Transcriptome Analysis of the Human Corneal Endothelium

Abstract

Purpose.

To comprehensively characterize human corneal endothelial cell (HCEnC) gene expression and age-dependent differential gene expression and to identify expressed genes mapped to chromosomal loci associated with the corneal endothelial dystrophies posterior polymorphous corneal dystrophy (PPCD)1, Fuchs endothelial corneal dystrophy (FECD)4, and X-linked endothelial dystrophy (XECD).

Methods.

Total RNA was isolated from ex vivo corneal endothelium obtained from six pediatric and five adult donor corneas. Complementary DNA was hybridized to the Affymetrix GeneChip 1.1ST array. Data analysis was performed using Partek Genomics Suite software, and differentially expressed genes were validated by digital molecular barcoding technology.

Results.

Transcripts corresponding to 12,596 genes were identified in HCEnC. Nine genes displayed the most significant differential expression between pediatric and adult HCEnC: CAPN6, HIST1H3A, HIST1H4E, and HSPA2 were expressed at higher levels in pediatric HCEnC, while ITGBL1, NALCN, PREX2, TAC1, and TMOD1 were expressed at higher levels in adult HCEnC. Analysis of the PPCD1, FECD4 and XECD loci demonstrated transcription of 53/95 protein-coding genes in the PPCD1 locus, 27/40 in the FECD4 locus, and 35/68 in the XECD locus.

Conclusions.

An analysis of the HCEnC transcriptome reveals the expression of almost 13,000 genes, with less than 1% mapped to chromosomal loci associated with PPCD1, FECD4, and XECD. At least nine genes demonstrated significant differential expression between pediatric and adult HCEnC, defining specific functional properties distinct to each age group. These data will serve as a resource for vision scientists investigating HCEnC gene expression and can be used to focus the search for the genetic basis of the corneal endothelial dystrophies for which the genetic basis remains unknown.

Transcriptomic analysis of human corneal endothelial cell gene expression identified age-dependent differential expression and candidate genes for the corneal endothelial dystrophies PPCD1, FECD4 and XECD.

Introduction

Corneal endothelial cells maintain corneal clarity by transcribing genes that are important in sustaining homeostatic levels of hydration and solute concentration in the corneal stroma. Human corneal endothelial cells (HCEnC) also express genes that encode extracellular structural proteins that give rise to the Descemet membrane and cell-adhesion molecules that support the semipermeable simple squamous epithelial morphology of the endothelium. In addition, HCEnC express genes that code for proteins involved in solute transport across the semipermeable endothelium, integral to maintaining corneal stromal solute concentrations.

Mutations in genes that encode extracellular matrix proteins, enzymes, and transcription factors have been associated with a number of inherited disorders of the corneal endothelium, collectively known as the corneal endothelial dystrophies. Posterior polymorphous corneal dystrophy (PPCD) mapped to chromosome 10 (PPCD3) has been associated with nonsense and frameshift mutations in the transcription factor ZEB1.^1–8 Congenital hereditary endothelial dystrophy (CHED) has been associated with mutations in the sodium-borate cotransporter SLC4A11,^9,10 while sequence variants in ZEB1,^11,12 SLC4A11,^13,14 COL8A2,¹⁵ TCF4,¹⁶ LOXHD1,¹⁷ and AGBL1¹⁸ have been associated with Fuchs endothelial corneal dystrophy (FECD). However, the genetic basis of at least three corneal endothelial dystrophies remains to be determined: PPCD mapped to chromosome 20 (PPCD1),^19–22 FECD mapped to chromosome 9 (FECD4),¹² and X-linked endothelial dystrophy (XECD).²³ Therefore, identification of the genes expressed in the human corneal endothelium that are mapped to these chromosomal loci would provide a list of positional gene candidates that would narrow the search for the genetic basis of these corneal endothelial dystrophies.

A complete characterization of the HCEnC transcriptome has been a goal of vision science researchers for the last decade, who have used serial analysis of gene expression (SAGE),^24,25 gene arrays,^25–27 and next-generation sequencing (RNA-seq) technologies.^28–30 We present a new approach and build on the previously published HCEnC transcriptome studies. Defining the normal and age-dependent HCEnC transcriptome will be useful for elucidation of the genetic basis of inherited disorders of the corneal endothelium and for facilitating the development of novel strategies for managing endothelial dysfunction, such as validating the transcriptome of cultured corneal endothelial cells prior to transplantation.

