Comparative Transcriptomic Analysis of Cultivated Limbal Epithelium and Donor Corneal Tissue Reveals Altered Wound Healing Gene Expression

Abstract

Purpose.: The improved surgical outcomes associated with transplantation of cultivated amniotic membrane expanded limbal epithelium (AMLE) compared to traditional donor methods has led to substantial adoption of this technique for treatment of limbal stem cell deficiency.

Methods.: The mRNA expression profiles of AMLE and CE were assayed using microarrays. Transcripts with a 1.5-fold change in either direction in addition to a Bonferroni adjusted P value < 0.05 were considered to be differentially expressed. Expression changes detected by microarray profiling and important corneal-limbal markers were assessed using quantitative real-time PCR (qRT-PCR) and immunofluorescence staining.

Results.: A total of 487 probe sets (319 upregulated and 168 downregulated) were found to be differentially expressed between AMLE and CE. Enrichment analysis revealed significant overrepresentation of multiple biological processes (e.g., response to wounding, wound healing, and regulation of cell morphogenesis) within the differentially expressed gene list. The expression of a number of genes that were upregulated (ABCG2, S100A9, ITGA5, TIMP2, FGF5, PDGFC, SEMA3A) and downregulated (KLF4, P63α) in AMLE was confirmed using qRT-PCR. Immunofluorescence confirmed that AMLE cultures were P63α, ABCG2, CK3, CK12, and E-cadherin (E-cad) positive.

Conclusions.: In this study, we have shown that genes associated with wound healing processes are upregulated in AMLE. These gene expression changes may contribute to corneal restoration and the positive outcomes associated with transplantation.

Introduction

The dynamic corneal epithelial surface requires continual regeneration if integral transparency, tensile strength, and pathogenic defense functions are to be maintained. ¹ Adequate populations of limbal epithelial stem cells residing in the pericorneal limbal niche are central to this regeneration process and are thought to divide both symmetrically and asymmetrically for self-renewal and production of daughter transit-amplifying cells that migrate and differentiate centrifugally toward the corneal apex. ^2–4 Limbal stem cell deficiency arises due to the destruction of these limbal epithelial stem cells and/or their niche microenvironment, resulting in a diminished capacity for self-renewal and ultimately an inability to adequately repopulate and repair the ocular surface. ^3,5 The deficiency may be congenital, systemic, or induced by injury, with symptoms including photophobia, corneal vascularization, ulceration, stromal scarring, recurrent corneal erosion, persistent pain, and loss of visual acuity that greatly reduces quality of life. ^6,7  

Standard surgical treatments for total limbal stem cell deficiency aim to replace damaged corneal epithelium with healthy donor tissue. ⁸ Application of corneal epithelium (keratoplasty) alone is ineffective ⁸ ; however, application of additional limbal tissue in dual keratoplasty-keratolimbal grafts offers improved outcomes, and success rates of ∼50% have been reported. ^8–12 While transplantation of donor corneal epithelium is generally well tolerated, application of additional limbal material increases the probability of immunological rejection. ^8,11,13  

Recent techniques pioneered by Pellegrini et al. ¹⁴ exploit the limbus by ex vivo limbal explant culture for production of stem cell–rich limbal epithelial sheets. ^8,14–18 Such cultivated limbal epithelium can be grafted to the debrided corneal surface without the need for limbal tissue, increasing success rates up to 70%. ^19–21 The reasons for these improved clinical outcomes are unclear; however, the carriers upon which the limbal epithelium is cultured do affect stem cell retention and therefore performance. ^13,18,22,23 Limbal epithelium has been cultured on a broad range of carriers including fibrin, soft contact lenses, biosynthetic collagen-based membranes, human anterior lens capsule, and human amniotic membrane. ^24,25 Human amniotic membrane carriers are often selected for clinical application due to the capacity of acellular amnion to aid ocular surface healing ²⁶ through reduction of inflammation, fibrosis, and neovascularization. ^27,28 Growth factors including keratinocyte growth factor (KGF), epidermal growth factor (EGF), hepatocyte growth factor (HGF), basic fibroblast growth factor (bFGF), and transforming growth factor-alpha (TGF-α), -beta1 (β1) and -beta2 (β2) present in amnion are thought to encourage proliferation and re-epithelialization, ²⁹ while the hypoimmunogenic effects of amniotic membrane are beneficial following graft surgery, where downregulation of key proinflammatory cytokines (e.g., interleukin-1 alpha [IL-1α] and beta [IL-1β] in addition to physical basement membrane sequestration of recipient immune cells likely aids engraftment. ^28,30,31 Application of acellular amnion in a technique termed amniotic membrane patching has also demonstrated utility by inhibition of proteinase activity and promotion of wound healing in alkali-burned cornea. This technique, however, is not sufficient to restore function for patients suffering from total limbal stem cell deficiency. ^8,32 These ameliorative effects and amnion's reported ability to support improved long-term survival of limbal epithelial stem cells make amniotic membrane an attractive carrier for limbal epithelial transplant. ^22,33,34  

A number of studies have compared the transcriptional profiles of uncultured limbal epithelium and corneal epithelium (CE) ^35–37 ; however, while limbal explants are used to generate cultivated limbal epithelium, the explants themselves may be removed during graft surgery ³⁸ and are not essential for achievement of successful transplant outcomes. ^8,14,15,39 This study therefore aimed to investigate the molecular basis for the clinical utility of amniotic membrane expanded limbal epithelium (AMLE) by comparing global gene expression profiles of AMLE and CE. From this analysis we observed overexpression of putative limbal epithelial stem cell markers (e.g., ABCG2, ITGA5, S100A9) and, for the first time, enrichment of response to wounding- and wound healing–associated genes and pathways in AMLE. 

