Identification of Cell Surface Markers Glypican-4 and CD200 That Differentiate Human Corneal Endothelium From Stromal Fibroblasts

Abstract

Purpose.: There is a lack of definitive cell surface markers to differentiate cultured human corneal endothelial cells (HCECs) from stromal fibroblasts, which could contaminate HCEC cultures. The aim of our study is to discover cell surface antigens on HCECs that can be used to identify and purify HCECs from stromal fibroblasts.

Methods.: RNA sequencing (RNA-seq) was used to find differentially overexpressed genes in HCECs and commercial antibodies against these overexpressed antigens were screened by immunofluorescence assay. Similarly, 242 commercial antibodies against cell-surface antigens also were screened. Selected antibodies were used to sort HCECs from stromal fibroblasts by fluorescence-activated cell sorting (FACS).

Results.: Two monoclonal antibodies, anti-GPC4 and anti-CD200, were identified to stain HCECs specifically. FACS was used successfully to sort HCECs away from stromal fibroblasts. Recovery efficiency of HCECs after sorting using anti-GPC4 antibody was higher compared to anti-CD200 antibody, but purity of HCECs culture using either antibody was comparable.

Conclusions.: Taken together, the anti-GPC4 and anti-CD200 antibodies can be useful for purification and identification of HCECs in cultures containing stromal fibroblasts.

Introduction

The human corneal endothelium is a monolayer of cells that functions as a barrier between the aqueous humor and corneal stroma via the formation of focal tight junctions to prevent excessive fluids from entering the stromal layer. ¹ The Na⁺/K⁺ and Mg²⁺ ionic pumps on the corneal endothelial cell surface regulate corneal hydration ^2,3 to maintain corneal transparency. As the human corneal endothelial cells (HCECs) do not regenerate in vivo, ^4,5 the integrity of the cellular monolayer is sustained through migration and enlargement of existing cells in the event of cell loss due to disease or surgical trauma. ⁶ However, decompensation of the endothelium occurs when corneal endothelial cell density falls below a critical threshold, resulting in its inability to maintain stroma deturgescence, thereby affecting visual acuity. ⁷  

Endothelial dysfunction is the second leading cause of visual blindness worldwide. ⁸ Endothelial keratoplasty involving dissected Descemet's membrane (DM) with the endothelium layer can be performed to restore vision. However, global supply of transplant-grade corneal tissues is limited and this restricts the number of corneal transplantation performed annually. ⁹ To circumvent the shortage of corneal tissues and reliance on human donor material, development of suitable tissue-engineered constructs using cultured HCECs from cadaveric donors ^10–13 or HCECs derived from multipotent progenitor cells are being explored by various groups as potential graft alternatives. ^14–16  

The isolation of HCECs from cadaveric donors involves peeling of the DM together with the corneal endothelium layer, and subsequent enzymatic digestion with collagenase to release the corneal endothelial cells from the DM. ¹¹ However, in some cases excessive manipulation may result in tearing into and co-isolation of small amounts of stromal tissue. Exposure to culture medium required for HCECs growth would transform these stromal keratocytes into fast-growing stromal fibroblasts, ¹⁷ which outgrow the less proliferative HCECs. ^18,19 Although an L-valine–free selection medium could be used to arrest growth of stromal fibroblasts, they could not be eliminated. ¹⁸ To abolish fibroblastic contamination, the culture would have to be passaged several times to dilute away the stromal fibroblasts. However, the selection media is not optimal for long-term cultivation of HCECs. ¹⁸ Alternatively, stromal fibroblasts could be depleted from contaminated cultures through a negative cell–selection strategy by magnetic affinity cell separation (MACS) using antifibroblast magnetic microbeads. ²⁰ A positive cell–selection approach, using magnetic microbeads or a more sensitive fluorescence-activated cell sorting (FACS), is plausible, but this is limited by the specificity of cell-surface markers for HCECs. For example, the two highly used markers reported for cultured HCECs are pump-associated protein Na⁺/K⁺-ATPase (NaK) and tight-junction protein ZO-1. In the cornea, the coexpression of these proteins indicates the presence of functional components on the corneal endothelium, but do not define the identity of cultivated HCECs because these proteins are expressed ubiquitously in the heart, ²¹ brain, ²² and kidney. ²³  

