lable at ScienceDirect

Biomaterials 32 (2011) 4076e4084

Contents lists avai

Biomaterials

journal homepage: www.elsevier .com/locate/biomateria ls

Human corneal endothelial cell growth on a silk fibroin membrane

Peter W. Madden a,b,*, Jonathan N.X. Lai a,c, Karina A. George a,d, Talia Giovenco a,d, Damien G. Harkin a,d,Traian V. Chirila a,b,e, f

aQueensland Eye Institute, 41 Annerley Road, South Brisbane, Queensland 4101, Australiab School of Medicine, University of Queensland, Herston, Queensland 4006, Australiac School of Medicine, The University of Melbourne, Melbourne, Victoria 3010, AustraliadMedical Sciences Discipline and Institute of Health and Biomedical Innovation, Queensland University of Technology, Brisbane, Queensland 4001, AustraliaeChemistry Discipline, Queensland University of Technology, Brisbane, Queensland 4001, AustraliafAustralian Institute for Bioengineering and Nanotechnology, University of Queensland, Brisbane, Queensland 4072, Australia

a r t i c l e i n f o

Article history:Received 14 November 2010Accepted 24 December 2010Available online 21 March 2011

Keywords:SilkFibroinCorneaEndotheliumCell cultureCollagen

* Corresponding author. Queensland Eye InstitutBrisbane, Queensland 4101, Australia. Tel.: þ61 7 3010

E-mail address: peter.madden@qei.org.au (P.W. M

0142-9612/$ e see front matter � 2011 Elsevier Ltd.doi:10.1016/j.biomaterials.2010.12.034

a b s t r a c t

Tissue engineering of the cornea could overcome shortages of donor corneas for transplantation andimprove quality. Our aim was to grow an endothelial layer on a substratum suitable for transplant.Silkworm (Bombyx mori) fibroin was prepared as 5 mm thick transparent membranes. The B4G12 cell linewas used to assess attachment and growth of human corneal endothelial cells on fibroin and comparethis with a reference substratum of tissue-culture plastic. To see if cell attachment and proliferation couldbe improved, we assessed coatings of collagen IV, FNC Coating Mix� and a chondroitin sulphateelamininmixture. All the coatings improved the final mean cell count, but consistently higher cell densities wereachieved on a tissue-culture plastic rather than fibroin substratum. Collagen-coated substrata were thebest of both groups and collagen-coated fibroin was comparable to uncoated tissue-culture plastic. Onlyfibroin with collagen coating achieved cell confluency. Primary human corneal endothelial cells werethen grown using a sphere-forming technique and when seeded onto collagen-coated fibroin they grewto confluency with polygonal morphology. We report the first successful growth of primary humancorneal endothelial cells on coated fibroin as a step in evaluating fibroin as a substratum for thetransplantation of tissue-constructs for endothelial keratoplasty.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

When the cornea is diseased or damaged, it can be surgicallyremoved and replaced with tissue from a deceased donor. Pene-trating keratoplasty, where the full thickness of the cornea isreplaced, has been the primary method to restore corneal clarityand sight. Unfortunately, penetrating keratoplasty weakens thestructural integrity of the eye and can adversely alter its opticalcharacteristics [1]. Moreover, it also involves transfer of epithelial,stromal cells and possibly antigen-presenting cells that can incitean immunological response [2]. The majority of transplants areperformed to replace solely a damaged endothelial layer [3], and soreplacing the other layers is ancillary and unnecessary. Conven-tional penetrating keratoplasty is therefore now being supplantedwith methods of keratoplasty that aim to replace the endothelial

e, 41 Annerley Road, South3302; fax: þ61 7 3010 3390.

adden).

All rights reserved.

cells alone (endothelial keratoplasty) [4]. At least 40,000 cornealtransplants are performed in the USA each year, and of these morethan 40% are now endothelial keratoplasty [5].

Transplantation of isolated endothelial cells has been performedin animals [6], but clinically a substratum is required to facilitatetransplantation of an organised monolayer of cells, and also toallow precise surgical insertion and fixation. This has been achievedin humans by the use of partial thickness transplants of donortissue [7]. There are two alternative methods: the first involvestransplanting with a thin layer of stroma, with disadvantagesincluding the transfer of other cell types and optical change; thesecond involves transplanting the endothelial layer attached onlyto its natural basement substratum, Descemet’s membrane [8],which is technically demanding. Both methods still require a donorcornea. Methods for the isolation and in vitro cultivation of humancorneal endothelial cells have been established [9], and trans-plantation of cells expanded in vitro may become an alternativetherapeutic option [10] compared with using cells directly froma donor. This would reduce the demand for donor tissue and allowready supply of material, with consistently high viable cell

P.W. Madden et al. / Biomaterials 32 (2011) 4076e4084 4077

numbers, as well as minimising the risk of microbial or priontransfer from the donor.

De-endothelialised corneas [11], isolated Descemet’s membrane[12], amniotic membrane [13], lens capsule [14], collagen I sheet[15], gelatine [16], chitosan [17] and lactide homo and copolymers[18], have all been used as substrata for corneal endothelial cellgrowth. The aim has been to provide a tissue-engineered layer thatcan be used instead of conventionally processed human donorcorneas.We are reporting here the use of a protein (fibroin) isolatedfrom the silk of the domesticated silkworm (Bombyx mori) asa substratum for viable human corneal endothelium. We havepreviously reported B. mori silk fibroin as a substratum for thegrowth of human corneal limbal epithelial cells [19,20], andtympanic cells [21].

