Comparative global gene expression profile of human limbal stromal cells, bone marrow mesenchymal stromal cells, adipose-derived mesenchymal stromal cells and foreskin fibroblasts

Background: Limbal epithelial stem cells (LESC) have great potential in treating the blindness caused by corneal damage. LESC are maintained in stem cell niche called Palisade of Vogt. Limbal stromal (LS) cells are critical component of LESC niche and help in their self renewal. These cells resemble mesenchymal stromal/stem cells with multilineage differentiation potential. However little is known about their gene expression profile compared to MSC derived from various sources. Methods: In this study, we compared the gene expression profile of LS cells expanded in two different culture conditions: basal media with bFGF, LIF and matrigel (LS-matrigel), as well as basal media with 10% FBS (LS-FBS). In addition, gene expression profile of LS-FBS cells were compared to bone marrow, adipose-derived MSC and foreskin fibroblasts. Total RNA was extracted from various cell types upon achieving confluency and subjected to microarray experiments (Agilent platform) using Human GE 8x60k gene chips. Data analysis was done with GeneSpring software. Results: LS cells cultured in matrigel system showed upregulation of 871 genes as compared to LS-FBS and 58 genes were consistently differentially expressed in LS-FBS as compared to other cell types. Despite many long intergenic non-coding RNA and function unknown genes, differentially expressed genes represent gene ontology for cell signaling molecules including various growth factors, cell metabolism and extracellular matrix components for various biological processes. Samples from the same source were closely clustered by hierachical clustering analysis. Conclusions: The two culture conditions used in the study affected the gene expression profile of LS cells significantly. The LS cells showed distinct molecular signature as compared to MSC from various other sources. This comparative study will help in understanding the limbal stem cell niche biology and the novel set of genes could be used as biomarkers for LS cells.


Introduction
Human cornea on the front surface of eye is very critical for vision. The corneal transparency, continuous regeneration and functionality of corneal epithelium play an important role in refraction of light on to the retina. Corneal epithelium is regenerated by unique population of stem cells called limbal epithelial stem cells (LESC) that are located in the basal region of limbus. LESC differ from the corneal epithelium due to the lack of corneo-specific differentiation keratins (K3/K12) expression [1][2][3], connexin 43-mediated gap junction intercellular communication [4][5][6], p63 nuclear transcription factor [7,8], cell cycle duration [9], and label retaining property [10]. The limbal stroma provides a unique stem cell niche or microenvironment which is important for the modulation of stemness as it is heavily pigmented, highly innervated and vascularized. Clinically, destruction of LESC or the limbal stromal niche can lead to a pathological stage of LESC deficiency with severe loss of vision [11]. Chronic inflammation in the limbal deficient stroma is sufficient to cause detrimental damage to the conjunctival limbal autograft transplanted to patients or experimental rabbits [12]. These findings suggest that the limbal stromal niche is critical in regulating the selfrenewal and the fate of LESC. Although the mechanism remains elusive, modulation of epithelial proliferation, differentiation, proliferation and apoptosis by the limbal stroma has been doi: 10.7243/2054-717X-1-1 reported to favor stemness [13]. Limbal stromal (LS) cells are very important component of limbal stromal niche that helps in self renewal of LESC. Recently, LS cells were shown to have multilineage differentiation potential [14][15][16][17]. In one of the studies, an ABCG2-expressing FACS sorted side population cells from limbal stroma were able to differentiate into chondrocytes and neurons following differentiation induction [14]. In other studies, multipotent cells were also found in corneal stroma [15] and limbal stroma [16][17]. Earlier, we have reported that an ex vivo expanded LS cells possess multipotent differentiation potential towards adipocytes, osteocytes and chondrocytes [18]. Other stromal cells such as mesenchymal stem/stromal cells (MSC) can also be isolated and expanded in vitro for tissue regeneration applications [19][20][21][22]. MSC were first identified from bone marrow aspirates [23,24] and subsequently in Wharton's jelly of human umbilical cords [25], adipose tissue [26], dental tissues [27,28] and skin [29]. Most of the stromal cells derived from various sources expressed the markers of MSCs such as CD44, CD73, CD90, CD105, STRO1 and do not express markers of hematopoietic lineage such as CD14, CD34, CD45 and HLA-DR [30].
The advent of microarray technology has enabled the monitoring of individual and global gene expression patterns across multiple cell populations. Numerous stu-dies have now examined the global gene expression profile of MSC derived from different tissues that exhibit varying levels of proliferation, secretome and differentiation potentials. These studies showed that stromal cells isolated from different sources have unique molecular signatures despite sharing some common genes belonging to MSC when compared to mature fibroblasts [31][32][33]. It was also shown that gene expression profile of MSC remained stable during ex vivo expansion and subculture and was proven more sensitive to define MSC [34]. Other studies have reported gene expression profile of LESC [35][36][37][38], but comparative gene expression analysis of LS cells that are critical component of limbal stem cell niche and help in maintaining self renewal of LESC has not been well studied. As far as genome-wide differential gene expression profiling of LS cells is concerned, there is only one report where analysis were performed on mesenchymal cells derived from limbal explants culture [39]. Gene expression profile of LS cells also has not been compared with other stromal cells such as adipose stromal cells and foreskin fibroblast.
In order to find out the specific molecular signature, cellular function and potential biomarkers of the LS cells, we compared the global gene expression profile including long non-coding RNA (lincRNA) of the expanded LS cells with the MSCs derived from bone marrow, adipose tissue and foreskin fibroblasts. In addition, we also evaluated the effects of two different culture conditions on the LS cells gene expression.

