Endothelial Cell Whole Genome Expression Analysis in a Mouse Model of Early-Onset Fuchs' Endothelial Corneal Dystrophy

Abstract

Purpose.: To investigate the endothelial gene expression profile in a Col8a2 Q455K mutant knock-in mouse model of early-onset Fuchs' endothelial corneal dystrophy (FECD) and identify potential targets that can be correlated to human late-onset FECD.

Methods.: Diseased or normal endothelial phenotypes were verified in 12-month-old homozygous Col8a2 ^Q455K/Q455K mutant and wild-type mice by clinical confocal microscopy. An endothelial whole genome expression profile was generated by microarray-based analysis. Result validation was performed by real-time PCR. Endothelial COX2 and JUN expression was further studied in human late-onset FECD compared to normal samples.

Results.: Microarray analysis demonstrated endothelial expression of 24,538 genes (162 up-regulated and 172 down-regulated targets) and identified affected gene ontology terms including Response to Stress, Protein Metabolic Process, Protein Folding, Regulation of Apoptosis, and Transporter Activity. Real-time PCR assessment confirmed increased Cox2 (P = 0.001) and Jun mRNA (P = 0.03) levels in Col8a2 ^Q455K/Q455K mutant compared to wild-type mice. In human FECD samples, real-time PCR demonstrated a statistically significant increase in COX2 mRNA (P < 0.0001) and JUN mRNA (P = 0.002) and tissue microarray analysis showed increased endothelial COX2 (P = 0.02) and JUN protein (P = 0.04).

Conclusions.: The present study provides the first endothelial whole genome expression analysis in an animal model of FECD and represents a useful resource for future studies of the disease. In particular endothelial COX2 up-regulation warrants further investigation of its role in FECD.

Introduction

Fuchs endothelial corneal dystrophy (FECD) is a common disease of the corneal endothelium. It has been demonstrated to rank among the leading indications (9.3%–23.8%) for corneal transplant surgery in numerous Western countries but has a lower impact (1.7%–3.9%) in populations from the Asian region. ^1–9 An early-onset type and a more common late-onset type can be differentiated. ¹⁰ The pathology of FECD is characterized by a progressive decrease in corneal endothelial cell (CEC) density over several decades and concomitant formation of posterior excrescences (guttae) and thickening of the Descemet membrane. The attenuated CEC monolayer is eventually unable to maintain corneal deturgescence, and stromal edema with sub- or intraepithelial bullae ensues. Recent investigations proposed that CEC oxidative stress and stress of the endoplasmic reticulum (ER) play critical pathogenetic roles and may result in endothelial apoptosis induction. ^11–14  

Recently, we have reported the development of the first transgenic knock-in mouse model of FECD harboring a point mutation in the Col8a2 gene that has previously been associated with early-onset human disease. ¹⁴ Using this mouse model, we have demonstrated that the Col8a2 Q455K mutation is sufficient to elicit an FECD-like endothelial morphology; to activate the unfolded protein response (UPR), a cytoprotective signaling cascade; and to induce CEC apoptosis. ¹⁴ These results were supported by our previous observation of ER stress, UPR activation, and apoptosis induction in human late-onset FECD. ¹³  

Studying FECD in an animal model offers important advantages, including the ability to obtain and investigate corneal tissues at earlier disease stages (compared to end-stage tissues retrieved after corneal transplant surgery in humans) and to use highly standardized testing conditions. In this respect, fresh tissues from an animal model can be processed with the exclusion of external factors like prolonged death-to-preservation time and exposure to other nonphysiological conditions like organ culturing that may bias gene expression, especially in normal control corneas. 

The present study sought to obtain a deeper insight into the pathophysiology of early-onset FECD and, based on the results from the animal model, to detect potential parallels in the more common late-onset human disease. After verification of the diseased or normal corneal endothelial phenotype by clinical confocal microscopy, we performed endothelial whole genome expression profiling in midaged 12-month-old Col8a2^Q455K/Q455K mutant FECD mice and wild-type controls by high resolution gene microarray analysis. The results from this study were validated by quantitative real-time PCR. The expression of two individual targets, cyclooxygenase 2 and jun-proto-oncogene, was further studied in our mouse model and in human samples of late-onset FECD. 

Materials and Methods

Animals

Homozygous mutant knock-in mice (MUT) harboring the Col8a2 Q455K point mutation and Col8a2 wild-type (WT) mice were generated as previously described. ¹⁴ Animals were maintained and treated under specific pathogen-free conditions. All experiments were performed according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and adhering to protocols approved and monitored by the Animal Care and Use Committee of the Johns Hopkins University School of Medicine. 

Human Samples

Studies using human tissues were approved by the Johns Hopkins Institutional Review Board and adhered to the tenets of the Declaration of Helsinki. Tissue microarray (TMA) studies were performed using triplicate 1-mm-diameter cores of formalin-fixed paraffin-embedded tissue samples from 50 FECD corneas, 5 keratoconus corneas, 10 normal corneas, and nonocular control tissue specimens as previously described. ¹⁵  

Unfixed corneal endothelial samples (Descemet membrane and adhering CECs) for immediate endothelial mRNA extraction were retrieved from corneal transplant surgeries in FECD patients (n = 13 eyes from 13 patients, mean age [± SEM] 68.6 ± 2.3, 7:6 male to female ratio) or retrieved directly from whole donor eyes without corneal pathologies or glaucoma (n = 11 eyes from 9 donors, mean age [± SEM] 70.4 ± 4.6, 3:6 male to female ratio, mean death to preservation time [± SEM] 12.2 ± 2.1 hours). Written informed consent was obtained from all patients. 