Silks are produced mainly by the larvae of certain species in theorders Lepidoptera (moths and butterflies) and Araneae (spiders).They are fibres of high-molecular-weight polypeptide compositesbelonging to a group of scleroproteins (fibrous proteins), which alsoincludes collagens, elastins, keratins, and myosins. The fibresconsist of two major components, namely fibroin and sericin, thelatter forming an outer glue-like coating. There is a substantialinterest in the use of silks as biomaterials and a number of other celltypes have been grown on fibroin [20,22]. Fibroin membranes arestrong to facilitate handling and, most importantly for light trans-mission in the eye, can also be manufactured transparent [19,23].Because of these characteristics, there is also particular interest inthe use of fibroin in other parts of the cornea [24].

Adult human corneal endothelium has low proliferative activityin vivo andmay not divide after childhood [10]; most cells are static,often for longer than 50 years. They normally respond to adjacentcell death only by enlarging over the area, rather than dividing. Invitro, however, some adult corneal endothelial cells can be stimu-lated to divide by the use of strong mitogens [25]. Getting the cellsof older donors to divide is difficult though, and a technique usingfree-floating spheres of cells has been found beneficial [6], and thismay also select for cells with younger characteristics [26]. Theprimary function of the endothelium is to regulate water move-ment and for this to occur the morphology of the cells has beenshown to be important [10,27]. Only those cells that appear‘‘differentiated’’, with polygonal shapes characteristic of theappearance in vivo, were able to deswell the cornea; elongated,fibroblast-like cells had poorer function. For transplant materialtherefore, the polygonal cell morphology is preferable.

In our study, a human corneal endothelial cell line was first usedto assess cell attachment and growth features on fibroin comparedwith those on commercially available tissue-culture plastic (TCP),which is a surface-treated polystyrene. We then used the sphere-forming technique to prepare human primary corneal endothelialcells that were seeded onto transparent fibroin membranes. Theaim was to produce a tissue-engineered construct that is suitablefor endothelial keratoplasty.

2. Materials and methods

2.1. Materials

The source of silk was B. mori cocoons that had been cut open and the larvaeremoved (Tajima Shoji Co. Ltd., Yokohama, Japan).

2.2. Preparation of silk fibroin substrata

Aqueous silk fibroin solutionwas prepared according to our previously reportedprocedure [19]. Firstly, in order to degum and separate the silk fibres, the cocoonswere cut into pieces and boiled in a 0.02 M sodium carbonate solution, then rinsedand dried. High purity water was used throughout preparation. The fibroin wasdissolved at 60 �C in a 9.3 M solution of lithium bromide for 4 h, and then filtered toremove particulates and repeatedly dialyzed against water (molecular mass cut-off3.5 kDa). After 3 days, a solution of fibroin remained that was approximately 3.5% by

weight. Membranes were prepared using a 1.78% w/v fibroin dilution. This was thencast onto an optically flat glass surface that had previously been coated with a nowdry hydrophobic cyclic olefin copolymer (TOPAS Advanced Polymers, Frankfurt,Germany) backing film cast from a cyclohexane solution. A motorised doctor bladeset at a height of 0.90mm from the surface was drawn across the backing film to castan even thickness of fibroin solution. This fibroin solution was evaporated atambient temperature until dry. The fibroin membrane and backing film were thendislodged from the glass by allowingwater to seep between the backing film and theglass. However, unlike the previous method that used methanol, structural stabili-sation was instead performed by treatment in a humidified vacuum to improvetransparency [23]. The membrane and backing film were transferred to a vacuumoven at�80 kPa. A humid atmospherewasmaintained with awater reservoir for 6 hat ambient temperature. This makes the membrane water-stable [23]. The sacrificialbacking filmwas removed, and, after extensive rinsing in water, the thickness of thefibroin membrane was measured with a micrometer to be 5.0 � 1.2 mm. Themembrane was dried at ambient temperature and stored dry until use. Whenrequired, fibroin discs were cut with a 16 mm trephine and placed in the bottom of16 mm diameter TCP wells (24-well plate, Iwaki, Japan), held in place by sterilesilicone O-rings (Ludowici Seals, Pinkenba, Australia). The surface of the membranethat had been the air-side of the liquid fibroin was used for cell growth. Prior to use,the fibroin membrane and control wells were sterilised using 70% ethanol for 2 hfollowed by 3 rinses with Hanks Balanced Salt Solution (HBSS, with calcium andmagnesium, Invitrogen, Mulgrave, Australia), and left overnight in fresh HBSS.

2.3. Measurement of light transmission

Light transmission through the fibroin membrane was measured with a micro-plate spectrophotometer (Paradigm Absorbance Detection, Beckman Coulter, Brea,USA), using 16 mm fibroin discs in 1 mL of 0.01 M phosphate buffered saline (PBS)(SigmaeAldrich, Castle Hill, Australia) in a 24-well plate. The measurements wererepeatedwith triplicate samples and thebackgroundwasdeterminedwithPBSalone.

2.4. Coating of silk fibroin membranes

FNC Coating Mix� (FNC) is a commercial mix of bovine fibronectin, collagen Iand albumin (Athena Enzyme Systems, Baltimore, USA). Using primary humancorneal endothelium, Engler demonstrated that FNC significantly enhanced thespreading of cells on TCP, and that FNC was superior at reducing cell loss due torinsing [28]. Chondroitin sulphate and laminin has also been used as an attachmentfactor with human corneal endothelium [29]. We therefore compared FNC againstcollagen IV, and a chondroitin sulphate combined with laminin mix, as coatings forfibroin and TCP. Sterile collagen type IV was used (SigmaeAldrich, Castle Hill,Australia) dissolved in 0.1 M acetic acid. This was evaporated at �70 kPa and 40 �C ina vacuum oven overnight to give 10 mg/cm2. Following the manufacturer’s methodfor FNC, the solution was applied without drying, and after 1 h the excess wasdiscarded. The same application method was used for the chondroitin sulphate(10 mg/ml) and (10 mg/ml) laminin (SigmaeAldrich).

