scaffolds for two typesof cell co-cultures and single cultures were prepared with the samemethods as the ESD tri-cultures but only contained the respectivecellular components.

2.11. Immunohistochemistry

The single cultured and co-cultured samples were fixed at days14 and 28. Samples were fixed in 4% paraformaldehyde in PBS(Affymetrix, Cleveland, OH) for 45 min and then treated with 5%BSA for 30 min. Cellular morphology was revealed with anti btubulin III staining. Keratocan was stained to reveal hCSSCs ECMsecretion while involucrin was stained to reflect the maturity ofhCECs. The dilution of antibodies is indicated in SupplementaryTable 1. The samples were treated with primary antibodies for12 h at 4 �C and then washed with PBS 3 times, 15 min each. Thesamples were stained with secondary antibodies for 8 h at 4 �C andwashed with PBS 3 times, 15 min each. DAPI was diluted 1:1000 in5% BSA solution at the same time as the primary antibodies. Imageswere taken on a BZX-700 microscope (Keyence Corporation, Itasca,IL) at 10X and 4X. Maximum intensity projection images weregenerated using Advanced 3D Analysis software (Keyence Corpo-ration, Itasca, IL). In order to compare transparency, the fixedsamples for ESD-LC, ESD-ALIC and the porcine cornea wereimmersed in glycerol for 1 h and placed over a paper with the word“SILK” printed. Photos were taken to indicate transparency.

2.12. Quantitative reverse transcript PCR (qPCR)

Gene expression levels for keratocan (KERA), lumican (LUM),smooth muscle actin (ACTA2) aldehyde dehydrogenase 3A1(ALDH3A1), involucrin (IVL), connexin 37 (GJA4), and cytokeratin 3(KRT3) were quantified by RT-PCR (qPCR) [23,35,36]. In brief, totalRNA was extracted using Trizol with single step acid-phenol gua-nidinium method [37], adsorbed onto a silica-gel membrane usingthe Qiagen RNeasy Kit protocol (Qiagen, Valencia, CA), eluted, andquantified [38]. The RNA extracted from the scaffolds was reversetranscribed to cDNA in a 20 ml reaction using high-capacity cDNA

Fig. 1. A) Side view schematics of human cornea and in vitro 3D corneal tissue model. Scale bars ¼ 3 mm. B) Pictures of porcine cornea, day 28 co-culture of hCECs, hCSSCs and DRGexplants in liquid (ESD-LC) and day 28 co-culture at air-liquid interface (ESD-ALIC) soaked in glycerol. Scale bars ¼ 3 mm.


reverse transcription kit (Thermo Fisher). Quantitative RT-PCR ofcDNA (~30 ng/ml) was performed using assays containing fluores-cent hybridization probes (Taq Man: Thermo Fisher). Reactionswere incubated at 95 �C for 10 min and amplification was carriedout on samples with 2 min incubation at 50 �C, followed by 50cycles of 15 s at 95 �C and 1 min at 60 �C. The reaction for RT-PCRwas processed in a 15 ml solution containing 1X Universal PCRMaster Mix (Thermo Fisher) with 6 ml cDNA samples. RNAexpression at day 28 was compared to day0 samples using 18s as areference gene.

2.13. Calcium imaging

Calcium imaging was conducted on day 28 ESD-ALIC scaffolds toobserve neuronal function. For the assay, 50 mg Fluo-4 (Invitrogen)was reconstituted with 50 ml pluronic F-127 (Invitrogen) and thendiluted in DMEM at a 1:1000 dilution. The scaffolds were incubatedin 1 mL of the DMEM-Fluo-4 solution for 1 h at 37 �C, and thenwashed with PBS solution to remove excess dye. The stimulationsolution consisted of 250 mM menthol in DMEM and was addeddropwise onto the scaffolds while images were taken every 60 mswith a BZX-700 microscope (Keyence Corporation, Itasca, IL) at 10Xmagnification.