Clinical Confocal Microscopy

Clinical confocal microscopy of the right eye was performed in all animals to confirm the diseased or normal endothelial phenotype, respectively. A ConfoScan3 microscope (Nidek, Fremont, CA) with a ×40 surface-contact objective was used as previously described. ¹⁴ Mice were anesthetized using isoflurane (Vedco, St. Joseph, MO) and euthanized by cervical dislocation. Whiskers were trimmed. Mice were placed on a customized platform and the head was fixed with the right eye pointing towards the objective. Lubricant eye gel (Genteal; Novartis, East Hanover, NJ) was used as an immersion fluid, and approximately 100 images of the central corneal endothelium were acquired. Mean CEC density was calculated from randomly selected microscopic images of the corneal endothelium using the ConfoScan software (Nidek). 

RNA Isolation and Gene Array Analysis

RNA was isolated from corneal endothelium of both eyes from three groups of MUT and three groups of WT mice. Each group consisted of two 12-month-old male animals (four corneas) of the respective strain (MUT or WT). Mice were euthanized, and clinical confocal microscopy was performed as already described. Corneal buttons from excised eyes were dissected for each group. Descemet membranes and adhering CECs were peeled off the corneal buttons using jewelers forceps under a dissecting microscope, pooled, and immediately disrupted by gentle pipetting in TRIzol Reagent (Invitrogen, Carlsbad, CA) at 4°C. RNA was extracted by combined TRIzol and RNeasy spin column (Qiagen, Valencia, CA) purification. Quantity and quality of the extracted RNA was measured by spectrophotometry (NanoDrop 2000; Thermo Scientific, Waltham, MA) and by bioanalyzer assessment (Bioanalyzer Series II Pico Chip; Agilent, Palo Alto, CA). 

Figure 1

 

Clinical confocal microscopy of the central corneal endothelium of 12-month-old Col8a2 ^Q455K/Q455K MUT (n = 12) and WT (n = 12) mice. MUT mice show loss of hexagonal shape (pleomorphism), irregularity of size (polymegethism), hyperreflective nuclei (open triangles), and guttae (closed triangles), whereas WT mice exhibit a normal endothelial morphology. The bar graph depicts loss of CEC density in MUT compared to WT animals as previously described. ¹⁴ Data are mean ± SEM; *P < 0.05.

Gene array analysis in endothelial RNA samples was performed on Mouse Exon 1.0 ST microarrays (Affymetrix, Santa Clara, CA). The Ovation Pico WTA system (NuGEN Technologies, San Carlos, CA) was used to generate SPIA-amplified cDNA from 10 ng of RNA. Sense transcript cDNA (ST-cDNA) was created from 3 μg of purified cDNA using the WT-Ovation Exon module (NuGen Technologies). Five micrograms of ST-cDNA were subsequently enzymatically fragmented, biotin labeled using the Encore Biotin module (NuGen Technologies), and hybridized onto Mouse Exon 1.0 ST arrays (Affymetrix) for 18 hours at 45°C with constant rotation at 60 rpm. Affymetrix Fluidics Station 450 was used to wash and stain the arrays, removing the nonhybridized target and incubating with a streptavidin–phycoerythrin conjugate to stain the biotinylated cDNA. Fluorescence was detected using the Affymetrix G3000 GeneArray Scanner and image analysis of each GeneChip was performed through the Affymetrix GeneChip Command Console version 2.0 (AGCC v2.0) software. 

For gene expression analysis, data sets were processed using PARTEK Genomics Suite (Partek, St. Louis, MO). Raw intensity levels of probe sets were normalized using the Robust Multichip Average (RMA) method. Two-way analysis of variance (ANOVA) was applied for statistical analysis of differential expression of individual genes in MUT compared to WT mice. Selection of significant genes was performed using a fold-change of >1.5 or <−1.5 and P < 0.05 as the cutoff criteria. 

To analyze the data quality among individual samples and between both biological groups, principal component analysis (PCA) was performed using the statistics package of R language (http://www.r-project.org). Unsupervised hierarchical clustering was performed using Cluster 3.0 (http://bonsai.hgc.jp/∼mdehoon/software/cluster/software.htm) and the average linkage method. ¹⁶ For gene ontology (GO) analysis, a Perl program was coded to identify significantly affected functional groups among the differentially expressed genes that showed enrichment of their respective GO term. The GO of all transcripts with a P value of <0.05 and a fold-change of >1.5 or <−1.5 was analyzed. 

