Mesenchymal Stem Cells in the Human Corneal Limbal Stroma

Abstract

Purpose.: Peripheral and limbal corneal stromal cells (PLCSCs), which contain keratocytes, have a complex phenotype. Knowledge of keratocyte cell properties, function, and origin is limited. Evidence available thus far has suggested both mesenchymal stromal and hematopoietic characteristics. Multipotent mesenchymal stromal cells (MSCs) are found in an increasing number of tissues and are the subject of considerable interest and investigation in the disciplines of tissue engineering, immunology, gene therapy, and oncology.

Methods.: Isolated PLCSCs were characterized by markers aldehyde dehydrogenase and keratocan, cultured, and analyzed against a set of criteria for the identification of MSCs developed by the International Society of Cellular Therapy (ISCT). PLCSCs were directly compared to fetal liver MSCs (flMSCs). Additional cell surface markers were also used to quantify differentiation, which was also performed on both cell types.

Results.: PLCSCs were found to be plastic adherent, displayed the correct profile and proportions of CSMs, and demonstrated trilineage potential in accordance with the ISCT guidelines. Furthermore, PLCSCs displayed a high degree of similarity to flMSCs and this likeness extended into the non-ISCT MSC cell surface markers and trilineage differentiation, which were often but not always comparable.

Conclusions.: Herein we report a novel observation that PLCSCs conform to all the ISCT criteria and are therefore MSCs. Furthermore, this study has identified the limbal stroma as yet another MSC niche and presents a new perspective on the role of the PLCSC.

Introduction

The cornea is the transparent window of the ocular surface. The stroma constitutes approximately 90% of the cornea and is sparsely populated with fibroblastic-like cells conventionally known as keratocytes. The keratocytes have a small body with extensive cytoplasmic processes that make contact with similar processes from adjacent cells. Under normal physiological conditions, keratocytes remain mitotically quiescent ^1,2 and maintain the highly organized collagen lamellae and proteoglycans, essential for providing corneal transparency. ^1,3,4 They are conventionally characterized by molecular markers including keratocan, aldehyde dehydrogenase (ALDH), crystallins, CD133, and CD34. ^2,5–9 In response to insult or injury, keratocytes become activated, lose expression of keratocyte markers, and adopt fibroblast and myofibroblast (α-smooth muscle actin positive) scar-forming phenotypes. ^10,11 This transition is reversible, ^12,13 indicating that keratocytes have an adaptive phenotype influenced by their environment. 

The corneal stroma is derived from the neural crest—a source of mesenchymal tissue in the head and neck. ¹⁴ Varied reports propose that keratocytes express markers, both mesenchymal (CD13, CD29, CD44, CD56, CD73, CD90, and CD105) ^2,15–17 and hematopoietic (CD11b, CD34, and CD133). ^7,9,18 Evidence also suggests cells from the corneal limbal stroma possess a degree of multipotency ¹⁵ and may be stem-like. ² Although no single report completely validates keratocytes as stem cells of mesenchymal, hematopoietic, or of any other lineage, the most convincing evidence suggests they are multipotent mesenchymal stromal cells (MSCs) also known as mesenchymal stem cells. Notably, Polisetty et al. ¹⁵ and Choong et al. ¹⁶ have both published work demonstrating some MSC characteristics of keratocytes isolated from the limbal stromal regions; however, their work falls short of validating the presence of MSCs according to the defined International Society for Cellular Therapy's (ISCT's) criteria. ¹⁹  

MSCs are known for their potency and plasticity, ^20,21 among other attributes. ^22–24 Furthermore, MSCs are being discovered in an increasing number of tissues throughout the human body and in extraembryonic tissues such as placenta and umbilical cord. ^22,24–26 Understanding of the role of MSCs in vivo is limited. Their in vitro properties are often extrapolated to ascribe in vivo functions. In the bone marrow, they function as stroma to which hematopoietic stem cells engraft in order to proliferate and differentiate. ^26–29 Therefore, it is plausible that MSCs in other tissues perform a similar function for other stem/progenitor cells. MSCs may also be involved in tissue repair, as circulating and engrafting MSC levels increase in response to cytokine signaling associated with injury. ^30–33  