Materials and Methods

Corneal Tissue Preparation and Culture

Normal human corneas were obtained from the Rocky Mountain Lions Eye Bank (Casper, WY, USA) at 3 to 10 days post mortem. Tissues were collected with informed consent of all subjects and in compliance with the guidelines of the Declaration of Helsinki. Ethical approval was granted by the Institutional Research Ethics Committee of the Royal Victoria Eye and Ear Hospital, Dublin, Ireland. Anonymized patient and processing information may be found in supplementary data (Supplementary Table S1). 

Corneal epithelium samples were isolated from donor corneal buttons. Upon receipt, the blunt edge of a scalpel blade was used to separate corneal epithelial tissue. Separation of the corneal epithelial sheet was confirmed using a Nikon Eclipse TS100 microscope (Nikon, Melville, NY, USA). Amniotic membrane expanded limbal epithelium was derived from corneal limbal tissue of six cadavers within 10 days of biologic death. Limbal rings were dissected and limbal explants (measuring ∼2 × 2 × 0.25 mm) cultured on fresh-frozen human amniotic membrane (Transplant Service Foundation, Barcelona, Spain; Code 30102-619). To prepare the amniotic membrane for limbal explant culture, cryopreserved amniotic membrane was thawed and freezing medium aspirated; nitrocellulose paper was removed, and membranes were washed in PBS and incubated with trypsin and EDTA (20 minutes, 37°C). Cell scrapers (Costar-Corning, Corning, NY, USA) were used to remove amniotic epithelium and mucus from the membrane. The amniotic membrane was held taut by wrapping around a sterile glass slide, and limbal explants were placed on the de-epithelialized surface. Explants were cultured in supplemented epithelial growth medium: 3:1 Dulbecco's modified Eagle's medium/Ham's F12 (Sigma-Aldrich Corp., St. Louis, MO, USA), 10% fetal calf serum (Invitrogen-Gibco, Grand Island, NY, USA), 5μg/mL insulin, 10ng/mL recombinant human epidermal growth factor, 100ng/mL cholera toxin, 0.4μg/mL hydrocortisone, 1 nM triiodothyronine (Sigma-Aldrich Corp.). All cultures were incubated at 37°C, 5% CO₂ with media aspiration every 2 to 3 days until confluent. TRI reagent (Sigma-Aldrich Corp.) was applied and RNA isolated according to manufacturer's instructions. 

Characterization of Cells in AMLE Cultures

To validate ex vivo expansion of AMLE, morphological inspection and immunofluorescent staining of the corneal epithelial differentiation markers E-cadherin (E-cad), KRT3, and KRT12 ⁴⁰ and limbal epithelial stem cell markers P63α and ABCG2 were carried out. ^40–43 Samples were washed with PBS and fixed using appropriate solvent (ice-cold methanol or methanol:acetone). Primary antibody solutions were then applied to samples: mouse anti-KRT3 (Millipore Corporation, Billerica, MA, USA), rabbit anti-KRT12 (Santa Cruz Biotechnology, Dallas, TX, USA), mouse anti-E-cadherin (BD Biosciences, Oxford, UK), rabbit anti-p63α (Santa Cruz Biotechnology), mouse anti-ABCG2 (Enzo Life Sciences, Exeter, UK). These were incubated at 4°C overnight and for 1 hour at 37°C where required. Following three wash steps (PBS/0.25% Tween 20), secondary antibody solutions (anti-mouse Alexa Fluor 488 and anti-rabbit Alexa Fluor 647 [Life Technologies, Carlsbad, CA, USA]) were applied and incubated for 1 hour at room temperature. Stained samples were washed (PBS/0.25% Tween 20) and cell nuclei counterstained using 300 mM 4′,6-diamidino-2-phenylindole (DAPI; Life Technologies). Samples were mounted using ProLong Gold antifade reagent (Life Technologies) and analyzed using confocal microscopy (Leica SP2 AOBS; Leica Microsystems, Meath, Ireland). 