Quantitative Real-Time PCR

RNA was extracted from three additional and independent groups of 12-month-old MUT and WT mice, respectively, and from human samples using TRIzol and RNeasy spin column (Qiagen) purification. RNA was reverse transcribed to cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions. The cDNA samples were preamplified using the Preamp Master Mix (Applied Biosystems) and Taqman inventoried gene expression assays (Table 1; Applied Biosystems). Preamplified cDNA was subjected to real-time PCR on a StepOne Plus Cycler (Applied Biosystems). Universal Master Mix, no UNG (Applied Biosystems), and the Taqman inventoried gene expression assays listed in Table 1 were used. Cycling conditions were 50°C for 2 minutes, 95°C for 10 minutes, 40 cycles at 95°C for 15 seconds, and 60°C for 1 minute. Target gene expression was normalized to expression of actin-beta (ActB/ACTB) in mouse and human samples. Fold changes between diseased tissues and controls were calculated using the comparative Ct Method and ExpressionSuite Software 1.0 (Applied Biosystems). 

Figure 2

 

Microarray analysis. (A) The first three principal components (PCs) of the PCA mapping account for 91.4% variance (PC1, 79.8%; PC2, 6.4%; PC3, 5.2%) of the data set. The endothelial samples are clearly arranged in clusters in conformity with their origin from Col8a2 ^Q455K/Q455K MUT (red) and WT (blue) mice. (B) The heat map and the dendrogram summarizing the unsupervised hierarchical clustering of all differentially expressed transcripts (fold change of >1.5 [red] or <−1.5 [green] with P < 0.05 from ANOVA) demonstrate the clear distinction between biological replicates from MUT and WT animals. (C) Volcano plot of significantly changed transcripts in MUT compared to WT animals with the x-axis representing the mean fold change (−10-fold to 10-fold change) and the y-axis representing the negative log10 of the P values. Horizontal threshold line, P = 0.05; vertical threshold lines, fold-change = −1.5 and 1.5, respectively.

Table 1. 

 

Assays Used for Quantitative Real-Time PCR Assessment of Mouse (m) and Human (h) Samples

Taqman Assay	Species	Gene Symbol	RefSeq	Gene Name
Mm00478374_m1	m	Ptgs2	NM_011198.3	Prostaglandin-endoperoxide synthase 2
Mm01213380_s1	m	Sod3	NM_011435.3	Superoxide dismutase 3, extracellular
Mm00495062_s1	m	Jun	NM_010591.2	Jun oncogene
Mm00472715_m1	m	Serp1	NM_030685.3	Stress-associated endoplasmic reticulum protein 1
Mm00459056_m1	m	Slc38a4	NM_027052.3	Solute carrier family 38, member 4
Mm00444330_s1	m	Slc5a3	NM_017391.3	Solute carrier family 5 (inositol transporters), member 3
Mm00442612_m1	m	Atp1b2	NM_013415.5	ATPase, Na+/K+ transporting, beta 2 polypeptide
Mm01329588_m1	m	Slc4a11	NM_001081162.1	Solute carrier family 4, sodium bicarbonate transport
Mm00607939_s1	m	Actb	NM_007393.3	Actin, beta
Hs00153133_m1	h	PTGS2	NM_000963.2	Prostaglandin-endoperoxide synthase 2
Hs01103582_s1	h	JUN	NM_002228.3	JUN oncogene
Hs99999903_m1	h	ACTB	NM_001101.3	Actin, beta

Figure 3

 

Enriched GO terms: bars in the diagram indicate false discovery rates (FDRs) of GO terms showing enrichment in the groups of the most up-regulated (white) and the most down-regulated (black) transcripts in Col8a2 ^Q455K/Q455K mutant compared to wild-type mice.

TMA Immunohistochemical Staining and Analysis

Four-micrometer sections were taken from the TMA paraffin block (for human tissue samples included on TMA see Methods subsection “Human Samples”). ¹⁵ Staining was performed through the Translational Pathology Core Laboratory, Department of Pathology and Laboratory Medicine, David Geffen School of Medicine at UCLA (COX2) and Translational Pathology Shared Resource at Vanderbilt University Medical Center (JUN). 

For COX2 staining, sections were deparaffinized and rehydrated through graded ethanol. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide in methanol for 10 minutes. Heat-induced antigen retrieval was carried out in 0.01 M citrate buffer, pH 6.00 at 95°C for 25 minutes. Rabbit antihuman COX2 antibody (Neomarkers, Clone SP21, concentration 1:100; Thermo Scientific) was applied for 45 minutes followed by incubation in secondary antibody (Dakocytomation Envision, System Labelled Polymer HRP antirabbit; Dako, Carpinteria, CA) for 30 minutes and 3,3′-diaminobenzidine (DAB) substrate-chromogen (Dako) for 10 minutes. Nuclei were counterstained with Mayer's hematoxylin. 

Figure 4

 

Validation of microarray data. (A) Four up-regulated and four down-regulated genes were selected for quantitative real-time PCR validation of Exon 1.0 ST microarray data in endothelial samples from Col8a2 ^Q455K/Q455K MUT (n = 6) and WT (n = 6) mice. Bar diagram shows fold change (MUT compared to WT) of individual transcripts as detected by the two different methods (*P < 0.05). (B) Whole corneas from MUT and WT animals (n = 3 each) were immunofluorescence double-labeled for cyclooxygenase 2 protein (COX2, green), which is encoded by Ptgs2, and for zonula-1 (ZO-1, red). Nuclei were counterstained with DAPI (blue). COX2-positive CECs within three microscopic fields at the given magnification were counted, and proportions of COX2-positive CECs per sample were calculated. Original magnification ×400. Data are mean ± SEM.