This increasing diversity of sources, properties, and potential applications, coupled with the lack of a specific MSC marker, necessitates standard, minimal criteria. The ISCT's position paper on the minimal criteria is now widely used as a standard in the MSC field. ¹⁹ It stipulates that MSCs must possess plastic adherence; be positive for cell-surface markers (CSMs) CD73, CD90, and CD105 (≥95% of the population), but not CD11b or CD14, CD19 or CD79α, CD34, CD45, and HLA-DR (≤2% of the population); and have the ability to undergo adipogenesis, osteogenesis, and chondrogenesis. To date keratocytes have not been characterized as MSCs in this way. We adopted these criteria in order to characterize keratocytes isolated from the peripheral cornea and limbus. Furthermore, we directly compared these cells to previously validated human fetal liver MSCs (flMSCs), ^34–37 using the ISCT criteria. Because cultured keratocytes quickly lost their characteristic markers in vitro, we have used the term peripheral and limbal corneal stromal cells (PLCSCs) to refer to isolated and expanded cells from the peripheral corneal and limbal stroma. 

Materials and Methods

Isolation of PLCSCs

Use of human donor tissue for research was approved by the local ethics research committee (No. 07/H0403/140) and in accordance with the tenets of the Declaration of Helsinki, following consent obtained from the donors or their relatives. The tissue was obtained from Manchester Eye Bank. 

Stromal cells were obtained from human corneal rims. A rim constitutes a full thickness, 360° ring, of the corneoscleral disc, without conjunctiva, from which the central 7.5 to 8 mm of cornea has been punched out for corneal grafting. Preparation of this transplant material was done as previously discussed. ³⁸ Upon receipt, the rim was transferred from the storage solution into a petri dish for dissection. During dissection the rim was kept moist by using storage solution. Excess sclera was trimmed, leaving approximately 1 mm around the limbus, ensuring the limbus remained intact. The tissue thus included the peripheral cornea and limbus. For the purpose of this study, cells obtained from this tissue were termed peripheral and limbal corneal stromal cells. The epithelial and endothelial layers of the rim were scraped off. The remainder of the limbus was cut into small pieces, to increase the surface area, and was placed in 0.1 mg/mL collagenase Type IA (Sigma Aldrich, Dorset, UK) in basal culture medium M199 (Sigma Aldrich) filter-sterilized by using a 0.20-μm filter (Minisart High-Flow, Sartorius Stedim Epsom, Surrey, UK). The tissue was incubated overnight (37.0°C, 5% CO₂, 95% humidity) for approximately 18 hours after which the solution was filtered with a 41-μm nylon net filter (Fisher Scientific, Loughborough, UK) to remove the remnants of any undigested tissue. 

The rim/collagenase filtrate was deactivated by using culture medium (50% vol/vol) before being centrifuged at 450g for 6 minutes. The supernatant was discarded and the cell pellet was resuspended in 7 mL culture medium and then cultured. 

Isolation of Fetal-Derived MSCs

Human fetal liver was provided by the MRC-Wellcome Trust Human Developmental Biology Resource in Newcastle-upon-Tyne, United Kingdom, with appropriate maternal consents following ethical approval by the Newcastle and North Tyneside Research Ethics Committee in accordance with Human Tissue Authority regulatory guidelines. Livers from a gestational age between 8 and 9 weeks, determined by using published Carnegie staging, were collected into RPMI culture medium (Sigma Aldrich) containing penicillin/streptomycin. Livers were transported on ice and processed within 24 hours of harvesting. Fetal livers were processed and the cell suspensions, frozen and thawed as previously described. ³⁹  

Fetal liver MSCs were isolated from thawed fetal liver cell suspensions on the basis of adherence to tissue culture plastic. Cells were cultured in 25-cm² culture flasks by using approximately 5 × 10⁶ first-trimester fetal liver cells per culture flask in 7 mL medium. 