2.5. Human corneal endothelium cell line

Primary cells can differ markedly between donors; therefore, to enable consis-tent and repeatable comparison between groups a human corneal endothelial celllinewas used. The B4G12 endothelial cell line, that was a gift fromDrMonika Valtink(University of Dresden, Germany), has been immortalised by SV40 transfection [30].It was used at passage 146. The cells were split at confluence by treating with celldissociation medium for 2 min, and then trypsinised (TrypLe, Invitrogen) fora further 2e5 min. The cell suspension was diluted with HBSS and pelleted bycentrifugation for 5 min at 100 relative centripetal force (rcf). Viable cell density wascalculated using a haemocytometer and trypan blue (SigmaeAldrich) staining withstandardisation to 20,000 cells/cm2.

Serum supplemented growth medium that was established for primary cellculture by Konomi [31] was used for all the cell line experiments. It consisted of:Opti-MEM-I supplemented with 8% FBS, 5 ng/mL EGF, 20 ng/mL NGF, 100 mg/mLbovine pituitary extract, 20 mg/mL ascorbic acid, 200 mg/mL CaCl2 and 0.08% chon-droitin sulphate (all from Invitrogen). A reduced concentration of antibiotic-anti-mycotic was used at 0.1% v/v. All experiments were repeated in triplicate.

2.6. Primary human corneal endothelial cells

Corneas from deceased donors and the corneoscleral rims remaining followingpenetrating keratoplasty were obtained from the Queensland Eye Bank (Brisbane,Australia), after specific consent for research use and ethical approval (HREC/07/QPAH/48). We did not select for donor age and the donors ranged from 22 to 86years old. Endothelial cells were prepared based on a sphere-forming method [32].Under a dissecting microscope, the Descemet’s membrane with the endotheliumattached was carefully peeled away. To stabilise the cells, they were first incubatedovernight in a defined medium, CnT-20 (CELLnTEC, Bern, Switzerland), which isdesigned to support epithelial progenitor cells. Themediumwas supplementedwith200 mg/mL calcium chloride (SigmaeAldrich) and 0.1% v/v antibiotic-antimycotic

P.W. Madden et al. / Biomaterials 32 (2011) 4076e40844078

solution. The following day, pieces of Descemet’s visible to the unaided eye werepicked out and digested at 37 �C for 90min in the samemedium supplementedwith1 mg/mL collagenase (type 1A) (SigmaeAldrich, from Clostridium histolyticum). Thecollagenase was removed by dilution and the cells centrifuged at 100 rcf, thentiturated by pipette into single or small groups of cells, before being resuspended inthe supplemented CnT-20 medium. During 3e7 days of incubation at 37 �C with 5%CO2, distinct round spheres of endothelial cells formed. For the experiments, sphereswere individually removed by pipette to minimise transfer of non-endothelial cellcontaminants.

To confirm that the passaged primary cells could also grow on fibroin, cells firstisolated by sphere-forming culture and then grown to confluence on TPC wereremoved with trypsin. The cells were divided into two and each half seeded onto anequal area of collagen IV coated fibroin and TCP control.

2.7. Microscopy for cell counting

To assess endothelial cell line attachment and growth, phase contrast micro-graphs at �100 magnification of five random fields of view were recorded for up to15 days and the number of cells counted manually using ImageJ software [33], andthen averaged. Floating cells were not counted; only cells that displayed signs ofattachment were counted. These signs included loss of refractile appearance of thecell margin, angular cell extensions, or a distinguishable nucleus [28]. Cell attach-ment effect was defined as day 1 results. Cell proliferation or growth was defined asday 15 results. Results were compared at day 1 and 15 using a 2-way ANOVA toanalyse the main effect of substratum with coating. 1-way ANOVA was used tocompare uncoated fibroin, uncoated TCP, and fibroin coated with the coating thatgave the highest mean cell count. A Tukey’s post hoc test was used within the 1-wayANOVA to determine which pairings gave significant differences. For all ANOVAanalyses, p < 0.05 was taken as significant.

For counts of primary cell density, at least 100 contiguous cells from a repre-sentative area were counted. A minimum of 2000 cells/mm2 was regarded astransplantation density.

2.8. Immunohistochemical staining

For immuno-staining, the cells were first fixedwith 4% v/v formalin, rinsed threetimes with phosphate buffered saline (PBS) and then permeabilized with 0.3% Tritonbefore being rinsed again. Primary antibody for anti-ZO-1 (BD Biosciences, NorthRyde) was used at a 1:500 dilution with PBS. After the primary antibody wasremoved, goat serum was used to block other antigens. A fluorescent marker ofAlexa Fluor 488 secondary antibody (Invitrogen) was used at 1:100 against theprimary antibody attachment.

3. Results

3.1. Transmission of light of fibroin membrane

The discs of fibroin in PBS showed high light transmission(>96%) for the full visible spectrum (Fig. 1).

3.2. Growth of BG412 human corneal endothelial cell line on silkfibroin

The cell line B4G12 did not growwell on uncoated fibroin. Somecells did attach, but theyweremore aggregated comparedwith TCP

0.90.910.920.930.940.950.960.970.980.99

1

350 400 450 500 550 600 650 700 750

Tran

smitt

ance

Wavelength / nm

Fig. 1. Light transmission through silk fibroin membrane hydrated in PBS. Error barsare standard deviation. Transmission in the visible range is greater than 96%.