Currently, no HCEC-specific cell surface antigens have been described. The goal of our study is to identify cell-surface antibodies expressed on the corneal endothelium and cultivated HCECs that can differentiate them from stromal keratocytes and fibroblasts. Additionally, cell-surface antibodies identified in our study were assessed in its efficacy in purifying cultivated HCECs from a culture contaminated with stromal fibroblasts using FACS. 

Materials and Methods

The following protocols conformed to the tenets of the Declaration of Helsinki, and written consent was acquired from the next of kin of all deceased donors regarding eye donation for research. The study was approved by the institutional review board of the Singapore Eye Research Institute/Singapore National Eye Centre. 

Isolation and Growth of HCECs

Cultivation of HCECs isolated from pairs of donor corneas were done as described previously, ⁸ with some modification. Briefly, the DM and corneal endothelial cells were peeled off from the stroma under a dissecting stereomicroscope. The DM-endothelial layer was treated with collagenase for 6 hours and dissociated further into smaller clumps with TrypLE Express (TE; Life Technologies, Carlsbad, CA). Isolated cells were cultured on fibronectin and collagen (FnC) coated culture wares in F99 medium (1:1 Ham's F12 and M199), supplemented with 5% fetal bovine serum (FBS), 20 μg/mL ascorbic acid, 1 × ITS, 1 × antibiotic/antimycotic, ²⁴ and 10 ng/mL basic fibroblast growth factor. At 90% confluency, HCECs were exposed to Endothelial-SFM supplemented with 5% FBS and 1 × antibiotic/antimycotic (AA) for 5 days before they were subcultured. Seeding density was approximately 1 × 10⁵ cells/cm². Cultured HCECs were used at the first, third, or fifth passage. Incubation and cell culture were done in a humidified 5% CO₂ incubator at 37°C. Cell culture medium was refreshed every other day. 

Isolation and Growth of Human Corneal Stromal Fibroblasts

An 8.5 mm stromal button was obtained by trephination after the DM-endothelial layer was peeled off. Remnant corneal epithelial layer was scraped off with a scalpel. Stromal buttons were washed in PBS-AA solution and digested in collagenase overnight. Stromal keratocytes released from the stromal button were washed with PBS, and cultured in F99 medium. Corneal stromal keratocytes were transformed into corneal stromal fibroblasts in serum-supplemented medium. Culture medium was refreshed every two days and subcultured at a 1:5 ratio. 

Sectioning of Cornea Tissue

Human donor cornea was rinsed in PBS twice and immersed in OCT, and frozen at −80°C until sectioning. Sections 8-μm thick were cut using a MicromHM550 cryostat (Thermo Scientific, Kalamazoo, MI), collected on glass slides, air dried, and stored at −80°C. 

RNA Extraction, Transcription, and Amplification for RNA Sequencing

Two different donor corneas were used. Isolated HCEC-DM and the remaining cornea containing the corneal stroma cells were rinsed in PBS, and placed in Trizol reagent. It should be noted that, although most of the corneal epithelium spontaneously sloughed off the cornea surface during the processing of the cornea, remnant basal corneal epithelium cells still might be present on the corneal stromal button used for RNA extraction. Tissues were homogenized before addition of chloroform. Total RNA was extracted with RNeasy kit (Qiagen, Inc., Valencia, CA) with DNAse digestion step. 

