Human corneal endothelial cells from older donors can be cultured and passaged on cell‐derived extracellular matrix

To investigate the effect of culturing human corneal endothelial cells (HCEnCs) from older donors on extracellular matrix (ECM) derived from human corneal endothelial cell line (HCEC‐12).


Introduction
Human corneal endothelial cells (HCEnCs) are the posterior hexagonal monolayer of neural crest-derived cells that maintain optimal corneal hydration (Joyce 2003). As a result, these cells are responsible for maintaining the transparency required for clear vision, and any damage or absence of endothelium can lead to corneal oedema (Joyce 2003). Relatively recent advances in the field of endothelial keratoplasty, techniques such as Descemet Stripping Automated Endothelial Keratoplasty (DSAEK) and Descemet Membrane Endothelial Keratoplasty (DMEK; Melles 2006), have efficiently replaced the damaged recipient endothelial layer with healthy donor endothelium (Parekh et al. 2013). However, these techniques rely heavily on donor material and are mostly limited to one donor per patient. A global survey has reported that the supply to demand ratio of corneal tissues is about 1:70, indicating a huge demand of corneal tissues worldwide (Gain et al. 2016). Therefore, to overcome this limitation, HCEnCs have been isolated from donor tissues and cultured in vitro using different techniques. Rho-kinase inhibitor (ROCK) supplemented cultures in particular have shown the most promise so far (Okumura et al. 2012). Although HCEnCs do not have regenerative capacity in vivo, HCEnCs can proliferate in vitro provided optimal conditions are available for cell growth (Joyce 2005). Many studies have suggested techniques to isolate and propagate these cells in vitro (Peh et al. 2011a(Peh et al. ,b, 2013a(Peh et al. ,b, 2015Parekh et al. 2016Parekh et al. , 2017Parekh et al. , 2019a with only one successful clinical trial reported so far (Kinoshita et al. 2018).
Having a higher endothelial cell density (ECD) compared to old aged donor tissues (>60 years of age), young donors seem to be an ideal choice for culturing HCEnCs (Parekh et al. 2017;Kinoshita et al. 2018). However, due to higher endothelial cell counts, young donor tissues are often selected for transplantation, which is the principle aim of corneal donation. It is therefore difficult to source these tissues for cell culture optimization and clinical trials. Simultaneously, old aged donor tissues are easy to obtain due to endothelial cell counts that are usually lower than the threshold required for transplantation. As the old aged donor cells go into cell senescence quickly, it becomes challenging to culture these cells (Parekh et al. 2017). However, if these cells can be efficiently cultured in vitro, then we may utilize available donor material more effectively. In the studies outlined here, we therefore used old-age donor HCEnCs to overcome the challenge associated with young donor tissue availability.
Several studies have shown that corneal endothelial cells strongly attach and interact with Descemet's membrane in vivo and their extracellular matrix (ECM) in vitro (Underwood & Bennett 1993;Joyce 2003;Engler et al. 2009). Human corneal endothelial cells have been shown to produce an ECM that helps in promoting cellular adhesion, proliferation, migration, morphogenesis and differentiation (Hsieh & Baum, 1985;Underwood & Bennett, 1993). ECM proteins found in Descemet's membrane include fibronectin, laminin, collagen (IV, VI and VIII) and proteoglycans containing heparan sulphate, dermatan sulphate or keratan sulphate (Yue et al. 1978;Hassell et al. 1980;Lee & Davison 1984;Nakayasu et al. 1986;Zimmermann et al. 1986;Gordon 1988;Marshall et al. 1991;Choi et al. 2013). The ECM laid by corneal endothelial cells may have all the proteins and growth factors that are required to promote cell proliferation and maintenance of HCEnCs in culture. Thus, in the studies outlined here, our aims were to investigate the effect of corneal endothelial cellderived ECM on HCEnC culture to optimize the culture of old aged donor cells and to evaluate the proteins in this ECM.