2.14. Neuronal extension measurement

Positive staining of b tubulin III was determined at 4X magni-fication. The images were collected from n ¼ 3 samples from 3independent experiments. All the stitched images were then con-verted into 8- bit tiff files using Image J (NIH). The neuron J routine[39] was then applied to measure axon length. The density of axonson 4X images was counted using Image J cell counter.

2.15. Statistical analysis

Data analysis was performed using Student's T test. The signif-icance level was set at p < 0.05. All experiments were run in at leasttriplicates for two independent experiments.

3. Results

3.1. Guidance of neuronal innervation

After 14 days of cultivation, the uniformly stamped silk filmsprovided higher axon density (112 ± 34 termini/mm2) than multi-circular (26 ± 5 termini/mm2) and radial stamped films (33 ± 8termini/mm2) (Supplementary Fig. 1). Thus, this strategy wasselected to guide neuronal innervation towards the center of thescaffolds in the remaining studies. After 28 days of cultivation theaxons were mostly located on the top surface of the scaffolds (Fig. 2A, B) indicating successful guidance of innervation. This guidancewas also studied at the ALIC (Fig. 2 C, D) where the top surface ofscaffolds had twice the density of axons than in the LC. The lengthand density of axons reached an average of 3 ± 0.7 mm and55 ± 10.61 termini/mm2 in the LC versus 4 ± 0.5 mm and99 ± 13.5 termini/mm2 in the ALIC. Thus, the combination ofstamped silk films and NGF loaded collagen supported the effectiveguidance of neuronal innervation towards the top center of thescaffolds.

3.2. Co-culture and single cultures in the liquid phase

The scaffolds for tri-culture in liquid remained intact andtransparent through 28 days (Fig. 1B). When hCECs were inner-vated in the ESD-LC and ED-LC groups, the morphology of thehCECs adopted a healthy polygonal epithelial cell morphology

Fig. 2. Immunohistochemistry staining of b III tubulin (green) showed axons being guided towards the top center of the scaffolds in the LCs and ALICs. A) The top surface of day 28D-LC sample. B) The bottom surface of day 28 D-LC sample. C) The top surface of day 28 D-ALIC sample. D) The bottom surface of day 28 D-ALIC sample. Scale bars ¼ 6 mm. (Forinterpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)


whereas in the E-LC group the cells were elongated (Fig. 3). HCECsaggregation was observed in the liquid phase single cultures andco-cultures (Fig. 3). The hCSSCs retained alignment at both culti-vation time points (Fig. 3). Innervation was developed in all the co-culture groups, with the SD-LC group resulting in axons that were~2 times longer than in the ED-LC group (Fig. 5 A). The innervationwas located on the top surface of scaffolds and between each layerof silk films (Fig. 3, Supplementary Fig. 4 B). There was no signifi-cant difference in length and density of axons between the SD-LCand ESD-LC (Fig. 5).

Fig. 3. Immunohistochemistry staining of DRGs, hCECs, and hCSSCs cultured alone (D-LC,hCSSCs, DRG tri-culture (ESD-LC) in liquid phase. b III tubulin was stained green in all the ssample. Keratocan was stained red in S-LC, SD-LC groups, middle and bottom layers of ESD

3.3. Co-culture and single cultures at the air-liquid interface

The integrity of the scaffolds and the transparency of the stackedfilms was maintained through 28 days of cultivation (Fig. 1B). Im-munostaining showed that the hCECs formed polygonal epithelialcell morphology (Fig. 4) and developed into multicellular layers inthe ESD-ALIC group (Supplementary Fig. 2). The presence of invo-lucrin in the day 28 ESD-ALIC group was highest among thedifferent conditions in ALIC and higher than the day 28 ESD-LCsamples (Figs. 3 and 4). The alignment of hCSSCs remained

E-LC, S-LC); hCECs, DRG co-culture (ED-LC); hCSSCs, DRG co-culture (SD-LC); hCECs,amples. Involucrin was stained red in E-LC, ED-LC groups and the top layer of ESD-LC-LC sample. The dashed boxes indicate the location of images. Scale bars ¼ 100 mm.