Microarray Profiling and Data Analysis

Using the tissue culture techniques outlined, we prepared three biological replicates for both donor CE and AMLE. Total RNA was isolated from cell samples using Qiagen RNeasy Mini Kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. The concentration of RNA was determined using a NanoDrop spectrophotometer (ND-1000; Labtech International, Uckfield, East Sussex, UK), and RNA integrity verified using a 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). Affymetrix GeneChip Human Gene 1.0 ST arrays were used for gene expression analysis according to the manufacturer's instructions (Affymetrix, Santa Clara, CA, USA). The resulting microarray data were preprocessed using the robust multichip average (RMA) algorithm via the BioConductor oligo package. ⁴⁴ High-level quality control of the microarray data was conducted in the R software environment (provided in the public domain by R Foundation for Statistical Computing, Vienna, Austria, available at http://www.r-project.org/) with hierarchical cluster analysis (HCA) using the Euclidean distance measure and Ward's clustering algorithm. Differential expression analysis was carried out to highlight transcripts that were up- or downregulated in the AMLE samples in comparison to the CE samples. Only those probe sets in which at least 1.5-fold change (FC) in either direction along with a Bonferroni adjusted P value < 0.05 were considered to be dysregulated. Enrichment analysis was carried out for the differentially expressed mRNAs via the DAVID interface (http://david.abcc.ncifcrf.gov [in the public domain]) for three separate pathway databases (Gene Ontology [GO], KEGG, and Panther). Only those biological processes yielding a Benjamini-Hochberg adjusted P value < 0.05 were considered to be significantly overrepresented within the differently expressed gene list. The microarray data used in the study are available from the National Center for Biotechnology Information (NCBI) GEO database (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=svsdwikkrtozdqx&acc=GSE56421 [in the public domain]). 

Quantitative RT-PCR Analysis

Primers for quantitative RT-PCR (qRT-PCR) were designed using NCBI primer blast (http://www.ncbi.nlm.nih.gov/tools/primer-blast/ [in the public domain]) against NCBI Entrez nucleotide mRNA sequences (http://www.ncbi.nlm.nih.gov/nuccore [in the public domain]). Primer sequences and characteristics are detailed in Supplementary Table S2. Following isolation, RNA quantity and quality were assessed using a NanoDrop spectrophotometer (ND-1000; Labtech International). RNA to cDNA transcription was performed using High Capacity RNA-to-cDNA Kit (Applied Biosystems, Foster City, CA, USA) and T3 thermocycler (Biometra, Göttingen, Germany) according to manufacturer's instructions. The PCR master mix consisted of KiCqStart SYBR Green qPCR ReadyMix (Sigma-Aldrich Corp.), 50 ng cDNA template, and 300 mM primer. The PCR thermal cycling conditions were as follows: ABI 7500 thermal cycler, Fast7500 mode, initial denaturation of 95°C for 30 seconds, followed by 38 cycles of 95°C for 5 seconds, 56°C for 15 seconds, and 72°C for 25 seconds (Applied Biosystems). 

Results

Characterization of AMLE

Immunofluorescence staining of AMLE was performed to validate the culture method prior to microarray analysis (Fig. 1). Once cells cultured from limbal explants reached confluence (17–19 days), they were characterized using antibodies against the corneal epithelial differentiation markers E-cad, KRT3, and KRT12 ⁴⁰ as well as the limbal-basal epithelial stem cell markers ABCG2 and P63α. ^40–43 Expanded AMLE cells formed stratified cell sheets two to three cells thick (Fig. 1A); cells on the apical layer exhibited larger squamous cell morphology (Figs. 1B–E) while cells on the basal layer exhibited small nuclei and cobble-shaped cell morphology (Figs. 1F–I). This variation in morphology of the cell layers was also reflected in the expression of the selected markers. The apical cell layer stained strongly for corneal epithelial differentiation markers KRT12 (Fig. 1C) and E-cad (Fig. 1D), while the basal cell layer demonstrated the strongest expression of limbal/basal epithelial stem cell markers P63α (nuclear localized) and of ABCG2 (membrane localized) (Figs. 1H, 1I). Cells throughout the AMLE layers stained positive for KRT3 (Figs. 1B, 1F), demonstrating their corneal epithelial lineage. 

Figure 1

 

Immunofluorescence characterization of AMLE. (A) Cross-sectional view demonstrates classic corneal epithelial stratification in AMLE cultures. Apical squamous cell layers (B–E) display strong staining for corneal epithelial differentiation markers (B) KRT3, (C) KRT12, and (D) E-cad and weak staining for limbal epithelial stem cell markers (D, E) P63α and ABCG2. The basal layer (F–I) is composed of tightly packed cobble-shaped cells strongly stained for limbal/basal epithelial stem cell markers (H, I) P63α (nuclear localized) and ABCG2 (membrane localized). (B, F) Cells throughout the AMLE layers stain positive for KRT3, demonstrating their corneal-epithelial lineage. Nuclei are DAPI counterstained (blue).

Microarray Analysis of AMLE and CE Tissue

Following tissue culture, RNA was extracted from triplicate AMLE and CE samples for microarray analysis. The resulting data were normalized, and an unbiased data quality control analysis using hierarchical cluster analysis was conducted to confirm that sample replicates grouped as expected and that there was distinct separation between the sample types (Supplementary Fig. S1). Differential expression analysis identified 487 probe sets that were significantly differentially expressed between these sample types (i.e., a 1.5-fold change in either direction with a Bonferroni adjusted P value < 0.05). In total, 319 probe sets (277 nonredundant annotated genes) were found to be upregulated and 168 probe sets (124 nonredundant annotated genes) downregulated in AMLE relative to CE samples (Fig. 2A; Supplementary Table S3). 