Materials and Methods
Ethical statement and donor characteristics Human donor corneas were shipped from Fondazione Banca degli Occhi del Veneto Onlus (FBOV, Venice, Italy) to UCL Institute of Ophthalmology (London, UK) with written consent for research use as they were not suitable for transplantation due to poor endothelial cell count (<2200 cells/ mm 2 ). The tissues were utilized and discarded as per standard Human Tissue Authority (HTA, UK) requirements. The experiments were approved by the UCL ethics committee (10/ H0106/57-2011ETR10) and were performed in accordance with the Declaration of Helsinki.

Endothelial cell evaluation
Donor endothelium of all the tissues was stained with trypan blue (0.25% wt/ vol), a vital dye that selectively stain dead cells (VisionBlue, D.O.R.C., Zuidland, The Netherlands), to detect the percentage of dead/necrotic cells. Approximately 100 µl of trypan blue was applied on the endothelial surface for 20 seconds and then washed with sterile phosphate-buffered saline (PBS). The endothelium was exposed to a hypotonic sucrose solution (1.8%) to swell the intercellular borders for counting the number of endothelial cells and to examine general morphology before processing. A reticule (10 9 10) was fixed to the eyepiece of an inverted microscope (Nikon Eclipse TS100, Nikon, Surrey, UK) to count the number of endothelial cells and trypan blue positive cells (TBPCs) manually. An average of five different counts was recorded (Parekh et al. 2019a).
Human corneal endothelial cell line culture and exposure of extracellular matrix (ECM) Human corneal endothelial cells (HCEC) from a certified cell line  were cultured on 75 cm 2 culture flasks (Nunc, Thermo Fisher Scientific, Rochester, NY, USA) until confluence using culture medium containing 1:1 Ham's F12 and Medium 199 (cat no: 31765027 and 41150020, respectively, Thermo Fisher Scientific) supplemented with 5% foetal bovine serum (FBS). The cells were trypsinized upon confluence and cultured on labtek II chamber slides (8 chambers, 25 9 75 mm, 0.7 cm 2 culture area, Thermo Fisher Scientific) to obtain the ECM required for primary endothelial cell culture. Upon confluence, medium was removed and the HCEC-12 cells were washed with sterile PBS. The cells were treated with 20 mM ammonium hydroxide (NH 4 OH, Thermo Fisher Scientific) in sterile PBS for 5 min (Gospodarowicz 1984;Todorovic et al. 2010). The resulting ruptured cells were washed away several times with sterile PBS followed by a single wash of double distilled water (ddH 2 O) to remove any remaining cell remnants. This resulted in the lab-tek II chambers coated with whole ECM proteins. Different time-points (days 1, 5 and 9) were evaluated for optimal conditions required for ECM production.

Scanning electron microscopy
Samples were prepared in a laminar flow hood, and extra care was taken to ensure the samples did not dry out prematurely. The samples were washed twice for 5 min in ddH 2 O, and then, osmium tetroxide (1%, aqueous) was added to the samples, followed by further washes with distilled water three times for 10 min each. The samples were then washed with ethanol in increasing order of concentrations, that is 30%, 50%, 70%, 90% and 100% three times for 10 min each, and dried using methanol twice for 10 min each. The samples were dipped in hexamethyldisilazane (HMDS) in a small glass jar thrice for 3 min each and allowed to dry under a petri-dish in the hood for at least 1 hr. The samples were finally mounted into a labelled stub using a conductive carbon disc and sealed at the edges with silver paint. Samples were stored in a drying cabinet and coated with 1.5 nm platinum (Cressington sputter coated) before imaging using Zeiss Sigma VP SEM (Oberkochen, Germany).