For JUN staining, slides were placed on the Bond Max IHC stainer (Leica, Deerfield, IL) and deparaffinized. Heat-induced antigen retrieval was performed using Epitope Retrieval 2 solution (Leica) for 20 minutes. Slides were incubated with anti-c-Jun (Clone 60A8; Cell Signaling Technology, Boston, MA) for 1 hour at a 1:400 dilution. The Bond Polymer Refine Detection system (Leica) was used for antibody detection. 

Immunolabeled TMA sections were analyzed by light microscopy. Evaluation was performed manually by one observer (MM) masked for all data (including diagnosis) from the donors' records. Antecedent-masked parallel scorings of multiple sections with a board-certified pathologist (CGE) ensured agreement in the grading procedure. 

Examples for individual TMA cores stained for COX2 and JUN from the FECD, KC, autopsy, and noncorneal control groups are shown in Figures 5A and 6A. For COX2, the endothelial staining intensity of each corneal core was evaluated using a ×100 oil immersion objective. Scoring was standardized on the basis of a four grade scoring system: 3, intense; 2, moderate; 1, weak; 0, negative (examples shown in Fig. 5B) applied to the most intense endothelial staining of each core section. For JUN, all endothelial nuclei per section were analyzed using a ×40 objective to determine the JUN-positive percentage of nuclei (example shown in Fig. 6B). Mean staining intensity scores (COX2) and mean percentages of positive nuclei (JUN) were calculated for each specimen and for the three respective corneal entities (FECD, KC, and autopsy/normal cornea) present on the TMA. Core sections with fewer than four evaluable CECs were excluded. 

Figure 5

 

COX2 expression in late-onset human FECD specimens. Sections of a TMA including triplicate cores from 50 FECD, 5 KC, and 10 normal autopsy corneas and additional noncorneal control cores as previously described were immunolabeled for COX2 protein. ¹⁵ (A) A representative core from each group (arrows indicate endothelial side, original magnification ×100). Staining intensities of core sections were evaluated using a four-grade scoring system (0 = negative; 1 = weak; 2 = moderate; 3 = intense; examples shown in [B], original magnification ×1000). Placenta tissue, also included on the TMA, served as positive control for the specificity of the staining (shown in [A, B]). Mean grades were calculated for each of the three groups (shown in [C], error bars indicate SEM). (D) Real-time PCR was used to assess relative PTGS2 transcriptional expression in FECD specimens (n = 13) compared to normal controls (n = 9). Horizontal bars indicate the mean value for each group. *P < 0.05.

Figure 6

 

JUN expression in late-onset human FECD specimens. TMA sections were immunolabeled for JUN protein. (A) A representative core from each group (arrows indicate endothelial side, original magnification ×100). (B) Representative endothelial sections (original magnification ×1000). Myometrium tissue, also included on the TMA, served as positive control for the specificity of the staining (shown in [A, B]). (C) The percentage of JUN-positive endothelial nuclei was determined in each core section. Mean percentages of positive nuclei were calculated for each of the three groups (shown in [C], error bars indicate SEM). (D) Real-time PCR was used to assess relative JUN transcriptional expression in FECD specimens (n = 13) compared to normal controls (n = 9). Horizontal bars indicate the mean value for each group. *P < 0.05.

Immunofluorescent Double-Labeling of Corneal Wholemounts

A modified protocol from Blitzer et al. ¹⁷ was used. Nine-month-old MUT and WT animals (n = 3 each) were euthanized, and clinical confocal microscopy was performed. Eyes were dissected and excised corneoscleral buttons were fixed in 0.5% paraformaldehyde for 30 minutes. Tissue samples were permeabilized in 0.5% Triton-X (Sigma-Aldrich, St. Louis, MO) in PBS for 15 minutes, and nonspecific binding sites were blocked by incubation in 5% goat serum for 30 minutes. Mouse anti-ZO-1 antibody (concentration 1:200; Invitrogen) and rabbit anti-COX2 antibody (concentration 1:500; Cayman Chemical, Ann Arbor, MI) diluted in 2% BSA served as primary antibodies. Alexa Fluor 555 goat antimouse and Alexa Fluor 488 goat antirabbit (both concentrations 1:500; Invitrogen) in 2% BSA served as secondary antibodies. Tissue samples were incubated at 37°C for 60 minutes in primary antibody solution and for 45 minutes in secondary antibody solution. Corneoscleral buttons were flatmounted in Prolong Gold Antifade Reagent with DAPI (Invitrogen) after making four radial incisions. Images were acquired with a LSM510 laser scanning confocal microscope (Carl Zeiss, Oberkochen, Germany) using a ×40 oil-immersion objective. COX2-positive CECs within three randomly selected microscopic visual fields were counted and used to calculate the proportions of COX2-positive CECs per sample. 

Statistical Analysis

Unless stated otherwise, PRISM4 software (Graphpad Software, La Jolla, CA) was used for statistical analyses applying the unpaired, two-tailed t-test. P < 0.05 was considered statistically significant. 

Results

Endothelial Phenotype of MUT and WT Animals

The endothelium of 12-month-old MUT mice exhibited a FECD-like phenotype with loss of hexagonal shape (pleomorphism) and irregularity of size (polymegethism) of the CECs as previously described and quantified. ¹⁴ None of these morphologic abnormalities were observed in WT endothelium. The CEC densities in representative cohorts of 12-month-old mice from each strain (measured in right corneas from n = 12 mice per strain) were 1050 ± 39/mm² in MUT animals (mean ± SEM) and 2086 ± 27/mm² in WT animals (P < 0.0001; Fig. 1). 