Figure 2

 

Differentially expressed transcripts between AMLE and CE samples are enriched for several biological processes and pathways. (A) Heat map of 487 differentially expressed probe sets shows 168 downregulated probe sets (red) and 319 upregulated probe sets (green) in AMLE relative to CE comparator samples. (B) DAVID bioinformatics enrichment tool identified enrichment in response to wounding, wound healing, and regulation of morphogenesis biological process categories and MAPK signaling pathway, regulation of actin cytoskeleton pathway, and integrin signaling pathway. P values are adjusted using the Benjamini-Hochberg method.

Enrichment Analysis of Differentially Expressed Genes Reveals the Overrepresentation of Multiple Biological Processes

In order to gain an appreciation of overrepresented biological processes within those genes found to be differentially expressed, we carried out enrichment analyses against GO biological processes, KEGG pathways, and Panther pathways. The most significantly enriched biological processes were response to wounding (P = 1.74 × 10⁻³), wound healing (P = 1.10 × 10⁻²), and regulation of cell morphogenesis (P = 3.88 × 10⁻²) (Fig. 2B). The most significant category, response to wounding, contains 30 genes including S100A9, SAAs, IL1B, CFB, CD59, FBN1, CD24, ITGB6, ITGA5, TGFB2, PKC, GNA12, and MAPKK1, which are overexpressed in AMLE compared to CE. Eight response to wounding genes were downregulated in AMLE (Table 1). All 17 genes (16 upregulated and one downregulated) in the wound healing biological processes category were also listed in the response to wounding category (Table 1). In addition, the regulation of cell morphogenesis GO category was found to be overrepresented and contained 13 genes, 11 upregulated and two downregulated in AMLE (Table 2). Several of these genes, including FYN, LIMK1, MAP1B, NCAM, and SEMA3A, are associated with neural or oligodendrocyte development. ^45–51 KEGG and Panther pathway analysis revealed the enrichment of wound healing–associated biological processes in the list of differentially expressed genes; enriched KEGG pathways included mitogen-activated protein kinase (MAPK) signaling pathway (P = 7.82 × 10⁻³) and regulation of actin cytoskeleton (P = 1.20 × 10⁻²), while Panther identified significant enrichment for the integrin signaling pathway (P = 2.22 × 10⁻²). 

Table 1

 

Microarray Analysis of AMLE and CE Reveals the Enrichment of Genes Associated With Wound Response and Healing