RNA sequencing (RNA-seq) libraries were prepared using an AB Demonstrated Protocol. ²⁵ For each sample, 100 pg of total RNA was transcribed with SuperScript III reverse transcriptase (Invitrogen Corp., Carlsbad, CA), using a poly(T) primer with a UP1 anchor sequence. Exonuclease I removed free primers. Subsequently, a poly(A) tail was added to the 3′ end of the first strand cDNA by terminal deoxynucleotidyl transferase. This product was used for the synthesis of the second strand cDNA, using a poly(T) primer with anchor sequence UP2. The product then was amplified separately using primers UP1 and UP2. Amplified cDNA was purified with QIAquick columns (Qiagen, Inc.), and re-amplified using NH2-UP1 and NH2-UP2 primers. This subsequently was pooled, purified, and size-selected (500 base pairs [bp] to 3 kilobase pairs [kb]) by gel purification. 

RNA Sequencing (RNA-seq)

The ABI standard library preparation protocol (Chapter 2 of AB SOLiD 4 system library preparation guide—AB PN 4445673, available in the public domain at http://www.lifetechnologies.com by Life Technologies) was used for nonbarcoded library preparation. The SOLiD fragment library barcoding kit module protocol (AB PN 4443045) was used for barcoded library preparation. The libraries were generated using the SOLiD fragment library construction (AB PN 4443471) and barcoding (AB PN 4444836) kit. 

Briefly, 1 μg of amplified cDNA was sheared to approximately 165 bp, end repaired, and P1, P2 adapters ligated. The product was nick translated and amplified (5 cycles), and size selected between approximately 240 to 270 bp. The libraries were processed using the AB SOLiD 4 system templated bead preparation guide (PN 4448378) and sequenced on SOLiD slide. 

The resulting RNA-seq reads were processed with ABI Bioscope pipeline. The reads were aligned to the hg19 human reference genome. Gene expression was measured by counting the number of reads mapping uniquely to both strands of each gene footprint, and normalized to the total uniquely mapped reads to the entire genome. Values were adjusted to show reads per million. 

Quantitative PCR (qPCR)

To measure the relative abundance of each gene transcribed in HCECs and stromal fibroblasts, RNA was extracted with mRNeasy kit (Qiagen, Inc.). One μg of total RNA from HCECs and stromal fibroblasts was used to synthesize cDNA (Fermentas; Thermo Scientific). Each real-time PCR reaction was performed with iQsyber green supermix (Bio-Rad Laboratories, Hercules, CA) in the CFX96 thermocycler (Bio-Rad Laboratories). Technical triplicates for each gene were performed. The Δ-CT value is normalized to Actin or GAPDH. The ΔΔ-CT value of each gene is normalized to values obtained from stromal fibroblasts. The average ΔΔ-CT value of each gene is calculated from 3 independent experiments. Primers to each gene were synthesized by 1st Base Pte Ltd. (Singapore). 

BD Lyoplate Screening

The BD Lyoplate human cell surface marker screening panel (BD Biosciences, Franklin Lakes, NJ) contains 242 purified monoclonal antibodies against human cell-surface antigens. The antibodies were used to stain cells in 96-well plates (Nunc Multiwell Plates; Sigma-Aldrich Corp., St. Louis, MO). Antibodies were reconstituted in 110 μL of sterile PBS and 20 μL of each antibody was used for each reaction. Both HCECs and stromal fibroblasts were fixed in 4% paraformaldehyde for 10 minutes, washed and incubated with the primary antibodies for 1 hour. The cells were washed in PBS and incubated with 1.25 μg/mL of the appropriate secondary antibodies for 1 hour. The cells were analyzed with InCell Analyzer 2000 (GE Healthcare Biosciences, Pittsburgh, PA). 