Human corneal endothelial cell culture from old aged donor tissues
Peeling and plating the cells at passage 0 The Descemet membrane-endothelial complex of the tissues was peeled in several pieces to ensure a quicker isolation process. The excised pieces were incubated in 2 mg/ml collagenase Type 1 (Thermo Fisher Scientific) solution for approximately 2 hr at 37°C and 5% CO 2 . The solution with cell clumps was then centrifuged for 5 min at 1000 rpm. The supernatant was removed, and the pellet was re-suspended with trypLE express (19) phenol red (Life Technologies, Monza, Italy) for 5 min at 37°C to further dissociate the pellet into single cells. The supernatant was removed, and the cells were re-suspended in 200 µl of the cell culture medium, a formulation of 1:1 Ham's F12:M199 (Sigma-Aldrich), 5% FBS, 20 µg/ml ascorbic acid (Sigma-Aldrich), 1% insulin transferrin selenium (Gibco), 10 ng/ml recombinant human FGF basic (Gibco), 10 µM ROCK inhibitor (Y-27632; Miltenyi Biotech) and 1% penstrep (Sigma-Aldrich; Peh et al. 2011aPeh et al. ,b, 2013aPeh et al. ,b, 2015Parekh et al. 2016Parekh et al. , 2017Parekh et al. , 2019a. The resulting cells were counted using a haemocytometer, and the number of plated cells was recorded for all the cultures. Control lab-tek II chamber slides were coated with 50 µl fibronectin-collagen (FNC) coating mix (US Biological Life Sciences, Salem, MA, USA) for at least 30-45 min at 37°C and 5% CO 2 . The residual coating was removed before plating cells. 200 µl of the cell suspension from each cornea was divided into two equal halves and plated on each chamber of (i) FNC coated slide, and (ii) ECM coated slide derived from HCEC-12 cells. The medium was replaced, and the cells were monitored every alternate day until confluence followed by end-stage characterization or further passaging.

Passaging the cells
Upon confluence, primary cells were briefly washed with PBS and incubated at 37°C and 5% CO 2 for 5-10 min in trypsin-EDTA solution. The reaction was stopped using culture medium. The cell suspension was collected in a 15-ml tube and centrifuged at 1000 rpm for 5 min. The supernatant was removed, and the pellet was re-suspended using the medium. The lab-tek II slides were precoated with FNC and ECM (as mentioned above), and the cells were further divided into two groups at P1: cells obtained from the FNC group were sub-cultured on FNC and ECM coated slides separately, and cells from the ECM group were sub-cultured on FNC and ECM coated slides separately. Similarly, at P2, the cells from each group were further sub-cultured on both FNC or ECM coated slides separately. This resulted in the following different matrix formats of culture/ sub-culture for P0-P1-P2, that is FNC-ECM-FNC; FNC-ECM-ECM; FNC-FNC-FNC; FNC-FNC-ECM; ECM-ECM-FNC; ECM-ECM-ECM; ECM-FNC-FNC; and ECM-FNC-ECM, respectively.
Hoechst, ethidium homodimer and calcein AM (HEC) staining to determine live/dead cells Cells at confluence were washed with PBS after preservation prior to the assay. Five microlitre of hoechst 33342 (H) (Thermo Fisher Scientific), 4 µl of ethidium homodimer EthD-1 (E) and 2 µl calcein AM (C) (live/dead viability/ cytotoxicity kit, Thermo Fisher Scientific) were mixed in 1 ml of PBS. 100 µl of the final solution was directly added to the cells and incubated at room temperature in dark for 45 min, followed by a single washing step. The walls of the lab-tek slides were detached, and the cells were mounted with cover slips without 4 0 ,6-diamidino-2-phenylindole (DAPI) mounting media. The zeiss LSM 700 confocal microscope (Carl Zeiss, Cambridge, UK) was used to image the cells that were captured using in-built zen software. Triple labelling showed the presence of ethidium homodimer stained in red representing dead cells, hoechst in blue representing nuclei and calcein AM in green marking viable cells (Pipparelli et al. 2011).
Immunostaining of zonula occludens-1 (ZO-1) tight junction protein Cells at confluence were washed with PBS and fixed in 4% paraformaldehyde (PFA) at room temperature (RT) for 20 min. The cells were permeabilized with 0.25% triton X-100 in PBS for 30 min. After blocking with 10% goat serum for 1 hr at RT, the cells were incubated overnight at 4°C with primary antibody anti-ZO-1, 1:200 (FITC-conjugated, Thermo Fisher Scientific). Hoechst (33342) was mixed (1:1000), and 100 ll of the solution was added to the cells to stain the nucleus. After each step, the cells were washed three times with PBS. After detaching the walls of the lab-tek slides, the cells were covered with mounting medium and cover slips. The expression of ZO-1 was examined using the LSM 700 confocal microscope (Carl Zeiss), and images were captured using in-built zen software.