Affymetrix ID	Entrez ID	Gene Symbol	Gene Title	FC	P Value	Adjusted P Value
8058765	2335	FN1*	Fibronectin 1	184.7	3.6 × 10−11	1.2 × 10−6
8180303	NA	NA	Unannotated	146.1	4.7 × 10−8	1.6 × 10−3
7938758	6288	SAA1*	Serum amyloid A1	65.6	1.8 × 10−8	5.9 × 10−4
8054722	3553	IL1B*	Interleukin 1, beta	33.8	1.8 × 10−7	6.1 × 10−3
7946983	6289	SAA2*	Serum amyloid A2	27.8	1.7 × 10−8	5.7 × 10−4
7963786	3678	ITGA5*	Integrin, alpha 5 (fibronectin receptor, alpha poly peptide)	25.0	6.1 × 10−10	2.0 × 10−5
8056184	3694	ITGB6	Integrin, beta 6	23.6	1.2 × 10−8	3.8 × 10−4
8089082	131566	DCBLD2*	Discoidin, CUB and LCCL domain containing 2	23.3	1.4 × 10−6	4.5 × 10−2
8177222	100133941	CD24	CD24 molecule	21.3	7.6 × 10−7	2.5 × 10−2
8178115	629	CFB	Complement factor B	17.7	8.6 × 10−7	2.9 × 10−2
8179351	629	CFB	Complement factor B	16.9	6.9 × 10−7	2.3 × 10−2
8118345	629	CFB	Complement factor B	16.6	8.2 × 10−7	2.7 × 10−2
8135069	5054	SERPINE1*	Serpin peptidase inhibitor, clade E (nexin, plas minogen activator inhibitor type 1), member 1	15.3	2.5 × 10−7	8.4 × 10−3
8048864	6364	CCL20	Chemokine (C-C motif) ligand 20	13.2	1.7 × 10−7	5.5 × 10−3
8038683	5653	KLK6*	Kallikrein-related peptidase 6	12.8	9.4 × 10−7	3.1 × 10−2
8126784	7941	PLA2G7	Phospholipase A2, group VII (platelet-activating factor acetylhydrolase, plasma)	10.4	1.4 × 10−8	4.7 × 10−4
8078350	7048	TGFBR2*	Transforming growth factor, beta receptor II (70/ 80 kDa)	8.0	2.0 × 10−7	6.7 × 10−3
7905571	6280	S100A9	S100 calcium binding protein A9	7.6	1.4 × 10−6	4.7 × 10−2
7946977	6291/ 100528017	SAA4/SAA2-SAA4/ SAA2-SAA4	Serum amyloid A4, constitutive///SAA2-SAA4  read-through///SAA2-SAA4 read-through	7.0	3.8 × 10−7	1.3 × 10−2
7947425	966	CD59*	CD59 molecule, complement regulatory protein	6.5	7.7 × 10−7	2.6 × 10−2
8091327	5359	PLSCR1*	Phospholipid scramblase 1	6.0	8.4 × 10−8	2.8 × 10−3
8106098	4131	MAP1B	Microtubule-associated protein 1B	5.0	4.6 × 10−7	1.5 × 10−2
8137865	2768	GNA12*	Guanine nucleotide binding protein (G protein) alpha 12	4.9	3.1 × 10−9	1.0 × 10−4
7989023	5873	RAB27A*	RAB27A, member RAS oncogene family	4.2	5.9 × 10−7	2.0 × 10−2
7940028	710	SERPING1*	Serpin peptidase inhibitor, clade G (C1 inhibitor), member 1	4.2	3.4 × 10−7	1.1 × 10−2
7984319	5604	MAP2K1	Mitogen-activated protein kinase kinase 1	3.7	9.8 × 10−7	3.3 × 10−2
7931930	5588	PRKCQ*	Protein kinase C, theta	3.4	6.0 × 10−7	2.0 × 10−2
8109283	3340	NDST1	N-deacetylase/N-sulfotransferase (heparan glucosa minyl) 1	3.2	9.9 × 10−7	3.3 × 10−2
7949503	30008	EFEMP2*	EGF containing fibulin-like extracellular matrix protein 2	3.1	5.5 × 10−8	1.8 × 10−3
8102311	839	CASP6	Caspase 6, apoptosis-related cysteine peptidase	2.7	4.4 × 10−7	1.5 × 10−2
7896707	NA	NA	Unannotated	−4.0	1.3 × 10−6	4.3 × 10−2
7935270	29760	BLNK	B-cell linker	−4.0	2.1 × 10−7	7.0 × 10−3
7896690	NA	NA	Unannotated	−4.1	1.2 × 10−7	4.0 × 10−3
7988414	2628	GATM*	Glycine amidinotransferase (L-arginine:glycine a midinotransferase)	−4.3	2.4 × 10−7	7.9 × 10−3
8145532	2053	EPHX2	Epoxide hydrolase 2, cytoplasmic	−4.3	1.7 × 10−7	5.5 × 10−3
7999364	2903	GRIN2A	Glutamate receptor, ionotropic, N-methyl D-aspar tate 2A	−7.1	7.7 × 10−7	2.6 × 10−2
7938390	133	ADM	Adrenomedullin	−9.74	3.7 × 10−7	1.2 × 10−2
8149927	1191	CLU	Clusterin	−19.38	3.9 × 10−8	1.3 × 10−3

 In total, 38 differentially expressed probe sets were associated with the GO response to wounding and/or wound healing biological processes (30 transcripts were upregulated and eight genes were downregulated in AMLE). The fold change for each probe set along with respective P value and Bonferroni adjusted P value is shown.

*  Genes present in both the response to wounding and wound healing GO biological processes.

Table 2

 

Microarray Analysis of AMLE and CE Reveals the Enrichment of Genes Associated With Regulation of Cell Morphogenesis

Affymetrix ID	Entrez ID	Gene Symbol	Gene Title	FC	P Value	Adjusted P Value
8128956	2534	FYN	FYN oncogene related to SRC,  FGR, YES	5.6	7.48 × 10−7	2.49 × 10−2
8133413	3984	LIMK1	LIM domain kinase 1	2.7	9.4 × 10−7	3.13 × 10−2
8153223	5747///100653146 ///100653024	PTK2/// LOC100653146/// LOC100653024	PTK2 protein tyrosine kinase 2///  uncharacterized LOC100653146///  uncharacterized LOC100653024	2.2	8.21 × 10−7	2.73 × 10−2
8106784	5921	RASA1	RAS p21 protein activator (GTPase  activating protein) 1	5.3	6.84 × 10−7	2.28 × 10−2
7952268	7070	THY1	Thy-1 cell surface antigen	3.5	4.54 × 10−7	1.51 × 10−2
8064904	55612	FERMT1	Fermitin family member 1	16.2	1.23 × 10−6	4.10 × 10−2
8058765	2335	FN1	Fibronectin 1	184.7	3.57 × 10−11	1.19 × 10−6
8137865	2768	GNA12	Guanine nucleotide binding protein  (G protein) alpha 12	4.9	3.06 × 10−9	1.02 × 10−4
8106098	4131	MAP1B	Microtubule-associated protein 1B	5.0	4.59 × 10−7	1.53 × 10−2
8142270	4897	NRCAM	Neuronal cell adhesion molecule	12.2	3.6 × 10−8	1.20 × 10−3
7910387	58480	RHOU	Ras homolog family member U	−14.0	4.1 × 10−7	1.37 × 10−2
8140668	10371	SEMA3A	Sema domain, immunoglobulin do main (Ig), short basic domain, se creted, (semaphorin) 3A	24.6	9.68 × 10−8	3.22 × 10−3
7917649	7049	TGFBR3	Transforming growth factor, beta  receptor III	−4.2	5.42 × 10−7	1.80 × 10−2

 Thirteen differentially expressed probe sets were associated with the GO regulation of cell morphogenesis biological process (11 genes were upregulated and two genes downregulated in AMLE relative to CE comparator). The fold change for each probe set along with respective P value and Bonferroni adjusted P value is shown.