Antibodies and Immunofluorescence

Antibodies were obtained from commercial sources (Table 1). HCECs and stromal fibroblasts were seeded on multi-test slides (MP Biomedicals, Solon, OH). Cells or tissue sections were blocked with 5% goat serum (Millipore, Billerica, MA) or 3% bovine serum albumin (Santa Cruz Biotechnology, Inc., Dallas, TX) and incubated with primary antibodies for 1 hour, and washed with PBS with 0.05% Tween (PBST). Subsequently, the cells or tissue sections were incubated with secondary antibodies for 1 hour and washed with PBST. Cells were counterstained with Hoeschst (Invitrogen Corp.). A mouse, rabbit, goat, or rat IgG isotype antibody was used as negative control (Santa Cruz Biotechnology, Inc.). Slides were viewed under a fluorescence microscope (Nikon, Inc., Melville, NY). 

Table 1. 

 

List of Primary Antibodies

Name	Company	Species and Type	Dilution Used
CD104 Integrin β4 Cat. No. 555719	BD Pharmingen, Franklin Lakes, NJ	Rat mAb	1:100
CD200 OX-2 membrane glycoprotein Cat. No. 552023	BD Pharmingen	Mouse mAb	1:100
CNTN6 Contactin-6 Cat. No. T3253	Epitomics, Burlingame, CA	Rabbit pAb	1:200
GPC4 Glypican-4 Cat. No. NBP1-45286	Novus Biologicals, Littleton, CO	Mouse mAb	1:250
Na+K+ATPase sodium-potassium pump Cat. No. SC71638	Santa Cruz Biotechnology, Inc.	Mouse mAb	1:40
PVRL3 Poliovirus receptor-related 3 Cat. No. SC28637	Santa Cruz Biotechnology, Inc.	Rabbit pAb	1:100
SLC4A4 Solute carrier family 4, sodium bicarbonate cotransporter, member 4 Cat. No. SC1622	Santa Cruz Biotechnology, Inc.	Goat pAb	1:100
SLC9A7 Solute carrier family 9 member 7 Cat. No. AB104868	Abcam, Cambridge, MA	Rabbit pAb	1:100
Z0-1 Zonaoccludens protein	BD Pharmingen	Mouse mAb	1:50

Flow Cytometry and FACS

Stromal fibroblasts were labeled with CellTracker green, 5-chloromethylfluorescein diacetate (CFMDA; Invitrogen, Corp.), before dissociation from culture according to manufacturer's protocol. Briefly, stromal fibroblasts were incubated with 5 μM of CFMDA in serum-free media at 37°C for 30 minutes. Subsequently, the CFMDA-containing media was replaced with fresh media for 30 minutes and the CFMDA would become a cell-impermeant, fluorescent green product. HCECs and stromal fibroblasts were dislodged from culture with cell dissociation buffer (Gibco, Carlsbad, CA). An approximately equal amount of HCECs and stromal fibroblasts were mixed together and blocked with 10% goat serum (Millipore) for 10 minutes at 4°C. The mixture was incubated with either anti-GPC4 or anti-CD200 antibody at the appropriate dilution (Table 1) for 20 minutes at 4°C. Cells were washed and incubated with anti-mouse antibody conjugated with Alexa Fluor 488 (Invitrogen Corp.) for 20 minutes at 1:1000 dilution. Thereafter, the cells were washed and resuspended in 200 μL of PBS. Cells either were analyzed by flow cytometry (Accuri C6; BD Biosciences) or sorted (MoFlo XDP; Beckman Coulter, Inc., Indianapolis, IN). Sorted cells were seeded onto a 48-well plate and observed under a microscope 24 hours later. Three independent sorting experiments were performed to obtain the average percentage of recovery and purity. Unpaired Student's t-test was applied to calculate statistical significance in recovery and purity between using anti-GPC4 and anti-CD200 antibody for sorting. 

Results

RNA-seq and Quantitative PCR of Relatively Overexpressed Genes in HCECs

RNA-seq was performed on two donor corneas to determine the relative expression of genes in human corneal endothelium and stromal keratocytes. Genes that were expressed at least 20-fold higher in the endothelium were crossed referenced with Genbank database to select those that code for membrane associated proteins. A selected list of candidate genes is shown in Table 2. In addition, quantitative PCR analysis of extracted RNA from cultured HCECs and stromal fibroblasts was performed to investigate if relative overexpression of some genes also would be reflected in cultured cells. The top 5 genes that were overexpressed in HCECs relative to stromal fibroblasts are shown in Table 3. 