(Fig. 2g and h). Coatings did improve attachment, but cells on TCPgenerally were flatter with a larger surface area than the corre-sponding coating on fibroin (Fig. 2aef). Fig. 3 shows a comparisonof cell attachment counts on day 1 for coated and uncoated fibroinagainst that for uncoated TCP. When using a fibroin substratum,although FNC and chondroitin sulphateelaminin made someimprovement, only with collagen IV was the mean cell count closeto that for uncoated TCP.

Statistical analysis by 2-way ANOVA for attachment (day 1),demonstrated no main effect for coating (p ¼ 0.22), but there wasamain effect for substratum (p¼ 0.04). Thus, therewas a significantdifference between fibroin and TCP groups, but coating in generaldid not change cell counts significantly for the substratum.

A confluent layer of cells was never achieved with uncoatedfibroin, or when coated with FNC or chondroitin sulphateelaminin;in the course of the experiment, some areas always remaineddevoid of cells. The uncoated fibroin gave very low cell counts(Fig. 4). However, with collagen coating, the cells not only attachedmore uniformly, but they also had better proliferation and reachedconfluence within 2 weeks, in about 70% of the experiments. The 2-way ANOVA at day 15, however, returned a similar pattern: nomaineffect for coating (p ¼ 0.11) but a main effect for substratum(p ¼ 0.00).

Since collagen IV coating gave the highest mean cell counts onfibroin (well above the mean counts with FNC and chondroitinsulphateelaminin), a 1-way ANOVAwas used to compare uncoatedfibroin, uncoated TCP and fibroin coated with collagen IV. At day 1,there was no significant difference between these three groups(p ¼ 0.29). However, at day 15, there was a significant difference.Analysis by Tukey’s post hoc test demonstrated a differencebetween uncoated TCP and uncoatedfibroin (p¼ 0.04). On the otherhand, there was no significant difference shown between fibroincoated with collagen IV and uncoated TCP at day 15 (p ¼ 0.60) i.e.fibroin coated with collagen IV was comparable to uncoated TCP.

3.3. Growth of primary human endothelial cells on silk fibroin

When primary endothelial cells prepared by a sphere-formingtechnique were transferred to serum supplemented media, theyattached to uncoated TCP or collagen-coated fibroin and grew intoa confluent monolayer, some with transplantation level density(Fig. 5). Staining with ZO-1 displayed polygonal cells withpronounced staining at the cell margins with fine interdigitation.This was consistent with ‘differentiated’ morphological character-istics, similar to in vivo (Fig. 6).

3.4. Growth of passage one primary human endothelial cells on silkfibroin

When endothelial cells that had been first grown on TCP wereremoved with trypsin and passaged onto TCP or collagen-coatedfibroin, they grew with similar gross morphology and trans-plantation level density on both substrata (Fig. 7).

4. Discussion

This study demonstrated that silk fibroin can be used asa substratum for the culture of human corneal endothelial cells.

The cornea is the transparent window at the front of the eye. Itmust allow free passage of all visible light and perform primaryfocussing without aberration. Substitutes for any of its layers arerequired to be not only transparent, but also biocompatible withcells of the eye and have the necessary mechanical and functionalabilities to perform their role. The purpose of the corneal endo-thelial layer is to act as a barrier to the movement of water and to

Fig. 2. Phase contrast light microscopy of human corneal endothelial cell line B4G12 after 1 day of growth on: a) fibroin þ collagen IV; b) TCP þ collagen IV; c) fibroin þ FNC; d)TCP þ FNC; e) fibroin þ chondroitin sulphateelaminin; f) TCP þ chondroitin sulphateelaminin; g) fibroin; h) TCP. Cell attachment is better with the TCP substratum compared withthe corresponding coating on fibroin.

P.W. Madden et al. / Biomaterials 32 (2011) 4076e4084 4079

ceaselessly remove water from the bulk of the cornea, the stromallayer. Without such action the stroma swells and becomes opaque,resulting in blindness. To perform its duties the corneal endothe-lium is established as a monolayer and has to ensure continuously

apposed cells. The Descemet’s membrane is secreted by theendothelium, and its role is to act as a stable substratum. Therequirement for a stable organisation of the endothelial cells cannotbe over-emphasised.

0

0.5

1

1.5

2

2.5

3

fibroin +collagen

fibroin +FNC

fibroin +chondroitin

/ laminin

fibroin TCP +collagen

TCP +FNC

TCP +chondroitin

/ laminin

TCPCel

l atta

chm

ent r

atio

rela

tive

to u

ncoa

ted

TCP

Fig. 3. Comparison of attachment of human corneal endothelial cell line B4G12 between coatings for fibroin and TCP at day 1. The vertical axis represents the mean cell countrelative to that of uncoated TCP. Error bars are standard error of the mean. Although the collagen IV coated fibroin gives the highest mean cell counts, there was nonetheless, nomain effect for coating (p ¼ 0.22), but there was a main effect for substratum (p ¼ 0.04).

P.W. Madden et al. / Biomaterials 32 (2011) 4076e40844080

The reluctance of adult human corneal endothelial cells todivide, and the requirement that they have a polygonalmorphologyto functionwell, means that these issues have to be overcome in theculture technique. The sphere-forming technique that we used hasadvantages: not only did the cells grow with the requiredmorphology, but it also allowed selection of endothelial cells toavoid contamination by other cell types, and also it may select forprecursor (progenitor) cells [6]. The isolation of the endothelialcells in our study used CnT-20 media, which is a serum and bovinepituitary extract-free formulation. Serum was required in the finalgrowth media, however, to ensure the cells came out of sphereorganisation and populated the fibroin. Further studies will benecessary to see if a defined media could also be used for this stage,

0 2 4 6 80

500

1000

1500

2000

2500

3000

3500

4000

Days in Cultur

Mea

n C

ell C

ount

(cel

ls /

mm

2 )

Fig. 4. Longer term comparison of endothelial cell line growth on silk fibroin membrane. Coa significant difference between uncoated TCP and uncoated fibroin (p ¼ 0.04), but there wasTCP at day 15 (p ¼ 0.60); fibroin coated with collagen IV was comparable to uncoated TCP

as this would reduce recipient risk from animal-derived infectiousagents.