Methods
The research protocol was approved by the Medical Research and Ethics Committee, Ministry of Health and the Medical Research Secretariat, UKM (University Kebangsaan Malaysia).

Establishment of limbal stromal cell culture
Corneoscleral rims from three cadaveric donors were obtained from post cornea graft transplantation with informed consent from the donor's relative. The rims were washed with phosphate buffer saline (PBS; Invitrogen Corporation, Carlsbad, CA) and then trimmed to remove the sclera. The limbal tissues were incubated at 37°C for 2 h with dispase (BD Biosciences, Mississauga, Canada) at a concentration of 5 mg/ mL. The limbal tissues were then cut into approximately 2 mm explants after washing with PBS. The limbal explants were cultured on matrigel (BD Biosciences, Mississauga, Canada) coated plates with complete medium containing Dulbecco's Modified Eagle's Medium (DMEM)/F12, 10% knockout serum replacement, 10 µg/mL insulin, 5 µg/mL transferrin, 5 µg/mL selenium-X, 100 IU/mL penicillin, 100 µg/mL streptomycin (all from Invitrogen Corporation, Carlsbad, CA, USA), 10 ng/ mL leukemia inhibitory factor (LIF) (Sigma-Aldrich Chemic, Steinheim, Germany) and 4 ng/mL basic fibroblast growth factor (bFGF; BD Biosciences, Mississauga, Canada) [17]. The expanded limbal stromal cells were subjected to fluorescenceactivated cells sorting (FACS) for the isolation of stage-specific embryonic antigen 4 (SSEA-4+) cells as reported previously [18]. The sorted SSEA-4+ cells were propagated on matrigel coated plate with the medium as mentioned previously. The limbal stromal cells that were maintained in this matrigel system were named as LS-matrigel. On the other hand, some of the sorted cells were maintained on normal plates with Dulbecco's Modified Eagle's Medium (DMEM)/F12 supplemented with 10% fetal bovine serum (FBS), 100 IU/mL penicillin, 100 µg/ mL streptomycin (all from Invitrogen Corporation, Carlsbad, CA, USA). These cells were identified as LS-FBS.

Human bone marrow mesenchymal stromal/stem cells (BM-MSC) culture
Bone marrow MSC from three different lots (Millipore, Billerica, MA) were propagated and cultured according to manufacturer's protocol. Briefly, cells at passage 4 were cultured on 0.1% gelatin coated plates with Mesenchymal Stem Cell Expansion Medium (Millipore, Billerica, MA) supplemented with 8 ng/ mL fibroblast growth factor-2 (FGF-2) (Millipore, Billerica, MA). When the cells were approximately 80% confluent, they were dissociated with trypsin-EDTA (Invitrogen Corporation, Carlsbad, CA) and passaged or alternatively frozen for later use. penicillin and 100 µg/mL streptomycin (all from Invitrogen Corporation, Carlsbad, CA). When the cells were approximately 80% confluent, they were dissociated with trypsin-EDTA (Invitrogen Corporation, Carlsbad, CA) and passaged or alternatively frozen for later use.