Table 2. 

 

Relative Fold Difference of RNA Expression in Cornea Tissue Section

Gene	Fold Difference by RNA Seq (HCEC Over Stromal Keratocytes)
PVRL3	1700
TRO	1508
SLC9A7	889
PARD6A	796
SCAMP5	776
CNTN6	636
GPC4	490
PTPRO	315
PTCH1	125
SLC4A4	88

Table 3. 

 

Relative Fold Difference of RNA in Cultured HCEC Over Stromal Fibroblast

Gene	Fold Difference by Quantitative PCR (HCEC Over Stromal Fibroblast)
	Normalized to Actin	Normalized to GAPDH
GPC4	262.0 ± 55.3	70.3 ± 12.2
CNTN6	163.0 ± 21.5	43.80 ± 6.3
SLC9A7	8.30 ± 1.52	2.2 ± 0.92
PVRL3	2.83 ± 0.79	0.76 ± 0.22
SLC4A4	2.33 ± 0.56	0.62 ± 0.10

Screening of Antibodies on Cell Culture

Quantitative PCR data indicated that these genes - CNTN6, GPC4, PVRL3, SLC4A4, and SLC9A7, were significantly more abundant in HCECs compared to stromal fibroblasts in culture (Table 3). Commercial antibodies to these proteins were used to stain cultures of HCECs and stromal fibroblasts to confirm their expression on HCECs. It was observed that only anti-GPC4 antibody stained HCECs, but not stromal fibroblasts (Fig. 1A). 

Figure 1. 

 

Immunofluorescence microscopy of HCECs and stomal fibroblasts. Cell cultures of HCECs and stromal fibroblasts seeded onto a 12-well multi-slide was stained with antibodies against markers identified in Table 3 (A) or from screening the BD Lyoplate (B). The primary antibodies were detected with the appropriate secondary antibodies conjugated to Alexa Fluor 488 (Invitrogen Corp.). Nuclei of cells were stained with Hoechst.

The BD Lyoplate human surface cell marker panel provided a total of 242 monoclonal antibodies for screening on HCECs and stromal fibroblasts. From the screen, only CD104 and CD200 appeared to stain HCECs specifically (Fig. 1B). There may be differences in gene expression between HCECs in culture and cells in the corneal endothelium within the cornea tissue (Tables 2, 3); therefore, these antibodies were used to stain cornea tissue sections, too. It was observed that anti-GPC4, CD200, and SLC4A4 antibodies stained the corneal endothelial layer specifically. The rest of the antibodies either were found to be positive on the corneal endothelium and stromal layers (SLC9A7, PVRL3 and CD104), or on neither of them (CNTN6, Fig. 2). It also was observed that the DM was stained by some of the antibodies (Fig. 2, white arrowhead). 

Figure 2. 

 

Staining of cornea tissue sections. Formaldehyde-fixed cornea tissue sections were incubated with various primary antibodies. The primary antibodies were detected with the appropriate secondary conjugated to Alexa Fluor 488. A white arrow indicates the corneal endothelial layer. A white arrowhead indicates the DM.

In all, 7 potential cell surface markers were identified (5 from RNA-seq and 2 from BD Lyoplate) that might be specific for HCECs only. Since CD200 was identified through BD Lyoplate screen, we asked if CD200 is expressed more abundantly in cultured HCECs than cultured stromal fibroblasts. Quantitative RT-PCR analysis for CD200 indicated that it was, indeed, expressed more abundantly in HCECS than in stromal fibroblasts (Supplementary Table S1). 

Taken together, only anti-GPC4 and CD200 antibodies specifically stained both cultured HCECs and the endothelium in tissue sections, and hence, they were used for subsequent experiments. 

