Targeting RhoA/ROCK Signaling to Modulate Extracellular Matrix Remodeling in Corneal Endothelial Dystrophies

Abstract

Purpose

Fuchs endothelial corneal dystrophy (FECD) involves pathological extracellular matrix (ECM) accumulation within Descemet's membrane, leading to guttae formation, endothelial dysfunction, and vision impairment. As current treatment is primarily surgical, we investigated whether Rho-associated protein kinase (ROCK) inhibition can reduce fibrotic ECM remodeling in FECD and related corneal endothelial diseases.

Methods

Endothelial cell–Descemet membrane specimens from FECD, pseudophakous bullous keratopathy (PBK), pseudoexfoliation keratopathy (PEX-K), and normal donor corneas were treated ex vivo with ripasudil or comparator ROCK inhibitors (netarsudil, Y-27632). Complementary in vitro studies used human corneal endothelial cells. ROCK activity, ECM expression, matrix metalloproteinase (MMP) activity, and TGF-β/Smad signaling were analyzed using molecular and imaging techniques.

Results

Relative to controls, FECD samples showed constitutive ROCK and TGF-β signaling activation. Ripasudil suppressed ROCK activity and downregulated fibrosis-associated ECM components (including collagens I/III, fibronectin, agrin, TGFBI, clusterin, and tenascin-C), while shifting the MMP–tissue inhibitor of metalloproteinases (TIMP) balance to enhance ECM turnover. These effects were accompanied by reduced ECM deposition and attenuation of TGF-β/Smad signaling, alongside broad transcriptional reprogramming related to fibrosis, inflammation, and cytoskeletal dynamics. Similar ECM-modulating effects were observed in PBK and PEX-K, with ripasudil showing the most pronounced effects among tested inhibitors.

Conclusions

ROCK activation contributes to pathological ECM alterations in FECD, and its pharmacologic inhibition with ripasudil suppresses pro-fibrotic ECM production, promotes ECM-remodeling pathways, and supports restoration of ECM homeostasis. These findings suggest ROCK inhibition as a promising non-surgical therapeutic strategy for FECD and related corneal endothelial disorders, warranting further validation in preclinical studies.

The corneal endothelium is a highly specialized, non-regenerative monolayer that maintains corneal transparency by regulating stromal hydration.¹ Corneal endothelial cells continuously secrete a specialized basement membrane, Descemet's membrane (DM), which provides structural support and biochemical cues essential for preserving endothelial phenotype and function.² Under physiological conditions, DM exhibits a tightly regulated composition of collagens, laminins, and proteoglycans, reflecting a finely balanced extracellular matrix (ECM) homeostasis required for endothelial function and optical clarity.^3,4 Disruption of this balance is a common feature of corneal endothelial disorders and is increasingly recognized as a key determinant of disease progression.⁵ In response to cellular stress, injury, or aging, endothelial cells may alter their secretory profile, leading to excessive or aberrant ECM deposition that compromises the functional properties of DM and ultimately results in endothelial dysfunction, corneal edema, and vision loss.^4,5 ECM dysregulation is therefore increasingly recognized as a shared feature of corneal endothelial disorders such as Fuchs endothelial corneal dystrophy (FECD), pseudophakous bullous keratopathy (PBK), and pseudoexfoliation keratopathy (PEX-K).^2,4,6

FECD is the most common corneal endothelial dystrophy and the leading indication for corneal endothelial transplantation worldwide.⁷ The disease is characterized by progressive thickening of DM and the formation of focal excrescences known as guttae, which represent early and clinically detectable hallmarks of the disorder.⁸ These structural changes induce light scattering, reduce contrast sensitivity, and impair visual quality even before the onset of manifest corneal edema.⁹ At the molecular level, FECD is increasingly recognized as a fibrotic disease of the corneal endothelium associated with oxidative stress, cellular stress, and senescence.^10,11 Endothelial cells in FECD undergo stress-induced phenotypic changes reminiscent of endothelial-to-mesenchymal transition (EndMT), accompanied by upregulation of pro-fibrotic matrix components such as collagens type I and III, fibronectin, agrin, transforming growth factor beta-induced (TGFBI), and other ECM proteins.^12–15 Elevated levels of TGF-β, particularly TGF-β2, in the aqueous humor of FECD patients are thought to be key drivers of this aberrant matrix synthesis.¹⁶ Despite significant advances in surgical techniques, including DM endothelial keratoplasty (DMEK) and Descemetorhexis without endothelial keratoplasty (DWEK), also termed Descemet stripping only (DSO), treatment options for FECD remain limited to surgical intervention.¹⁷ Given the global shortage of donor corneas, pharmacological therapies targeting the underlying fibrotic pathology are urgently needed.

The RhoA/Rho-associated protein kinase (ROCK) signaling pathway is a central regulator of cytoskeletal dynamics and cellular contractility, controlling processes such as adhesion, migration, proliferation, differentiation, and apoptosis.¹⁸ Upon activation, RhoA stimulates ROCK1 and ROCK2, which phosphorylate downstream targets such as myosin light chain, leading to actin cytoskeletal reorganization. Dysregulated RhoA/ROCK signaling has been implicated in fibrotic diseases across multiple organ systems, including lung, kidney, liver, heart, skin, and eye.^19–22 In all fibrotic conditions, ROCK activation promotes myofibroblast differentiation, stress fiber formation, and transcriptional activation of pro-fibrotic ECM genes, often in cooperation with TGF-β signaling pathways.^23,24 Pharmacological inhibition of ROCK has therefore emerged as a promising anti-fibrotic strategy, capable of reducing ECM synthesis and promoting remodeling of established fibrotic changes in experimental models.^20,22 In ophthalmology, ROCK inhibitors have already demonstrated therapeutic potential through both anti-fibrotic and regenerative effects. They have been reported to suppress ECM deposition and tissue stiffening in vitreoretinal diseases and glaucoma^25–29 and to promote corneal endothelial cell adhesion, proliferation, migration, wound healing, and functional recovery in regenerative strategies, including cell-injection therapies and postoperative rescue following DWEK/DSO.^30–35 However, their potential effects on fibrotic ECM remodeling in corneal endothelial disorders remain largely unexplored.