Cell Culture

Culture medium for both cell types contained M199 medium supplemented with 20% vol/vol heat-inactivated fetal bovine serum (Fisher Scientific), 2.5 μg/mL Plasmocin (Autogen Bioclear, Wiltshire, UK), 0.02 μg/mL gentamicin, 0.5 ng/mL amphotericin B (combination, Gibco Invitrogen, Paisley, UK), and 1.59 mM l-glutamine (Sigma Aldrich). M199 is a well-established MSC medium and has been shown to encourage expression of MSC characteristics when compared to DMEM. ^40–44  

Post isolation, PLCSCs and flMSCs were treated identically. Cells were cultured in 25-cm² culture flasks (Fisher Scientific), incubated at 37°C, 5% CO₂ vol/vol, and 95% humidity, and the medium was changed every 2 days until 80% to 90% confluent before sequential passaging up to passage 9, with cells cultured up to the first passage considered passage 0. Briefly, culture medium was removed and adherent cells were washed with Dulbecco's phosphate-buffered saline (PBS, Sigma Aldrich), incubated with TripLE Xpress (Gibco, Invitrogen) until cells detached. The resultant cell suspension was centrifuged for 10 minutes at 200g and the supernatant discarded. The cell pellet was then either resuspended in 21 mL culture medium and divided into three aliquots of 7 mL for further culture, or prepared for flow cytometry, differentiation, or gene analysis. Corneal epithelial cells (CECs) were cultured by using previously described protocols. ⁴⁵  

Sample Preparation for Flow Cytometry Analysis and Cell Sorting

Cells were detached as previously mentioned and the cell suspension was diluted with culture medium to 10⁶ cells per milliliter, 1 mL was added to flow cytometry tubes, and centrifuged for 10 minutes at 200g. For analysis only, cells were fixed for 5 minutes in 1 mL 3% vol/vol formaldehyde (Sigma Aldrich) in PBS. Cells were washed before incubation with the appropriate primary conjugated antibodies for 30 minutes at room temperature. For each sample a corresponding isotype control was incubated in a separate tube. A tube of cells was identically prepared but without antibody for the purposes of gating. After 30 minutes, cells were washed, centrifuged at 200g for 10 minutes, and resuspended in 1 mL PBS before analysis. 

Antibodies were as follows: CD11b, CD13, CD19, CD29, CD34, CD44, CD45, CD49b, CD49d, CD49e, CD105, HLA-ABC, and HLA-DR (Beckman Coulter, London, UK); CD49f, CD104, CD106, and cytokeratin (CK) 14 (AbD Serotec, Oxford, UK); CD73 (R&D Systems, Foster City, CA); CD90 (BD Pharmingen, Oxford, UK); CD133/2 and CD271 (Miltenyi Biotec, Surrey, UK); ABCG2 (Santa Cruz; Inside Biotechnology, Middlesex, UK); Cytokeratin 19 (CK19) and vimentin (Abcam, Cambridge, UK); Stro-1 (Biolegend, Cambridge, UK); and keratin 3/76 (Millipore; Fisher Scientific). 

Flow Cytometry Analysis and Cell Sorting

Samples (n ≥ 3) were analyzed by flow cytometry with the Epics Altra Flow Cytometer (Beckman Coulter, London, UK). Cells with specific marker phenotypes were isolated by fluorescence-activated cell sorting (FACS) with MoFlo XDP (DakoCytomation, Fort Collins, CO) and cells showing increased fluorescence were collected ² and cultured up to P3 before further analysis was done. Data obtained were further analyzed on a dot plot and histogram by using WEASEL software (version 3.0), with cellular debris being excluded from the analysis through relevant gating of events according to the forward and side scatter. Isotype controls were used as negative controls to set the threshold (0.5%) for the percentage of positive cells. ^46,47 The percentage of positivity for each cell surface marker was obtained by comparing the isotype controls to the test sample to obtain the reported values. 

Differentiation and Histologic Staining

At passage 3, flMSCs and PLCSCs (n ≥ 3 for each) were seeded at a cell density of 1.04 × 10⁴/cm² in 6-well plates. Once grown to 90% confluence, culture medium was switched to either adipogenic or osteogenic differentiation medium, following a protocol previously described by Sottile et al. ⁴⁸ in 2002, or chondrogenic differentiation medium (Invitrogen) following a protocol provided by the company. Negative controls, for which cells were maintained in culture medium without differentiation, were cultured alongside where all other conditions were the same. Cells were stained by following their respective protocols and using Alizarin Red (Sigma Aldrich), Oil Red O (Sigma Aldrich), and Alcian Blue (TCS Biosciences, Buckingham, UK) for osteoblasts, adipocytes, and chondroblasts, respectively. Cells were analyzed under an inverted light microscope (Nikon Eclipse TS100 Light Microscope, Surrey, UK), and images captured with the digital camera (Nikon D70s). 