Flow Cytometry With Anti-GPC4 and Anti-CD200 on Cultured Cells

Flow cytometry with HCECs specific antibodies potentially could be used to purify HCECs from contaminating fibroblasts. To test if these antibodies are amenable for FACS, HCECs and stromal fibroblasts were labeled with anti-GPC4 and anti-CD200 antibodies in separate experiments, and analyzed by flow cytometry. Consistent with the staining results, both antibodies labeled HCECs only, but not stromal fibroblasts, causing a shift in the histogram when compared to the isotype controls (Fig. 3A). In contrast, the anti-NaK antibody caused a smaller shift in the histogram with HCECs relative to the isotype controls, and it also caused a slight shift with stromal fibroblasts (Supplementary Fig. S1). This suggests that anti-NaK antibody may not be suitable for sorting. The anti–ZO-1 antibody did not cause any shift in the histogram in either cell type relative to the isotype control (Supplementary Fig. S1). 

Figure 3. 

 

Flow cytometry analysis of antibody-labeled HCECs and stromal fibroblasts. (A) HCECs or stromal fibroblasts were labeled with anti-GPC4 antibody, anti-CD200 antibody, or a mouse IgG isotype control antibody. Anti-mouse antibody conjugated to Alexa Fluor 488 (AF488) was used to detect the primary antibodies. The histogram for HCECs or stromal fibroblasts labeled with the isotype control antibody (arrow, black) is overlaid with the histogram for HCECs or stromal fibroblasts labeled with either anti-GPC4 or anti-CD200 antibodies (pink). (B) CFMDA-labeled stromal fibroblasts were mixed with HCECs in a 1:1 ratio before incubation with either anti-GPC4 of anti-CD200 antibodies. The mixture subsequently was incubated with anti-mouse antibody conjugated to Alexa Fluor 647 (AF647). (I) and (IV) show detection of fluorescence produced by CFMDA, and (II) and (V) show the detection of AF647. M1 and M2 indicate the percentage of HCECs and stromal fibroblasts in the mixture, respectively. (III) and (VI) show a dot plot of all the events on a CFMDA versus AF647 graph. Each graph is divided into 4 quadrants, and the percentage of events captured in each quadrant is indicated.

Since all data until now were derived from HCECs in their third passage, FACS analysis was performed on HCECs in their first passage to ensure that GPC4 and CD200 were present. Consistent with previous observations, both antibodies caused a shift in the histogram when compared to the isotype control. (Supplementary Fig. S2). 

Next, we asked if anti-GPC4 and anti-CD200 antibodies were able to identify differentially HCECs within a mixed population of cells. Stromal fibroblasts were prelabeled with CFMDA before mixing with the HCECs at a 1:1 ratio. This was done to differentiate them from HCECs during the analysis. The mixed population of cells was incubated with either anti-GPC4 or anti-CD200 antibodies. Although both antibodies were able to stain HCECs, labeling with anti-GPC4 antibodies resulted in a better histogram separation (Figs. 3BII, 3BV). FACS analysis showed that within the mixed population, 1% of the anti-GPC4 positive cells and 1.7% of the anti-CD200 positive cells also were CFMDA positive, indicating a low level background staining of these antibodies on stromal fibroblasts (Figs. 3BIII, 3BVI). 