Fig. 4. Immunohistochemistry staining of DRGs, hCECs, and hCSSCs cultured alone (D-ALIC, E-ALIC, S-ALIC); hCECs, DRG co-culture (ED-ALIC); hCSSCs, DRG co-culture (SD-ALIC);hCECs, hCSSCs, DRG tri-culture (ESD-ALIC) at air-liquid interface. b III tubulin was stained green in all the samples. Involucrin was stained red in E-ALIC, ED-ALIC groups and the toplayer of ESD-ALIC sample. Keratocan was stained red in S-ALIC, SD-ALIC groups, middle and bottom layers of ESD-ALIC sample. The dashed boxes indicate the location of images.Scale bars ¼ 100 mm.

Fig. 5. Quantification of density and length of axons in day28 DRG single cultures in liquid phase (D-LC) and at air-liquid interface (D-ALI); DRG and hCSSCs co-culture in liquidphase (SD-LC) and at air-liquid interface (SD-ALIC); DRG co-culture with hCECs in liquid phase (ED-LC) and at air-liquid interfaces (ED-ALIC); DRG neuron, hCSSCs and hCECs tri-culture in liquid phase (ESD-LC) and at air-liquid interfaces (ESD-ALIC). The air-liquid interface culture supported significantly longer axons compared to the liquid cultures. D-ALICand ESD-ALIC groups provided the densest axons among the groups reaching an average of ~100 termini/cm2. Data was collected from n > 3 from three independent experiments.Standard deviation of each group is indicated as error bars. ***P < 0.0001; **P < 0.00; *P < 0.01.



through the cultures in the S-ALIC, SD-ALIC and ESD-ALIC systems(Fig. 4). The secretion of keratocan was also observed in the hCSSCssingle cultures and co-cultures. The DRGs cultured alone at the ALIChad an axon density that was approximately two-fold higher thanin the LC (Fig. 5). Innervation was observed at the top surface andbetween the silk film layers in the scaffolds (Fig. 4 SD-ALIC, ED-ALIC, ESD-ALIC, Supplementary Fig. 4 A). The dense innervationappeared in the ESD-ALIC group (80 termini/mm2), which was 3and 2 times higher than the ESD-LC and D-LC groups respectively(Fig. 5 B). Calcium imaging (Supplementary Fig. 3) showed morefiring when menthol was added into the medium, supportingneuron function in the tissue constructs.

3.4. Q-PCR analysis

In order to investigate the impact of innervation on the hCECs,expression of mRNA for involucrin (IVL), connexin 37 (GJA4), andcytokeratin-3 (KRT3) was quantified by qPCR. IVL is a marker formaturity whereas GJA4 documents accumulation of cell-cell junc-tions in the epithelial layer. IVL and GJA4 were expressed in allgroups, whereas KRT3 was only expressed in ED-ALIC, ESD-LC andESD-ALIC samples (Fig. 6 A). ED-ALIC had significantly higherexpression of IVL and GJA4 compared to ED-LC. Also, the ESD-LCand ESD-ALIC groups had significantly higher IVL and GJA4expression compared to ED-LC and ED-ALIC groups. The expressionof mRNA for keratocan (KERA), lumican (LUM), aldehyde dehy-drogenase (ALDH3A1) and smooth muscle actin (ACTA2) wasquantified to analyze the functional state of the hCSSCs (Fig. 6 B).KERA and LUM are corneal stroma ECM proteins that were previ-ously shown to be expressed by keratocytes differentiated fromhCSSCs [36]. ALDH3A1 is highly expressed in both epithelium andstromal cells of differentiated cornea and important for corneltransparency [40]. Smooth muscle actin (ACTA2) is a marker ofmyofibroblasts, cells involved in secretion of fibrotic non-transparent corneal tissue [41]. The expression of ACTA2 wasdown regulated in all the culture groups (Fig. 6 B). The S-LC grouphad the highest KERA and LUM expression compared to all theother groups. When hCSSCs were innervated (SD-ALIC), theexpression of KERA and LUMwere not different from the SD-LC andwere significantly higher than the S-ALIC. The expression of

Fig. 6. Gene expression of hCSSCs (A) and hCECs (B). The expression levels of mRNA for invPCR and normalized with day 0 samples of hCECs. The expression levels of mRNAs for kermuscle actin (ACTA2) were quantified through Q-PCR and normalized with day 0 samples oco-cultures compared to hCECs single cultures. hCSSCs expressed the most keratocan whexpression compared to the S-ALIC groups. The data were collected from n > 3 from thre***P < 0.0001; **P < 0.001; *P < 0.01.