Methods

Ribonucleic Acid Isolation and Purification

Eleven corneas from six pediatric (4, 6, 10, 11, 17, and 18 years old) and five adult (53, 56, 57, 64, and 70 years old) donors were obtained from various eye banks affiliated with the Vision Share consortium of eye banks (Vision Share, Apex, NC, USA) with the following characteristics: mean central corneal endothelial cell density of 2973 cells/mm² (range, 2114–3937 cells/mm²); mean death to preservation time of 6 hours 57 minutes (range, 3 hours 5 minutes–11 hours 49 minutes); and mean death to removal of Descemet membrane and corneal endothelium time of 2.9 days (range, 2–4 days) (Table 1). Total RNA was isolated from the corneal endothelium using TriReagent (Life Technologies, Grand Island, NY, USA) and subsequently purified with the RNeasy Clean-Up Kit (Qiagen, Valencia, CA, USA). The integrity of the isolated RNA was analyzed using the Agilent 2100 Electrophoresis Bioanalyzer System (Agilent Technologies, Inc., Santa Clara, CA, USA) and found to be of sufficient quantity (approximately 1 μg) and quality (RNA integrity number in the range of 8.3–9.0) for analysis using the Affymetrix gene chip arrays.

Table 1.

Donor Tissue Information

Age	Age Group	Cause of Death	Endothelial Cells/mm2	Storage Medium	Reason Not Suitable for Transplantation?	Death to Cooling	Death to Preservation	Death to Harvest
4	Preschooler	Brain death	3937	Life4C	PA in respiratory cx	1 h 49 min	8 h 34 min	4 d
6	Preschooler	Drowning	3067, 3390	Life4C	PA in respiratory cx	NA	3 h 5 min	4 d
10	Preadolescent	Seizure	2924, 2725	Life4C	PA in respiratory cx	NA	3 h 59 min	2 d
11	Preadolescent	Probable OD	2923, 3115	Optisol GS	SA-positive sputum	3 h 35 min	8 h 12 min	3 d
15	Adolescent	Cardiac arrhythmia	3096	Optisol GS	Contact lens attached	2 h 48 min	11 h 13 min	2 d
17	Adolescent	Severe head injury	3184, 3184	Optisol GS	Submersion under water	3 h 27 min	2 h 21 min	3 d
53	Adult	CHF	2959, 2793	Life4C	Possible IV drug use	NA	5 h 15 min	3 d
56	Adult	SDH	2923	Optisol GS	Cocci (staph?) in blood	1 h 11 min	4 h 26 min	2 d
57	Adult	Liver cancer	2114, 2183	Optisol GS	Paracentral tear in DM	2 h 20 min	8 h 10 min	2 d
64	Adult	Probable MI	2793, 2762	Optisol GS	Possible HRB*	3 h 57 min	11 h 49 min	3 d
70	Adult	MOSF	3215, 3205	Optisol GS	Possible hepatitis A†	6 h 28 min	9 h 29 min	4 d

OD, overdose; CHF, congestive heart failure; SDH, subdural hematoma; MI, myocardial infarction; MOSF, multiple organ system failure; PA, Pseudomonas aeruginosa; cx, microbiological culture; SA, Staphlococcus aureus; HRB, high-risk behavior.

^* Next of kin were unable to complete high-risk behavior questionnaire relating to the deceased.

^† Next of kin indicated resolved hepatitis A infection, but were unable to confirm infection status.

Transcriptome Analysis

Samples of total RNA from each donor cornea were processed and analyzed separately (i.e., samples were not pooled). Total RNA from pediatric and adult normal endothelium was hybridized to the Affymetrix GeneChip 1.1ST array (Affymetrix, Inc., Santa Clara, CA, USA), and raw intensity values stored in CEL files were imported into Partek Genomics Suite software (PGS) (Partek, Inc., St. Louis, MO, USA). Raw values were normalized using the Robust Multi-array Average (RMA) method with background correction, and the adjusted intensity values were output as log2 transformed values.^31,32 An additional background threshold was calculated by averaging the log2 values for genes predicted to have no transcription in HCEnC. Subsequently, transcripts corresponding to normalized probe intensity values at or below the background value of 3.8 were considered as not expressed. Principal component analysis (PCA) and hierarchical clustering (HC) were performed to identify variation and relationships among samples. Principal component analysis and HC analysis revealed that the pediatric samples clustered into three distinct groups that are hereafter referred to as preschooler (4 and 6 years old), preadolescent (10 and 11 years old), and adolescent (17 and 18 years old). The CEL files containing the raw signal intensity values and the RMA normalized and log2 transformed signal intensity values are available from the GEO DataSets database (accession number GSE58315; National Center for Biotechnology Information [NCBI], Bethesda, MD, USA).

