Extracellular Matrix Alterations in Late-Onset Fuchs' Corneal Dystrophy

Abstract

Purpose.: To characterize the alterations of extracellular matrix proteins in Descemet's membranes (DM) of patients with late-onset Fuchs' corneal dystrophy (FCD) and to differentiate them from nonspecific alterations in pseudophakic bullous keratopathy (PBK).

Methods.: Human DM–endothelial cell complexes were obtained from patients with late-onset FCD (n = 40), PBK (n = 6), and control eyes (n = 5). Gene expression profiles of endothelial cells were compared using a commercial real-time PCR array and quantitative real-time PCR assays for confirmation of differentially expressed genes. A total of 24 extracellular matrix proteins were also localized in cryosections of corneal specimens from FCD (n = 10), PBK (n = 4), and control eyes (n = 5) by immunohistochemistry.

Results.: Polymerase chain reaction array analysis revealed a significant upregulation of 27 out of 84 extracellular matrix–related genes including collagens, proteoglycans, glycoproteins, cell adhesion molecules, and matrix metalloproteinases in FCD specimens as compared to normal controls, which could be partly confirmed and quantified by real-time PCR. Comparative analysis of FCD and PBK specimens showed a significant and consistent FCD-specific upregulation of collagen types I, III, and XVI; fibronectin; agrin; clusterin; transforming growth factor beta-induced (TGFBI); and integrin α4 (3- to 18-fold, P < 0.05). Immunohistochemistry revealed an increased labeling of collagen (types III, VII, XV, XVI), agrin, fibulin-2, TGFBI, versican, and clusterin in the DM of FCD specimens compared to PBK specimens.

Conclusions.: The findings provide evidence for a specific upregulation, production, and deposition of collagen types III and XVI, agrin, TGFBI, and clusterin in late-onset FCD and thus point to the importance of matrix alterations in the pathophysiology of FCD.

Introduction

Fuchs' corneal dystrophy (FCD) is the most common hereditary disease of the corneal endothelium, leading to decreased, blurred vision and, if untreated, to blindness and pain. In early-onset FCD, different mutations in the COL8A2 (collagen type VIII α2) gene have been found. ^1–4 In late-onset FCD, which is far more common, four chromosomal loci (FCD1–4) have been detected by analysis of families with autosomal dominant inheritance of FCD. ^5–8 Recently, polymorphisms in the TCF4 (transcription factor 4) and CLU (clusterin) genes have been associated with late-onset FCD ⁹ as well as missense mutations in the LOXHD1 (Lipoxygenase homology domains 1), ¹⁰ the SLC4A11 (solute carrier family 4, sodium borate transporter, member 11), ¹¹ and the AGBL1 (ATP/GTP-binding protein-like 1) genes. ¹² Apart from genetic factors, the eminent role of oxidative stress and apoptosis of corneal endothelial cells (EC) in the pathophysiology of FCD has been highlighted in recent years. ^13–20 Nonetheless, currently there is no cure for FCD apart from corneal transplantation, which creates the need to better understand the pathophysiology of FCD in order to find more specific therapeutic targets. 

A characteristic and early clinical sign of both early- and late-onset FCD is the formation of mushroom- or wart-like excrescences of Descemet's membrane (DM), so called guttae, which is accompanied by decrease of EC density. These alterations start usually in the center of DM with single guttae and spread gradually to the periphery. ^21,22 The formation of guttae as the most characteristic and pathognomonic feature of FCD, which may reflect an abnormal extracellular matrix (ECM) metabolism of corneal EC, has not been investigated in detail. 