FACS With a Mixed Population of Cells

The mixed cell population of CFMDA prelabeled stromal fibroblasts and unlabeled HCECs, mixed at a ratio of 1:1, was incubated with either anti-GPC4 or anti-CD200 antibodies. GPC4- and CD200-positive cells were sorted (Figs. 4AI, 4AIV) and seeded into wells of a 48-well plate (Figs. 4AIII, 4AVI). Consistent with earlier results (Fig. 3B), cells labeled with anti-GPC4 antibodies were better separated from stromal fibroblasts (Fig. 4AI) than cells labeled with anti-CD200 (Fig. 4AIII). On average from three independent experiments, approximately 20% of HCECs were not recovered when sorted with anti-CD200 antibody (Figs. 4AIV, 4BI). In contrast, 96% of the cells were recovered when sorted with anti-GPC4 antibody (Figs. 4AII, 4BI). The difference in recovery between using anti-GPC4 or anti-CD200 antibody is significant (P ≤ 0.05). Flow cytometry analysis of cells sorted by either antibodies indicated that a small percentage of the cells were CFMDA-positive (Figs. 4AII, V). Low numbers of CFMDA-positive cells were, indeed, observed in culture after FACS (Figs. 4AIII, 4VI), indicating that some stromal fibroblasts were co-sorted with the HCECs. There is no significant difference (P ≥ 0.05) in terms of the purity (>98%) of the cells when sorted by the two antibodies (Fig. 4BII). 

Figure 4. 

 

FACS of HCECs from stromal fibroblasts. (A) A representative sorting experiment with anti-GPC4 and anti-CD200 antibodies. Anti-GPC4 or anti-CD200 labeled cells were sorted away from CFMDA labeled stromal fibroblasts (I) and (IV). Sorted cells (from [I] and [IV]) were gated and plotted on a histogram to quantify the percentage of CFMDA positive cells (indicated by arrows) in (II) and (V). Sorted cells subsequently were seeded on 12-well multi-slide and viewed under fluorescence microscope (III) and (VI). (B) The average recovery yield of HCECs and proportion of CFMDA-positive stromal fibroblasts from 3 independent sorting experiments using either anti-GPC4 or anti-CD200 antibody were tabulated and plotted as a bar graph (*) indicates significant difference, P ≤ 0.05). Mean values with SDs are indicated in or above the bars.

Cells Retain Surface Markers After Sorting

To determine if the sorted HCECs express markers indicative of the human corneal endothelium, sorted cells were plated and probed for their expression of Na⁺/K⁺-ATPase pump, tight junctions Z0-1, GPC4, and CD200 proteins. After sorting and plating, these cells would be in their fourth passage. Positive staining observed for all these antibodies indicated that HCECs retained the expression of these proteins after sorting (Fig. 5). These markers also are retained on HCECs that were cultured for five passages without undergoing any sorting procedure (Supplementary Fig. S3). 

Figure 5. 

 

Immunofluorescence staining of cells after FACS. After the HCECs were sorted, they were seeded onto a 12-well multi-test slide. Anti-GPC4, CD200, Z0-1, and Na⁺/K⁺-ATPase (NaK) antibodies were used to probe the cells. Anti-mouse conjugated to Alexa Fluor 647 was used to detect the primary antibodies.

Discussion

Expansion of HCECs from cultured donor corneal endothelium could provide an alternative source of transplantable material for endothelial keratoplasty. However, isolation of HCECs for in vitro culture potentially can be contaminated by stromal keratocytes/fibroblasts during the isolation process. If HCECs-specific cell-surface markers are identified in the context of the cornea, they can be used to identify and purify HCECs from cultures contaminated with stromal fibroblasts. Hence, cell-surface antibodies that are specific for HCECs would be valuable tools in this regard. 

Through gene expression profiling and screening of a panel of commercial antibodies, we identified two cell-surface antibodies that are specific for the corneal endothelium and cultivated HCECs. Although a recent transcriptomic study also has identified 4 markers for adult HCECs, they are intracellular proteins, and it was not demonstrated if these markers could be used to differentiate HCECs from stromal fibroblast cells in culture. ²⁶ Moreover, being intracellular proteins, they cannot be used in cell separation procedures to purify HCECs from cultures contaminated with stromal fibroblasts. 