Given that DM thickening and guttae formation represent early pathological events in FECD,⁸ therapeutic strategies targeting abnormal ECM metabolism may represent a promising approach to mitigate pathological ECM accumulation and potentially slow down disease progression. We therefore hypothesized that stress- or TGF-β–induced activation of RhoA/ROCK signaling drives fibrotic ECM alterations in FECD, and that pharmacological ROCK inhibition can attenuate these pathological changes. To address this hypothesis, we investigated the effects of the ROCK inhibitor ripasudil on ECM synthesis, turnover, and associated signaling pathways in corneal endothelial cells and native tissues from normal donors and FECD patients, thereby evaluating ROCK inhibition as a potential anti-fibrotic therapeutic strategy for FECD and related corneal endothelial disorders.

Materials and Methods

Human Tissues and Study Approval

Central endothelial cell–Descemet membrane (EDM) lamellae, 8 mm in diameter, were prepared by manual stripping from FECD patients (n = 659) during DMEK surgery. Mean patient age was 68.7 ± 9.2 years (273 males, 386 females). For comparative analyses, EDM specimens were additionally collected from patients with PBK (n = 45) and PEX-K (n = 44). Mean patient ages were 77.1 ± 8.6 years for PBK and 73.6 ± 7.9 years for PEX-K. The specimens were collected from January 2023 to January 2025 from consecutive patients undergoing DMEK at our institution. Immediately after excision, the EDM scrolls were dissected into halves and incubated for different periods ranging from 15 minutes to 168 hours in 4 mL of CorneaMax storage medium (Eurobio Scientific, Les Ulis, France) without or with a single dose of 30-µM ripasudil hydrochloride dihydrate (K-321; Kowa Company Ltd., Nagoya, Japan) at 37°C. For comparative analyses, EDM specimens were incubated in ROCK inhibitors Y-27632 (30 µM) and netarsudil (3 µM) (Selleck Chemicals, Houston, TX, USA) in CorneaMax for 24 hours. ROCK inhibitor concentrations were based on previously published studies on corneal endothelial cells in vitro (i.e., ripasudil, 10 µM; Y-27632, 10 µM; netarsudil, 1 µM) and adapted for the respective ex vivo applications.^35–37 Initial dose–response analyses identified 30-µM ripasudil as the most effective concentration for modulation of ECM-associated gene expression in native tissue specimens. To allow comparison among different ROCK inhibitors under identical experimental conditions, Y-27632 and netarsudil were used at 30 µM and 3 µM, respectively.

Normal donor corneas (n = 45) unsuitable for transplantation with appropriate research consent were obtained from CorneaGen (Seattle, WA, USA) in Optisol solution (Bausch & Lomb, Rochester, NY, USA) and shipped at 4°C. Mean donor age was 58.4 ± 9.8 years (24 males, 21 females). There was a mean storage time of 9 days (9.3 ± 3.1). EDM complexes were prepared by manual stripping from intact corneas. For the assessment of dose-dependent effects, five EDM complexes were dissected into four equal parts, combined into sets of four pools, and incubated with 0-, 10-, 30-, or 100-µM ripasudil for 24 hours. Following incubation, the paired EDM specimens were subjected to RNA or protein extraction. The number of FECD and control specimens used for each experimental analysis is summarized in Supplementary Table S1.

Handling of donor tissues was in accordance with the tenets of the Declaration of Helsinki for experiments involving human tissues and samples. Ethics approval for this laboratory investigation was obtained from the Institutional Review Board of the Medical Faculty of the University of Erlangen-Nürnberg (No. 138_18B), and informed research consent was obtained from all patients.

Cell Culture

The corneal endothelial cell line HCEC-12 was obtained from DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany) and cultured in Medium 199/Ham's F-12 medium containing 5% fetal calf serum (FCS) and 1% penicillin/streptomycin/amphotericin B (PSA) solution (PAN-Biotech, Aidenbach, Germany). To test the effect of ripasudil, cells were grown on pre-coated (iMatrix-511; Takara Bio, Shiga, Japan) cell culture ware until confluency and incubated with 10-µM ripasudil in serum-deprived medium (0.1% FCS) for 6 to 72 hours. A concentration of 10 µM was used in accordance with previous studies demonstrating robust ROCK inhibition and biological activity under cell culture conditions.^35–37 In additional experiments, cells were stimulated with 5 ng/mL TGF-β1 (R&D Systems, Minneapolis, MN, USA) with or without ripasudil for up to 24 hours.

For transmission electron microscopy, HCEC-12 cells were cultivated without or with 10-µM ripasudil on plastic coverslips for 7 days, fixed in 2.5% glutaraldehyde in 0.1-M phosphate buffer, post-fixed in 2% buffered osmium tetroxide, and embedded in epoxy resin according to standard protocols. Ultrathin horizontal and cross sections were stained with uranyl acetate and lead citrate and were examined with a transmission electron microscope (EM 906E; Carl Zeiss Microscopy, Jena, Germany).

Real-Time PCR

RNA isolation from EDM specimens and cultured cells was performed using the RNeasy Micro Kit or the RNeasy Mini Kit (QIAGEN, Hilden, Germany), respectively, including an on-column DNase digestion step. First-strand cDNA synthesis was performed using 50 ng of RNA from EDM specimens or 1 µg of RNA from cultured cells, 200 U Superscript II reverse transcriptase, and 200 ng random primers (Thermo Fisher Scientific, Waltham, MA, USA) in a 20-µL reaction volume. Quantitative real-time PCR (qRT-PCR) was performed using the CFX Connect thermal cycler and software (Bio-Rad Laboratories, Hercules, CA, USA). PCR reactions were run in duplicate and contained diluted first-strand cDNA, 0.4 µM each of upstream- and downstream-primer, and 1× SsoFast EvaGreen Supermix (Bio-Rad Laboratories), according to manufacturer recommendations. Primer sequences (Eurofins, Anzing, Germany), designed using Primer 3 software, and reaction conditions are given in Supplementary Table S2. Expression levels were normalized to the housekeeping genes GAPDH, HPRT1, and RPLP0.

For pathway-focused gene expression profiling, cDNA generated from FECD–EDM specimens treated with or without ripasudil was analyzed using the Human Extracellular Matrix and Adhesion Molecules PCR Array or the TGF-β/BMP Signaling Pathway PCR Array (QIAGEN), following the manufacturer's protocols. Data were normalized to the housekeeping genes GAPDH, HPRT1, RPLP0, and B2M and analyzed using the RT2 Profiler PCR Array data analysis tool, version 4.0 (QIAGEN).