Corneal EC produce various ECM components of DM as a specialized basement membrane. During fetal development, an anterior banded layer is formed, covered posteriorly by a nonbanded layer that gradually increases in thickness during the lifetime. Descemet's membrane is composed of different collagens (types VI, VIII, XII, XVIII), glycoproteins (e.g., fibronectin, laminin, osteonectin), and proteoglycans (e.g., versican). ²³ The ECM components are distributed inhomogeneously within DM, with some ECM proteins concentrated on the stromal side (e.g., fibronectin, vitronectin) and others mainly on the endothelial side of DM (e.g., osteonectin). ²³ Beyond that, the composition of ECM components in DM changes during life ²⁴ and in various pathologic conditions involving any disturbance of or damage to the corneal endothelium, such as pseudophakic bullous keratopathy (PBK). ^25,26 Histopathologic manifestation of corneal endothelial impairment is the formation of an additional fibrous layer, also termed posterior collagenous layer (PCL), which is deposited on the posterior surface of DM proper. ²⁶ Although an altered composition of DM with increased deposition of collagen types IV and VIII, fibronectin, and laminin has also been described in corneal specimens from FCD patients, ^27–31 these alterations may reflect merely nonspecific changes of a corneal endothelial damage during later stages of the disease, instead of disease-specific genetically determined alterations. So far, only one study has differentiated between specific ECM alterations in FCD and PBK, but it focused on immunohistochemical studies of a few DM components only. ³² Therefore, the purpose of our study was to perform a comprehensive analysis of the ECM composition of DM on the mRNA and protein level and to identify FCD-specific alterations by comparing ECM expression levels with those of PBK and normal corneas. 

Patients and Methods

Tissue Specimens

For RNA extraction with on-column DNase digestion step (RNeasy Micro kit; Qiagen, Hilden, Germany), DM–EC complexes were obtained from patients with late-onset FCD (n = 40, mean age 70.3 ± 9.7 years) or PBK (n = 6, mean age 70.2 ± 11.3 years) during Descemet's membrane endothelial keratoplasty (DMEK). After paracentesis, DM was stripped from the corneal stroma with a hook (Price Endothelial Keratoplasty Hook; Moria SA, Antony, France) as previously described.³³ Ten of 40 FCD specimens (mean age 69.1 ± 10.3 years) were processed in toto; 30 of 40 FCD specimens (mean age 70.7 ± 9.6 years) were divided into a central 5-mm portion and a peripheral rim and analyzed separately in order to detect differences in expression levels between the corneal center and periphery. Due to the low amount of RNA in the divided specimen, the central and peripheral samples of 10 individuals with FCD were pooled (n = 3). The specimens were directly transferred into RNA extraction buffer (RLT; Qiagen) and snap frozen in liquid nitrogen. 

Two different tissue samples were used as controls (n = 5, mean age 70.6 ± 10.1 years). Firstly, DM–EC complexes were obtained from two eyes with posterior malignant melanoma of the uvea and intact anterior segment by DM stripping after enucleation. Secondly, three donor eyes that could not be used for corneal transplantation due to contraindications (dementia, malignoma of the hematopoietic system) were obtained, and DM was stripped using the same technique used to prepare the grafts for DMEK. ³³  

There was no significant difference in the age of the patients of the three study groups (FCD, PBK, control; Mann-Whitney U test P = 0.87). 

For immunohistochemistry, corneal specimens were obtained during penetrating keratoplasty from patients with FCD (n = 10, mean age 73.3 ± 7.9 years), from those with PBK (n = 4, mean age 77.7 ± 5.4 years), and from normal donor eyes (n = 5, mean age 74.5 ± 5.7 years) without any history of ocular disease, then cryopreserved in optimum cutting temperature compound (Sakura Finetec, Staufen, Germany). 

The study was approved by the institutional review board and followed the tenets of the Declaration of Helsinki. Informed consent was obtained from the subjects after explanation of the nature of the study. 

Real-Time RT-PCR

First-strand cDNA synthesis and preamplification of cDNA for 12 cycles was performed using 50 ng total RNA and the RT² PreAMP cDNA synthesis kit (Qiagen). Expression of endothelial mRNAs was investigated using the RT² Profiler PCR-Array Human Extracellular Matrix and Adhesion Molecules (Qiagen) according to the manufacturer's recommendations, and analyzed using RT² Profiler PCR Array Data Analysis Tool version 3.2. 