Differential Gene Expression Analysis

Differential gene expression was determined for all genes using PGS. Three separate statistical (P value) and fold-change (FC) settings were used to obtain lists of significant differentially expressed genes between pediatric and adult endothelium. In order to correct for multiple comparisons and reduce the number of false positives, a false discovery rate (FDR) adjusted P value of 0.05 was used in combination with a FC > 3. To obtain a larger number of genes for validation by digital molecular barcoding technology (NanoString Technologies, Seattle, WA, USA), a non-FDR P value of 0.005 and FC > 2.9 were applied. The third settings were the least stringent, with a non-FDR P value of 0.05 and FC > 2, and were used to generate a list of differentially expressed genes for gene ontology (GO) analysis.

Digital Molecular Barcoding Analysis

A custom multiplexed NanoString nCounter Elements expression assay (NanoString Technologies) comprising the nine differentially expressed genes was assembled. All reagents and software were from NanoString Technologies, unless stated otherwise. The procedure was performed according to the NanoString nCounter Single Cell Gene Expression and Elements Gene Expression Assay Protocols. Briefly, 1 ng total RNA was used for the amplification (10 cycles) of nine separate transcripts using Multiplexed Target Enrichment oligonucleotides (Integrated DNA Technologies, San Diego, CA, USA). Subsequently, 5 μL amplification reaction was mixed with 20 and 5 μL nCounter reporter probes and biotinylated nCounter capture probes (Integrated DNA Technologies), respectively. Hybridization was performed at 65°C for approximately 16 to 20 hours. Excess probe was washed away using a two-step magnetic bead–based purification technique on the nCounter Prep Station. Biotinylated capture probe–bound samples were immobilized and recovered on a streptavidin-coated cartridge using the nCounter PrepStation. To determine transcript species (i.e., barcode pattern) and abundance (i.e., fluorescence intensity), a max-density scan (encompassing 555 fields of view) was performed using the nCounter Digital Analyzer. Data analysis was performed using the nSolver and GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA, USA).

Gene Ontology Analysis

Gene ontology analysis was performed for the identification of significantly enriched functional groups using the GO Enrichment option available in the PGS software (Partek, Inc.). A list of differentially expressed genes between pediatric and adult corneal endothelium was generated utilizing a non-FDR P value of 0.05 and FC cutoff of 2. Significantly enriched GO functional groups were defined as having an enrichment score equal to or greater than 3 (P value > 0.05) and were required to be composed of at least three genes.

Corneal Endothelial Dystrophy Disease Locus Analysis

The genes located within the PPCD1 locus (bordered by the D20S182 and D20S106 genetic markers), the FECD4 locus (bordered by the D9S1681 and D9S1684 genetic markers), and the XECD locus (bordered by the DXS8057 and DXS1047 genetic markers) were identified using the most recent build of the human genome (Annotation 105) and the Map Viewer resource available through the NCBI website (http://www.ncbi.nlm.nih.gov/mapview/ [in the public domain]). These lists of genes were then compared to the list of expressed genes corresponding to the PPCD1, FECD4, and XECD loci obtained using the PGS software to identify positional candidate genes for each of these corneal endothelial dystrophies.

Statistical Analysis

The two-tailed unpaired t-test was utilized for identifying a significant difference in the means of the expression levels (barcode counts) of transcripts in pediatric versus adult corneal endothelium as determined by digital molecular barcoding technology. Differences in the means of gene expression in the different age groups of ex vivo endothelium were considered significant (P ≤ 0.05) by rejection of the null hypothesis. One-way analysis of variance (ANOVA) was used to identify significantly differentially expressed genes in the microarray data of the pediatric versus adult samples.

Results

Multidimensional Transcriptome Analysis

Principal component analysis and HC showed distinct relationships between the pediatric and adult samples (Fig. 1). An initial analysis of all of the samples demonstrated that the six pediatric samples segregated into three distinct groups. We observed that the groupings were not random and that they depended on the age of the donor. As such, the six pediatric samples were categorized into three distinct age groups, which were defined as preschooler (4 and 6 years old), preadolescent (10 and 11 years old), and adolescent (17 and 18 years old). The two adolescent samples clustered near the adult samples, suggesting that the transcriptomes of the two adolescent samples were more similar to the adult than to the other two pediatric groups. Therefore, they were subsequently excluded from the pediatric group for the additional analyses (Figs. 1A, 1B). An analysis of the remaining samples demonstrated that the first three principal components (PC1: 23.7%; PC2: 18.3%; PC3: 14.4%) accounted for 56.4% of the variance observed between the gene expression data sets (Fig. 1C). Principal component analysis demonstrated three distinct groups that correlated with age. Hierarchical clustering recapitulated the results obtained with PCA, with gene expression correlated to age (Fig. 1D). Differential gene expression analysis was performed on two HCEnC groups: pediatric (preschooler and preadolescent) and adult.