RNA-Sequencing and Pathway Analysis

For transcriptome-wide analysis, RNA sequencing was performed on pooled FECD–EDM specimens (n = 3, each containing eight specimens) treated with or without 30 µM ripasudil for 48 hours. Pooling was performed to ensure sufficient RNA input for library preparation and sequencing. Pools were generated by random allocation of consecutively collected FECD–EDM specimens. Demographic characteristics were comparable across pools without any significant differences in mean age and sex distribution (Supplementary Table S3). Library preparation and sequencing were conducted at the Core Unit for Next Generation Sequencing (Institute of Human Genetics, Friedrich-Alexander University Erlangen-Nürnberg, Germany) using a HiSeq 2500 platform (Illumina, San Diego, CA, USA). After pooled single-end sequencing, reads were demultiplexed and converted to FASTQ format according to their indices. The demultiplexed reads averaged 42.3 million per sample and were consistently of excellent quality, requiring only minimal trimming, with an average of 99.9% of raw reads retained. Reads were aligned to the GRCh38 reference genome using STAR Aligner 2.7.8a. Uniquely mapped reads aligning to exons of human genes were counted using featureCounts 2.0.1 based on the Ensembl Genes 103 annotation, and gene-level counts were obtained. Differential gene expression analysis was performed using the DESeq2 1.30 R package (R Foundation for Statistical Computing, Vienna, Austria), with an adjusted P < 0.05 considered statistically significant. Samples were analyzed using a paired design. Heatmaps and pathway visualizations were generated using R 4.0.3–based visualization tools.

Pathway enrichment analysis was performed using Ingenuity Pathway Analysis (IPA; QIAGEN). IPA uses the activation z-score algorithm to predict the activation states of pathways, biological functions, and upstream regulators based on differentially expressed gene (DEG) lists. The z-score estimates pathway activation by comparing observed gene expression patterns in the dataset with expected patterns derived from the literature for each annotation. DEGs with |log₂ fold change| > 0.5 and adjusted P < 0.05 were used for downstream analysis. IPA results with overlap log(P value) > 1.3 were considered statistically significant. To identify biological processes significantly over-represented in the DEG list, Gene Ontology (GO) analysis was performed using the open-source web tool InnateDB 5.4.

Western Blot Analysis

Total protein was isolated from FECD–EDM specimens using radioimmunoprecipitation assay (RIPA) buffer (Sigma-Aldrich, St. Louis, MO, USA). Protein concentration was measured using the Micro-BCA Protein Assay kit (Thermo Fisher Scientific). Then, 2 to 9 µg of total protein was separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions (6% dithiothreitol) using Mini-PROTEAN TGX Stain-Free Precast Gels (Bio-Rad Laboratories) and transferred onto nitrocellulose membranes with a semidry blotting unit (Trans-Blot Turbo; Bio-Rad Laboratories). Membranes were blocked with 5% bovine serum albumin (BSA; Sigma-Aldrich), 3% non-fat milk powder (Sigma-Aldrich), or Superblock (Thermo Fisher Scientific) in phosphate-buffered saline (PBS) with 0.1% Tween 20 (PBST; Sigma-Aldrich) for 1 hour and incubated overnight at 4°C in primary antibodies as indicated in Supplementary Table S4. Equal loading of samples was verified with an antibody against GAPDH. In negative control experiments, the primary antibody was replaced by PBS. Immunodetection was performed with horseradish peroxidase–conjugated goat-anti-rabbit (1/10.000; Cell Signaling Technology, Danvers, MA, USA), goat-anti-mouse (1/15.000; BioLegend, San Diego, CA, USA), or donkey-anti-goat (1/10.000; SouthernBiotech, Birmingham, AL, USA) secondary antibodies diluted in 1% BSA/PBST and the Super Signal West Femto ECL kit (Thermo Fisher Scientific). Specific protein bands were quantitatively analyzed with the LAS-3000 chemiluminescence detection system (Fujifilm, Tokyo, Japan) and software (Multi Gauge V3.0; Fujifilm). For normalization of protein expression levels, protein ratios relative to the housekeeping gene GAPDH were calculated; for phosphorylated proteins, protein ratios relative to the unphosphorylated protein and GAPDH were calculated.

Matrix Metalloproteinase Activity Assays

Matrix metalloproteinase (MMP) activity was analyzed in culture supernatants from EDM specimens following ripasudil treatment for 48 to 72 hours. For gelatin zymography, culture supernatants were pooled (n = 5, each containing three specimens), because the amount of active MMPs released by individual specimen halves was insufficient to reliably detect gelatinolytic bands. Supernatants were concentrated using 30-kDa centrifugal filters (Amicon Ultra; Merck Millipore, Burlington, MA, USA) and normalized to total protein content as determined by Bradford assay. Equal protein amounts were separated on SDS–polyacrylamide gels containing 1-mg/mL gelatin under non-reducing conditions. Following electrophoresis, gels were washed in renaturation buffer and incubated overnight at 37°C in developing buffer to allow enzymatic digestion. Gels were subsequently stained with Coomassie Brilliant Blue (Bio-Rad Laboratories), and gelatinolytic activity appeared as clear bands against a dark background. Densitometric analysis of gelatinolytic bands corresponding to pro- and active forms of MMP2 and MMP9 was performed using ImageJ 1.54.

To quantify total MMP activity, a fluorescence-based MMP activity assay (Abcam, Cambridge, UK) was employed according to the manufacturer's instructions. Supernatants from individual EDM specimens (n = 3) were incubated with a quenched fluorogenic peptide substrate, and fluorescence intensity was measured kinetically every 5 minutes over a 2-hour period using a microplate reader. Fluorescence values were normalized to total protein content.

Statistical Analysis

Statistical analyses were performed using Prism 9.5.1 (GraphPad Software, Boston, MA, USA) software. Data are expressed as mean ± standard deviation (SD). Gaussian distributions of datasets were tested using the Kolmogorov–Smirnov or Shapiro–Wilk normality tests. Significance of differences between groups was calculated by Student's t-test, one-sample t-test, or Mann–Whitney U test. P < 0.05 was considered statistically significant.

Results

Constitutive Activation and Pharmacological Inhibition of RhoA/ROCK Signaling in FECD

EDM specimens from FECD patients and normal donor corneas were first analyzed for constitutive expression of key components of the RhoA/ROCK signaling pathway. Quantitative PCR revealed no significant differences in basal mRNA expression levels of RHOA, ROCK1, or ROCK2 between FECD and donor specimens (Supplementary Fig. S1A). Consistently, western blot analysis demonstrated comparable protein levels of RhoA, ROCK1, and ROCK2 in both groups (Supplementary Fig. S1B). In contrast, phosphorylation of myosin light chain 2 (MLC2), a direct downstream target of ROCK activity, was significantly increased in FECD compared with normal specimens, indicating constitutive activation of ROCK signaling in corneal endothelial cells of FECD patients (Fig. 1A).