ALDH3A1 was significantly greater in groups with innervationcompared to non-innervated samples.

4. Discussion

Compared to collagen-based corneal tissue models with inner-vation [15,42], the silk protein provided tunable materials to matchthe mechanical properties of human cornea [17,26,43]. Films andsponges prepared from silk supported aligned hCSSCs growth andimproved neuronal extensions [23,24]. In collagen-based tissuemodels, NGF was loaded into hydrogels to create a concentrationgradient to guide neuronal growth [12,32]. However, the densityand length of innervation were not quantified, and collagen un-dergoes consistent contraction over time that impacts cell func-tions. In our present study of corneal tissue models, NGF loadedcollagen gels were combined with NGF stamped silk films to guidethe axons towards the top center of the scaffolds, while concur-rently avoiding the problem of material contraction. Interactionsbetween cells sourced from chickens in contact with human cellshas been previously reported [44]. Dorsal root ganglion neuronswere used to mimic corneal innervation, due to its roles as a criticalcomponent as sensory neurons [45]. The average terminal densityand axon length reached 100 termini/mm2 and 4 mm in the ALICs.The guided, dense, and long axons establish an essential foundationto for innervated corneal tissue models. Further, these systemsremained functional for at least one month in culture, supportingsustained cultivation to allow both acute and chronic studies withthese new corneal tissues.

Previously, an air-liquid interface was achieved by culturingtissue constructs in trans-wells [14,46,47]. HCSSC survival in trans-wells was not robust in some of our preliminary experiments. Thus,PDMS shelves were designed to maintain a fluid environment forthe stroma while the epithelium was positioned at the ALIC. As aresult, the hCSSCs survived well in these systems.

After the scaffold design for neuronal innervation guidance andair-liquid cultivation was completed, the hCSSCs and hCECs wereincluded in the cultures. In our previous studies, we developedpatterned corneal stoma construct using RGD functionalized silkfilm stacks [26], which supported the secretion of aligned ECMfrom hCSSCs. In a more recent study, 2D co-culture systems with

olucrin (IVL), connexin 37 (GJA4) and cytokeratin 3 (KRT3) were quantified through Q-atocan (KERA), lumican (LUM), aldehyde dehydrogenase 3A1 (ALDH3A1) and smoothf hCSSCs. The expression level of IVL and GJA4 were significantly higher in ED and ESDen cultured in the liquid phase. SD-ALIC groups had significantly higher keratocane independent experiments. Standard error of each group is indicated as error bars.


silk films and silk collagen hydrogels improved DRG axonal devel-opment when co-cultured with hCSSCs. In order to include theepithelium, an important barrier layer for the cornea [3], HGF, KGFand EGFwere stamped on the top silk film layer and hCECs survivedthrough 28 days of cultivation in the LC and ALIC. Multilayer growthof hCECs was achieved in the ALIC systems, reflecting the impor-tance of the air-liquid environment to generate suitable outcomesfor these cornea tissues.

The expression of IVL, GJA4 and the number of epithelial cellularlayers in the innervated and air-liquid interface cultured sampleswere significantly higher than in the non-innervated samplescultivated in liquid phase, suggesting innervation and the air liquidinterface contributes towards achieving cell and tissue maturity ofthe corneal epithelium.

During the sustained cultivation in LC, the survival of hCSSCsappeared to decrease when hCECs were included in the system.This outcome was likely due to the hCECs remaining proliferativethroughout the cultivation which created competition for nutri-ents. This issue did not appear in the ALICs, which again supportedthe key role of this environment maintaining a healthy epitheliumand stroma. However, when hCSSCs were cultured alone, the LCprovided better KERA expression compared to the ALIC, indicatingthe liquid environment enhanced the secretion of ECM in thestroma.