Differentially expressed genes were confirmed by specific PCR assays. First-strand cDNA synthesis was performed using 50 ng total RNA, 200 U Superscript II reverse transcriptase (Invitrogen, Karlsruhe, Germany), and 200 ng random primers (Invitrogen) in a 20-μL reaction volume. Quantitative real-time PCR was carried out with the MyIQ Thermal Cycler and software (Biorad, Munich, Germany). The PCR reactions (20 μL) contained 2 μL cDNA, 3 mM MgCl₂, 0.8 μM each upstream and downstream primer, and 0.3 μM universal probe (Roche, Mannheim, Germany) in 1× TaqMan Probe Mastermix (Roche). All samples were analyzed in duplicates with a program of 95°C for 10 minutes and 40 cycles of 95°C for 10 seconds, 60°C for 30 seconds. For quantification, standard curves using serial dilutions (10²–10⁷ copies) of plasmid-cloned amplicons were run in parallel. For normalization of gene expression levels, ratios relative to the housekeeping gene GAPDH were calculated. Primer sequences (Eurofins, Anzing, Germany) are given in Table 1. 

Table 1

 

Primers Used for Quantitative Real-Time PCR

Gene	Accession No.	Product	Probe	Sequence, 5′–3′
GAPDH	NM_002046	66 bp	No. 60	AGCCACATCGCTCAGACAC
				GCCCAATACGACCAAATCC
Clusterin	NM_001831	65 bp	No. 71	CAGATGTACTGCAATGGAACAAA
				CATGTGGACTTTGCTACACACC
Col4A2	NM_001846	74 bp	No. 29	TCCCTATAGACCACTGGGTTTG
				GCAAGGCTGACAATGATGTCTA
Col5A1	NM_000093	88 bp	No. 6	TTTGTCCTTTTCTCCTGTCATTT
				AGAACGGGACACATTTTGAAG
Col6A1	NM_001848	76 bp	No. 18	CAGACATAAATCTCGGCGACT
				CTGTAGGGCCAAGGTCCA
Col8A2	NM_005202	65 bp	No. 13	TCACCCCAGCCAGGTATC
				GCGTGCACCTTGTTCAGAG
Fibronectin	NM_002026	97 bp	No. 39	TTGCTCTTTTCTAACCATTGTAATTCT
				TATTTCCCTTGCAGGCAATC
ITGAL	NM_002209	84 bp	No. 45	GGGAACCACGTCTGCTAACT
				TGGACAGAATTTCACATTTATTGG
ITGA4	NM_000885	64 bp	No. 32	TGATTTTGAAATTTAACTGCTCTGG
				CAGATTTCATAAGTCTGCCTTGATT
TGFBI	NM_000358	77 bp	No. 78	ACAGTTTTTGTAAAGCCCTTGC
				CATTTGACAGAACATTTCAACTCAT
VCAN	NM_004385	73 bp	No. 12	ATGTTTAAAGAAAAACCTGTAATGGA
				CTCTTCTTCAAGTTGCTCTAAACTGA

Immunohistochemistry

Immunohistochemistry was performed as previously described. ²³ Cyrosections 5 μm thick were incubated with primary antibodies against agrin (T. Sasaki, Erlangen, Germany), clusterin (Merck Millipore, Darmstadt, Germany), fibrillin-1 (Merck Millipore), fibronectin (Sigma-Aldrich, St. Louis, MO, USA), fibulin-1 (T. Sasaki), fibulin-2 (T. Sasaki), collagen I (SouthernBiotech, Birmingham, AL, USA), collagen III (Merck Millipore), collagen IV (SouthernBiotech), collagen V (Merck Millipore), collagen VI (Merck Millipore), collagen VII (D. Reinhardt, Montreal, Canada), collagen VIII (Santa Cruz Biotechnology, Dallas, TX, USA), collagen XII (Assay Biotech, Sunnyvale, CA, USA), collagen XV (T. Sasaki), collagen XVI (Assay Biotech), collagen XVII (L. Bruckner-Tudermann, Freiburg, Germany), collagen XVIII (T. Sasaki), laminin-1 (M. Paulsson, Cologne, Germany), osteopontin (Abcam, Cambridge, UK), tenascin-C (Dako, Glostrup, Denmark), TGFBI (Sigma-Aldrich), versican (Acris, San Diego, CA, USA), and vitronectin (Merck Millipore) overnight at 4°C. Alexa 488-conjugated secondary antibodies (goat anti-rabbit/-mouse; Molecular Probes, Eugene, OR) were applied for the detection of antibody binding, and cell nuclei were stained using propidium iodide. In negative control experiments, the primary antibody was replaced by PBS or equimolar concentrations of an irrelevant primary antibody. 