Gene expression profiling by microarray experiments
Genome-wide expression profiles of all the samples were analyzed using Agilent SurePrint G3 8x60K arrays (Agilent Technologies, Santa Clara, CA) that combined both coding and long intergenic non-coding RNA (lincRNA) for human genome. Prior to Cy3 labeling, 2uL of Agilent One-Color Spike Mix dilution was added to 100ng of total RNA for each sample. The total RNA was converted to cDNA and then to Cy3-labeled cRNA using Agilent One-Color RNA Spike-In Kit as per the manufacturer's protocol. The labeled cRNA was purified and quantitated prior to hybridization in hybridization oven at 65°C for 17 hr.

Microarray image and data analysis
Microarray image analysis was done using Feature Extraction version 10.7 and data analysis was done by using GeneSpring 11.5 (both from Agilent Technologies, Santa Clara, CA). The threshold was set to intensity value of 1.0. Normalization was done by 75 percentile shift. Baseline transformation was based on the median of samples. The data were further filtered by probeset on flags and expression less than 20. The data has been deposited in Gene Expression Omnibus (GEO) with accession number GSE38947. Unpaired Student's t test was used for statistical analysis. Genes up or downregulated by two-fold change were selected for further analysis. The false discovery rate (FDR) of 5% was estimated with the Benjamini-Hochberg method.
The gene expression profile of LS-FBS and LS-matrigel was compared. LS-FBS were chosen for the subsequent comparisons to other lineages. Hierarchical clustering was performed for LS-FBS versus BM-MSC, AD-MSC and HFF using Pearson Centered and Average-linkage clustering algorithm. Venn diagrams were drawn for the genes upregulated or downregulated in LS-FBS as compared to other lineages. Gene Ontology (GO) analysis was carried out for the upregulated genes and downregulated genes. Significant pathway analysis was also performed wherever possible. Gene functional classification was further carried out by DAVID software [40].

Real time RT-PCR
First strand cDNA was synthesized with Transcriptor First Strand cDNA synthesis kit (Roche Applied Science, Nonnenwald, Penzberg, Germany) as per manufacturer's protocol. Then, quantitative real time polymerase chain reaction (RT-PCR) was performed by using a LightCycler instrument (Roche Diagnostics, Nonnenwald, Penzberg, Germany). Primers for the panel of genes used in this study are listed in Table 1. Products of PCR amplification were detected through intercalation of the SYBR green dye from LightCycler FastStart DNA Master SYBR Green 1 kit (Roche Diagnostics, Nonnenwald, Penzberg, Germany). The amplification cycles were as follows: 95°C for 10 min, followed by 45 cycles at 95°C for 15 s, 62°C for 5 s and 72°C for 20 s. The concentration of MgCl 2 in all cycling reactions was 2.4 mM. Gene specific products were confirmed by melting curve analysis. Expression of the genes was normalized with the expression of GAPDH and The table details the primers sequences, their accession number and expected product size after RT-PCR.

Gene expression profiling
A total of 871 entities were found upregulated in LS-matrigel compared to LS-FBS (p<0.05, fold change ≥2). The differentially expressed genes (fold change >10) of LS-matrigel versus LS-FBS are depicted in Table 2. Hierarchical cluster analysis was performed to determine the relationship of the four different cell types (LS-FBS, BM-MSC, AD-MSC and HFF). The dendrogram in Figure 2 demonstrates that MSC isolated from the same source were clustered together. A total of 340 significant differentially expressed genes (p<0.05, fold change ≥2) were identified between LS-FBS and BM-MSC. Whereas, 399 and 146 differentially expressed genes were identified for AD-MSC the expression ratio was calculated by REST software [41].