The highly expressed ALDH3A1 and KERA in the innervatedstroma showed the essential role of innervation on corneal stromaltransparency and function [48]. In humans, it was observed that theimpairment of corneal innervation can cause corneal ulcers (neu-rotrophic keratitis) [49]. Patients with neurotrophic keratitis pre-sent decreased corneal sensitivity with alterations in cornealepithelium, nerve, keratocyte, and endothelium. Our findingscorroborate that corneal sensory nerves play a critical role inmaintaining the vitality, metabolism, and replenishment of cornealcells [50]. While preliminary in outcome in the present study, thedata suggest that this new corneal 3D tissue model has potential tohelp to explore and address these types of corneal diseases.

The interaction between corneal cells and innervationwas alsoobserved through the morphology. The axons developed from thebottom of silk sponge and grew towards the top of the scaffold insingle culture and co-culture groups. In the tri-cultures, the axonsbranched at the edge of scaffold and sprouted thin and long axonsthat grew in between stromal layers and on the epithelial layer.Epithelial innervation developed through cultivation formedclose connections with hCECs in the ALIC. In stromal layer, theaxons were guided by the pattern on the silk film and grew par-allel with hCSSCs. Through the longer and denser axons in theESD-ALIC compared to the ED-ALIC, a synergistic effect by thehCSSCs was revealed. This finding is in agreement with our pre-vious study which illustrated the importance of collagen type Iand BDNF secreted by hCSSCs in improving neuronal extensions[16]. Interestingly, we observed longer axons in all the ALICgroups. A similar effect was shown in other DRG neuron singlecultures and co-cultures with skin tissue at the air-liquid inter-face [51,52]. The results indicate the importance of the ALIC insupporting long and dense neuronal innervation at in vitroenvironment.

Cultivation time for current cornea tissues models was limitedto 1e2 weeks [15,42]. Cornea development takes 2 months in thehuman embryo [6]. Interactions between innervation and cornealtissue in longer-term cultivation are critical in this study, therobust silk protein based scaffold retained integrity and trans-parency through 28 days of cultivation. These results support theutility of these scaffold designs and bioreactor support for sus-tainable cultivation of 3D cornea tissues for a range of studies inthe future.

5. Conclusions

Innervated silk-based corneal tissue models were developedwhich supported long and dense neuronal innervation with multi-layer hCECs for the epithelium and aligned hCSSCs for the stromallayers. The impact of innervation on the corneal stroma andepithelium in sustained culture was demonstrated. With theseadvances in relevance, innervation, and time in culture, thesecorneal tissue models can be considered as supplements to animalmodels in areas such as drug development, disease interventionand in general corneal physiology research.

Acknowledgment

This work was supported by the National Institutes of Health[Grant number: R01 EY02085]. J Funderburgh was supported byR01 EY016415. The authors thank Erica P Kimmerling, James DWhite, and Nicole Raia for the insightful discussions.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.biomaterials.2016.09.030.

References

[1] P.G. McMenamin, C. Steeleand, C.N. McGhee, Cornea: Anatomy, Physiologyand Healing. Excimer Lasers in Ophthalmology, Martin Dunitz, London, 1997,pp. 41e63.

[2] D.W. DelMonte, T. Kim, Anatomy and physiology of the cornea, J. CataractRefract. Surg. 37 (3) (2011) 588e598.

[3] J.P. Whitcher, M. Srinivasan, M.P. Upadhyay, Corneal blindness: a globalperspective, Bull. World Health Organ. 79 (3) (2001) 214e221.

[4] D. Pascolini, S.P. Mariotti, Global estimates of visual impairment: 2010, Br. J.Ophthalmol. 96 (5) (2011) 614e618.

[5] B.S. Shaheen, M. Bakir, S. Jain, Corneal nerves in health and disease, Surv.Ophthalmol. 59 (3) (2014) 263e285.

[6] C.F. Marfurt, J. Cox, S. Deek, L. Dvorscak, Anatomy of the human cornealinnervation, Exp. Eye Res. 90 (4) (2010) 478e492.