Comparative Analysis of Expression Profiles

Pooled endothelial mRNA samples of MUT and WT mice with an RNA integrity number (RIN) value of at least 7.5 were used for hybridization onto EXON 1.0 ST microarrays. One strong point of the EXON 1.0 ST microarray platform is the high-resolution analysis of the whole transcript by approximately 40 probes per gene. In total, we could detect transcriptional expression of 24,538 genes. 

PCA and hierarchical clustering were employed to visualize and assess the differences in gene expression profiles between the tested endothelial MUT and WT samples. Figure 2A shows a PCA map explaining 91.4% of the variance and demonstrating a distinct grouping of the respective MUT and WT samples. Also, the data analysis by unsupervised hierarchical clustering demonstrated a clear distribution of all biological replicates in the respective WT and MUT group as presented in the dendrogram in Figure 2B. A volcano plot of significantly changed transcripts is shown in Figure 2C. 

Using a positive or negative fold change of 1.5 and a P value of 0.05 as the cutoff, 162 (0.66%) genes were up-regulated, and 172 (0.7%) genes were down-regulated. A complete list of differentially expressed genes is presented as Supplementary Material (see Supplementary Tables S1 [up-regulated genes] and S2 [down-regulated genes]). The 35 genes with the highest differential expression in either direction in MUT compared to WT mice are presented in Table 2. 

Table 2. 

 

Top 35 Differentially Regulated Genes in MUT Compared to WT Animals*

	Gene Symbol	RefSeq	Fold Change, MUT versus WT	P Value, MUT versus
1	Bard1	NM_007525	23.71	0.0012
2	Rnps1	NM_009070	8.78	0.0113
3	Stoml1	NM_026942	7.65	0.0249
4	Magix	NM_018832	−6.34	0.0015
5	Tbl3	NM_145396	5.37	0.0258
6	Cdv3	NM_175565	−5.25	0.0075
7	Gabrb2	NM_008070	−5.16	0.0042
8	Lman1	NM_001172062	5.16	0.0474
9	1700029I01Rik	NM_027285	−5.13	0.0085
10	3110048L19Rik	NR_003549	−5.09	0.0398
11	Tmtc4	NM_028651	−5.04	0.0010
12	B230220N19Rik	AK053483	−4.89	0.0215
13	Shisa7	NM_172737	−4.45	0.0353
14	LOC100046744	XM_001475765	4.27	0.0006
15	Ccnd2	NM_009829	4.26	0.0011
16	D730045A05Rik	NR_045390	−4.22	0.0468
17	Timp1	NM_001044384	4.12	0.0011
18	Ptgs2	NM_011198	4.00	0.0029
19	Med28	NM_025895	3.97	0.0361
20	Osmr	NM_011019	3.83	0.0038
21	Tmem175	NM_001163531	−3.70	0.0289
22	Wars	NM_001164314	3.61	0.0252
23	Gm10590	XR_001565	−3.54	0.0299
24	Slc38a4	NM_027052	−3.48	0.0000
25	Sorcs3	NM_025696	−3.43	0.0202
26	1700021N20Rik	AK006223	−3.40	0.0282
27	Lect1	NM_010701	−3.35	0.0025
28	Ebf1	NM_007897	−3.30	0.0042
29	Bptf	NM_176850	−3.30	0.0434
30	Cnn3	NM_028044	−3.30	0.0497
31	Gm6455	NR_003596	−3.30	0.0220
32	Hmox1	NM_010442	3.22	0.0001
33	Rragc	NM_017475	3.20	0.0246
34	Gnpnat1	NM_019425	3.18	0.0035
35	Tulp4	NM_054040	−3.11	0.0143

*  Fold-change values of >1.00 indicate up-regulation and values of <−1.00 indicate down-regulation.

Gene Ontology of Differentially Expressed Genes

Statistically significant enrichment of at least 15 functional groups was detected (false discovery rate <0.05, Fig. 3). 

Real-Time PCR and Immunofluorescence Validation of Microarray Data

Real-time PCR analysis investigating examples of four overrepresented and four underrepresented targets (Tables 1, 3, 4) in the six original (n = 3, MUT and WT samples, respectively) and in six independent endothelial samples (n = 3, MUT and WT samples, respectively) from 12-month-old mice was performed to validate the microarray data. Real-time PCR analysis confirmed relative mRNA expression levels compared with the microarrays and demonstrated a consistency of the data retrieved by both methods (Fig. 4A). 

Table 3. 