Figure 1.

Principal component analysis and hierarchical clustering of ex vivo HCEnC gene expression data sets. (A) PCA demonstrated three distinct groupings, with the adolescent (17 and 18 years, red spheres) HCEnC group having a closer relationship to adult samples (53–70 years, blue spheres) compared to preadolescents (10 and 11 years, green spheres) and preschoolers (4 and 6 years, violet spheres). (B) Hierarchical clustering confirmed the relationships observed with PCA, with the adolescent HCEnC samples (red bars) demonstrating a closer relationship to adult HCEnC samples (blue bars). (C) After removal of the HCEnC samples from the adolescent group, PCA demonstrated three distinct groupings, two of which were analyzed together as the pediatric group (preschooler and preadolescent, 4–11 years, blue spheres) and the adult group (53–57 years, red spheres). Subsequent data analysis was performed on two groups defined by the colored ovals (pediatric and adult). (D) Hierarchical clustering confirmed the relationships observed in the PCA analysis, with the preschooler (violet bar) and the preadolescent (green bar) HCEnC samples having clustered separately from the adult (blue bar) HCEnC.

Whole-Genome Gene Expression

The pediatric and adult samples were combined and analyzed for whole-genome gene transcription. Annotation with the RefSeq Transcripts database (NCBI) demonstrated 20,253 unique gene symbols, of which 12,596 were considered expressed in the corneal endothelium with an expression value above the background threshold of 3.8 (Table 2, top 100 unique probeset IDs comprising 83 unique genes). Four of the six genes that have been associated with the endothelial corneal dystrophies (COL8A2, SLC4A11, ZEB1, and TCF4) were expressed in the corneal endothelium, while two (LOXHD1 and AGBL1) were not (Table 3).

Table 2.