RNA Extraction, cDNA Synthesis, and Quantitative PCR (qPCR)

RNA was isolated from three different cell lines for both flMSCs and PLCSCs (n ≥ 3 for each) and each sample analyzed in triplicate in accordance with the manufacturer's recommendations. Differentiated and control cells were lyzed in RLT and their RNA extracted by using the RNeasy Mini Kit (Qiagen, Surrey, UK). Approximately 1 ng/μL extracted RNA was used for cDNA synthesis with QuantiTect Reverse Transcription Kit (Qiagen) as directed by the manufacturer. 

Inventoried TaqMan assays (Applied Biosystems, Warrington, UK) were used for the genes as follows—adipogenesis: fatty acid synthase (FASN), peroxisome proliferator activated receptor γ (PPARG), and perilipin (PLIN); osteogenesis: bone morphogenetic protein (BMP) 4, BMP-6, and osteoclastogenesis inhibitory factor (OPG); chondrogenesis: cartilage oligomeric matrix protein (COMP), Sry-related HMG box 9 (SOX9), and aggrecan (ACAN); keratocyte markers: keratocan and ALDH. All reactions were performed in triplicates for all three donors, with the final reaction volume of 20 μL for each. The individual components in each reaction included 10 μL TaqMan Gene Expression Master Mix (Applied Biosystems), 4 μL RNAse-free water (Promega, Southampton, UK), 5 μL cDNA (1:5 dilutions), and 1 μL TaqMan probe. Amplification was performed with the Mx2005P multicolor 96-well PCR system (Agilent Technologies, Santa Clara, CA) with parameters of 95°C for 10 minutes (1 cycle) followed by 95°C for 30 seconds, 55°C for 1 minute, and 72°C for 30 seconds (the latter three run 40 cycles combined). Data analysis was done with MxPro version 4.2 software (Stratagene, Milton Keynes, UK) to measure the threshold cycle (Ct) for every reaction. A mean Ct value was established by using triplicate Ct values and analysis was completed by using the Δ-ΔCt method. ⁴⁹  

Results

Keratocyte Markers

PLCSCs were analyzed at passage P0 by qPCR and were shown to express keratocyte markers CD34, keratocan, and ALDH. This expression was significant when compared to that of CECs at P0 (Figs. 1a–1c).The relative gene expression of CD34 was measured over sequential passages from P0 to P9; expression levels decreased from P0 until there were none detectable by P3 (Fig. 1d). The percentage of PLCSCs positive for CD34 was measured by using flow cytometry over sequential passages from P0 to P9. This experiment demonstrated a decrease in CD34⁺ PLCSCs until at P3 they fell below detectable levels (Fig. 1e). 

Figure 1. 

 

Characterization marker expression in peripheral and limbal corneal stromal cells (PLCSCs). Comparative gene expression, by quantitative PCR analysis, of keratocan (a), ALDH (b), and CD34 (c) in PLCSCs and CECs. CD34 gene expression (d) and percentage of cells expressing protein (e) in sequential passages P0 to P9 in PLCSCs; flow cytometry analysis of PLCSCs showing numbers of CD34⁺ (gray fill) and CD34⁻ (patterned) cells in sequential passages P0 and P3 (f), showing compressed y-axis to scale.

Separately, CD34⁺ PLCSCs were isolated by FACS and counted at P0 and cultured until P3, at which time total CD34⁺ and CD34⁻ PLCSCs were counted. The total number of CD34⁺ PLCSCs remained relatively consistent between passages 0 and 3 (Fig. 1f). Cultured PLCSCs were plastic adherent and appeared morphologically identical to flMSCs at all passages (data not shown). 