 

Genes Associated with Top Three Up-Regulated GO Terms in MUT Compared to WT Animals*

Gene Symbol	RefSeq	Response to Stress	Catalytic Activity	Protein Metabolic Process	Fold Change, MUT versus WT	P Value, MUT versus WT
Bard1	NM_007525	✓	✓		23.71	0.0012
	NM_011198	✓	✓		4.00	0.0029
Wars	NM_001164314		✓	✓	3.61	0.0252
Hmox1	NM_010442	✓	✓		3.22	0.0001
Rragc	NM_017475		✓		3.20	0.0246
Gnpnat1	NM_019425		✓		3.18	0.0035
Pde3a	NM_018779		✓		3.04	0.0076
Tyrobp	NM_011662		✓		2.86	0.0084
Serpine1	NM_008871	✓	✓		2.80	0.0074
Exosc6	NM_028274		✓		2.63	0.0299
Itch	NM_008395		✓	✓	2.60	0.0210
Ero1lb	NM_026184		✓	✓	2.45	0.0109
	NM_011435	✓	✓		2.30	0.0002
Manba	NM_026968		✓		2.25	0.0058
Ptpra	NM_008980		✓	✓	2.24	0.0488
Dusp6	NM_026268		✓	✓	2.23	0.0026
Lyz2	NM_017372	✓	✓		2.20	0.0235
Fanca	NM_016925	✓			2.17	0.0465
Dnajb13	NM_153527			✓	2.15	0.0465
Ddb1	NM_015735	✓			2.03	0.0131
Gls	NM_001081081		✓		2.03	0.0070
Pcsk7	NM_008794		✓	✓	2.02	0.0020
	NM_010591	✓			1.92	0.0001
B4galt6	NM_019737		✓		1.89	0.0435
Cpe	NM_013494		✓	✓	1.89	0.0089
Ahr	NM_013464	✓			1.85	0.0375
Ugcg	NM_011673		✓		1.82	0.0045
Ephx3	NM_001033163		✓		1.78	0.0008
Plat	NM_008872	✓	✓	✓	1.77	0.0328
Mtif2	NM_133767		✓	✓	1.75	0.0349
Fut10	NM_134161		✓	✓	1.74	0.0250
Pdia4	NM_009787		✓		1.73	0.0051
Ctsc	NM_009982		✓	✓	1.72	0.0088
Ppil3	NM_027351		✓	✓	1.69	0.0071
Ccl2	NM_011333	✓			1.69	0.0036
Ddit3	NM_007837	✓			1.66	0.0163
Clec2d	NM_053109	✓			1.66	0.0378
Hdac5	NM_001077696	✓	✓		1.65	0.0012
Uba5	NM_025692		✓	✓	1.64	0.0024
Stk39	NM_016866		✓	✓	1.63	0.0041
Cd74	NM_001042605	✓		✓	1.62	0.0057
Prss23	NM_029614		✓	✓	1.62	0.0065
Jak1	NM_146145		✓	✓	1.60	0.0339
Rhobtb3	NM_028493		✓		1.60	0.0223
Iars	NM_172015		✓	✓	1.59	0.0012
Acsl4	NM_207625		✓		1.59	0.0197
Scd1	NM_009127	✓	✓		1.58	0.0190
Smok4a	NR_030763		✓	✓	1.58	0.0183
Myof	NM_001099634	✓			1.56	0.0103
Cct6a	NM_009838			✓	1.55	0.0288
Pgm3	NM_028352		✓		1.54	0.0015
Steap1	NM_027399		✓		1.54	0.0248
Ece2	NM_025462		✓	✓	1.53	0.0212
Entpd7	NM_053103		✓		1.53	0.0321
	NM_030685	✓		✓	1.52	0.0401

*  Differential expression of bold genes was validated by real-time PCR.

Table 4. 

 

Genes Associated with Top Three Down-Regulated GO Terms in MUT Compared to WT Animals*

Gene Symbol	RefSeq	Transporter Activity	Carbohydrate Metabolic Process	Establishment of Localization	Fold Change, MUT versus WT	P Value, MUT versus WT
Gabrb2	NM_008070	✓		✓	−5.16	0.0042
	NM_027052	✓		✓	−3.48	0.0000
Cacng7	NM_133189	✓		✓	−2.63	0.0023
Mgat5	NM_145128		✓		−2.48	0.0427
Pglyrp4	NM_001165968		✓		−2.45	0.0181
	NM_017391	✓	✓	✓	−2.39	0.0085
Txnip	NM_023719			✓	−2.36	0.0088
Gpm6a	NM_153581	✓			−1.97	0.0022
	NM_013415	✓		✓	−1.91	0.0069
Trpm3	NM_001035240	✓		✓	−1.86	0.0004
Gdpd2	NM_023608		✓		−1.82	0.0132
Ogdh	NM_010956		✓		−1.82	0.0289
Man2a1	NM_008549		✓		−1.80	0.0237
Ttpal	NM_181734	✓		✓	−1.78	0.0419
Igf2	NM_010514		✓		−1.73	0.0040
Lrp1b	NM_053011			✓	−1.68	0.0293
Syt9	NM_021889	✓		✓	−1.67	0.0142
Hyal1	NM_008317		✓		−1.66	0.0316
Vps4b	NM_009190			✓	−1.61	0.0047
	NM_001081162	✓		✓	−1.61	0.0245
Cox7c	NM_007749	✓			−1.60	0.0412
B3galt2	NM_020025		✓		−1.60	0.0230
Atp2a2	NR_027838	✓		✓	−1.60	0.0007
Fras1	NM_175473			✓	−1.58	0.0139
Grik5	NM_008168	✓		✓	−1.53	0.0038
Atp8a1	NM_001038999	✓		✓	−1.53	0.0137
Cacna1g	NM_009783	✓		✓	−1.53	0.0050
Crabp2	NM_007759	✓		✓	−1.52	0.0066
Slc6a1	NM_178703	✓		✓	−1.52	0.0041
Exph5	NM_176846			✓	−1.51	0.0341
Pitpnc1	NM_145823	✓		✓	−1.50	0.0145

*  Differential expression of bold genes was validated by real-time PCR.