[7] M.A. Al-Aqaba, U. Fares, H. Suleman, J. Lowe, H.S. Dua, Architecture and dis-tribution of human corneal nerves, Br. J. Ophthalmol. 94 (6) (2010) 784e789.

[8] J. He, N.G. Bazan, H.E. Bazan, Mapping the entire human corneal nerve ar-chitecture, Exp. Eye Res. 91 (4) (2010) 513e523.

[9] L.J. Müller, C.F. Marfurt, F. Kruse, T.M. Tervo, Corneal nerves: structure, con-tents and function, Exp. Eye Res. 76 (5) (2003) 521e542.

[10] S. Madduri, M. Papaloïzos, B. Gander, Synergistic effect of GDNF and NGF onaxonal branching and elongation in vitro, Neurosci. Res. 65 (1) (2009) 88e97.

[11] L. You, S. Ebner, F.E. Kruse, Glial cellederived neurotrophic factor (GDNF)eInduced migration and signal transduction in corneal epithelial cells, Investig.Ophthalmol. Vis. Sci. 42 (11) (2001) 2496e2504.

[12] L. You, F.E. Kruse, H. Volcker, Neurotrophic factors in the human cornea,Investig. Ophthalmol. Vis. Sci. 41 (3) (2000) 692e702.

[13] Y. Okada, P.S. Reinach, A. Kitano, K. Shirai, W.W.-Y. Kao, S. Saika, Neurotrophickeratopathy; its pathophysiology and treatment, Histol. Histopathol. 25 (4)(2010) 771.

[14] S. Reichl, J. Bednarz, C. Müller-Goymann, Human corneal equivalent as cellculture model for in vitro drug permeation studies, Br. J. Ophthalmol. 88 (4)(2004) 560e565.

[15] E.J. Suuronen, M. Nakamura, M.A. Watsky, P.K. Stys, L.J. Müller, R. Munger,N. Shinozaki, M. Griffith, Innervated human corneal equivalents as in vitromodels for nerve-target cell interactions, FASEB J. 18 (1) (2004) 170e172.

[16] S. Wang, C.E. Ghezzi, J.D. White, D.L. Kaplan, Coculture of dorsal root ganglionneurons and differentiated human corneal stromal stem cells on silk-basedscaffolds, J. Biomed. Mater. Res. Part A 103 (10) (2015) 3339e3348.

[17] B.A. Roeder, K. Kokini, J.E. Sturgis, J.P. Robinson, S.L. Voytik-Harbin, Tensilemechanical properties of three-dimensional type I collagen extracellularmatrices with varied microstructure, J. Biomech. Eng. 124 (2) (2002) 214e222.

[18] A. Elsheikh, D. Wang, M. Brown, P. Rama, M. Campanelli, D. Pye, Assessment ofcorneal biomechanical properties and their variation with age, Curr. Eye Res.32 (1) (2007) 11e19.

[19] H.J. Jin, J. Park, V. Karageorgiou, U.J. Kim, R. Valluzzi, P. Cebe, D.L. Kaplan,Water-stable silk films with reduced b-sheet content, Adv. Funct. Mater. 15(8) (2005) 1241e1247.

[20] X. Hu, K. Shmelev, L. Sun, E.-S. Gil, S.-H. Park, P. Cebe, D.L. Kaplan, Regulationof silk material structure by temperature-controlled water vapor annealing,Biomacromolecules 12 (5) (2011) 1686e1696.

[21] Y. Zeng, J. Yang, K. Huang, Z. Lee, X. Lee, A comparison of biomechanical


properties between human and porcine cornea, J. Biomech. 34 (4) (2001)533e537.

[22] J.A. Last, S.J. Liliensiek, P.F. Nealey, C.J. Murphy, Determining the mechanicalproperties of human corneal basement membranes with atomic force mi-croscopy, J. Struct. Biol. 167 (1) (2009) 19e24.