Most Highly Expressed Genes in Normal Human Corneal Endothelial Cells*

Probeset ID	Gene Symbol	Gene Type	Mean Intensity
7917645	RN28S1	rRNA	12.35
8165665	COX2	Protein coding	12.34
8165690	CYTB	Protein coding	12.23
7978905	SDCCAG1	Protein coding	12.22
8165661	COX1	Protein coding	12.21
8165686	ND5	Protein coding	12.05
8170360	FTH1	Protein coding	11.94
8165653	ND1	Protein coding	11.89
8165669	COX3	Protein coding	11.87
8165648	C7orf11	Protein coding	11.84
7953385	GAPDH	Protein coding	11.81
7938777	LDHA	Protein coding	11.78
7912198	ENO1	Protein coding	11.70
8165676	ND4	Protein coding	11.69
7941272	MALAT1	ncRNA	11.68
7942791	RN28S1	rRNA	11.59
7986765	RPL5	Protein coding	11.58
8165674	SH3KBP1	Protein coding	11.54
7954310	RN18S1	rRNA	11.50
8158952	EEF1A1	Protein coding	11.37
7967563	UBC	Protein coding	11.35
8050548	LAPTM4A	Protein coding	11.35
8138531	EEF1A1	Protein coding	11.34
7981978	SNORD116-15	snoRNA	11.28
8109821	RPL10	Protein coding	11.28
7952325	HSPA8	Protein coding	11.20
7965467	RPL41	Protein coding	11.11
7982129	RPL41	Protein coding	11.11
7961514	MGP	Protein coding	11.09
8061136	PTMA	Protein coding	11.08
7922301	MYOC	Protein coding	11.08
8137008	C7orf11	Protein coding	11.08
8076511	RPL5	Protein coding	10.99
8165672	RFC1	Protein coding	10.92
8116520	GNB2L1	Protein coding	10.91
8102587	C4orf31	Protein coding	10.90
8154305	SELT	Protein coding	10.88
7915472	SLC2A1	Protein coding	10.87
8075691	RPL41	Protein coding	10.80
8107470	PTMA	Protein coding	10.78
8069644	APP	Protein coding	10.77
8173941	TSPAN6	Protein coding	10.77
8025395	RPS28	Protein coding	10.76
7904254	ATP1A1	Protein coding	10.74
8165658	ND2	Protein coding	10.74
7942824	RPS28	Protein coding	10.73
8092970	APOD	Protein coding	10.73
8005471	RPS28	Protein coding	10.73
8066262	SNORA71D	snoRNA	10.71
7942586	RPS3	Protein coding	10.70
7921916	RGS5	Protein coding	10.67
8178435	IER3	Protein coding	10.65
7989501	CA12	Protein coding	10.64
7981976	SNORD116-14	snoRNA	10.63
8113120	TOB2	Protein coding	10.62
8165707	TOB2	Protein coding	10.62
8109750	RPLP0	Protein coding	10.62
8064613	SLC4A11	Protein coding	10.60
8120249	RN7SK	snRNA	10.57
8102831	C4orf49	Protein coding	10.56
7977775	DAD1	Protein coding	10.52
8159642	TUBB2C	Protein coding	10.51
8081431	ALCAM	Protein coding	10.51
8164165	HSPA5	Protein coding	10.50
8117377	HIST1H1E	Protein coding	10.49
8034512	SNORD41	snoRNA	10.48
8061364	RPL41	Protein coding	10.48
8105432	RPL41	Protein coding	10.48
8042335	VDAC2	Protein coding	10.47
7981988	SNORD116-20	snoRNA	10.46
8067955	CXADR	Protein coding	10.44
8110522	CANX	Protein coding	10.41
8086752	SNORD13	snoRNA	10.39
8038086	RPL18	Protein coding	10.36
8142468	TPM3	Protein coding	10.35
7902398	SNORD45A	snoRNA	10.35
7981982	SNORD116-17	snoRNA	10.34
7981986	SNORD116-17	snoRNA	10.34
7929816	SCD	Protein coding	10.33
8000217	SMG1	Protein coding	10.32
7978644	NFKBIA	Protein coding	10.31
7958130	HSP90B1	Protein coding	10.31
7919193	NUDT4P1	Pseudogene	10.30
8035829	RPL34	Protein coding	10.30
8179704	IER3	Protein coding	10.30
8124848	IER3	Protein coding	10.30
8117995	TUBB	Protein coding	10.30
8177858	TUBB	Protein coding	10.30
8179174	TUBB	Protein coding	10.30
8146216	VDAC3	Protein coding	10.30
8022814	HNRNPA1	Protein coding	10.26
7938329	SNORA23	snoRNA	10.26
7899480	SNORA73A	snoRNA	10.26
8109222	RPL7	Protein coding	10.25
8119993	HSP90AB1	Protein coding	10.22
8116929	RPL15	Protein coding	10.22
8030980	ZNF525	Protein coding	10.22
8012110	GABARAP	Protein coding	10.21
7982597	THBS1	Protein coding	10.21
8005547	SNORD3A	snoRNA	10.21

rRNA, ribosomal RNA; ncRNA, non-coding RNA; snoRNA, small nucleolar RNA; snRNA, small nuclear RNA.

^* Top 100 transcript IDs by level of expression comprising 83 genes.

Table 3.

Expression in Normal Corneal Endothelium of Genes Associated With Corneal Endothelial Dystrophies

Probeset ID	Gene Symbol	Cytoband	Corneal Dystrophy	Mean Intensity*
8064613	SLC4A11	20p13	CHED, FECD	10.60
8023415	TCF4	18q21.2	FECD	9.44
7914880	COL8A2	1p34.3	FECD	7.28
7926916	ZEB1	10p11.22	PPCD3, FECD	6.13
8023080	LOXHD1	18q21.1	FECD	3.41†
7985741	AGBL1	15q25.3	FECD	3.13†

^* Mean intensity is given in log2 values.

^† Intensity values are below the background threshold of 3.8.

Identification of Differentially Expressed Genes

Of the 12,596 genes that were identified as expressed by HCEnC, a total of 60 genes (Fig. 2) demonstrated significant differential expression (non-FDR P value < 0.05, >2 fold change) between pediatric and adult corneal endothelium. A total of 30 genes had higher transcript abundance in pediatric endothelium, and 30 genes had higher transcript abundance in adult endothelium (Supplementary Table S1). After more stringent filtering criteria were applied, 9 genes demonstrated highly significant differential expression (non-FDR P value < 0.005 and >2.9 fold change) between pediatric and adult HCEnC (Table 4). Only HIST1H3A, which was expressed at higher levels in the pediatric HCEnC, survived the most stringent criteria (FDR P value < 0.05 and >3 fold change). All 9 genes demonstrated a fold change greater than 2.5 by digital molecular barcoding analysis, while statistical analysis confirmed the significant differential expression of 7 of the 9 genes (Fig. 3). HIST1H4E and HSPA2 did not show statistical significance, but did demonstrate a trend toward significance with P values of 0.08 and 0.06, respectively.