Quantitative RT-PCR Analysis of Known Stem Cell Markers and Validation of Microarray-Detected Expression Changes

Quantitative RT-PCR was used to quantify expression of known limbal epithelial stem cell markers P63α (an important splice variant of TP63 not present on the microarray) and ABCG2, which is known to be highly expressed in culture-expanded limbal epithelium. ^41–43 We found ABCG2 expression to be higher in AMLE than in CE (FC = 8.5, P = 5.25 × 10⁻³). P63α was expressed in both AMLE and CE; however, we observed a relatively small decrease in P63α expression in AMLE (FC = −2.2, P = 4.24 × 10⁻²). 

Quantitative RT-PCR was used to validate differential expression of seven genes identified during microarray analysis (Fig. 3). In general, the qRT-PCR results were in agreement with the direction of expression change between AMLE and CE detected by microarray. Although not found to be differentially expressed on the microarray, the ABCG2 gene (selected for its role as a limbal epithelial stem cell marker) ⁴³ was found by qRT-PCR to be upregulated. The remaining genes analyzed by qRT-PCR were found to correlate with the microarray analysis. For instance, we confirmed AMLE overexpression of wound healing–associated TIMP2 (FC = 16.9, P = 3.09 × 10⁻⁴), FGF5 (FC = 41.5-fold, P = 7.35 × 10⁻³), PDGFC (FC = 20.3, P = 1.01 × 10⁻³), SEMA3A (FC = 28.6, P = 7.55 × 10⁻³), and putative limbal epithelial stem cell markers ITGA5 (FC = 2.8, P = 1.91 × 10⁻²) and S100A9 (FC = 33.6, P = 1.4 × 10⁻²). Amniotic membrane expanded limbal epithelium downregulation of embryonic stem cell/corneal differentiation marker KLF4 (FC = −66.7, P = 2.74 × 10⁻⁵) further validated microarray expression. 

Figure 3

 

Quantitative RT-PCR confirmation of microarray-detected gene expression changes (FGF5, PDGFC, SEMA3A, TIMP2, S100A9, and KLF4) and limbal epithelial stem cell marker expression (P63α and ABCG2). All samples were normalized to endogenous beta-actin gene expression for the AMLE and CE samples. Fold change differences between AMLE and CE were calculated using the 2^−ΔΔCt method. All fold changes shown are statistically significant (P < 0.05). The direction of gene expression change was consistent with that observed by microarray for FGF5, PDGFC, SEMA3A, TIMP2, S100A9, and KLF4.

Discussion

Standard surgical treatments for total limbal stem cell deficiency that replace damaged tissue using dual keratoplasty-keratolimbal allografts yield success rates of ∼50%. ^8–12 Recently developed culture-expanded limbal epithelial graft procedures are associated with improved clinical outcomes ^19,20 and have been shown to be effective for treatment of even the most severe and difficult-to-manage cases of limbal stem cell deficiency that respond poorly to traditional treatment. ^19,52 Yet despite the tremendous advantages afforded by use of AMLE, the reasons for such superior clinical outcomes are unclear. 

A consensus regarding the best method for culturing transplant-suitable limbal epithelium has yet to be reached. ⁵³ Multiple approaches using modified isolation techniques, ^{14,19,39,54,55} media preparations, ^{21,22,56–59} feeder layers, ^60–62 and growth substrates ⁸ continue to be investigated, improving culture outcomes ^56,63 and reducing risk of interspecies pathogen transfer. ^53,58 With such dynamic technical developments and broadly ranging patient pathologies, ^8,53 direct comparison of clinical outcomes can be challenging. ⁶⁴ Regardless of the culture approach used, however, identification of cultivated tissue often relies on morphological inspection, as well as staining for limbal epithelial stem cell markers including ABCG2 and P63. ^65,66 Limbal epithelial stem cells are known to be critical for the maintenance of healthy corneal epithelium in vivo. ⁶ In cultivated limbal epithelium, higher stem cell content (as determined by P63 bright staining) is associated with improved transplant outcomes. ^63,67 In this study, by microarray analysis we found a number of stem cell–associated markers to be differentially expressed. The classic mesenchymal stem cell markers CD90 and CD73 were found to be upregulated in AMLE; a recent study, however, found no correlation between gene expression of these markers and corresponding cell surface proteins in culture-expanded limbal epithelium. ⁶⁸ KLF4, traditionally classified as an embryonic stem cell marker, was shown to be downregulated in AMLE. KLF4 appears to have an opposing function in cornea, however, inducing differentiation and provision of epithelial barrier functions by upregulation of tight junctions, ^69,70 suggesting that KLF4 expression may be indicative of corneal differentiation. Immunofluorescent staining showed AMLE tissues to be rich in limbal epithelial stem cell marker P63α; however, a small decrease was observed at the gene expression level in AMLE compared to CE. This reduced expression may be attributable to the culture process, and one of the challenges of limbal epithelial stem cell culture is to encourage limbal epithelial cell expansion while minimizing stem cell/progenitor loss and differentiation. ⁷¹ Amnion mimics the stromal niche and encourages stem cell retention during culture; however, differentiation does occur, and reduced P63 staining has been shown to correlate with increased distance from limbal explants. ^72,73 Putative limbal epithelial stem cell markers ABCG2, S100A9, and ITGA5 were shown by qRT-PCR to be overexpressed in AMLE. ABCG2 is expressed by stem cell–rich limbal epithelial holoclones and has previously been shown to be overexpressed in cultivated limbal epithelium relative to native corneal epithelium. ^43,71 S100A9 is highly expressed in healthy limbal epithelial crypts and demonstrates reduced expression in acutely inflamed limbi. ⁷⁴ ITGA5 is expressed by limbal epithelial stem cells and associated with increased colony-forming efficiency. ⁷⁵ The upregulation of ABCG2, S100A9, and ITGA5 observed in this study suggests that the expression of these genes in AMLE is indicative of healthy limbal epithelium. 