Figure 1.

Constitutive activation of ROCK signaling in FECD and its inhibition by ripasudil. Representative western blot analyses and quantitative densitometry of phosphorylated myosin light chain 2 (pMLC2) as a readout of ROCK activity in corneal endothelial cells. (A) Comparison of pMLC2 levels in EDM specimens from FECD patients and normal donor corneas (n = 8; unpaired two-tailed Student's t-test). (B) Time-dependent effects of ripasudil (30 µM) on ROCK activity in FECD–EDM specimens after 6, 12, and 24 hours of incubation (n = 8; one-sample t-test). (C) Validation of time-dependent effects of ripasudil (10 µM) in cultured HCEC-12 (n = 8; one-sample t-test). Total MLC2 and GAPDH were used as loading controls. Quantitative data are expressed as pMLC2/MLC2 normalized to GAPDH. Bars represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001.

The effect of the ROCK inhibitor ripasudil on RhoA/ROCK signaling was subsequently examined in FECD specimens only. Although protein expression levels of RhoA, ROCK1, and ROCK2 remained unchanged after 48 and 72 hours of ripasudil treatment (Supplementary Fig. S1C), MLC2 phosphorylation was significantly reduced after 6, 12, and 24 hours, demonstrating potent inhibition of ROCK signaling by ripasudil (Fig. 1B). This inhibitory effect on MLC2 phosphorylation was further confirmed in HCEC-12 cells in vitro (Fig. 1C).

ROCK Inhibition Suppresses Pathological ECM Gene and Protein Expression

To determine an effective ripasudil concentration for ECM modulation, dose–response experiments (10, 30, and 100 µM) were initially performed using EDM specimens from normal donor corneas (n = 5). These specimens were analyzed for expression of ECM genes previously shown to be specifically upregulated in FECD, including AGRN (agrin), COL1A1 (collagen type I, alpha 1), COL3A1 (collagen type III, alpha 1), FN1 (fibronectin 1), TGFBI (transforming growth factor beta–induced), and CLU (clusterin).¹³ Cytoskeleton-related genes ACTA2 (actin alpha 2, smooth muscle) and ACTB (actin beta) served as positive control genes.³⁵ Although all ripasudil concentrations resulted in significant downregulation of ECM gene expression, treatment with 30-µM ripasudil produced the most pronounced suppressive effects (Supplementary Fig. S2A). Consequently, 30-µM ripasudil was selected for subsequent experiments using EDM specimens, whereas 10-µM ripasudil was applied in cell culture studies.

Expression analysis of FECD-associated ECM genes in a larger cohort of EDM specimens from normal donor corneas (n = 20) and FECD patients (n = 50) revealed significant downregulation of AGRN, COL1A1, COL3A1, FN1, TGFBI, and CLU, together with the positive control genes, following treatment with a single dose of ripasudil (30 µM) for 24 to 72 hours (Fig. 2A). Consistently, ripasudil-induced downregulation of agrin, collagen type I, and fibronectin in EDM specimens from FECD patients was confirmed at the protein level by western blot analysis (n = 8) after 48 and 72 hours of incubation (Fig. 2B).

Figure 2.

Ripasudil suppresses fibrotic extracellular matrix gene and protein expression in FECD. (A) The qRT-PCR analysis of ECM-related gene expression in EDM specimens from normal donor corneas (n = 20, top) and FECD patients (n = 50, bottom) following treatment with ripasudil (30 µM) for 24, 48, and 72 hours. Significant downregulation of fibrosis-associated genes, including AGRN, COL1A1, COL3A1, FN1, TGFBI, and CLU, as well as cytoskeleton-associated control genes (ACTA2 and ACTB), was observed compared with untreated controls (unpaired two-tailed Mann–Whitney U test). (B) Representative western blot analyses and corresponding densitometric quantification of agrin (top), collagen type I (COL1A1, COL2A1), and fibronectin 1 (bottom) protein expression in FECD–EDM specimens after ripasudil treatment for 48 and 72 hours (n = 8; one-sample t-test). GAPDH was used as loading control. (C) The qRT-PCR analysis of additional selected ECM-related genes, including BGN, EMILIN1, FBLN4, FBN1, HSPG2, LTBP1, LTBP2, SPARC, SPP1, TNC, VCAN, and VTN, in FECD–EDM specimens following treatment with ripasudil (30 µM) for 24 hours (n = 23; unpaired two-tailed Mann–Whitney U test). Data are presented as relative mRNA or protein expression normalized to GAPDH. Bars represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001.

To identify additional ECM-associated genes regulated by ROCK signaling, a Human Extracellular Matrix and Adhesion Molecules PCR Array comprised of 84 genes involved in matrix metabolism was applied to FECD specimens (n = 4) incubated with or without ripasudil (30 µM) for 24 hours. This broader analysis confirmed significant downregulation of FN1 and COL1A1 and revealed additional ripasudil-induced expression changes showing ≥1.5-fold up- or downregulation in COL4A2 (collagen type IV, alpha 2), COL5A1 (collagen type V, alpha 1), COL6A2 (collagen type VI, alpha 2), SPP1 (osteopontin), THBS3 (thrombospondin 3), TNC (tenascin), VCAN (versican), and VTN (vitronectin) (Supplementary Table S5). Moreover, the array indicated regulation of various cell adhesion molecules, including integrins, and proteolytic enzymes, including MMPs, as further targets of ripasudil action.

Based on these findings, ECM-related genes showing ripasudil-induced expression changes in the PCR array, together with additional candidate genes previously shown to be associated with FECD,^15,38–40 were subsequently analyzed in a larger sample cohort using targeted PCR assays. Analysis of candidate genes encoding different collagen types confirmed significant downregulation of COL1A1 and COL3A1, which are established biomarkers of fibrosis, whereas expression of COL4A1, COL5A1, COL6A1, COL7A1, COL8A1, COL8A2, and COL16A1 was not affected by ripasudil treatment (Supplementary Fig. S2B). However, examination of selected glycoproteins and proteoglycans further revealed significant downregulation of FBN1 (fibrillin-1), LTBP1 (latent TGF-β binding protein 1), LTBP2 (latent TGF-β binding protein 2), SPARC (osteonectin), TNC, and VCAN, which all have been implicated in fibrotic tissues undergoing pathological ECM remodeling.⁴¹ In contrast, expression levels of other ECM-related candidate genes, including BGN (biglycan), EMILIN1 (emilin-1), FBLN4 (fibulin-4), HSPG2 (perlecan), SPP1, and VTN, remained unchanged following ripasudil treatment (Fig. 2C). Notably, the significant expression changes in FBN1, LTBP1, LTBP2, SPARC, TNC, and VCAN were largely maintained up to 48 and 72 hours of ripasudil incubation (Supplementary Fig. S2C).