Statistical Analysis

Data are presented as means ± SD. Statistical evaluation of significant differences between groups of eyes was performed with Student's t-test for pairwise comparison. P < 0.05 was considered statistically significant. 

Results

Differential Gene Expression

As a first step to investigate the molecular mechanisms underlying FCD, we used real-time PCR array technology to simultaneously monitor mRNA expression levels of 84 genes related to ECM metabolism in DM specimens from patients with late-onset FCD (n = 4) compared with normal controls (n = 4). Genes were considered as differentially expressed when their expression levels exceeded a 2-fold increase over control in all four FCD specimens. 

The majority of the gene products tested (70%) displayed equal expression in diseased and normal specimens. Genes that were significantly upregulated in FCD specimens included collagens (collagen types IV, V, VI, VII; upregulation: 3- to 42-fold, P < 0.05), glycoproteins (fibronectin 1, laminin α2 chain, laminin γ1 chain, osteonectin; upregulation: 2- to 400-fold, P < 0.05), proteoglycans (versican; upregulation: 351-fold, P < 0.05), cell adhesion molecules (neural cell adhesion molecule 1, integrin α1, α3, α4, αL, β1, β3, β4; upregulation: 2- to 24-fold, P < 0.05), matrix metalloproteinases (MMP) and inhibitors (dysintegrin and matrix metallopeptidase with thrombospondin motif 1 and 8, MMP 9, 10, 11, 14, tissue inhibitor of MMP 1; upregulation: 2- to 22-fold, P < 0.05), and profibrotic growth factors (connective tissue growth factor; upregulation: 5-fold, P < 0.05). Transforming growth factor beta-induced (TGFBI) showed a 3-fold upregulation with a borderline significance (P = 0.06). Highest expression levels were observed for fibronectin (400-fold) and versican (351-fold). Differentially expressed gene products are summarized in Table 2. 

Table 2

 

Real-Time PCR Array: Differentially Expressed Gene Products in Endothelial Specimens From FCD Patients as Compared to Unaffected Controls (n = 4)

Gene Product	Gene Symbol	Fold Change	P Value
A dysintegrin and matrix metallopeptidase with thrombospondin motif 1	ADAMTS1	2.02	0.04
A dysintegrin and matrix metallopeptidase with thrombospondin motif 8	ADAMTS8	2.09	0.05
Collagen type 4 α2	COL4A2	37.50	0.02
Collagen type 5 α1	COL5A1	42.15	0.05
Collagen type 6 α1	COL6A1	3.77	0.02
Collagen type 7 α1	COL7A1	3.02	0.05
Versican	VCAN	351.24	0.03
Connective tissue growth factor	CTGF	4.71	0.04
Fibronectin	FN1	398.50	0.05
Integrin α1	ITGA1	2.34	0.02
Integrin α3	ITGA3	6.12	0.01
Integrin α4	ITGA4	5.97	0.06
Integrin αL	ITGAL	7.14	0.01
Integrin β1	ITGB1	2.41	0.01
Integrin β3	ITGB3	23.96	0.01
Integrin β4	ITGB4	3.96	0.02
Laminin α2	LAMA2	6.91	0.03
Laminin γ1	LAMC1	2.38	0.04
Matrix metalloproteinase 10	MMP10	21.51	0.01
Matrix metalloproteinase 11	MMP11	3.43	0.01
Matrix metalloproteinase 14	MMP14	2.61	0.02
Matrix metalloproteinase 9	MMP9	21.58	0.01
Neural cell adhesion molecule 1	NCAM1	2.71	0.05
Sarcoglycan	SGCE	3.31	0.04
Osteonectin	ON	2.28	0.04
Transforming growth factor beta-induced	TGFBI	3.21	0.06
Tissue inhibitor of matrix metalloproteinase 1	TIMP1	1.80	0.04
C-type lectin domain family 3, member B	CLEC3B	4.73	0.01

 Twenty-seven gene products showed significant upregulation (P < 0.05), and one target showed borderline significance (TGFBI, P =  0.06).