Forfibroin to be successful as a temporary substitute for a corneallayer, we need to consider its transparency, biocompatibility,mechanical abilities and functional abilities. Regarding trans-parency, silk fibroin can be made optically transparent. Ourmembranes at 98.0%� 1.4 SD at 500 nmhave transmission of visiblelight greater than the central corneal stroma which achieves87.1%� 2.0 SD, and less than this further towards the periphery [34].

According to Williams [35], biocompatibility “refers to the abilityto perform as a substrate that will support the appropriate cellularactivity, including the facilitation of molecular and mechanical sig-nalling systems, in order to optimise tissue regeneration, without

10 12 14 16e

TCP +collagenTCP +FNCTCP +chondroitin/lamininTCP

fibroin +collagenfibroin +FNCfibroin +chondroitin/lamininfibroin

unts for the TCP substrata are consistently greater than those for fibroin, with at day 15no significant difference shown between fibroin coated with collagen IV and uncoated

.

Fig. 5. Phase contrast light micrographs of primary human endothelial cells on silk fibroin membrane. a) Day 0. A single sphere of endothelial cells 4e6 cells in diameter is shownattached to the fibroin (white arrow). b) Day 11. The endothelial cells spread from this and other spheres. c) Day 17. The cells proliferate and density increases until at, d) Day 25,a confluent monolayer is formed. Folds formed from air bubbles trapped under the silk fibroin membrane act as reference point (white stars).

Fig. 6. Fluorescence microscopy of confluent primary endothelial cells on silk fibroinmembrane, ZO-1 pronounced staining at margins. The cells display ‘differentiated’morphological characteristics, similar to in vivo with polygonal shaped cells and fineinterdigitation.

P.W. Madden et al. / Biomaterials 32 (2011) 4076e4084 4081

eliciting any undesirable local or systemic responses in the eventualhost”. Virgin silk, composed of fibroin and sericin, has a long recordof use as a surgical suture dating back to antiquity [36]. Virgin silksutures can elicit inflammatory responses [37], which have beenattributed to the allergenic activity of sericin, mainly following pre-sensitisation [38]. With pure fibroin, allergenic activity is very rare,although there has been a single report showing that it is possibleto be hypersensitive to fibroin itself [39], again through sensitisa-tion by prior exposure to surgical silk. Despite this, silk has beenused very extensively, even in the eye [40]. During the last twentyfive years synthetic sutures have been in more widespread use, andso the risk of prior sensitisation is not only small, but should also bedecreasing in the population. Fibroin is eventually broken down tosoluble peptides which are capable of being phagocytosed forfurther metabolism by the cell [37], and so its degradation productscan be cleared without local cell damage.

Engler has shown, with primary human corneal endothelium,that both collagen I and FNC significantly enhanced the spreadingof cells on TCP, and that FNC was superior at reducing cell loss dueto rinsing [28]. Engler suggested that the most physiologicapproach to endothelial attachment would be to investigatecollagen VIII, rather than collagen I, because collagen VIII is abun-dant in Descemet’s membrane, and it can reduce loss of other celltypes from shear stress. We can see the value of this approach, butinstead, we used collagen IV for a number of reasons: (1) In base-ment membranes in general, type IV collagen is the most abundantprotein along with laminin, nidogen/entactin, and perlecan [41]. (2)

Fig. 7. Phase contrast light microscope images of primary human endothelial cells after 20 days in passage 1. a) Cells on fibroin coated with collagen IV and, b) on TCP. Grossmorphology is similar for the different substrata.

P.W. Madden et al. / Biomaterials 32 (2011) 4076e40844082

Collagen IV is also present in Descemet’s membrane, and indeed,unlike type VIII, it is located on the endothelial side of themembrane, not only in the infant, but also in the adult [42]. (3)Human limbal corneal epithelial cells that have stem cell propertiescan be partially enriched by their adhesiveness to collagen IV [43].(4) Both type IV and VIII collagen can self-assemble into a networkforming suprastructure, but with type IV this is a crucial part ofbasement membrane stability [41]. (5) Collagen IV has also beenshown to support the formation of a corneal endothelial monolayerwith a morphology similar to that in vivo [44]. This ‘differentiated’[25] endothelium would be expected to offer greater opportunityfor normal function than endothelium that does not show thesecharacteristics. (6) Furthermore, in posterior polymorphous cornealdystrophy, where division of the endothelium increases unnatu-rally, the localisation and distribution of both collagen IV and VIIIincreases in Descemet’s membrane, and it has been suggested thatthis may contribute to the stimulation of its proliferative capacity[45]. We suggest that, if actively dividing cells produce morecollagen IV, then it may be appropriate to use this collagen fora construct where spreading and dividing cells are desirable.

We have been able to show that fibroin, when coated withcollagen IV, can support the growth of a confluent human cornealendothelium, which is an indication of biocompatibility with thesecells. We now plan to confirm more systemic biocompatibility invivo, in an appropriate animal model.