Cell Surface Markers

Assessment of the ISCT profile of CSMs revealed that ≥95% of cells expressed CD73, CD90, and CD105 (Figs. 2–4) but were CD11b, CD19, CD34, CD45, and HLA-DR negative (≤2%) (Figs. 2–4). Furthermore, the mean fluorescence intensity of these markers demonstrated comparable levels of expression between PLCSCs and flMSCs (Fig. 3). CK3/12, CK14, and CK19 were used as controls against epithelial contamination of PLCSCs. These were negative in all samples (Fig. 2). 

Figure 2. 

 

Marker profiling of PLCSCs and flMSCs. Flow cytometry characterization cell phenotypes for cell surface markers related to the ISCT criteria and additional literature-based MSC and corneal epithelial markers. Data are summarized in tabular form (a) and then as bar charts according to ISCT (b), positive non-ISCT (c), and negative non-ISCT (d) marker profiles. In (a), ✓ denotes that the cells have fulfilled the ISCT criteria; ✗ denotes a value of ≤0.5% expression indicating negativity, while positive marker expression of additional markers is represented by actual experimental percentage expression. Bar charts (b–d) illustrate the error bars of the standard error of the mean.

Figure 3. 

 

Comparative expression levels of ISCT cell surface markers in flMSCs and PLCSCs. Histogram of the ISCT markers CD73 (a), CD90 (b), and CD105 (c) analyzed with flow cytometry. PLCSCs (blue), flMSCs (red), and overlapping fluorescence between flMSCs and PLCSCs (purple).

Figure 4. 

 

Comparative percentage levels of positive and negative ISCT cell surface markers in flMSCs and PLCSCs. Flow cytometry dot plot analysis for the ISCT markers CD73 (a, b), CD90 (c, d), CD105 (e, f), CD11b (g, h), CD19 (i, j), CD34 (k, l), CD45 (m, n), and HLA DP, DQ, DR (o, p) in flMSCs (a, c, e, g, i, k, m, o) and PLCSCs (b, d, f, h, j, l, n, p). Positive (a–f, >0.5%) and negative (g–p, <0.5%) levels of expression are indicated.

Additional CSMs selected on the basis of their reported association to MSCs in the published literature were used to further characterize these cells (Fig. 2). Three markers (CD29, CD44, and HLA-ABC) were almost ubiquitously (≥95%) expressed in both PLCSC and flMSC cell populations (Fig. 2). CD49f, CD104, CD133, and Stro-1 were negative (≤2%) in both. Two markers were differentially expressed by PLCSCs and flMSCs, which resulted in them falling into separate thresholds. CD106 was ≤2% in PLCSCs (whilst still showing some expression) but ≥2% in flMSCs. CD271 was uniformly absent in the PLCSC population but was present in 6.5% of flMSCs (Fig. 2). Values for all other additional markers (CD13, CD49b, CD49d and CD49e, CD106, ABCG2, and vimentin) fell between the >2% and <95% thresholds (Fig. 2). 

Trilineage Potential

Conventional staining showed that PLCSCs and flMSCs possess the ability to differentiate into osteoblast, chondroblast, and adipocyte lineages (Fig. 5). 

Figure 5. 

 

Characterization of flMSC and PLCSC multipotent differentiation potential. Comparative quantitative PCR analysis (a–i) and histologic staining (j–o) of flMSC (MC = control, MS = stimulated) and PLCSC (CC = control, CS = stimulated) multipotency. Gene expression and staining for adipogenesis (a = PPARG, d = PLIN, g = FASN, j and k), osteogenesis (b = BMP4, e = BMP6, h = OPG, l and m), and chondrogenesis (c = ACAN, f = COMP, i = SOX9, n and o) were assessed before (MC, CC) and after (MS, CS) differentiation. P values for gene expression are indicated, and statistically significant values are represented (*P < 0.1; **P < 0.01; ***P < 0.001). Positive adipogenic, osteogenic, and chondrogenic staining is represented by presence of lipids (pink), extracellular calcium deposits (red), and proteoglycans (green), respectively. Images taken are at ×400 magnification.