Data analysis time-points
For SEM analysis, the images were obtained at days 1, 5 and 9. The data for HCEnCs cultured on both FNC and whole ECM coatings were obtained at days 1, 3, 5, 7 and 9 for P0 and at days 1, 5 and 9 for P1 and P2.

Confluence and cell doubling of the cultured HCEnCs
The percentage confluence was measured every alternate day using a 10 9 10 reticule (0.1 mm 2 ) attached to the eyepiece of an inverted microscope (Nikon Eclipse TS100; Nikon). The number of endothelial cells per mm 2 was counted using the same reticule to determine the cell numbers in the given area and the doubling time.
Determining viable/dead cells, hexagonality, polymorphism and cell area All the measurements and data analysis were performed using ImageJ (FIJI) bundled with 64-bit Java 1.8.0 112. Viability of cells was measured as the number of calcein AM positive cells compared with number of Hoechst positivity. Similarly, dead cells (ethidium homodimer positive) were counted using ImageJ and converted to a percentage (Parekh et al. 2019a,b). For ZO-1 analysis, the images were converted to overlay masks using predetermined macroinstructions to define the parameters of both hexagonality and polymorphism within a particular area (Parekh et al. 2019a,b). The images were auto-converted, and the total number of cells in the investigated area was counted using the macros for ZO-1. The hexagonal and polymorphic cells were counted manually depending on the cellular structure comprising six borders per cell for hexagonal cells and less than four borders for severely polymorphic cells in the investigated area. Cell area (µm 2 ) was measured by marking the borders of the cell using the free-hand tool and then using the area measurement tool.
Liquid chromatography-mass spectrometry ECM analysis by mass spectrometry Cell lines from human conjunctival epithelial cells (HCjE-Gi; Gipson et al. 2003) and human telomeraseimmortalized corneal epithelial cells (hTCEpi; Robertson et al. 2005) were cultured as controls for comparative protein analysis and were treated the same way as HCEC-12 to obtain the ECM (as described previously) for mass spectrometry. HCjE-Gi cells were cultured using keratinocyte serum-free media (KSFM; Gibco, Thermo Fisher Scientific), supplemented with 0.2% (V/V) bovine pituitary extract (BPE), 0.2 ng/ml of epithelial growth factor (EGF) and 1% (V/V) penstrep (Sigma-Aldrich, Haverhill, UK) and 0.4 mM of CaCl 2 (Sigma-Aldrich). The hTCEpi cells were cultured in KSFM with CaCl 2 (Gibco), supplemented with 0.2% (V/V) BPE, 0.23 ng/ml EGF (all supplied with the medium) and CaCl 2 to a final concentration of 0.13mM. Basal cell protein proteolysis and extraction were performed on HCjE-Gi, hTCEpi and HCEC-12 cells according to the method described by Gospodarowicz (1984) and Todorovic et al. (2010) with minor modifications (50 mM of NH 4 HCO 3 instead of 100 mM, 0.33% of trypsin gold instead of 0.5 lg and instead of storing the samples the protocol for Mass Spec was continued immediately). A solution of 50 mM NH 4 HCO 3 (Sigma-Aldrich) in ddH 2 O, pH 8.0, containing: 0.02% (V/V) ProteaseMax surfactant (Promega, Southampton, UK), 0.33% (W/V) trypsin gold (Mass Spec grade; Promega) was added to the exposed ECM, and the samples were digested for 1 hr at 37°C in a humidified chamber. The peptides (resulting solution) were transferred into low-binding tubes (Eppendorf LoBind, Sigma-Aldrich). The enzymatic reaction was stopped by adding acetic acid (Sigma-Aldrich) setting the pH to 3-4; 1,4-dithiothreitol (DTT, Sigma-Aldrich; 1% (V/V) of 0.5M) was added for 20 min at 56°C; and 2.5% (V/V) of 0.55 M iodoacetamide (Sigma-Aldrich) was added for 15 min in the dark at room temperature. 0.3125% (W/V) of trypsin gold and 0.00625% (V/V) ProteaseMax surfactant were added, and the solution was heated at 37°C for 3 hr for protein digestion. Trifluoroacetic acid (TFA, Sigma-Aldrich) was added to a final concentration of 0.5% (V/V).
LC-MS analysis was performed at Dublin City University, Ireland, using an ultimate 3000 RSLCnano system (Dionex, Thermo Fisher Scientific) coupled to a hybrid linear ion trap/ Orbitrap mass spectrometer (LTQ Orbitrap XL; Thermo Fisher Scientific). The digested samples were sonicated, and 1 lg of digested proteins was loaded onto a C18 trap column (C18 PepMap, 300µm i.d. 9 5mm, 5 µm particle size, 100 µm pore size; Dionex) and desalted for 3 min at a flow rate of 25 ll/min using 2% acetonitrile containing 0.1 M TFA. The trap column was then switched online with the analytical column (PepMap C18, 75 µm i.d. 9 500 mm, 3 µm particle, and 100 µm pore size; Dionex), and peptides were eluted in a 180-min gradient at a flow rate of 300 nl/min using 2% acetonitrile with 0.1% formic acid (FA) to 50% acetonitrile containing 0.08% FA. Mass spectrometry data was acquired with Xcalibur software, version 2.0.7 (Thermo Fisher Scientific). The mass spectrometer was operated in data-dependent mode and externally calibrated. MS1 survey scan (m/z 400-1200) was set at a resolution of 30 000 in the Orbitrap, followed by ten MS2 scans using collision-induced dissociation (CID) activation mode in the ion trap. Dynamic exclusion was enabled with the following settings: repeat count, 1; repeat duration, 30 seconds; exclusion list size, 500; and exclusion duration, 40 seconds. The activation time was 30 mseconds, with an isolation width of 2 Da for ion trap mobility spectrometry (ITMS); the normalized activation energy was 32%; and the activation (q) was 0.25 (Henry & Meleady 2017).