PTGS2 (COX2), which encodes the enzyme cyclooxygenase 2 (COX2), was chosen for further evaluation in mouse and human tissues due to its comparably high transcriptional overexpression in MUT mice (4.3-fold, P = 0.001) and the previously established role of cyclooxygenases in injured corneal endothelium. ¹⁸ By immunofluorescence labeling, 4.5 ± 1.6% COX2-positive CECs were detected in the endothelium of 9-month-old MUT mice, whereas no COX2 expression was found in the endothelium of WT mice (Fig. 4B). 

COX2 and JUN in Human Samples

Two TMA sections, each including triplicate cores from 50 FECD, 10 autopsy, and 5 KC corneas, respectively, were investigated per protein (data presented as COX2/JUN, mean ± SEM): adequate tissue quality (at least one evaluable core) on the immunolabeled TMA sections was present for 42/45 FECD corneas (mean value of 2.8 ± 0.3/3.6 ± 0.3 evaluable cores per cornea), 5/5 KC corneas (mean value of 3.0 ± 0.5/4.6 ± 0.7 evaluable cores per cornea), and 8/8 autopsy corneas (mean value of 3.3 ± 0.7/3.0 ± 0.7 evaluable cores per cornea). The mean donor ages (and male to female ratios) of these evaluable corneal specimens were 69.2 ± 1.3/69.6 ± 1.3 (14:28/15:30) for the FECD group, 58.6 ± 8.6/58.6 ± 8.6 (2:3/2:3) for the KC group, and 62.1 ± 3.0/62.1 ± 3.0 (5:3/5:3) for the autopsy group. The mean scores according to the COX2 four grade scoring system (Fig. 5B)/the mean percentages of JUN positive cells were 1.5 ± 0.1/35.3 ± 0.1 for the FECD group, 0.6 ± 0.2/4.8 ± 1.0 for the KC group, and 0.7 ± 0.3/15.2 ± 4.3 for the autopsy group (Figs. 5C, 6C). The difference between the FECD and autopsy groups (for COX2, P = 0.02 and for JUN, P = 0.04) and the difference between the FECD and KC groups (for COX2, P = 0.03 and for JUN, P = 0.01) were statistically significant for both targets. 

Real-time PCR showed statistically significant transcriptional up-regulation of PTGS2 and JUN (4.00- and 1.92-fold up-regulated, respectively, in MUT mouse endothelium) in human endothelial FECD samples (n = 13 eyes from 13 patients) compared to normal controls (n = 11 eyes from nine donors; in the case of two available eyes from the same donor, the expression values from both eyes were averaged). Up-regulation for COX2 was 3.1-fold (P < 0.0001; Fig. 5D) and for JUN was 3.1-fold (P = 0.002; Fig. 6D). 

Discussion

In the present study we used high-resolution microarrays to detect endothelial gene expression changes in a Col8a2 Q455K mutant mouse model of early-onset FECD. Our results demonstrate distinct changes in the endothelial transcriptome of mutant animals, providing further insight into the pathophysiology of early-onset FECD. Cox2 and Jun were selected from the subset of up-regulated genes for further evaluation. In human late-onset FECD endothelium, elevated COX2 and JUN expression was demonstrated by TMA analysis and real-time PCR. 

We have recently reported the development of the first transgenic mouse model of FECD. ¹⁴ The point mutation harbored by this animal model causes a glutamine to lysine substitution at amino acid position 455 (Q455K) of the Col8a2 gene, and the equivalent mutation in the human COL8A2 gene has been associated with early-onset FECD. ^14,19 Homozygous Col8a2 ^Q455K/Q455K mutant mice exhibit an endothelial phenotype comprising important characteristics of the human disease. These include progressive loss of CEC density, alterations in endothelial cell morphology, and basement membrane guttae. ¹⁴  

Our microarray analysis revealed 334 differentially regulated genes (162 up-regulated, 172 down-regulated) in the corneal endothelium of 12-month-old mutant mice. Real-time PCR experiments reinvestigating four respective up- or down-regulated transcripts in Col8a2 mutant and wild-type animals validated our results and demonstrated good accordance with the microarray data. Response to Stress, Protein Metabolic Process, Protein Folding, Regulation of Apoptosis, and Transporter Activity were among the functional groups that showed enrichment in the GO analysis. A complete list of all affected GO terms is presented in Figure 3. 