Mechanically, human Descemet’s membrane has a wetmeasured tensile strength of 0.14 MPa [46]. Fibroin that has notbeen water annealed has a wet strength of 13.8 MPa [47]. Waterannealing increases tensile strength and elasticity [23]. Althoughwe did not measure the tensile strength of our samples, the fibroinis expected to be at least an order of magnitude stronger than theDescemet’s membrane. The latter is approximately 6 mm thick at 2years of age and further thickens with age, reaching 16.5 mm by 90years of life [48]. Descemet’s membrane can be used alone totransplant the endothelial layer [8], and fibroin appears mechan-ically suitable for such a role too, employing conventional surgicaltechniques. Additionally, because fibroin can be cast and manu-factured in many forms, then it could be engineered for better easeof handling compared with Descemet’s membrane.

The primary functional requirement of the endothelium is toremove water from the stroma. Normal endothelial pump functionis 25e40 mL/cm2/hr (measured in rabbit) [49] and so the perme-ability of fibroin at 52.5 mL/cm2/hr [50] can accommodate sucha flow. An advantage of tissue-engineered constructs is the abilityof fibroin to degrade at a relatively slow rate in a biological envi-ronment [51], which should allow cells time to first establish and

deposit their own substratum. In our study, we chose fibroin of5 mm thickness, but we can also manufacture membranes of otherthickness, including those as thin as 2 mm. To improve permeabilityto water, ions or molecules these thinner membranes can also bemade with micron diameter pores. The endothelial cells can alsogrow on these thinner, porous membranes (data not shown).

Despite fibroin’s transparency, biocompatibility, strength andpermeability, we found that, without coating, it did not supportgood attachment and growth of the human corneal endothelial cellline B4G12, or primary endothelial cells. This differs from otherhuman cells; for example with fibroblasts the initial cell attach-ment and growth ratios were similar for collagen (type not stated,but assumed to be type I) compared with fibroin [52]. Along withprevious studies [28,29], we have shown that FNC, chondroitinsulphateelaminin and collagen can all improve attachment andgrowth on TCP of human corneal endothelium. Furthermore, in ourstudy, growth on fibroin only approached that on TCP if collagenwas used as a coating. In Engler’s study of B4G12 on TCP [28], FNCwas found to be significantly better as an attachment factor thancollagen I. We did not find FNC superior to collagen IV. However,with the fibroin, we did not assess the level of attachment of FNC.Cell adhesion to polymeric substrates is mediated by extracellularmatrix proteins. Hydrophobic polymer surfaces usually allowbetter adsorption of proteins, which then facilitate cell adhesion[53]. Water annealed fibroin is not hydrophobic [23], and it may bethe case that binding of the FNC and chondroitinelaminin to fibroindid not occur, or that there was aggregation. Assembly of collagenIV differs depending upon whether the substratum is hydrophobicor hydrophilic [54]. With vascular endothelial cells, this in turn hasalso been shown to alter cell interaction, with hydrophobic surfacesresulting in aggregation of collagen that was not as favourable forcell growth as the more uniform deposit on a hydrophilic surface[54]. With fibroin, we found that deposition of collagen IV wassufficient for sizeable cell growth, but it may be the case that byadjusting fibroin hydrophobicity this could be further improved.

Coating with collagen IV did allow the best attachment andgrowth on silk fibroin. Any coating, however, introduces additionalantigens and adverse reaction to collagen can occur [55]. It istherefore desirable to avoid the use of coatings, particularlymammalian-derived coatings. In our study we used fibroin fromthe domesticated silkworm B. mori. This substratum has beenshown for fibroblasts to give cell attachment and growth that isclose to that with collagen [52]. In this fibroblast study though,uncoated fibroin from another species of moth, Antheraea pernyi,displayed higher cell attachment and growth compared with thatfrom B. mori. This has been attributed to the presence of the

P.W. Madden et al. / Biomaterials 32 (2011) 4076e4084 4083

tripeptide sequence ArgeGlyeAsp (RGD) in A. pernyi fibroin. TheRGD sequence is well-recognised as being a specific interaction sitefor cell attachment, and it will be valuable to see if fibroin from analternative moth species can support adequate corneal endothelialgrowth and avoid the need for a coating.

In our study, not only the B4G12 cell line at confluence onfibroin, but also the primary cells exceeded minimum trans-plantation density. With primary cells, growth varied betweendonors as in previous studies [10] and it depends upon donorselection. For this reason primary cell density was not rigorouslycalculated using multiple experiments with multiple counts fromindividual samples as was the case with the cell line. Endothelialcell counts used by the Eye Banks to assess donor suitability areroutinely based upon single-area sampling from the central corneaof less than 100 cells. Minimum cell density suitable for trans-plantation is not determined by the major Eye Banking organisa-tions, which leave the decision to individual Eye Banks or surgeons.Few surgeons would accept <2000 cells/mm2 though, and studiesof transplant success may be limited to initial counts of>2300 [56].Cell density with our primary cells reached transplantation density,but further studies will be required to quantify how best to achievethe highest density consistently. Similarly, although the cells hadappropriate morphology, the ability of the endothelial pump tomove water is an important parameter to measure in future.

5. Conclusions

We have shown that silk fibroin can be prepared as a trans-parent membrane that supports the growth of human cornealendothelium with gross normal morphology. These attributesfurther the potential application of fibroin for clinical cornealendothelial transplantation.

Acknowledgements

This work was supported by the Prevent Blindness Foundationat the Queensland Eye Institute, Brisbane, Australia. The AdvancedMedical Science course, Melbourne University, Australia provideda placement grant for JL. Corneas were kindly provided by theQueensland Eye Bank, Queensland Health, Brisbane, Australia. Wethank Dr Monika Valtink of the University of Dresden, Germany forproviding as a gift the B4G12 endothelial cell line.

References

[1] Lee WB, Jacobs DS, Musch DC, Kaufman SC, Reinhart WJ, Shtein RM. Desce-met’s stripping endothelial keratoplasty: safety and outcomes: a report by theAmerican Academy of Ophthalmology. Ophthalmology 2009;116:1818e30.