Cell culture
The LS cells were established from corneoscleral rim tissues and cultured in two different conditions as mentioned in the methods. Cell outgrowths were observed after a few days of plating and the cells reached confluence in about 2-3 weeks. The LS cells appeared to be fibroblastic, elongated and spindle shape growing pattern ( Figure 1A). LS-matrigel cells have more elongated feature compared to LS-FBS (as shown in Supplement figure S1). The LS-matrigel cells could be cultured up to 10 passages or more. The LS cells derived from the samples using both methods were used in the subsequent experiments. The BM-MSC, AD-MSC and HFF showed spindle and fibroblastic morphology when cultured and expanded (Figures 1B-1D).
and HFF when compared to LS-FBS respectively. Venn diagram (Figure 3) shows that among the differentially expressed genes, 23 entities were upregulated (including one lincRNA) in LS-FBS (Table 3) whereas 35 entities (including of 4 lincRNA) were downregulated in LS cells versus the other three cell types ( Table 4). Among these were several genes contributing to the cell signaling receptors such as Frizzled family receptor 5 (FZD5) involved in Wnt signaling, extra cellular matrix proteins (SPON 2), transmembrane glycoprotein (GPMNB) involved in G protein coupled receptor activity and genes related to cell signaling and metabolism such as phosphatidylserine binding (SCIN and SCARB1) or lipid binding proteins (SCIN, SCARBI, NPC2, FABP3, and HS1BP3). In addition, transcription factor involved in sequence specific DNA binding including transcription factor AP-2 beta (TFAP2B) was also upregulated. However, we found different set of upregulated or downregulated genes when LS-matrigel was compared to the BM-MSC, AD-MSC and HFF. The common upregulated and downregulated genes in LS-matrigel are listed in Appendix 1 and 2. Venn diagrams of the comparison are shown in Supplement figures S2 and S3. In addition, genes that are differentially expressed when LS-matrigel was compared to BM-MSC, AD-MSC and HFF separately are shown in Appendix 3,4 and 5 respecti-vely. Some of the common genes that upregulated in LS-FBS and LS-matrigel when compared to BM-MSC were insulin-like growth factor binding 5 (IGFBP5), fibrillin 2 (FBN2), proprotein convertase subtilisin/kexin type 9 (PCSK9) and adrenomedullin (ADM). On the other hand, common genes that upregulated in LS-matrigel and LS-FBS when compared to HFF were Xg blood group (Xg), insulin-like growth factor binding protein 2 (IGFBP2) and laminin alpha 3 (LAMA 3). However, common upregulated genes could not be found in LS-FBS and LS-mat-rigel when compared to AD-MSC.
Further functional classification of genes that were highly expressed (more than 10-15 folds) in LS-FBS was done by using DAVID software. The differentially expressed genes as shown in Table 5 were classified as secreted proteins, signaling protein, extra cellular matrix, cell differentiation, basement membrane and cell adhesion protein. Some of the secreted proteins that were highly expressed when compared to BM-MSC and AD-MSC were EGF-like-domain, multiple 6 (EGFL6), angiopoietin-like 7 (ANGPTL7), insulin-like growth factor binding protein 2 (IGFBP2) and insulin-like growth factor binding 5 (IGFBP5). These proteins might play important role in proliferation and self renewal of LESC in paracrine fashion in the stem cell niche. Most of the Hox gene family members were downregulated as they are involved in differentiation. This also shows that LS-cells maintain multipotent phenotype by suppressing Hox genes.

Table 3. List of common upregulated genes in LS-FBS versus BM-MSC, AD-MSC and HFF.
f.c=fold change; N/A=not available.

Confirmation of differential expression by RT-PCR analysis
Differential expression of six regulated genes and a housekeeping gene, GAPDH was determined by RT-PCR as presented in Figure 4. RT-PCR confirmed the differential expression as observed in four cell the types. Specific amplification was confirmed by melting curve analysis. The tendency of differential expression in LS-FBS versus the rest was consistent between microarray data and real-time RT-PCR data in all the genes tested.