While GO analysis of microarray differential expression data did not identify enrichment in specific stem cell maintenance or progenitor categories, a number of statistically significant biological process categories indicated that wound response–related gene expression is enriched in AMLE. The GO biological process interrogation revealed response to wounding (P = 8.31 × 10⁻⁷), wound healing (P = 1.06 × 10⁻⁵), and regulation of cell morphogenesis (P = 5.67 × 10⁻⁵) enrichment in the differentially expressed gene list. Of the 38 genes identified in the response to wounding category, 30 genes (including ITGA5, ITGB6, SERPINE1, SERPING1, and FBN1) were found to be upregulated in AMLE. Integrin αvβ6 contributes to laminin production during corneal wound repair, allowing reassembly of basement membranes and mature hemidesmosomes following keratectomy. ⁷⁶ SERPINE1 gene overexpression has been associated with accelerated epithelial wound closure, while increased relative abundance of SERPING1 protein has been observed in exudate of healing wounds compared to nonhealing wounds. ^77,78 FBN1 (fibronectin) displayed the highest upregulation in gene expression in AMLE (FC = 184.7, P = 1.2 × 10⁻⁶) and is well known to be associated with increased wound healing. ⁷⁹ In the cornea, fibronectin is expressed following injury and serves as a matrix for epithelial migration during wound healing. ⁸⁰ Its potent therapeutic effects have long been exploited via autologous fibronectin-containing eye drops for treatment of nonhealing corneal epithelial defects; more recently, synthetic fibronectin peptide sequence (PHSRN) has demonstrated migration-stimulatory effects and considerable therapeutic benefit for patients suffering from persistent corneal epithelial defects. ^81,82 Expression of genes such as these, which affect the rate of wound repair, may be important, as delays to the wound healing process often result in poorer long-term outcomes including neovascularization, chronic inflammation, and nonhealing wounds. ^83,84 Eight wound-associated genes were downregulated in AMLE. Unexpectedly, CLU, a gene associated with corneal epithelial cell growth promotion and increased colony-forming efficiency, ⁸⁵ was found to be downregulated; however, few corneal-specific associations are known for other significantly downregulated genes (ADM, GRIN2A, EPHX2, GATM, BLNK). 

Of the 13 differentially expressed genes in the regulation of cell morphogenesis category, 11 genes were found to be upregulated in AMLE. Morphogenesis occurs during development, altering cell shape and spatial arrangement for production of functional organs and tissues. ⁸⁶ Many of these same cellular processes and pathways are exploited during wound healing for regeneration and reconstruction of healthy tissue architecture. ^87,88 In our study, we noted several regulation of cell morphogenesis genes that affect neuronal and oligodendrocyte morphology in particular. The cornea is one of the most highly innervated tissues in the body, and communication between epithelial and neuronal cells has been shown previously to mediate corneal wound response. ⁸⁸ Reduced corneal sensitivity and improper nerve regulation are associated with poor corneal epithelial wound healing in patients suffering from diabetes. ⁸⁹ Corneal denervation has also been shown to reduce the extent and to increase the duration of wound closure and is directly associated with severe loss of corneal-limbal stem cells. ⁹⁰ Several groups have demonstrated the powerful restorative effects of topical neurotrophic factors, including nerve growth factor for treatment of corneal neurotrophic defects, ^91,92 and neuroprotectin D1 for stimulation of neuroregeneration following lamellar keratectomy. ⁹³ In this study, we validated AMLE overexpression of SEMA3A by qRT-PCR. SEMA3A functions as a chemoattractant for leading apical dendrites ⁹⁴ and is highly expressed during pericorneal nerve ring formation. ⁵¹ It is also secreted in newly healed rat corneal basal epithelium, further supporting a role for such secreted neuronal-associated proteins in corneal wound repair. ⁹⁵  