[23] J. Wu, J. Rnjak-Kovacina, Y. Du, M.L. Funderburgh, D.L. Kaplan,J.L. Funderburgh, Corneal stromal bioequivalents secreted on patterned silksubstrates, Biomaterials 35 (12) (2014) 3744e3755.

[24] M.D. Tang-Schomer, J.D. White, L.W. Tien, L.I. Schmitt, T.M. Valentin,D.J. Graziano, A.M. Hopkins, F.G. Omenetto, P.G. Haydon, D.L. Kaplan, Bio-engineered functional brain-like cortical tissue, Proc. Natl. Acad. Sci. 111 (38)(2014) 13811e13816.

[25] A.J. Coulombre, J.L. Coulombre, The role of intraocular pressure in the devel-opment of the chick eye: IV. Corneal curvature, AMA Arch. Ophthalmol. 59 (4)(1958) 502e506.

[26] G.J. Orssengo, D.C. Pye, Determination of the true intraocular pressure andmodulus of elasticity of the human cornea in vivo, Bull. Math. Biol. 61 (3)(1999) 551e572.

[27] X. Wang, J.A. Kluge, G.G. Leisk, D.L. Kaplan, Sonication-induced gelation of silkfibroin for cell encapsulation, Biomaterials 29 (8) (2008) 1054e1064.

[28] B.D. Lawrence, J.K. Marchant, M.A. Pindrus, F.G. Omenetto, D.L. Kaplan, Silkfilm biomaterials for cornea tissue engineering, Biomaterials 30 (7) (2009)1299e1308.

[29] E.S. Gil, B.B. Mandal, S.-H. Park, J.K. Marchant, F.G. Omenetto, D.L. Kaplan,Helicoidal multi-lamellar features of RGD-functionalized silk biomaterials forcorneal tissue engineering, Biomaterials 31 (34) (2010) 8953e8963.

[30] E.S. Gil, S.H. Park, J. Marchant, F. Omenetto, D.L. Kaplan, Response of humancorneal fibroblasts on silk film surface patterns, Macromol. Biosci. 10 (6)(2010) 664e673.

[31] D. Yao, S. Dong, Q. Lu, X. Hu, D.L. Kaplan, B. Zhang, H. Zhu, Salt-leached silkscaffolds with tunable mechanical properties, Biomacromolecules 13 (11)(2012) 3723e3729.

[32] S. Basu, A.J. Hertsenberg, M.L. Funderburgh, M.K. Burrow, M.M. Mann, Y. Du,K.L. Lathrop, F.N. Syed-Picard, S.M. Adams, D.E. Birk, Human limbal bio-psyederived stromal stem cells prevent corneal scarring, Sci. Transl. Med. 6(266) (2014), 266ra172-266ra172.

[33] Y. Du, M.L. Funderburgh, M.M. Mann, N. SundarRaj, J.L. Funderburgh, Multi-potent stem cells in human corneal stroma, Stem Cells 23 (9) (2005)1266e1275.

[34] P.G. Leclere, E. Norman, F. Groutsi, R. Coffin, U. Mayer, J. Pizzey, D. Tonge,Impaired axonal regeneration by isolectin B4-binding dorsal root ganglionneurons in vitro, J. Neurosci. 27 (5) (2007) 1190e1199.

[35] J. Wu, Y. Du, M.M. Mann, J.L. Funderburgh, W.R. Wagner, Corneal stromal stemcells versus corneal fibroblasts in generating structurally appropriate cornealstromal tissue, Exp. Eye Res. 120 (2014) 71e81.

[36] J. Wu, Y. Du, M.M. Mann, E. Yang, J.L. Funderburgh, W.R. Wagner, Bioengi-neering organized, multilamellar human corneal stromal tissue by growthfactor supplementation on highly aligned synthetic substrates, Tissue Eng.Part A 19 (17e18) (2013) 2063e2075.

[37] P. Chomczynski, N. Sacchi, Single-step method of RNA isolation by acid

guanidinium thiocyanate-phenol-chloroform extraction, Anal. Biochem. 162(1) (1987) 156e159.