To verify the PCR array data and to quantify the expression levels of ECM-related genes, quantitative real-time PCR assays were performed for a subset of genes using specimens from patients with late-onset FCD (n = 10) and unaffected controls (n = 5). Specimens from patients with PBK (n = 6) were analyzed in parallel to exclude nonspecific findings due to endothelial damage. Apart from genes that had shown a significant and consistent differential expression in the PCR array, we included the following groups of target genes in the specific PCR analysis: (1) agrin, collagen type XVI, and fibulin-2, which had previously been identified as specific basement membrane components of the limbal niche ^34,35 ; (2) collagen type VIII, which has been genetically associated with early-onset FCD ^1–4 ; (3) collagen type I, which showed a 13-fold, though nonsignificant upregulation in the PCR array (P = 0.14) and represents an important structural protein of the corneal stroma ³⁶ ; (4) collagen type III, which is known to be involved in corneal wound healing ^36,37 ; and (5) the extracellular chaperone clusterin, which has been shown to be upregulated in FCD and localized in guttae. ^9,29  

This approach confirmed a significant upregulation of collagen type IV α2 (30-fold, P < 0.0001), collagen type V α1 (17-fold, P < 0.0001), collagen type VI α1 (4-fold, P < 0.02), and versican (6-fold, P < 0.0001) in FCD as compared to controls (Table 3). However, these genes were also upregulated to the same or to an even higher extent in PBK specimens (3- to 160-fold, P < 0.001 to P < 0.05) as compared to controls, possibly reflecting a nonspecific endothelial dysfunction. In contrast, agrin (6-fold, P < 0.001), clusterin (3-fold, P = 0.002), collagen type I α1 (18-fold, P = 0.008), collagen type III α1 (13-fold, P = 0.04), collagen type XVI α1 (8-fold, P = 0.03), fibronectin 1 (6-fold, P < 0.001), integrin α4 (4-fold, P < 0.005), and TGFBI (3-fold, P = 0.006) displayed a FCD-specific upregulation as compared to both PBK and control specimens (Table 3; Fig. 1). Collagen type VIII α2 and fibulin-2 displayed no differential expression in FCD, PBK, and control specimens. 

Figure 1

 

Real-time PCR. Quantitative determination of mRNA levels in corneal endothelial specimens from patients with Fuchs' corneal dystrophy (FCD, n = 5), pseudophakic bullous keratopathy (PBK, n = 3), and control specimens (n = 4) using real-time PCR technology. Data were normalized to GAPDH and are expressed as molecules of interest per molecules GAPDH, together with the relative fold change in FCD or PBK as compared with control specimens (expression levels of TGFBI: ×10⁴; others: ×10³; P < 0.05).

Table 3

 

PCR Assay: Differentially Expressed Gene Products in Endothelial Specimens From FCD Patients as Compared to Unaffected Controls (Co) and PBK (PCR Assay)

Gene Product	Fold Change FCD/Co	P Value	Fold Change FCD/PBK	P Value
Agrin	5.6	0.0001	4.2	0.001
Clusterin	3.1	0.002	11.2	0.001
Col 1A1	18.3	0.008	20	0.023
Col 3A1	13.4	0.04	8.4	0.04
Col 4A2	30.3	0.001	1.2	NS
Col 5A1	17.3	0.001	0.2	NS
Col 6A1	4.3	0.002	0.03	0.01
Col 8A2	1.1	NS	1.42	NS
Col 16A1	8.3	0.03	1.9	0.04
Fibronectin	5.9	0.001	23.4	0.003
Fibulin-2	3.7	NS	1	NS
ITGA4	3.7	0.005	6.6	0.033
TGFBI	3	0.006	6.3	0.002
Versican	5.8	0.001	0.29	NS

 FCD-specifically upregulated genes are shown in bold font.

Differences in Gene Expression Between Central and Peripheral Cornea

Because the clinical signs of guttae are often more marked in the central cornea, ²² we performed a refined differential expression analysis of the central (5 mm diameter) and peripheral portions of DM (pooled RNA from 10 specimens, n = 3). Polymerase chain reaction analysis of the target genes for clusterin, TGFBI, collagen type I, collagen type III, fibronectin, agrin, and integrin α4, which had shown a FCD-specific upregulation in the former analyses, revealed no significant differences in expression levels between the central and peripheral regions of DM in late-onset FCD but only a general tendency for elevated expression levels in the center of DM (Fig. 2). 