Figure 2.

Hierarchical clustering analysis was performed for genes that demonstrated significant (non-FDR P value = 0.05, FC > 2) differential expression between pediatric (preschooler and preadolescent) and adult endothelium.

Table 4.

Highly Differentially Expressed Genes Between Pediatric and Adult HCEnC

Probeset ID	Gene Symbol	Fold Change*	P Value*	Cytoband
Upregulated in pediatric HCEnC
8117330	HIST1H3A	5.19082	4.0E-06	6p22.1
8117402	HIST1H4E	3.90932	2.1E-04	6p22.1
7975076	HSPA2	3.11007	2.5E-03	14q24.1
8174527	CAPN6	2.92365	3.4E-03	Xq23
Upregulated in adult HCEnC
8134420	TAC1	7.10805	4.3E-03	7q21-q22
8146794	PREX2	3.89636	2.9E-04	8q13.2
7969861	ITGBL1	3.61241	7.3E-05	13q33
8156706	TMOD1	2.98584	2.5E-03	9q22.3
7972601	NALCN	2.95067	4.9E-04	13q32.3

^* Differential gene expression between pediatric and adult HCEnC that showed a high probability of rejecting the null hypothesis (P ≤ 0.005), a fold change greater than or equal to 2.9, and an absolute expression intensity value greater than background intensity (3.8) was considered significant.

Figure 3.

Differentially expressed genes were validated by digital molecular barcoding technology. The fold change (FC) for each gene was greater than 2. Transcript levels for the TAC1, PREX2, ITGBL1, TMOD1, and NALCN genes were higher in adult samples (black bars), while transcript levels for the HIST1H3A, HISTH4E, HSPA2, and CAPN6 genes were higher in pediatric samples (white bars). (*P < 0.05, ***P < 0.001).

Gene Ontology Analysis

Gene transcripts that showed significant differential expression between the pediatric and adult group were analyzed for functional group enrichment. The 30 genes with significantly higher transcript levels in pediatric samples showed significant enrichment for six functional groups, of which four (extracellular matrix, proteinaceous extracellular matrix, calcium ion binding, and blood coagulation) were unique to pediatric endothelium (Table 5). The 30 genes that showed significantly higher expression in adult samples demonstrated significant enrichment for seven functional groups, of which five (heparin binding, axon, positive regulation of apoptosis, intracellular signal transduction, and signal transduction) were unique to adult endothelium (Table 5).

Table 5.

Gene Ontology in Pediatric and Adult Corneal Endothelium

Function	GO ID	Type	Enrichment Score	Genes
GO functional groups enriched in pediatric HCEnC
Extracellular matrix	31012	Cellular component	6.80	CRISPLD2,FBN1,IGFBP7
Proteinaceous extracellular matrix	5578	Cellular component	5.85	FBN1,SPOCK2,WNT5A
Extracellular region	5576	Cellular component	4.51	CRISPLD2, FBN1, HIST1H3A, IGFBP7, SEMA3E, SPOCK2, WNT5A
Calcium ion binding	5509	Molecular function	4.44	FBN1,GCH1,S100A10,SPOCK2
Extracellular space	5615	Cellular component	3.92	FBN1, IGFBP7, SEMA3E, WNT5A
Blood coagulation	7596	Biological process	3.72	CD177,HIST1H3A,PRKCH
GO functional groups enriched in adult HCEnC
Extracellular region	5576	Cellular component	9.80	CCK, F13A1, GDF15, HBA1, IGFBP5, ITGBL1, MAP2K6, NXPH4, PLA2G5, SERPINA5, TAC1
Heparin binding	8201	Molecular function	7.01	ABI3BP,PLA2G5,SERPINA5
Axon	30424	Cellular component	6.32	DST,EPHB1,TAC1
Positive regulation of apoptosis	43065	Biological process	6.29	ALDH1A2,CCK,MAP2K6
Extracellular space	5615	Cellular component	5.13	CCK, CPB2, GDF15, SERPINA5, TAC1
Intracellular signal transduction	35556	Biological process	4.53	IGFBP5,PPP1R1B,PREX2
Signal transduction	7165	Biological process	3.45	CCK,GDF15,IGFBP5,MAP2K6,PP1R1B

Bold text represents functional groups that were unique to the respective age group.