Discussion
In this study, we compared the gene expression of stromal cells derived from different sources namely limbal stromal cells (LS-FBS and LS-matrigel), bone marrow mesenchymal stem cells (BM-MSC), adipose-derived mesenchymal stem cells (AD-MSC) and human foreskin fibroblasts (HFF). Morphologically, these cells resembled the fibroblasts with a slight difference in their size and shape. The MSCs derived from various sources are known for their multipotential differentiation towards adipocytes, osteocytes and chondrocytes [42][43][44]. However, they differ in terms of growth factor, cytokine secretion and immunomodulatory properties [45]. atures despite sharing some common genes that are highly expressed compared to other MSC as shown in Tables 2 and 5 and Appendix 3,4 and 5. Most of the differentially expressed genes in LS-matrigel are involved in the extracellular components such as collagen, type XXI, alpha 1 (COL21A1), matrix metallopeptidase 27 (MMP27), cartilage oligomeric matrix protein (COMP), collagen, type XV, alpha 1 (COL15A1), collagen, type III, alpha 1 (COL3A1), collagen type IV, alpha 6 (COL4A6) and collagen type V, alpha 1 (COL5A1). The results demonstrated that when LS cells were cultured with FBS without matrigel, the expression of these matrix proteins was downregulated. The matrigel provided an efficient culture microenvironment supporting the production of ECM. Our findings concurred with others that culturing method can have influence on the gene expression profile of stem cells [46]. Higher expression of ECM proteins in LS-matrigel as compared to LS-FBS might mimic the stem cell niche environment for LS cells and might be useful in the maintenance of the limbal epithelial stem cells. Different culture conditions have effect on cell characteristic and gene expression. We believe this maybe an adaptive response to stimuli during damage or pathogenesis of limbal epithelial stem cell niche. Due to this adaptive response, LS cells may generate necessary paracrine factors and ECM proteins to help in recovery process. In addition, LIF has been reported to play a role in self renewal and differentiation of human and mouse stem cells [47]. Murine embryonic stem cells for instance depend strictly on LIF for self renewal and maintenance of pluripotency but LIF is not able to maintain human embryonic stem cells. However, our result showed that both LIF and matrigel were not able to induce pluripotency of the SSEA-4+ LS cells.
Although cell culture conditions, growth factors and even FBS affect the gene expression of the cultured cells, there is still no standard culture protocol for MSC derived from various sources. The characteristics of MSC are always confirmed by immunophenotyping and differentiation assay towards adipocytes, osteocytes and chondrocytes [30]. However, the ex vivo expanded MSC are normally heterogenous. Therefore, a systematic ex vivo global molecular characterization of MSC is needed in the future to define MSC. Thus, gene expression profiling provides an important tool for comparison and characterization of stromal cells from various sources.
In this study, LS-FBS and LS-matrigel were compared to BM-MSC, AD-MSC and HFF cultured in FBS. This study demonstrates a set of novel differentially expressed genes in LS-FBS compared to BM-MSC, AD-MSC and HFF. We also found different set of common genes that were highly expressed by LS-matrigel compared to BM-MSC, AD-MSC and HFF cultured in FBS. This might be due to the culture media components such as LIF, bFGF and matrigel. For LS-FBS, the highest expressed gene, SCIN is a Ca 2+ -dependent actin severing and capping protein [48] which is presumed to regulate exocytosis by affecting the organization of the microfilament network underneath the plasma membrane. This may play an important role in secretion of various growth factors required for maintenance and self renewal of LESC. It also regulates chondrocytes proliferation and differentiation. The second highly expressed
gene, Ras-related GTP binding D is a monomeric guanine nucleotide-binding protein, or G protein. The G proteins act as molecular switches in numerous cell processes and signaling pathways (supplied by OMIM). The intracellular fatty acid binding protein 3 (FABP3) is another highly expressed gene in LS cells. The fatty acid binding proteins (FABP) belong to a multigene family. FABP are thought to participate in the uptake, intracellular metabolism and/or transport of longchain fatty acids. They might be responsible in regulating cell growth and proliferation. One of the FABP genes, FABP4 has been reported to be upregulated during adipogenesis of MSC [31,49]. Fourthly, the transcription factor AP-2 beta (TFAP2B) is involved in the regulation of cell proliferation and differentiation during embryonic development. This protein functions both as transcriptional activator and repressor