Figure 2

 

Comparison of gene expression in central and peripheral cornea. Comparison of gene expression in the endothelial cells from the central versus peripheral cornea (pooled RNA from 10 FCD corneas, n = 3): no significant differences in the gene expression between central and peripheral endothelial cells.

Immunolocalization of ECM Proteins

To visualize the distribution of 24 selected ECM proteins (collagen types I, III, IV, V, VI, VII, VIII, XII, XV, XVI, XVII, XVIII; laminin-1; fibronectin; vitronectin; agrin; osteopontin; fibrillin-1; fibulin-1/2; tenascin-C; versican; TGFBI; clusterin) including those specifically upregulated in FCD (Table 3), indirect immunofluorescence was performed using corneal buttons from patients with FCD and PBK and normal donor eyes as controls. 

The DM of normal corneas showed a positive linear staining for collagen types III, IV, VIII, XII, XV, XVIII, laminin-1, fibronectin, agrin, fibrillin-1, and fibulin-1 along the endothelial side and for collagen types IV, VI, VII, XVI, XVII, fibronectin, vitronectin, osteopontin, fibrillin-1, fibulin-2, TGFBI, versican, and clusterin along the stromal side of DM. No immunostaining of DM was found for collagen types I and V as well as tenascin-C. 

In FCD specimens, increased labeling could be observed for collagen types I, III, IV, V, VI, VII, XII, XV, XVI, XVIII, laminin-1, fibronectin, vitronectin, agrin, tenascin-C, fibrillin-1, fibulin-1/2, versican, TGFBI, and clusterin in the PCL of DM, whereas the guttae proper did not reveal any particular staining pattern. In PBK specimens, increased staining was found on the endothelial face of DM (PCL) for collagen types I, III, IV, V, XII, XVIII, clusterin, fibrillin-1, fibronectin, fibulin-1, fibulin-2, laminin-1, osteopontin, tenascin-C, TGFBI, and vitronectin; staining on the stromal face of DM was found for collagen types VII, XVII, and versican; no staining was observed for agrin and collagen types VI, VIII, XV, XVI. Labeling intensity was generally stronger in the central regions than in the peripheral regions of DM. The staining reaction in the central area of DM was considerably more pronounced for collagen types III, VII, XV, XVI, agrin, fibulin-2, TGFBI, versican, and clusterin in FCD specimens compared to PBK specimens (Fig. 3). A patchy immunoreaction for collagen type III was additionally observed within DM proper in FCD specimens only (Fig. 3). 

Figure 3

 

Immunohistochemistry. Immunolocalization of nine ECM components (agrin, clusterin, collagen types III, VII, XV, XVI, fibulin-2, TGFBI, and clusterin) in control, FCD, and PBK corneas that showed an overexpression in FCD compared to PBK.

With the molecular biological and immunohistologic data combined, the ECM components collagen types III and XVI, agrin, TGFBI, and clusterin showed a FCD-specific upregulation in EC on both the mRNA and protein levels. 

Discussion

The findings of this study provide evidence for an increased disease-specific expression of ECM components in late-onset FCD. Alterations in the structure and composition of the ECM within DM in FCD, for example, an increased immunolabeling for collagen type IV, laminin, and fibronectin, have been previously described. ^27,31 In addition, excessive deposition of collagens forming abnormal posterior banded and fibrillar layers has been reported. ³⁰ In both FCD and other pathologic conditions like PBK, a PCL mainly consisting of collagen types IV and VIII and fibronectin is produced and deposited on the posterior surface, leading to thickening of the DM complex. This PCL is thought to be produced by EC that underwent transdifferentiation into myofibroblasts and to be the result of a common final pathway in endothelial dysfunction. ^25,26,38  

However, little is known about the mRNA expression of ECM-related genes in EC of corneas with FCD. Serial analysis of gene expression (SAGE) performed by Gottsch et al. ³⁹ detected a number of differentially expressed genes related to apoptosis and oxidative stress, but did not specifically focus on ECM metabolism, which has been addressed in the present study. 