ROCK Inhibition Modulates MMP Expression and Activity in FECD Endothelial Cells

Expression levels of MMPs and their tissue inhibitors (TIMPs), identified as potential targets of ripasudil in the array analysis, were further examined in FECD specimens. ROCK inhibition by ripasudil resulted in significant upregulation of MMP1, MMP3, MMP10, and MMP12 after 24, 48, and 72 hours of incubation (Fig. 3A). In contrast, MMP9 expression was transiently reduced at 24 hours only, whereas MMP2 and MMP14 levels remained unchanged; MMP7 and MMP8 expression was below the detection limit. TIMP expression was largely unaffected by ripasudil treatment, with the exception of a significant and sustained downregulation of TIMP1 across all time points (Fig. 3A). Collectively, these changes indicate that ROCK inhibition shifts the tightly regulated MMP–TIMP balance toward a predominance of matrix-degrading MMP activity.

Figure 3.

Ripasudil modulates MMP expression and and activity in FECD. (A) The qRT-PCR analysis of MMPs and TIMPs in EDM specimens from FECD patients following treatment with ripasudil (30 µM) for 24, 48, and 72 hours. Ripasudil induced significant upregulation of MMP1, MMP3, MMP10, and MMP12, but expression of TIMP1 was significantly reduced (n = 23; unpaired two-tailed Mann–Whitney U test). (B) Representative gelatin zymography of supernatants from FECD–EDM specimens incubated with or without ripasudil (30 µM) for 72 hours. Gelatinolytic activity corresponding to pro- and active forms of MMP2 (∼65/60 kDa) and MMP9 (∼97/80 kDa) is indicated by arrows. Ripasudil-treated samples showed a trend toward increased gelatinolytic activity compared with controls (n = 5; one-sample t-test). Molecular weight markers are indicated on the left. (C) Quantitative analysis of total MMP activity using a fluorescence-based assay. Supernatants from FECD specimens demonstrated a time-dependent increase in fluorescence intensity with a trend toward higher overall MMP activity in ripasudil-treated samples compared with untreated controls (n = 3). Data are presented as relative mRNA expression or fluorescence intensity (mean ± SD). *P < 0.05, **P < 0.01, ***P < 0.001.

In addition to mRNA expression analyses, enzymatic activity of MMPs was assessed in supernatants from EDM specimens obtained from FECD patients using substrate gel electrophoresis (zymography). Pooled EDM specimens (n = 5, each containing three specimens) were incubated in the absence or presence of a single dose of ripasudil (30 µM) for 72 hours. Supernatants were subsequently concentrated using 30-kDa Amicon filters, and protein concentrations were normalized to total protein content as determined by Bradford assay. Gelatin zymography revealed a predominant band of gelatinolytic activity at approximately 65 kDa, consistent with pro-MMP2. Additional bands corresponding to activated MMP2 (∼60 kDa), pro-MMP9 (∼97 kDa), and activated MMP9 (∼80 kDa) were also detected (Fig. 3B). Densitometric analysis demonstrated a modest increase in gelatinolytic activity of both pro- and active forms of MMP2 and MMP9 in ripasudil-treated samples compared with controls, reaching statistical significance for MMP9 only. In contrast, zymography using casein- and collagen I–containing gels to detect caseinolytic (MMP3, MMP7, MMP10) and collagenolytic MMPs (MMP1, MMP8, MMP13)⁴² did not reveal detectable enzymatic activity (data not shown).

In addition, total MMP activity in supernatants from EDM specimens was quantified using a fluorescence-based MMP activity assay with a quenched fluorogenic peptide substrate. In its intact state, the substrate exhibits minimal fluorescence; cleavage by active MMPs releases the fluorophore from quenching, resulting in a measurable increase in fluorescence intensity proportional to overall MMP activity. Individual EDM specimens obtained from FECD patients (n = 3) were incubated in the absence or presence of a single dose of ripasudil (30 µM) for 48 hours. Supernatants were subsequently collected and incubated with the fluorogenic substrate, and fluorescence was monitored kinetically at 5-minute intervals over a 2-hour period. Total MMP activity increased over time in all samples, with a trend toward higher activity in ripasudil-treated specimens compared with untreated controls; however, this difference did not reach statistical significance (Fig. 3C). Collectively, these findings suggest that ripasudil-induced upregulation of MMP expression and activity shifts the MMP–TIMP balance toward increased ECM turnover and matrix-remodeling capacity of endothelial cells.

Sustained Effects of ROCK Inhibition on ECM Remodeling

The long-term effects of a single dose of ripasudil (30 µM) on the expression of selected ECM-related genes (AGRN, COL1A1, COL3A1, FN1, TGFBI, CLU, FBN1, SPARC, TNC, VCAN, MMP1, MMP3, MMP10, MMP12, TIMP1) were analyzed in FECD specimens (n = 23) for up to 7 days. Suppression of all examined matrix genes, with the exception of VCAN, was sustained for at least 4 days and, in some cases (TGFBI and TNC), throughout the entire observation period (Fig. 4A). In parallel, upregulation of MMP1, MMP3, and MMP10 persisted for 4 to 5 days, whereas suppression of TIMP1 expression was maintained for up to 6 days.

Figure 4.

Sustained effects of ripasudil on extracellular matrix gene expression and ultrastructural matrix deposition in FECD. (A) Time-course analysis of ECM-related gene expression in EDM specimens from FECD patients following a single dose of ripasudil (30 µM) over a period of up to 7 days. Significant and sustained downregulation of fibrosis-associated genes, including AGRN, COL1A1, COL3A1, FN1, TGFBI, CLU, FBN1, SPARC, TNC, and VCAN, was observed. In parallel, expression of matrix metalloproteinases (MMP1, MMP3, MMP10, MMP12) remained elevated, whereas TIMP1 expression was persistently reduced (n = 23; one-sample t-test). (B) Transmission electron microscopy of HCEC-12 cultured without (top) or with (bottom) ripasudil (10 µM) for 8 days. Asterisks indicate amorphous, basement membrane–like material, and arrows highlight bundles of microfibrils within the intercellular space. Data are presented as relative mRNA expression normalized to control samples (mean ± SD). *P < 0.05, **P < 0.01, ***P < 0.001.