by binding to different promoters. Mutation in TFAP2B gene has been reported in autosomal dominant Char syndrome indicating its function in the differentiation of neural crest cell derivatives. Since cornea is one of the derivative of neural crest during embryo development, the over expression of this gene is significant. In addition, angiopoietin-like 7 (ANGPTL7) was highly expressed in LS cells. The protein acts as a negative regulator of angiogenesis besides playing a role as morphogen in inducing a corneal phenotype in vivo [50]. Another highly expressed gene related to the eye was tetraspanin 10 (TSPAN10) [51] and the function of this protein doi: 10.7243/2054-717X-1-1 is unknown. The transmembrane glycoprotein (GPNMB) was also reported to be expressed in the lowly metastatic human melanoma cell lines and xenografts and it could add in melanogenesis as the limbus of the human eyes is rich in melanin. Spondin 2 (SPON2) which was also highly expressed in LS cells is a cell adhesion protein that promotes adhesion and outgrowth of hippocampal embryonic neurons while inhibit angiogenesis [52]. The antiangiogenesis properties of this protein might help in maintaining transparency of cornea by inhibiting neoangiogenesis. Polisetti et al., recently reported a different set of differentially expressed genes in LS cells related to extracellular doi: 10.7243/2054-717X-1-1 components, cell adhesion molecules (microfibrillar associated protein 5, syndecan 2, matrix-remodelling associated 5, chondroitin sulfate proteoglycan 4 and collagen 8 alpha 1) as compared to BM-MSC [36]. However, by using different approaches such as single color versus dual color hybridization and isolation methods, our results revealed novel genes involved in the biological processes and molecular functions that have not been reported for ex vivo cultured LS cells. Besides extracellular related genes (LAMA3, IGFBP2, IGFBP5) and cell adhesion molecules (SPON2, JAM2, EGFL6, ITGA8), most of the genes significantly overexpressed belonged to the various biological process and molecular functions. In addition, many of the downregulated genes encode the class of transcription factors called homeobox genes that are involved in differentiation (HOXD8, HOXA9, HOXA7, HOXC10, HOXA10,  and HOXA11). This shows that the LS cells are multipotent and suppressing the expression of HOX gene family members. These DNA-binding transcription factors may regulate gene expression, morphogenesis, and differentiation. The HOX genes have been reported to play a role in both the early stem cell function as well as in the later stages of hematopoietic differentiation, and perturbation of HOX genes expression can be leukomogenic [53].
The Wnt pathway has been implicated in the regulation of self-renewal and cell fate determination in embryogenesis and stem cells [54]. Optimal level of Wnt/β catenin signalling increased the proliferation and colony-forming efficiency of primary LESC while maintaining the stem cell phenotype of these cells [55]. This was also shown in Dkk2 knockout mice where they displayed epidermal differentiation on the ocular surface [56]. PAX6 expression was also lost in the corneal epithelial cells of these mice, suggesting it is downstream of Dkk2 [56]. Deficiencies in PAX6 leads to aniridia resulting in impaired corneal epithelial function and eventual LESC failure, which may be due to altered niche development. Interestingly, we found high expression of frizzled family receptor 5 (FZD5) in LS-FBS compared to the other stromal cells. High level of Wnt signaling pathway in LS-FBS and LS-matrigel cells (as shown in Table 3 and Appendix 3) may help to modulate the proliferation, self-renewal and differentiation of LESC. This demonstrates the importance of limbal niche control over LESC fate during development.
In this study, we also reported a set of unknown genes and long intergenic non-coding RNAs (lincRNAs) which are highly expressed in LS cells (Tables 2 and 5). LincRNAs have been shown to act in the circuitry controlling self renewal and differentiation of stem cells by epigenetic modifications [57,58]. Further studies would be needed to elucidate the functions of these lincRNAs.

Conclusion
We report a novel set of genes that are consistently highly expressed in LS cells compared to the bone marrow MSC, adipose-derived MSCs and foreskin fibroblasts. The LS cells have unique molecular signature compared to other MSC lineages. Thus, the highly upregulated genes in LS cells could be used as biomarkers by using real time RT-PCR which is less labourious and quicker as compared to microarray analysis. The knowledge gained can help us to improve our understanding of the cellular signaling pathways involved in LESC self-renewal, survival and differentiation, and may aid in the development of strategies to improve the tissue regeneration potential of these cells.