Using a commercial PCR array for targets of ECM metabolism and specific PCR assays, a wide range of genes could be identified that were consistently upregulated in corneal EC of FCD patients compared to normal corneas. Comparison of FCD specimens to PBK specimens using quantitative PCR assays revealed that only part of these genes, that is, collagen types I, III, and XVI, TGFBI, agrin, fibronectin, clusterin, and integrin α4, were specifically upregulated in FCD. These findings could be largely confirmed by immunohistochemistry showing increased immunoreactivity for agrin, collagen type III, collagen type XVI, TGFBI, and clusterin in FCD corneas compared to PBK and normal corneas. However, it cannot be decided definitely from the present investigations whether the observed matrix alterations are primary or secondary to additional comodulating factors, such as oxidative stress, hypoxia, and EC loss. 

Subgroup analysis of gene expression levels showed no significant differences between central and peripheral parts of the cornea, but a trend for increased expression of target genes in the center of DM. These findings support the clinical observation that guttae formation is usually more pronounced in the center than in the periphery of the cornea. 

Some targets (collagen types IV, V, VI, versican) showed a nonspecific upregulation in EC of specimens from both FCD and PBK patients. These nonspecific findings of our PCR assays might be explained by a common final pathway in endothelial dysfunction. There are possible confounders influencing the results obtained in this study: 

1.  
   Interindividual differences due to different stages of the disease: Due to improved surgical techniques for corneal transplantation, ⁴⁰ patients with FCD undergo corneal transplantation if they are handicapped by blurred vision due to central guttae without corneal edema. In contrast, patients with PBK are not symptomatic in the beginning and thereby do not undergo transplantation before a relevant corneal edema and epithelial bullous keratopathy have developed.
2.  
   Interindividual differences due to different genetic etiologies causing FCD might lead to a differential ECM pattern within the patients with FCD.

Immunohistochemistry confirmed specific upregulation of agrin, collagen type III, collagen type XVI, TGFBI, and clusterin on the protein level in FCD specimens, particularly within the PCL, whereas collagen type I, fibronectin, and integrin α4 showed similar expression patterns in FCD and PBK specimens. Possible explanations for this discrepancy in the expression on mRNA and on protein level might be the following: (1) RNA analysis requires vital cells, whereas deposited ECM components can be detected by immunohistochemistry even if there is no active cell metabolism left at the time of examination; (2) vital EC in the corneal periphery with normal ECM production outbalance the minority of remaining, damaged EC in the central cornea. Therefore, the RNA of vital peripheral EC is overrepresented in the PCR analysis; (3) the heterogeneous nature of late-onset FCD with four known chromosomal loci might result in different expression patterns. 

Fuchs' corneal dystrophy–specific upregulation was found for collagen types III and XVI, TGFBI, agrin, and clusterin on both RNA and protein level. Notably, collagen type III could be immunolocalized not only to the PCL, but also to DM proper and to guttae, indicating that collagen type III is produced during earlier stages of the disease than the other ECM proteins. 

In agreement with our findings, an increased expression of TGFBI and clusterin by the corneal endothelium of FCD patients has been previously reported: Jurkunas et al. ^28,29 provided evidence for an increased expression of clusterin on the mRNA level and for a deposition of clusterin protein within guttae. In contrast, Kuot et al. ⁹ found clusterin staining only in the stromal aspect of DM, not in the posterior nonbanded layer or in guttae, whereas in our specimens the staining was most intense within the PCL. Similar discrepancies for the localization of a DM component have been reported for TGFBI, which was localized on the stromal side of guttae by Jurkunas et al., ²⁸ on the posterior aspect of DM by Kuot et al., ⁹ and within the PCL in our study. The functional role of overexpressed clusterin in FCD might be its association with apoptosis. ^28,29 Furthermore, it mediates aggregation of cells under stress conditions and formation of cell–cell junctions. ⁴¹ Transforming growth factor beta-induced induces cell adhesion by interaction with ECM components, for example, collagens, fibronectin, and integrins. ^9,28  

In addition to the established involvement of clusterin and TGFBI in FCD, the present study provided evidence for an overexpression of collagen types III (13-fold) and XVI (8-fold) as well as agrin (6-fold), which has not been described so far. 