Cellular stress, particularly oxidative stress and ER stress, have been recently assigned central roles in the pathogenesis of FECD. ^12,13 Our microarray data also support a pathogenesis involving cellular stress–associated mechanisms in our mouse model based on the detection of a significant number of up-regulated stress response–related targets in the endothelium of Col8a2 ^Q455K/Q455K mutant mice. It is noteworthy that this finding is paralleled by increased transcriptional expression of genes related to the functional groups of Protein Metabolic Process and Protein Folding like the chaperone-encoding genes Cct6a and Ppil3 or the stress associated endoplasmic reticulum protein 1 (Serp1). We consider this as potential further evidence for endothelial ER stress resulting in activation of the UPR, supporting results from our previous studies in the same mouse model and in a population of genetically heterogeneous late-onset FECD patients. ^13,14 The hypothesis that the misfolding or aggregation of proteins in FECD contributes to cellular stress and apoptosis induction in CECs has also been proposed in studies investigating other genes affected in late-onset FECD like SLC4A11 and LOXHD1. ^20,21  

Increased transcriptional regulation of other important stress-associated genes such as Ptgs2 (Cox2), Hmox1, Serpine1, and Sod3 (complete list in Table 3) was found in the corneal endothelium of mutant animals. Although these markers point at least in part to the presence of endothelial oxidative stress, we observed no down-regulation of the antioxidant defense system, which has been demonstrated by previous studies in human late-onset end-stage FECD specimens. ¹² The underlying reasons for this inequality may be among other possibilities differences in the species (mouse versus human), the entities (early-onset versus late-onset FECD) or the stages of the disease (midstaged samples from 12-month-old mice versus end-staged tissues from human patients). 

The corneal endothelial monolayer maintains corneal deturgescence through a “pump-leak” mechanism. Previous studies have demonstrated decreased transcriptional expression of pump function–related genes in human late-onset FECD endothelium ²² and reduced ATPase pump site density in the mid- and end-stage phase of the disease. ^23,24 Our study revealed the down-regulation of Slc4a11 and of numerous other transport-related genes like the Na/K ATPase encoding gene Atp1b2 in Col8a2 mutant animals. Mutations and down-regulation of Slc4a11 have been associated with human late-onset FECD before. ^20,22 Moreover, it was recently shown that the knock-down of SLC4A11 causes a reduction of cellular growth and proliferation in HeLa cells and that a depletion of the SLC4A11 gene alone may already lead to apoptosis induction in CECs. ^25,26  

Prostaglandin-endoperoxide synthase 2 (PTGS2, COX2) was further evaluated in the early-onset FECD mouse model and in human late-onset FECD. PTGS2 was chosen due to its 5-fold endothelial transcriptional up-regulation in mutant animals, the previously established corneal endothelial expression of cyclooxygenases and its modulation capability with readily available drugs like COX2 inhibitors. Increased expression of cyclooxygenases was previously detected in CECs involved in endothelial wound closure. ²⁷ Cyclooxygenases influence changes of CECs in morphology, mitosis, and migration through their most important product, the eicosanoid prostaglandin E2 (PGE2). ^18,27–29 Jumblatt et al. ¹⁸ have shown that injury-induced PGE2 synthesis exhibits increased sensitivity for diclofenac and indomethacin but not for aspirin, suggesting a central role of COX2. To our knowledge, studies on the expression of COX2 in FECD endothelium have not yet been performed. The present study demonstrates transcriptional and translational overexpression of COX2 in both the Col8a2 mutant mouse model and human late-onset FECD samples. Whether COX2 dysregulation has a protective or detrimental impact in FECD will be the subject of future studies. 

The expression of another stress-responsive target, jun proto-oncogene (JUN), which had shown moderate but significant up-regulation in the mouse model, was investigated in human FECD tissue to further pursue the transferability of the mouse model data to the human disease. Jun is a protein of the AP-1 complex with a complex spectrum of functional properties being involved in cellular processes such as proliferation, cellular survival or death, and differentiation. ^30,31  

We demonstrate that up-regulation of JUN mRNA and protein occur in the human disease and further corroborate the parallels between the two systems. However, comprehensive expression studies are needed to identify in more detail the extent of shared genes related to endothelial cell stress and death in the early-onset FECD mouse model and the late-onset human disease. 

In conclusion, the present study provides the first endothelial whole genome expression analysis in an animal model of early-onset FECD. It delineates important parallels and differences to previous findings in human late-onset FECD endothelium. Response to Stress, Protein Metabolic Process, Protein Folding, Regulation of Apoptosis, and Transporter Activity are among the differentially expressed functional groups of genes. The results may serve as an important guideline for the future application of the Col8a2 mutant mouse as a model of endothelial cell death and warrant further investigation of the role of COX2 up-regulation in FECD. 

Supplementary Materials

pdf S1 

Acknowledgments

The authors thank Clara M. Magyar, PhD, Anthony Frazier, Rhonda Grebe, BS, and Mary E. Pease, MS, for technical support and Shannath Merbs, MD, PhD, and Ray Enke, PhD, for providing human corneal control samples. 

Footnotes

 Supported by Deutsche Forschungsgemeinschaft (DFG MA 5110/2-1 [MM]); Richard Lindstrom/Eye Bank Association of America Research Grant (MM); National Institutes of Health (NIH EY019874 [ASJ], EY001765 to Wilmer Microscopy Core Facility); Medical Illness Counseling Center (ASJ); grants from Stanley Friedler, MD, Diane Kemker, Jean Mattison, and Lee Silverman (ASJ); Research to Prevent Blindness (to Wilmer Eye Institute); and National Cancer Institute (NCI Cancer Center Support Grant P30CA068485 to Vanderbilt Translational Pathology Shared Resource).

Footnotes

 Disclosure: M. Matthaei, None; J. Hu, None; H. Meng, None; E.-M. Lackner, None; C.G. Eberhart, None; J. Qian, None; H. Hao, None; A.S. Jun, None