Results from RNA-seq indicated that some genes were overexpressed in the corneal endothelium relative to the stroma. The relative expression of these genes in cultured HCECs and stromal fibroblasts subsequently were quantified by quantitative PCR, since expression levels could have changed during propagation in vitro. Discrepancies between the RNA-seq and qPCR data are expected, since these are two different methods used on two different types of samples. Differences in gene expression did not always translate to differences in protein expression as shown by immunohistochemistry. Nonetheless, immunohistochemistry on cultured HCECs and corneal tissue sections identified anti-GPC4 and anti-CD200 antibodies that specifically stain HCECs and the corneal endothelium. 

To our knowledge, GPC4 and CD200 have not been described as markers for HCECs. Glypicans are heparan sulfate proteoglycan bound to the outer surface of plasma membrane by a glycosyl-phosphatidylinosiol anchor. Glypicans can be released into the extracellular environment by an extracellular lipase known as Notum through the cleavage of the GPI anchor. ²⁷ This could explain why a signal was observed on the DM when corneal sections were stained for GPC4 (Fig. 2). It also has been reported that glypicans are associated with lipid rafts and the basolateral membranes of polarized cells. ²⁸ Members of this family are known to regulate Wnts signaling in Drosophila, ²⁹ zebra fish embryos, ³⁰ and mammalian cells, ³¹ as well as signaling of hedgehogs, fibroblast growth factors, and bone morphogenetic proteins. ³² CD200, previously known as OX2, is a membrane glycoprotein belonging to the immunoglobulin superfamily. It interacts with a receptor, CD200R, on myeloid lineage cells and downregulates myeloid lineage proliferation. ^33,34 Upregulation of this protein is associated with certain B-cell lymphoproliferative disorders, such as chronic lymphocytic leukemia and small lymphocytic lymphoma. ³⁵ A truncated isoform of CD200 resulting from alternative mRNA splicing has been found to be soluble, ³⁶ and this could explain the staining observed on the DM (Fig. 2). Functions of these two proteins in the HCECs have not been investigated. 

These two cell-surface antibodies are not just valuable tools for discriminating corneal endothelium/HCECs from stromal fibroblasts, but they also could be used for purifying HCECs contaminated with stromal fibroblasts. Previously, a negative depletion method using antifibroblast magnetic beads was used to purify cultured HCECs from stromal fibroblasts, since no cell-surface antibody specific for HCECs were described. ²⁰ In our study, we adopted a positive selection strategy to isolate HCECs from a mixed population of HCECs and stromal fibroblasts by FACS. Results in our study showed that cells separated by FACS were viable and retained pump-associated marker Na⁺/K⁺-ATPase, tight junction protein ZO-1, as well as GPC4 and CD200. Although purity of the eventual culture sorted by either antibody was comparable, sorting with anti-CD200 was found to be less efficient. Hence, anti-GPC4 antibody would be better suited for sorting by FACS. A strategy to sort cells using both markers simultaneously might enhance the purity of HCECs. Finally, these two markers are more useful than the other commonly used markers, such as NaK and Z0-1. Flow cytometry analysis showed that anti-NaK and Z0-1 antibodies would not be optimal reagents for sorting a heterogeneous culture (Supplementary Fig. S2). 

In conclusion, two cell-surface markers, GPC4 and CD200, previously not described in the corneal endothelium to our knowledge, were identified in our study. These markers could be used to differentiate HCECs from stromal keratocytes and corneal stromal fibroblasts. We demonstrated the use of these antibodies to separate and purify HCECs from corneal stromal fibroblasts by FACS. These antibodies also potentially might be useful as part of a panel of reagents for identifying and characterizing a population of putative HCECs derived from other cell sources, such as multipotent corneal progenitors ¹⁴ or pluripotent stem cells. 

Supplementary Materials

pdf S1 

pdf S2 

pdf S3 

pdf S4 

Acknowledgments

Supported by the Biomedical Research Council, Agency for Science, Technology and Research, Singapore. 

Disclosure: Y.K. Cheong, None; Z.X. Ngoh, None; G.S.L. Peh, None; H.-P. Ang, None; X.-Y. Seah, None; Z. Chng, None; A. Colman, None; J.S. Mehta, None; W. Sun, None