[38] J. Chen, G.H. Altman, V. Karageorgiou, R. Horan, A. Collette, V. Volloch,T. Colabro, D.L. Kaplan, Human bone marrow stromal cell and ligamentfibroblast responses on RGD-modified silk fibers, J. Biomed. Mater. Res. Part A67 (2) (2003) 559e570.

[39] E. Meijering, M. Jacob, J.C. Sarria, P. Steiner, H. Hirling, M. Unser, Design andvalidation of a tool for neurite tracing and analysis in fluorescence microscopyimages, Cytom. Part A 58 (2) (2004) 167e176.

[40] J.V. Jester, Corneal crystallins and the development of cellular transparency,Semin. Cell Dev. Biol. 19 (2) (2008) 82e93.

[41] Y. Du, D.S. Roh, M.L. Funderburgh, M.M. Mann, K.G. Marra, J.P. Rubin, X. Li,J.L. Funderburgh, Adipose-derived stem cells differentiate to keratocytesin vitro, Mol. Vis. 16 (2010) 2680e2689.

[42] E.J. Suuronen, C.R. McLaughlin, P.K. Stys, M. Nakamura, R. Munger, M. Griffith,Functional innervation in tissue engineered models for in vitro study andtesting purposes, Toxicol. Sci. 82 (2) (2004) 525e533.

[43] D.N. Rockwood, R.C. Preda, T. Yücel, X. Wang, M.L. Lovett, D.L. Kaplan, Ma-terials fabrication from Bombyx mori silk fibroin, Nat. Protoc. 6 (10) (2011)1612e1631.

[44] H. Salehi, K. Karbalaie, S. Razavi, S. Tanhaee, M. Nematollahi, M. Sagha, M.-H. Nasr-Esfahani, H. Baharvand, Neuronal induction and regional identity byco-culture of adherent human embryonic stem cells with chicken notochordsand somites, Int. J. Dev. Biol. 55 (3) (2011) 321e326.

[45] C.-C. Chen, S. England, A.N. Akopian, J.N. Wood, A sensory neuron-specific,proton-gated ion channel, Proc. Natl. Acad. Sci. 95 (17) (1998) 10240e10245.

[46] J.-E. Chang, S.K. Basu, V.H. Lee, Air-interface condition promotes the formationof tight corneal epithelial cell layers for drug transport studies, Pharm. Res. 17(6) (2000) 670e676.

[47] S. Tegtmeyer, I. Papantoniou, C.C. Müller-Goymann, Reconstruction of anin vitro cornea and its use for drug permeation studies from different for-mulations containing pilocarpine hydrochloride, Eur. J. Pharm. Biopharm. 51(2) (2001) 119e125.

[48] A. Lambiase, M. Sacchetti, A. Mastropasqua, S. Bonini, Corneal changes inneurosurgically induced neurotrophic keratitis, JAMA Ophthalmol. 131 (12)(2013) 1547e1553.

[49] F. Semeraro, E. Forbice, V. Romano, M. Angi, M. Romano, M. Filippelli, R. DiIorio, C. Costagliola, Neurotrophic keratitis, Ophthalmologica 231 (4) (2014)191e197.

[50] T. Blanco-Mezquita, C. Martinez-Garcia, R. Proença, J.D. Zieske, S. Bonini,A. Lambiase, J. Merayo-Lloves, Nerve growth factor promotes cornealepithelial migration by enhancing expression of matrix Metalloprotease-9NGFpromotes epithelial migration, Investig. Ophthalmol. Vis. Sci. 54 (6) (2013)3880e3890.

[51] M. Gingras, J. Bergeron, J. D�ery, H.D. Durham, F. Berthod, In vitro developmentof a tissue-engineered model of peripheral nerve regeneration to studyneurite growth, FASEB J. 17 (14) (2003) 2124e2126.

[52] G. Bryce, R. Ribchester, Culture of isolated embryonic chick dorsal root gangliaat an air-liquid interface: a simple method for studying the mechanism andcontrol of neurite outgrowth, J. Neurosci. Methods 48 (1) (1993) 89e97.

in vitro 3d corneal tissue model with epithelium, stroma ... · in vitro 3d corneal tissue model...

Documents