The long-term effects of ripasudil on matrix deposition were further examined by transmission electron microscopy of HCEC-12 cells cultured in the absence or presence of a single dose of ripasudil (10 µM) for 8 days. Tangential and cross sections of untreated control cells revealed abundant deposition of amorphous, basement membrane-like material and prominent bundles of microfibrils within the intercellular spaces (Fig. 4B, top). In contrast, ripasudil-treated cells exhibited a marked reduction in ECM deposition, with only occasional small bundles of microfibrils detectable between cells (Fig. 4B, bottom). These findings demonstrate that a single dose of ripasudil induces sustained suppression of fibrotic ECM expression and deposition, indicating durable remodeling of pathological ECM accumulation in corneal endothelial cells.

Ripasudil Modulates ECM Remodeling Via TGF-β/Smad Signaling

We hypothesized that the suppressive effects of ROCK inhibition on ECM gene expression in endothelial cells are mediated, at least in part, through modulation of TGF-β signaling. To address this, EDM specimens from FECD patients and normal donor corneas were analyzed for expression of key components of the profibrotic TGF-β pathway by qPCR, including TGFB1 (transforming growth factor beta 1), TGFB2, TGFB3, TGFBR1 (transforming growth factor beta receptor 1), and TGFBR2. Basal mRNA expression levels of TGF-β isoforms and their receptors did not differ between FECD and donor specimens (Supplementary Fig. S2D). In contrast, phosphorylation of Smad2, a central mediator of TGF-β signaling, was significantly increased in FECD specimens, indicating enhanced pathway activation in corneal endothelial cells of FECD patients (Fig. 5A). Although ripasudil did not alter expression of TGF-β ligands or receptors (Supplementary Fig. S2D), it markedly affected TGF-β signaling activity. Treatment of FECD specimens with ripasudil (30 µM) significantly suppressed TGF-β1–induced Smad2 phosphorylation at 15 minutes, 1 hour, and 6 hours, demonstrating potent inhibition of TGF-β signaling downstream of receptor activation (Fig. 5B).

Figure 5.

Ripasudil attenuates TGF-β/Smad signaling and modulates TGF-β/BMP-associated gene expression in FECD. (A) Representative western blot analysis and quantitative densitometry of phosphorylated Smad2 (pSmad2) in EDM specimens from normal donor corneas and FECD patients, demonstrating increased Smad2 phosphorylation in FECD (n = 8; unpaired two-tailed Student's t-test). (B) Time-dependent effects of ripasudil (30 µM) on Smad2 phosphorylation in FECD–EDM specimens. Ripasudil treatment resulted in a significant reduction of pSmad2 levels at early time points (15 minutes, 1 hour, and 6 hours) compared with untreated controls (n = 5; one-sample t-test). (C) The qRT-PCR analysis of selected TGF-β/BMP pathway–associated genes (ATF4, BMP2, BMP4, CDKN2B, DCN, FOS, GADD45B, IFRD1, IGFBP3, IL6, INHA, JUN, PDGFB, PLAU, RUNX1, and SERPINE1) in FECD–EDM specimens treated with 30-µM ripasudil (n = 23; unpaired two-tailed Mann–Whitney U test). Data are presented as relative protein or mRNA expression normalized to control samples (mean ± SD). *P < 0.05, **P < 0.01, ***P < 0.001.

To further characterize the broader impact of ROCK inhibition on TGF-β/bone morphogenetic protein (BMP)-related pathways, we analyzed gene expression profiles in FECD specimens treated with or without ripasudil (30 µM) using a TGF-β/BMP Signaling Pathway PCR Array (n = 4). Ripasudil induced a broad downregulation of stress-responsive transcription factors (ATF4, FOS, JUN), cell-cycle and injury-associated mediators (CDKN2B, GADD45B, PLAU, SERPINE1), inflammatory signaling components (IL6, IFRD1), and multiple pro-fibrotic or mesenchymal-associated TGF/BMP ligands and regulators (BMP2, BMP4, BMP6, BMP7, BMPER, LTBP1, NODAL, RUNX1), suggesting attenuation of stress, inflammatory, and fibrotic signaling (Supplementary Table S6). In contrast, selective upregulation of DCN, INHA, IGFBP3, GDF3, and PDGFB suggested increased expression of TGF-β inhibitory components and growth factor–associated transcripts. Differential expression of selected genes was subsequently validated by targeted qPCR analyses in a larger cohort of FECD specimens (n = 23). These analyses confirmed significant downregulation of ATF4 (activating transcription factor 4), BMP2 (bone morphogenetic protein 2), CDKN2B (cyclin-dependent kinase inhibitor 2B), GADD45B (growth arrest and DNA damage–inducible beta), IFRD1 (interferon-related developmental regulator 1), IL6 (interleukin 6), and RUNX1 (RUNX family transcription factor 1), together with significant upregulation of PDGFB (platelet-derived growth factor beta), a growth factor implicated in corneal endothelial cell proliferation (Fig. 5C). Together, these findings demonstrate that ROCK inhibition by ripasudil suppresses profibrotic ECM regulation in FECD primarily through attenuation of TGF-β/Smad signaling and downstream transcriptional networks.

Validation and Generalization of ROCK-Dependent ECM Regulation

To validate the findings obtained from ex vivo EDM specimens, expression of ECM-related genes was additionally analyzed in HCEC-12 cells cultured in serum-depleted medium and treated with ripasudil (10 µM) for 24 to 72 hours. In vitro, ripasudil treatment recapitulated the ex vivo effects, resulting in significant downregulation of AGRN, COL1A1, COL3A1, FN1, TGFBI, CLU, FBN1, SPARC, TNC, VCAN, ACTA2, and ACTB, together with significant upregulation of MMP1 and MMP3 at all time points examined (Fig. 6A). In addition, ripasudil significantly counteracted TGF-β1-induced upregulation of COL1A1, COL3A1, FN1, TGFBI, CLU, FBN1, SPARC, TNC, and VCAN in HCEC-12 cells after 24 hours, whereas its effects on TGF-β1–regulated expression of AGRN and MMP1 were not statistically significant (Fig. 6B).