ECM protein data analysis
Proteome Discoverer (PD) version 2.1.0.81 (Thermo Scientific) was used to perform the database search against the human sequences in the UniProt Swiss-Prot protein database (version January 2016 with 20 151 entries) for the mass spectrometry raw data files. The search engines SEQUEST-HT and Mascot (version 2.4.0) were utilized in PD. The search parameters used were as follows: 20ppm tolerance for precursor ion masses, 0.6 Da for fragment ion masses analysed by ion trap. A total of two missed cleavages were permitted for fully tryptic peptides. Carbamidomethylation of cysteines (+57.0215 Da) was set as a fixed modification, and variable modifications of methionine oxidation (+15.9949 Da) and N-terminal acetylation (+42.0106 Da) were allowed. The false discovery rate (FDR) was determined by using a target-decoy search strategy using Percolator in PD2.1. The sequence database contains each sequence in both forward and reverse orientations, enabling FDR estimation. The FDR was set to 0.01 at both the peptide and the protein levels. A higher percentage coverage was considered better, which was calculated by dividing the number of amino acids in all found peptides by the total number of amino acids in the entire protein sequence of the samples (Henry & Meleady 2017).

Statistical analysis
A two-tailed non-parametric Wilcoxon signed rank for paired test with 95% confidence interval was used to check the statistical difference between the cells cultured on ECM and FNC, respectively, at passage 0. Non-parametric Kruskal-Wallis test without Gaussian approximation with post hoc Dunn's test with a significance level of Alpha = 0.05 (95% confidence intervals) was used to compare the data at passage 1 and 2 using GraphPad Prism 5.01 (GraphPad Software, San Diego, CA, USA) software. A probability value of < 0.05 was deemed statistically significantly different.

Donor characteristics
Corneal tissues (n = 40) used in these studies were unsuitable for transplantation due to their low endothelial cell counts. Average endothelial cell density (ECD) before the isolation of cells was found to be 1885 (AE203) cells/mm 2 with 1.4 (AE0.9) % TBPCs. No other complications or indications were noted from the donor database. The mean age of the donors was found to be 70.9 (AE7.6) years comprising 22 males and 18 females and post-mortem time of 11.3 (AE7.2) hours. Tissues were stored in tissue culture medium at 37°C for 25 (AE4.4) days in the eye bank before use.
Scanning electron microscopy of the ECM SEM images of the culture flasks were obtained without any cells as the control group (Fig. 1A) and of the laid ECM after days 1, 5 and 9 (n = 3). Collagen-like fibrillary structures were noticed in the ECM after day 1 of culture ( Fig. 1B). At day 5, when cells were approximately 90% confluent, slightly longer fibres were observed (Fig. 1C). At day 9, most of the cellular debris was washed off without compromising the long fibrillary structures and maintaining the core ECM proteins (Fig. 1D). HCEC-12 cells were therefore cultured for 9 days to obtain ECM for the subsequent experiments.