[2] Hamrah P, Dana MR. Corneal antigen-presenting cells. Chem Immunol Allergy2007;92:58e70.

[3] Ghosheh F, Cremona F, Rapuano C, Cohen E, Ayres B, Hammersmith K, et al.Trends in penetrating keratoplasty in the United States 1980e2005. IntOphthalmol 2008;28:147e53.

[4] Price MO, Price Jr FW. Endothelial keratoplasty e a review. Clin ExperimentOphthalmol 2010;38:128e40.

[5] Eye banking statistics for 2008. Press release. Eye Bank Association of Amer-ica; 2009.

[6] Mimura T, Yamagami S, Yokoo S, Yanagi Y, Usui T, Ono K, et al. Sphere therapyfor corneal endothelium deficiency in a rabbit model. Invest Ophthalmol VisSci 2005;46:3128e35.

[7] Tan DT, Anshu A, Mehta JS. Paradigm shifts in corneal transplantation. AnnAcad Med Singap 2009;38:332e8.

[8] Melles GR, Lander F, Rietveld FJ. Transplantation of Descemet’s membranecarrying viable endothelium through a small scleral incision. Cornea 2002;21:415e8.

[9] Engelmann K, Bednarz J, Valtink M. Prospects for endothelial transplantation.Exp Eye Res 2004;78:573e8.

[10] Joyce NC. Proliferative capacity of the corneal endothelium. Prog Retin Eye Res2003;22:359e89.

[11] Gospodarowicz D, Greenburg G, Alvarado J. Transplantation of culturedbovine corneal endothelial cells to rabbit cornea: clinical implications forhuman studies. Proc Natl Acad Sci U S A 1979;76:464e8.

[12] Chen K-H, Azar D, Joyce NC. Transplantation of adult human corneal endo-thelium ex vivo: a morphologic study. Cornea 2001;20:731e7.

[13] Ishino Y, Sano Y, Nakamura T, Connon CJ, Rigby H, Fullwood NJ, et al. Amnioticmembrane as a carrier for cultivated human corneal endothelial cell trans-plantation. Invest Ophthalmol Vis Sci 2004;45:800e6.

[14] Yoeruek E, Saygili O, Spitzer MS, Tatar O, Bartz-Schmidt KU, Szurman P.Human anterior lens capsule as carrier matrix for cultivated human cornealendothelial cells. Cornea 2009;28:416e20.

[15] Mimura T, Yamagami S, Yokoo S, Usui T, Tanaka K, Hattori S, et al. Culturedhuman corneal endothelial cell transplantation with a collagen sheet ina rabbit model. Invest Ophthalmol Vis Sci 2004;45:2992e7.

[16] Hsiue GH, Lai JY, Chen KH, Hsu WM. A novel strategy for corneal endothelialreconstruction with a bioengineered cell sheet. Transplantation 2006;81:473e6.

[17] Gao X, Liu W, Han B, Wei X, Yang C. Preparation and properties of a chitosan-based carrier of corneal endothelial cells. J Mater Sci Mater Med 2008;19:3611e9.

[18] Hadlock T, Singh S, Vacanti JP, McLaughlin BJ. Ocular cell monolayers culturedon biodegradable substrates. Tissue Eng Part A 1999;5:187e96.

[19] Chirila TV, Barnard Z, Zainuddin, HarkinDG, Schwab IR,Hirst L.Bombyxmori silkfibroin membranes as potential substrata for epithelial constructs used in themanagement of ocular surface disorders. Tissue Eng Part A 2008;14:1203e11.

[20] Chirila TV, Barnard Z, Zainuddin, Harkin D. Silk as substratum for cellattachment and proliferation. Mater Sci Forum 2007;561e565:1549e52.

[21] Ghassemifar R, Redmond S, Zainuddin, Chirila TV. Advancing towards a tissue-engineered tympanic membrane: silk fibroin as a substratum for growinghuman eardrum keratinocytes. J Biomater Appl 2010;24:591e606.

[22] Wang Y, Kim H-J, Vunjak-Novakovic G, Kaplan DL. Stem cell-based tissueengineering with silk biomaterials. Biomaterials 2006;27:6064e82.

[23] Jin HJ, Park J, Karageorgiou V, Kim UJ, Valluzzi R, Cebe P, et al. Water-stablesilk films with reduced b-sheet content. Adv Funct Mater 2005;15:1241e7.

[24] Lawrence BD, Marchant JK, Pindrus MA, Omenetto FG, Kaplan DL. Silk filmbiomaterials for cornea tissue engineering. Biomaterials 2009;30:1299e308.

[25] Joyce NC, Zhu CC. Human corneal endothelial cell proliferation: potential foruse in regenerative medicine. Cornea 2004;23(8 Suppl.):S8e19.

[26] Mimura T, Yamagami S, Yokoo S, Usui T, Amano S. Selective isolation of youngcells from human corneal endothelium by the sphere-forming assay. TissueEng Part C 2010;16:803e12.

[27] Engelmann K, Drexler D, Bohnke M. Transplantation of adult human orporcine corneal endothelial cells onto human recipients in vitro. Part I: cellculturing and transplantation procedure. Cornea 1999;18:199e206.

[28] Engler C, Kelliher C, Speck CL, Jun AS. Assessment of attachment factors forprimary cultured human corneal endothelial cells. Cornea 2009;28:1050e4.

[29] Engelmann K, Bohnke M, Friedl P. Isolation and long-term cultivation ofhuman corneal endothelial cells. Invest Ophthalmol Vis Sci 1988;29:1656e62.