Figure 6.

Validation and generalization of ROCK-dependent ECM regulation. (A) The qRT-PCR analysis of ECM-related gene expression in HCEC-12 treated with ripasudil (10 µM) for 24, 48, and 72 hours. Ripasudil induced significant downregulation of fibrosis-associated genes, including AGRN, COL1A1, COL3A1, FN1, TGFBI, CLU, FBN1, SPARC, TNC, and VCAN, as well as cytoskeleton-associated genes (ACTA2 and ACTB), while increasing expression of matrix metalloproteinases (MMP1 and MMP3) (n = 4; unpaired two-tailed Student's t-test). (B) Effects of ripasudil on TGF-β1-induced ECM gene expression in HCEC-12 cells. Stimulation with TGF-β1 (5 ng/mL) significantly increased expression of profibrotic ECM genes, which was markedly attenuated by co-treatment with ripasudil (n = 4; unpaired two-tailed Student's t-test). (C) Comparative analysis of different ROCK inhibitors (ripasudil, netarsudil, and Y-27632) on expression of major ECM genes (AGRN, COL1A1, COL3A1, FN1, ACTA2, and ACTB) in FECD-derived EDM specimens (n = 18; one-sample t-test). (D) Effects of ripasudil on ECM-related gene expression in EDM specimens from patients with PBK and PEX-K (n = 15; unpaired two-tailed Mann–Whitney U test). Data are presented as relative mRNA expression normalized to control samples (mean ± SD). *P < 0.05, **P < 0.01, ***P < 0.001; #P < 0.05, ##P < 0.01, ###P < 0.001 (for comparisons between indicated groups).

The effects of ripasudil (30 µM, 24 hours) on expression of major FECD-associated ECM genes were further compared with those of other ROCK inhibitors (i.e., Y-27632, 30 µM; netarsudil, 3 µM) in FECD–EDM specimens (n = 18) ex vivo. Whereas all three ROCK inhibitors induced significant downregulation of ACTB, they displayed divergent transcriptional effects on ECM-associated genes, with ripasudil showing the most consistent and pronounced suppression of AGRN, COL1A1, COL3A1, FN1, and ACTA2 expression (Fig. 6C).

To assess whether the ECM-modulating effects of ripasudil extend beyond FECD, EDM specimens from patients with PBK (n = 15) and PEX-K (n = 15) were analyzed. Treatment with ripasudil (30 µM) for 24 hours resulted in consistent downregulation of all examined ECM-related genes and significant upregulation of MMP1 and MMP3 in both disease entities (Fig. 6D). These findings indicate that the observed anti-fibrotic and ECM-remodeling effects represent an inherent property of ROCK inhibition and may therefore be relevant across a broader spectrum of corneal endothelial pathologies.

Transcriptome-Wide Reprogramming of Matrix and Signaling Networks by ROCK Inhibition

To identify novel ECM-related genes regulated by ripasudil in an unbiased manner, bulk RNA sequencing was performed on pooled FECD specimens (n = 3; 67.1 ± 11.1 years; 12 males, 12 females) treated with or without ripasudil (30 µM) for 48 hours. Demographic characteristics were comparable across pools without any significant differences in mean age and sex (pool A: 65.3 ± 11.2 years, four males and four females; pool B: 72.8 ± 6.6 years, three males and five females; pool C: 63.0 ± 12.1 years, five males and three females) (Supplementary Table S3). Principal component analysis (PCA) revealed that PC1 and PC2 explained 45% and 32% of the total variance, respectively (Supplementary Fig. S3A). Control and ripasudil-treated EDM specimens showed clear separation, mainly along PC2, indicating treatment-associated differences in transcriptional profiles.

Differential expression analysis identified 592 significantly regulated genes (adjusted P < 0.05), including 82 genes with a ≥2-fold change (|log₂FC| ≥ 1) and 379 genes with a ≥1.4-fold change (|log₂FC| ≥ 0.5) (Fig. 7A). Heatmap visualization of these 82 candidate genes revealed clear separation between control and ripasudil-treated samples (Supplementary Fig. S3B). Ripasudil suppressed expression of genes involved in ECM remodeling (FN1, ODAPH, MMP9, MMP19, PLAT, PLAU) and inflammatory signaling (CCL7, CXCL5, CXCL3, IL32, IL18, IL1R1, NLRP10) while inducing genes associated with negative regulation of TGF-β and Wnt signaling (e.g., INHBA, KLF15, NKD2, SFRP4, DKK4) and ECM turnover (FNDC1, MMP3, ADAMTS9-AS1, EMIP).

Figure 7.

Transcriptome-wide effects of ripasudil on ECM regulation and signaling pathways in FECD. (A) Volcano plot showing differential gene expression in FECD–EDM specimens treated with ripasudil (30 µM) compared with untreated controls (n = 3). Significantly upregulated genes are shown in red and downregulated genes in blue. Selected differentially expressed genes are highlighted. (B) Validation of selected ECM-related genes (ARHGAP31, DKK4, ECM1, FNDC1, FNDC4, INHBA, LAMC2, MATN3, NKD2, NTN4, ODAPH, SERPINB1, SFRP4, TCF4, and TGM2) by qRT-PCR (n = 23; unpaired two-tailed Mann–Whitney U test). (C) Pathway enrichment analysis (IPA) of differentially expressed genes. Ripasudil modulated pathways related to wound healing and fibrosis, matrix metalloproteinase regulation, cytoskeletal organization, and Wnt/β-catenin signaling. Bars represent significance as −log(P-value), with predicted activation states indicated by color coding. Data are presented as relative mRNA expression normalized to control samples (mean ± SD). *P < 0.05, **P < 0.01, ***P < 0.001.

Focused analysis of a curated set of 53 ECM-related genes confirmed consistent downregulation of established ECM components, including SPARC, FBN1, FN1, and LTBP1, as well as a number of newly identified ECM genes such as ECM1 (extracellular matrix protein 1), MATN3 (matrilin-3), NTN4 (netrin-4), LAMC2 (laminin subunit gamma 2), FNDC4 (fibronectin type III domain containing 4), and ODAPH (odontogenesis associated phosphoprotein), together with TGM2 encoding the crosslinking enzyme transglutaminase 2 (Supplementary Fig. S4A). Expression of genes encoding collagens or proteolytic enzymes showed mixed regulation. Notably, the FECD-associated transcription factor TCF4 was upregulated following ripasudil treatment. DEGs were visualized according to their functional categories, revealing broad suppression of ECM genes, regulation of matrix-degrading enzymes, and involvement of multiple signaling pathways (Wnt, NF-κB, TGF-β, retinoic acid, and Hedgehog signaling) in response to ripasudil (Supplementary Fig. S4B).