Morphological analysis of primary HCEnCs on FNC and ECM coated base
Human corneal endothelial cells cultured on both FNC and whole ECM coatings at P0 showed compactly arranged hexagonal cells by day 9 (n = 40; Fig. 2A). Interestingly, the cells started showing hexagonal morphology and sharp intercellular junctions when cultured on whole ECM ( Fig. 2B marked with dotted circle) compared to those cultured on FNC by day 5. This indicates that ECM proteins may be providing an optimum environment facilitating cell growth and supporting the development of tight junction proteins as also shown later by immunostaining. At P1 ( Fig. 2C; n = 24), the cells were compactly organized and showed limited polymorphism when they were switched from ECM to a FNC coated base. However, morphology deteriorated at passage 2 (n = 12), but the cultures still showed confluence by day 9 (Fig. 2D) due to polymegathism.

Confluence
At P0 (n = 40), confluence was found to be statistically higher (p < 0.05) at days 1, 5 and 7 for ECM compared to FNC cultures (Fig. 3A). At P1 (n = 24), FNC cultured cells which were sub-cultured from P0 ECM cultures yielded statistically significant higher confluence at days 1, 5 and 9 of the culture period (Fig. 3B). At P2 (n = 12), all sub-cultures which had been grown from ECM cultures at P0 (as opposed to FNC) showed significantly higher confluence (Fig. 3C).

Cell doubling rate and time
Exponential growth in cell numbers was observed at P0 in both sets of cultures, for ECM and FNC coatings when 90 866 (AE8230) cells were plated in FNC group and 90 756 (AE6980) cells were plated in the ECM group at P0 (Fig. 3D). Significantly higher cell numbers from ECM cultures were observed than those grown on FNC at P0. Indeed, in subsequent passages, cultures which had been grown on ECM at P0 maintained significantly higher cell numbers at P1 (Fig. 3E) and P2 (Fig. 3F). By the end of the experiment at P2, collectively, cell numbers were found to be 563 457 (AE52 000) cells in FNC group and 639 047 (AE52 516) cells in the ECM group (p = 0.0355).
Cell doubling time was found to be significantly less in the ECM group compared with the FNC group at P0 (Fig. 3G; n = 40). Cell doubling time was not found to be significantly different at P1 (n = 12; Fig. 3H). However, at P2 (n = 12) the doubling time was significantly less between cells that were cultured on ECM-ECM-FNC at P0-P1-P2 compared with FNC-ECM-FNC (Fig. 3I). It was observed that in general the cell doubling times at P0 were higher compared to those at P1 and P2).
Live/dead staining HEC staining showed good viability of cells at P0 (n = 8; Fig. 4A), P1 (n = 6; Fig. 4B) or P2 (n = 6; Fig. 4C) without any significant difference between FNC and ECM groups at P0 (Fig. 4D), P1 (Fig. 4E) or P2 (Fig. 4F). However, a significant reduction (p = 0.0014) in the percentage of viable cells was observed between P0 and P2 in general. Simultaneously, the number of dead cells (ethidium homodimer positive) significantly increased with subsequent passaging from P0 to P2 (p = 0.0005; Fig. 4) with no significant difference observed between the FNC and ECM groups at either passages P0 (Fig. 4G), P1 (Fig. 4H) or P2 (Fig. 4I).  Fig. 5C), it was less strongly expressed at P1 and P2 compared with P0. Cultures from the ECM group showed significantly higher numbers of hexagonal cells compared with FNC at P0 (Fig. 5D). At P1, the ECM-FNC (P0-P1) group showed a significantly higher percentage of hexagonal cells compared with both P1 groups that were subcultured from P0 FNC cultures (FNC-   FNC and FNC-ECM; Fig. 5E). P2 cultures that were grown on ECM at P0 showed a significantly higher number of hexagonal cells compared to those P2 cultures grown on FNC at P0 (Fig. 5F). The results for polymorphism were inversely proportional to those for hexagonality at all passages as would be expected, with those cultures grown on ECM at P0 showing significantly less polymorphism (Fig. 5G-I). In terms of cell area, no significant difference was observed at P0 between the groups (Fig. 5J). However, P1 cultures grown from the P0 ECM cultures had significantly smaller cell areas (Fig. 5K). A similar trend was found in P2 cultures derived from P0 ECM cultures compared to those from P0 FNC cultures (Fig. 5L).