[30] Valtink M, Gruschwitz R, Funk RHW, Engelmann K. Two clonal cell lines ofimmortalized human corneal endothelial cells show either differentiated orprecursor cell characteristics. Cells Tissues Organs 2008;187:286e94.

[31] Konomi K, Zhu C, Harris D, Joyce NC. Comparison of the proliferative capacityof human corneal endothelial cells from the central and peripheral areas.Invest Ophthalmol Vis Sci 2005;46:4086e91.

[32] Yokoo S, Yamagami S, Yanagi Y, Uchida S, Mimura T, Usui T, et al. Humancorneal endothelial cell precursors isolated by sphere-forming assay. InvestOphthalmol Vis Sci 2005;46:1626e31.

[33] Rasband WS. Imagej. Bethesda, Maryland, USA: U.S. National Institutes ofHealth, http://rsbinfonihgov/ij/; 1997e2009.

[34] Doutch J, Quantock AJ, Smith VA, Meek KM. Light transmission in the humancornea as a function of position across the ocular surface: theoretical andexperimental aspects. Biophys J 2008;95:5092e9.

[35] Williams DF. On the mechanisms of biocompatibility. Biomaterials 2008;29:2941e53.

[36] De Methodo Medendi Galen, Mackenzie D. From translation of the first twobooks. In: Hankinson RJ, editor. Recorded at the Scottish Society of History ofMedicine 68th generalmeeting 1971. Oxford: Clarendon Press; 1991. p. AD150.

[37] Altman GH, Diaz F, Jakuba C, Calabro T, Horan RL, Chen J, et al. Silk-basedbiomaterials. Biomaterials 2003;24:401e16.

[38] Soong HK, Kenyon KR. Adverse reactions to virgin silk sutures in cataractsurgery. Ophthalmology 1984;91:479e83.

[39] Kurosaki S, Otsuka H, Kunitomo M, Koyama M, Pawankar R, Matumoto K.Fibroin allergy IgE mediated hypersensitivity to silk suture materials. J NipponMed Sch 1999;66:41e4.

[40] Chatterjee S. Comparative trial of DEXON (polyglycolic acid), collagen, and silksutures in ophthalmic surgery. Br J Ophthalmol 1975;59:736e40.

[41] LeBleu VS, Macdonald B, Kalluri R. Structure and function of basementmembranes. Exp Biol Med (Maywood) 2007;232:1121e9.

[42] Kabosova A, Azar DT, Bannikov GA, Campbell KP, Durbeej M, Ghohestani RF,et al. Compositional differences between infant and adult human cornealbasement membranes. Invest Ophthalmol Vis Sci 2007;48:4989e99.

[43] Li DQ, Chen Z, Song XJ, de Paiva CS, Kim HS, Pflugfelder SC. Partial enrichmentof a population of human limbal epithelial cells with putative stem cellproperties based on collagen type IV adhesiveness. Exp Eye Res 2005;80:581e90.

P.W. Madden et al. / Biomaterials 32 (2011) 4076e40844084

[44] Engelmann K, Friedl P. Optimization of culture conditions for human cornealendothelial cells. In Vitro Cell Dev Biol 1989;25:1065e72.

[45] Merjava S, Liskova P, Sado Y, Davis PF, Greenhill NS, Jirsova K. Changes in thelocalization of collagens IV and VIII in corneas obtained from patients withposterior polymorphous corneal dystrophy. Exp Eye Res 2009;88:945e52.

[46] Danielsen CC. Tensile mechanical and creep properties of Descemet’smembrane and lens capsule. Exp Eye Res 2004;79:343e50.

[47] Tang Y, Cao C, Ma X, Chen C, Zhu H. Study on the preparation of collagen-modified silk fibroin films and their properties. Biomed Mater 2006;1:242.

[48] Murphy C, Alvarado J, Juster R. Prenatal and postnatal growth of the humanDescemet’s membrane. Invest Ophthalmol Vis Sci 1984;25:1402e15.

[49] Dikstein S. Efficiency and survival of the corneal endothelial pump. Exp EyeRes 1973;15:639e44.

[50] Minoura N, Tsukada M, Nagura M. Physico-chemical properties of silk fibroinmembrane as a biomaterial. Biomaterials 1990;11:430e4.

[51] Wang Y, Rudym DD, Walsh A, Abrahamsen L, Kim H-J, Kim HS, et al. In vivodegradation of three-dimensional silk fibroin scaffolds. Biomaterials 2008;29:3415e28.

[52] Minoura N, Aiba S, Higuchi M, Gotoh Y, Tsukada M, Imai Y. Attachment andgrowth of fibroblast cells on silk fibroin. Biochem Biophys Res Commun1995;208:511e6.

[53] Campillo-Fernandez AJ, Unger RE, Peters K, Halstenberg S, Santos M, SalmeronSanchez M, et al. Analysis of the biological response of endothelial andfibroblast cells cultured on synthetic scaffolds with various hydrophilic/hydrophobic ratios: influence of fibronectin adsorption and conformation.Tissue Eng Part A 2009;15:1331e41.

[54] Coelho NM, Gonzalez-Garcia C, Planell JA, Salmeron-Sanchez M, Altankov G.Different assembly of type IV collagen on hydrophilic and hydrophobicsubstrata alters endothelial cells interaction. Eur Cells Mater 2010;19:262e72.

[55] Lynn AK, Yannas IV, Bonfield W. Antigenicity and immunogenicity of collagen.J Biomed Mater Res Part B 2004;71:343e54.

[56] Lass JH, Gal RL, Dontchev M, Beck RW, Kollman C, Dunn SP, et al. Donor ageand corneal endothelial cell loss 5 years after successful corneal trans-plantation. Specular microscopy ancillary study results. Ophthalmology2008;115:627e32.