Transcription of the adipocyte markers PPARG (P < 0.01), PLIN (P < 0.01), and FASN (P < 0.1) were all significantly upregulated in both flMSCs and PLCSCs when stimulated for adipogenesis (Fig. 5). Osteogenic stimulation of both cell types significantly upregulated BMP-4 (P < 0.01) and BMP-6 (P < 0.001), while OPG (P < 0.6) was upregulated but not significantly (Fig. 5). After chondrogenic stimulation of both cell types, ACAN (P < 0.001) and COMP (P < 0.1) displayed significant upregulation, whereas SOX9 (P < 0.1) was significantly downregulated (Fig. 5). 

Discussion

PLCSCs displayed the typical keratocyte markers keratocan, ALDH, and CD34 when isolated; however, the cultured PLCSC population progressively lost any observable CD34 expression by passage 3 (Fig. 1). CD34 is a characteristic marker of quiescent keratocytes in the limbus ⁵⁰ and previous studies have shown that most keratocytes in the human cornea are CD34⁺. ⁹ According to our data, not all cells of the limbal and peripheral stroma express CD34, and the proportions are highly variable between donors (<1%–70%; data not shown). In culture as the cells proliferate, the percentage of cells expressing CD34 markedly drops. Significantly, the absolute number of CD34⁺ cells remains relatively unchanged, indicating that the apparent loss of CD34 is due to the CD34⁺ cells being outcompeted by their CD34⁻ progeny. It is this resultant population that fulfills the ISCT criteria, as demonstrated in this study. CD34⁺ cells are known to be lost rapidly in culture, ⁷ a finding comparable to ours. CD34⁺ progenitors of MSCs have previously been described in liver ⁵¹ and bone marrow. ⁵² In this article we have demonstrated that limbal and peripheral keratocytes produce the MSCs at passage 3. 

PLCSCs are plastic-adherent cells with a morphology indistinguishable from that of flMSCs. They also conform to the required ISCT profile of CSMs at passage 3. With regard to non-ISCT CSMs, a degree of heterogeneity was observed within and between PLCSCs and flMSCs. CD13, CD49b, and CD49d showed large variation between samples from different donors of the same cell source. CD271 may highlight an important difference between PLCSCs and flMSCs, as it is the only marker we found that is expressed in the latter cell type but not in the former. CD271 expression varies in MSCs from different sources, ranging from a low of 0.4% in bone marrow MSCs to 6% to 35% in placental tissues. ⁵³ This heterogeneity between MSC sources could form a basis for defining tissue-specific subpopulations. 

The variability and incomplete penetrance of many of the additional CSMs within the two groups of MSC serves to highlight the consistency with which both cell types conform to the ISCT CSM criteria. However, our data suggest that even within this pool of cells, different subpopulations exist. 

PLCSCs fulfilled the third and final criteria of multipotency. The subjectivity associated with the selection and interpretation of the intensity of staining is an obvious drawback of the staining method. However, quantitative analysis of gene expression for key indicators validated this. The assay also demonstrated a greater overall response for adipogenesis and osteogenesis by PLCSCs than by flMSCs, which had a greater response for chondrogenesis. Interestingly, the widely reported chondrogenesis marker SOX9 ^54–56 was downregulated in both cell types following differentiation. However, SOX9 is a marker of early chondrogenesis and is downregulated post 14 days' stimulation, ⁵⁷ which is comparable to our findings. Secondly, SOX9 expression is lower in monolayer rather than three-dimensional cultures. ^55–57 Finally, SOX9 is expressed in undifferentiated MSCs, ⁵⁵ and reduced expression from MSC to chondroblast does not signify SOX9 is not active. 

Conclusion

This work is conclusive evidence that the corneal limbal stroma contains MSCs. Furthermore, it demonstrated a keratocyte origin for MSCs in the cornea. Analysis of additional CSMs suggests a degree of heterogeneity within and between both PLCSCs and flMSCs. 

Acknowledgments

The corneal tissue was kindly provided by Manchester Eye Bank. 

Footnotes

 Supported by the National Eye Research Council (MJB).

Footnotes

¹  These authors are joint first authors.

Footnotes

²  These authors are joint senior authors.

Footnotes

 Disclosure: M.J. Branch, None; K. Hashmani, None; P. Dhillon, None; D.R.E. Jones, None; H.S. Dua, None; A. Hopkinson, None