Mass spectrometry
A total of 1307 proteins were found in the corneal endothelial ECM (n = 3) using modified LC-mass spectrometry, out of which 92 proteins were found to be matrix oriented. 72 of these matrix oriented proteins overlapped with HCjE-Gi (n = 3) and hTCEpi (n = 3) derived ECM proteins leaving 20 specific to corneal endothelial ECM ( Fig. 6; Table 1).

Discussion
Out of all corneal transplants carried out in the UK, 40% are due to endothelial disease (NHSBT data). Although corneal transplantation is the mainstay of treating symptomatic corneal endothelial dysfunction, it is restricted due to the limited availability of donor tissues around the globe. Although alternative solution to endothelial keratoplasty has been identified, cell culture techniques are further constrained as healthy and high number of endothelial cells are difficult to obtain (Shima et al. 2011;Kimoto et al. 2012;Zhu et al. 2012;Okumura et al. 2013;Peh et al. 2013a). Therefore, optimizing proliferation from older donor tissue-derived corneal endothelial cells (Khaireddin et al. 2003;Joyce & Zhu 2004;Coster & Williams 2005), especially those that are discarded due to poor endothelial cell counts, would be ideal considering the availability of these tissues.
Previous studies have used denuded human amniotic membrane, collagen type I matrices, silk fibroin, fish scale scaffolds etc. as potential substrates for the propagation of HCEnCs (Ishino et al. 2004;Gruschwitz et al. 2010;Madden et al. 2011;Palchesko et al. 2015;Parekh et al. 2018). Several   6. A Venn diagram showing the total number of proteins extracted from the corneal epithelial, conjunctival epithelial and corneal endothelial extracellular matrix (ECM). Out of these, 235 proteins were exclusively found from the corneal endothelial ECM and 20 of these were found to be ECM oriented. specific ECM coatings including collagen I/IV, FNC, laminin and chondroitin sulphate have been shown to influence cellular adhesion, proliferation, morphology, molecular function, and the development and expression of markers for HCEnCs (Choi et al. 2013). Although several ECM proteins have already been investigated for the culture of HCEnCs in vitro, the culture of HCEnCs on HCEnC derived whole ECM has not been studied previously. Using SEM, we have shown that long collagen-like fibrillary structures can be seen in the whole ECM deposited by corneal endothelial cells. These appear at day 5 of culture and at day 9 there is wide coverage of the plasticware with these fibrils. FNC has been used commonly for HCEnC cultures (Peh et al. 2011a(Peh et al. ,b, 2013a(Peh et al. ,b, 2015Parekh et al. 2016Parekh et al. , 2017Parekh et al. , 2019a previously, and this was therefore used as a control to whole ECM we investigated here. Moreover, in the studies described here, we have shown that HCEnCs from older donor tissues that have suboptimal endothelial cell counts for transplantation, can be cultured with success on whole ECM. Morphologically, we have observed from our previous studies that HCEnCs cultured on FNC coated plastic at P0 show confluence by day 9. This trend was also observed when cells were cultured on whole ECM. However, systematic arrangement of hexagonal morphology and the development of intercellular tight junctions was observed in HCEnCs on whole ECM (compared to FNC) by day 5, further indicating that the ECM proteins and growth factors may provide an optimum environment facilitating the development of tight junction proteins at an earlier stage of culture.
Expansion of HCEnCs by passaging them in culture from P0 to P2 was performed to understand the proliferative capacity of these cells on whole ECM. From our previous experience and that of others in the field, it is difficult to sub-culture HCEnCs as cells often become senescent (Parekh et al. 2017). Therefore, our aim was to investigate the effect of whole ECM on the proliferative capacity and subculturing capability of HCEnCs. This was investigated in comparison with FNC and by switching between whole ECM and FNC at each passage to determine the effects of these different matrices on cell morphology and proliferation. Interestingly, we observed that cells showed better morphology and cell proliferation when passaged and switched from one base coat to the other provided they were all cultured on whole ECM at