Identification of Genes Expressed in Corneal Endothelial Dystrophy Loci

Posterior Polymorphous Corneal Dystrophy-1.

The PPCD1 locus as described by Yellore et al.²¹ in a large North American Caucasian pedigree is composed of 204 genes, 94 of which are protein coding and 110 of which are noncoding (pseudogenes, micro RNAs, lincRNAs, and so on). All of the protein-coding genes and 26% (29/110) of the pseudogenes and noncoding RNAs are represented by oligonucleotide probes on the Affymetrix GeneChip 1. array. Fifty-six percent (53/94) of the protein-coding genes and 34% (10/29) of the represented pseudogenes and noncoding RNAs demonstrated signal intensity greater than 3.8 (Supplementary Table S2).

Fuchs Endothelial Corneal Dystrophy-4.

The FECD4 locus is composed of 99 genes, 41 of which are protein coding and 58 of which are noncoding. Of the protein-coding genes, 98% (40/41) were represented on the array, while 19% (11/58) of the noncoding genes had probes on the 1. array. Forty percent (16/40) of the represented protein-coding genes and 55% (6/11) of the represented pseudogenes and noncoding RNAs demonstrated signal intensity greater than background (Supplementary Table S2).

X-Linked Endothelial Corneal Dystrophy.

The XECD locus is composed of 181 genes, of which 68 are protein coding and 113 are putative or noncoding. All of the protein-coding genes and 20% (23/113) of the putative or noncoding genes are represented by oligonucleotide probes on the 1. array. Fifty-one percent (35/68) of the protein-coding genes and 35% (8/23) of the represented putative or noncoding genes demonstrated signal intensity greater than background (Supplementary Table S2).

Discussion

Transcriptome analysis of the corneal endothelium has been previously reported in in vitro²⁷ and ex vivo^24–26 HCEnC. Recent reports demonstrate the utility of next-generation sequencing (NGS) technology for transcriptome analysis of human corneal cells.^28–30 However, many of these studies were performed using a low number of samples or samples pooled from multiple donor corneas, which limited the ability to perform robust statistical analyses. In addition, obtaining tissue from different sources (academic pathology department and eye bank in one study³⁰) may introduce a batch effect from the different tissue handling and storage conditions.

To address the technical variables that may have negatively impacted the results of prior studies and to build upon their findings, we utilized an experimental design that controlled for several technical factors, including tissue handling and processing, and used a larger sample size to enable robust statistical analysis. First, we obtained both pediatric and adult corneas from commercial eye banks, which use a standardized protocol for tissue recovery and storage. Second, we obtained high-quality RNA from HCEnC of one or both corneas from a single donor by using a recently developed technique for removing the corneal endothelium and Descemet membrane (DM) from the donor cornea, thus eliminating the need to create pooled samples from corneas obtained from multiple individuals.³³ Third, we used the Affymetrix GeneChip Human Gene 1.1 ST array, which provides comprehensive whole-transcript coverage of well-annotated protein-coding genes. We do, however, acknowledge the limitations of array-based transcriptome analysis, including that the detection of genes is dependent on the design of the array and that only genes known at the time of the array design are detectable, preventing the identification of novel genes. However, as gene annotation is essentially complete, the inability to identify novel genes is not expected to significantly impact the ability of microarray analysis to accurately characterize the corneal endothelial transcriptome.^34,35

After accounting for several technical variables, we performed a whole-genome gene transcription analysis and identified a total of 12,596 expressed genes. This represents approximately 62% (12,596/20,253) of the primarily protein-coding genes in the human genome, which is consistent with a recent report showing that 60% to 70% of protein-coding genes are expressed in human and mouse tissues.³⁶ These data were then analyzed for age-associated differential gene expression between pediatric and adult HCEnC to reveal genes that play temporal but central roles in HCEnC biology.