Pathway interrogation also identified enrichment of a number of pathways that support a wound healing genotype in AMLE tissues, including MAPK signaling pathway, regulation of actin cytoskeleton, and integrin signaling pathway. Mitogen-activated protein kinases regulate a broad spectrum of wound-relevant processes including migration, proliferation, differentiation, and apoptosis. ^96–99 Transmembrane integrins provide traction and facilitate directional migration at the wound edge by connecting intracellular actin filaments with extracellular matrix proteins. ^100,101 Integrins have also been implicated in stem cell homeostatic functions due to their governance of growth factor tyrosine kinase receptor availability and high density in basal transit-amplifying and limbal epithelial stem cells. ^102,103  

The term “wound healing” describes a complex, highly coordinated overlapping series of inflammatory, migratory, proliferative, and remodeling events that mediate tissue repair. ^104–106 Local and recruited stem cells are thought to play a role in this process by secretion of proliferation-stimulatory cytokines and provision of cells with high proliferative capacity. ^{2,105,107–109} This close relationship between wound healing and stem cells is seen in the cornea also, where immediately following injury, limbal epithelial stem cells in the limbus are stimulated and undergo a rapid 8- to 9-fold rise in proliferative activity. ^67,110 It is thought that limbal epithelial stem cell division may be asymmetric, resulting in two types of daughter cells—one that maintains stem cell properties identical to those of the parent and another transit-amplifying type that migrates from the limbus, undergoing limited cell division and differentiation until wound closure is complete. ^4,111 Patients suffering from limbal stem cell deficiency have inadequate limbal stem cell populations from which to mount an injury response, ⁶ and it has been shown that transplantation of ex vivo expanded limbal epithelium can restore the corneal epithelial surface. ^{8,14,15,19–21,52,63,112–114} If restoration is to be permanent, however, limbal epithelial stem cell populations in the limbus must also be restored. ^1,8  

Studies have shown that cells that repopulate the corneal surface following donor keratolimbal allograft surgery can display genetic profiles of both the donor and recipients, ^115,116 while cells that repopulate recipients' limbi following transplantation of cultivated limbal epithelium appear to be autologous. ²⁰ Several research groups have demonstrated the beneficial effects of corneal epithelial secreted factors including basic fibroblast growth factor, epidermal growth factor, tumor necrosis factor alpha, and interleukin-1 on corneal wound healing. ^67,117,118 These results suggest that secreted factors produced by AMLE may too be able to affect corneal wound healing, perhaps by encouraging self-renewal of recipients' residual limbal epithelial stem cells. In this study we found that mRNAs for the PDGFC and FGF5 secreted proteins, which are associated with enhanced proliferation and dermal wound healing, ^119–122 were upregulated in AMLE tissue. PDGFC is also highly expressed in developing fetal mouse ocular tissue ¹²³ while FGF5 is expressed during neural development, acting to delay differentiation of transit-amplifying cells and balancing the rate at which progenitor cells are renewed and differentiation occurs. ^124,125 Such opposing roles may reflect stage-specific activity observed in the wider FGF family, with FGFs having been shown to affect self-renewal of stem cells but differentiation in more committed progenitors. ^126,127 The overexpression of PDGFC and FGF5 in AMLE tissues suggests that these secreted factors may also have the capacity to affect wound healing and/or stem cell processes in the cornea. 

In conclusion, our study has, for the first time, investigated variations in the global transcriptomes of AMLE and CE tissue. We identified upregulation of a number of putative limbal epithelial stem cell markers in agreement with other reports in the literature; however, the utility of this study is exemplified by the enrichment and overexpression of wound healing genes in AMLE. Upregulation of these genes may contribute to the efficacy associated with transplantation of cultivated limbal epithelium, improving postsurgical response and ultimately longer-term outcomes. 

Supplementary Materials

docx S1 

docx S2 

tif S3 

xlsx S4 

Acknowledgments

The authors thank William Murphy, MD, FRCPEdin, FRCPath, Irish Blood Transfusion Service for advice and provision of clinical material. We also thank Carmel Walsh, RGN, Mater Misericordiae Hospital, and Ann Prunty, RGN, Royal Victoria Eye & Ear Hospital, Dublin, Ireland, for facilitating and coordinating primary tissue sample collection. 

Supported by the Royal Victoria Eye and Ear Hospital, Dublin Research Foundation/National Council for the Blind of Ireland, Health Research Board Partnership Award (PA/2007/08), 3U Biomedical Research (DCU-NUI Maynooth-RCSI), and the Pharmacia-Upjohn Irish College of Ophthalmologists Fellowship award, 2001. 

Disclosure: C. Gallagher, None; C. Clarke, None; S.T. Aherne, None; K.R. Katikireddy, None; P. Doolan, None; V. Lynch, None; S. Shaw, None; A. Bobart-Hone, None; C. Murphy, None; M. Clynes, None; W. Power, None; F. O'Sullivan, None 

Footnotes

 CG, CC, STA, WP, and FO contributed equally to the work presented here and should therefore be regarded as equivalent authors.