P0. Thus, in general at P1, we observed that proliferation was best for ECM-FNC (P0-P1 matrix) compared to ECM-ECM or FNC-FNC. We noted limited polymorphism when cells were transferred from ECM to FNC coated base, indicating better morphology. In the same way, cells cultured on whole ECM at P0 showed significantly better results at P2. This shows HCEnCs must be provided with optimum conditions at P0 for them to survive subsequent passaging. We also found that higher passages lead to more dead cells, but this is without compromising cell shape, size and area of the viable cells. It is already known that tight junction proteins lose their ability to form continuously (Joyce 2005) leading to lower expression of ZO-1 at every subsequent passage (Kundu et al. 2013). We observed a similar trend in the expression of ZO-1 at different passages.
We also investigated the protein content exclusive to the whole ECM laid down by HCEC-12 cells but not found in the ECM of HCjE-Gi and hTCEpi cells. These proteins showed specificity towards cell adherence, migration, proliferation and cell cycle control and many of which have not been used previously as coatings for HCEnC culture. These include the following: 1 SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily B member 1 plays an important role in cellular proliferation, differentiation and cell cycle control that causes cell cycle arrest in G0/G1 (Kalimuthu & Chetty 2016). Interaction with this protein could be one of the indirect causes of corneal endothelial cell cycle arrest and limited proliferative capacity in vivo. 2 Insulin-like growth factor-binding protein 5 has been shown to modulate the growth-promoting effects of insulin-like growth factors (IGFs) on cell culture by altering the interaction of IGFs with their cell surface receptors (Jones et al. 1993). 3 Vitronectin is a glycoprotein found in serum, ECM and bone and binds to integrin alpha-V beta-3 further promoting cell adhesion and migration (Boron Walter & Boulpaep Emilie 2012) and therefore may play an integral role in the development of focal adhesions of HCEnCs. 4 Chondroitin sulphate proteoglycan 4 (CSPG4) has a role in cell proliferation and migration. It promotes retraction fibre formation and cell polarization through Rho GTPase activation. We have used Rho-kinase inhibitor for our cells and therefore believe that CSPG4 could be one of the important proteins involved in the migration of cells on the ECM coated base (Michelacci 2003) in combination with Rho-kinase inhibitor. 5 Thrombospondin-4 is an adhesive glycoprotein that mediates cell-to-cell and cell-to-matrix interactions and is therefore an important protein required for HCEnCs culture (Lawler et al. 1995).
From our investigations, we believe that it is important to provide an environment close to physiologic conditions for culturing HCEnCs in vitro. We observed that for successful culture and passaging, maintenance of cellular morphology and expression of core proteins, the normal whole ECM plays a key role. It is not only that the cells must adhere and proliferate, but also that the cells show appropriate phenotypic characteristics and expression of core proteins in order for them to be considered suitable for transplantation. The 'de-roofed' culture system investigated here in the form of whole ECM not only provides a surface structure that enables optimal cell-matrix interactions (adhesion and migration), but also provides a pool of soluble growth factors trapped within the ECM. These cues have been shown to regulate central cell processes (Palchesko et al. 2015), including cell differentiation. We thus investigated a culture system, that mimics the cellular microenvironment, to be able to induce favourable HCEnC proliferation with good morphological and phenotypical behaviours for use in tissue engineering of the corneal endothelium (Koo et al. 2014).
In summary, we show in the studies outlined here that HCEnCs from older donors that are not suitable for conventional endothelial transplantation can be cultured on whole ECM and passaged successfully. Moreover, the HCEnCs cultured initially on whole ECM at P0 show better morphology and growth, than those grown on FNC at P0. In addition, analysis of this ECM by mass spectrometry identifies ECM proteins specific to HCEnCs that are of interest and warrant further study.