The eye, and in particular the cornea, undergoes substantial changes during a person's lifetime. The initial postnatal reduction in corneal endothelial cell density, a result of the increase in the size of the cornea coupled with low or no HCEnC mitotic activity, subsequently slows to approximately 0.6% per year.^37,38 However, a recent report has proposed a role for UV-induced oxidative stress in the loss of HCEnC.³⁹ Thus, activation of cellular processes in response to such extrinsic/intrinsic stimuli will invariably lead to de novo gene transcriptional activity. Results from our differential gene expression analysis have demonstrated that many genes are uniquely expressed in pediatric and adult HCEnC, while GO analysis identified functional groups associated with those genes. The functional groups enriched in pediatric HCEnC demonstrate an important role for cell–matrix and cell–cell interactions in these cells distinct from adult HCEnC. Thus, pediatric HCEnC may be in the process of establishing an intact and integral endothelial monolayer mediated by cell–cell adhesion molecules such as cadherins.⁴⁰ In contrast, the adult HCEnC data set was enriched for functional groups indicating that these cells have developed characteristics of senescent cells and have become sensitized to apoptotic stimuli. The observation that adult HCEnC have reduced proliferative capacity may be attributed to the role that heparins, which have been shown to reduce corneal endothelial growth factor availability, play in the extracellular concentration of mitogenic factors.⁴¹

A recent study that used NGS technology for transcriptome analysis showed a large number of gene expression changes between fetal and adult human corneal endothelium.³⁰ The findings identify key biological functions for two anatomically and functionally distinct tissues. Although the fetal corneal endothelium can be identified as cuboidal epithelium lining the undeveloped posterior cornea, given the absence of DM, which at this stage of development is simply a discontinuous homogenous acellular layer, it is not surprising that the fetal corneal endothelial transcriptome differs significantly from that of the adult. Herein we build upon previous reports by comparing gene expression of pediatric (preschooler and preadolescent) and adult corneal endothelium, which are anatomically and morphologically similar to each other and presumably functionally indistinct. We identified nine highly significantly differentially expressed genes between pediatric and adult corneal endothelium. NALCN, TAC1, and TMOD1 (upregulated in adult corneal endothelium) have been associated with neuronal function.^42–44 As neurons are generally regarded as a terminally differentiated nonproliferating cell type, the increased expression of these genes in adult HCEnC may be representative of the terminally differentiated state of adult HCEnC. PREX2 (upregulated in adult corneal endothelium) has been associated with increased cell proliferation by inhibiting PTEN,⁴⁵ yet its role in G-protein–coupled receptor signaling suggests a potential role in a wider array of functions not related to cell proliferation. Knockdown of CAPN6 message has been demonstrated to promote cell migration and spreading,⁴⁶ while the core histone genes HIST1H3A and HIST1H4E are highly expressed during the S1 phase of the cell cycle and play a positive role in cell division.^47,48 Thus, downregulation of these genes in the adult corneal endothelium could in part explain the increase in HCEnC polymegethism and pleomorphism seen with increasing age. In addition, the downregulation of heat shock gene HSPA2 may remove a critical protein involved in protection against various stressors.^49–51 Taken together, the upregulation of neuron-associated genes (NALCN, TAC1, and TMOD1) and adhesion-associated genes (ITGBL1), in combination with the downregulation of genes associated with cell migration (CAPN6), cell division (HISTH1H3A, HIST1H4E), and cell survival (HSPA2) in the adult corneal endothelium, may prove to be a characteristic gene expression profile of HCEnC that are in functional decline and sensitized to mediators of cell death.

Determining the gene expression profile of cells that show a prominent disease phenotype is crucial in the process of identifying the genetic basis for the disease and elucidating the mechanisms that result in the disease phenotype. With regard to the PPCD3 locus on chromosome 10, only 20 of the 69 positional candidate genes are expressed in the HCEnC. Not only is ZEB1 one of these 20 genes, it is one of the most highly expressed. Thus, had the expression data that we report been available prior to the identification of ZEB1 mutations in PPCD3, ZEB1 would have been among only a handful of candidate genes selected for screening in families with PPCD3. It is therefore reasonable to expect that the genes involved in the pathogenesis of PPCD1, FECD4, and XECD are also expressed in HCEnC and that corneal endothelial expression should be used as a filtering strategy when evaluating coding region variants identified in positional candidate genes with NGS.

In conclusion, we have identified a number of differentially expressed genes in HCEnC, each of which is associated with one of two HCEnC age groups (pediatric and adult). These results have provided, for the first time, insight into the different functional characteristics of anatomically and morphologically indistinct HCEnC from two donor age groups. We have presented transcriptomic evidence that pediatric and adult HCEnC possess a set of distinct functional properties, but follow-up investigations are needed to determine the full functional roles of the genes identified in this study. On the basis of these results, we recommend that investigators studying gene expression in diseased tissue consider using age-matched controls for the most accurate comparison. In addition, we encourage investigators who have reported families with FECD4, XECD, and PPCD1 linked to the aforementioned loci to use the expression data that we present to prioritize screening of the positional candidate genes for each of these corneal endothelial dystrophies.