Collagen type III, a component of the corneal stroma, ³⁶ is known to be upregulated in corneal stromal wound healing. ³⁷ Deposition of collagen type III in the PCL has been found in corneas with PBK, ⁴² but not in FCD up to now. Its upregulation in EC in FCD might be an early sign of the fibroblastic transdifferentiation of EC, resulting in a fibrotic reaction triggering a focal “wound healing” associated with upregulation of collagen type III. 

Collagen type XVI belongs to the fibril-associated collagens with interrupted triple helices (FACIT) and is produced by skin fibroblasts as well as intestinal myofibroblasts. ^43,44 It has been described as a specialized ECM component of the corneal limbal stem cell niche. ³⁴ We found an upregulation of collagen type XVI in corneal EC. The functional impact of this overexpression remains unclear so far. In fibrotic skin diseases, collagen type XVI is upregulated, too. ⁴⁵ Therefore, upregulation of collagen type XVI in FCD might be a sign of a fibrotic retrocorneal reaction. Inhibition of collagen type XVI reduces the integrin-mediated adhesion and thereby the invasiveness of cancer cells. ⁴⁶ Hypothetically, the cell–cell interaction and adhesion of EC in FCD might be influenced by the upregulation of collagen type XVI that we found. 

Agrin is a proteoglycan that is well known as a mediator of synaptogenesis and formation of neuromuscular junctions: It plays a crucial role not only in the development of the central nervous system, ⁴⁷ but also in the formation of synapses in the embryonic retina. ⁴⁸ Furthermore, agrin has nonneuronal functions and is expressed by mesenchymal stem cells and osteoblasts in the bone marrow ⁴⁹ and is involved in the activation of lymphocytes via the receptor α-dystroglycan. ⁵⁰ The removal of agrin from the synaptic membranes is mediated by MMP 3, ⁵¹ and agrin is target for degradation by MMP 1, 7, 11, and 14. ⁵² The interaction between agrin and MMPs, which were both upregulated in DM-EC of FCD patients in the present study, might be part of the pathophysiology of FCD. In the human eye, agrin has been identified in basement membranes of the sclera, choroid, and retina, especially in the internal limiting membrane and in retinal blood vessels, ^53,54 and was found to be upregulated in diabetes. ⁵⁵ Since oxidative stress is involved in the pathophysiology of both diabetes ⁵⁶ and FCD, ^13–18 similar pathways might result in the upregulation of agrin in both conditions. Agrin has been also shown to be part of the basement membrane of the limbal epithelium, ^34,35 but its expression by corneal EC or its presence in the PCL of DM has not been reported before. 

At present, the functional impact of increased deposition of collagen type III, collagen type XVI, and agrin on corneal EC can only be speculated about and may involve effects on cell proliferation, cell migration, and cell function. In the central nervous system, an inhibition of α3-Na⁺-K⁺-ATPase (sodium, potassium, and adenosine triphosphatase) by agrin has been reported. ^57,58 Na⁺-K⁺-ATPase is abundant in the corneal endothelium and necessary to maintain its pump function. Since reduced pump function of the corneal endothelium is a hallmark of FCD, increased expression and deposition of agrin might be associated with this alteration. On the other hand, upregulation of agrin may be an attempt to compensate for reduced endothelial pump function and the resulting corneal edema, since the long glycosaminoglycan side chains of agrin have a high water-binding capacity. ⁵⁵  

Despite the current lack of any functional data, the present study provides evidence for a FCD-specific upregulation of agrin, collagen types III and XVI, clusterin, and TGFBI in corneal EC in late-onset FCD, which supports the notion of an involvement of ECM alterations in the pathophysiology of FCD. On the basis of the findings described, further studies are required to elucidate the functional effect of these ECM components on corneal EC, their signaling pathways, and their interactions with other pathogenetic factors such as oxidative stress using appropriate in vitro models. 

Acknowledgments

The authors thank Elke Meyer and Angelika Mößner for technical assistance. 

Disclosure: J.M. Weller, None; M. Zenkel, None; U. Schlötzer-Schrehardt, None; B.O. Bachmann, None; T. Tourtas, None; F.E. Kruse, None