Validation by targeted qPCR in an independent cohort of FECD specimens (n = 23) confirmed significant downregulation of ARHGAP31 (Rho GTPase activating protein 31), FNDC4, and ODAPH, along with upregulation of INHBA and FNDC1, whereas changes in ECM1 and TCF4 expression could not be independently confirmed (Fig. 7B). TCF4 was included in the analysis because of its established role in FECD pathogenesis⁸; however, its expression was not significantly altered by ROCK inhibition, suggesting that the observed ECM-modulating effects are unlikely to be mediated through changes in TCF4 expression.

Subsequent pathway analysis using IPA was performed on the 379 DEGs (adjusted P < 0.05 and |log₂FC| ≥ 0.5). This analysis identified 129 significantly enriched canonical pathways (P < 0.05) between control and ripasudil-treated samples. Ten pathways were selected for presentation based on biological relevance (Fig. 7C). Based on IPA activation z-score predictions, pathways related to wound healing and fibrosis signaling were predominantly activated (red), whereas pathways associated with matrix metalloprotease inhibition, Wnt signaling, and cytoskeletal organization and adhesion were predicted to be inhibited (blue). Some pathways did not reach the threshold for directional activation state prediction. Collectively, these results suggest that ripasudil modulates gene networks involved in tissue repair, fibrotic processes, and cytoskeletal dynamics. GO analysis revealed that extracellular matrix organization, extracellular matrix disassembly, and collagen catabolic process were among the most significantly enriched biological processes, further supporting a role for ripasudil in promoting extracellular matrix turnover in corneal endothelial cells (Supplementary Fig. S4C).

Detailed RNA-sequencing results, including sequencing quality metrics, list of DEGs, and the complete GO and IPA outputs, are provided in Supplementary Table S7. Overall, transcriptome-wide analysis revealed that ripasudil induces a coordinated reprogramming of corneal endothelial cells characterized by suppression of fibrotic ECM gene networks, activation of matrix-degradative pathways, and modulation of multiple signaling cascades consistent with enhanced ECM turnover.

Discussion

Progressive structural alterations of DM are hallmark pathological features of FECD and other corneal endothelial disorders and have been associated with EndMT-like phenotypic changes that promote aberrant ECM production.^5,8,14 Increasing evidence indicates that these changes are not merely a consequence of endothelial degeneration but reflect an active fibrotic response of stressed and senescent endothelial cells.^10,11 Consistent with this concept, gene expression analyses demonstrated enrichment of pathways related to inflammation, oxidative and metabolic stress, EndMT, and fibrosis,^43–45 suggesting that ECM alterations in FECD are embedded within a broader self-perpetuating stress–fibrosis network. Supporting this view, transcriptomic and proteomic analyses have revealed abnormal accumulation of ECM components in FECD–EDM tissues, including fibrillar and structural collagens (e.g., collagens type I, III, VI, and VIII), glycoproteins (e.g., fibronectin, laminin, matrilin), proteoglycans (e.g., agrin, biglycan, versican), and matricellular proteins such as TGFBI, tenascin-C, and clusterin.^{12,13,15,38–40,46} These alterations contribute to DM thickening and guttae formation, modify its biochemical and biomechanical properties, and disrupt endothelial cell–matrix interactions, ultimately promoting endothelial dysfunction and apoptosis.^47–49 Recent omics and genetic studies further place ECM dysregulation at the center of FECD pathogenesis, alongside oxidative stress, mitochondrial dysfunction, and immune activation, including the identification of variants in laminin- and collagen-related genes.^40,50,51

The functional relationship between endothelial cells and DM has been conceptualized as a bidirectional “soil and seed” interaction, in which stressed endothelial cells produce abnormal ECM, and the altered structural, biochemical, and biomechanical properties of diseased DM further impair endothelial cell function and survival.^4,8 Beyond FECD, comparable mechanisms of pathological ECM accumulation have been described in other corneal endothelial disorders, including PBK and PEX-K, where progressive DM thickening is frequently accompanied by formation of a posterior fibrillar collagenous layer, indicative of fibrotic processes as a common downstream pathway of endothelial stress, injury, and dysfunction.^6,52–54 Together, these observations suggest that effective regenerative approaches for corneal endothelial disorders must target not only endothelial cell dysfunction but also the pathological ECM environment.⁴ In this context, RhoA/ROCK signaling may represent a targetable regulatory pathway linking cellular stress, cytoskeletal remodeling, and pathological ECM responses in corneal endothelial disease.³⁴

RhoA/ROCK signaling is a central regulator of fibrosis and has been implicated in pathological ECM remodeling in multiple organs.^19–22 A major mechanism underlying these effects is cross-talk with TGF-β signaling, whereby ROCK activity promotes profibrotic gene expression through both cytoskeletal tension-dependent and Smad-dependent mechanisms.^23,55 Accordingly, pharmacological ROCK inhibition has emerged as an anti-fibrotic strategy capable of reducing ECM accumulation and attenuating fibrotic tissue remodeling in a variety of experimental models.^20,22,56 However, fibrotic remodeling reflects not only increased ECM synthesis but also impaired ECM degradation. Accordingly, dysregulation of the MMP–TIMP balance is a common mechanism of progressive fibrosis across tissues, and restoration of this balance is increasingly recognized as a prerequisite for fibrosis regression.^57–59 In FECD, alterations in MMP and TIMP expression or activity have been described, including reduced levels of specific MMPs involved in ECM degradation and an overall shift toward a decreased MMP/TIMP ratio, suggesting impaired ECM turnover in DM.⁶⁰ Experimental studies further suggest that pharmacologic ROCK inhibition can modulate MMP expression or activity in a context-dependent manner, thereby influencing ECM remodeling and fibrotic processes.^61,62

In ocular fibrotic processes, ROCK inhibitors have been shown to limit fibroblast activation, reduce ECM accumulation, and attenuate postoperative scarring after glaucoma filtration surgery by suppressing key profibrotic markers including collagen I and fibronectin.^28,29 In the cornea, inhibition of TGF-β/RhoA/ROCK signaling has been shown to reduce stromal fibrosis, corneal haze, and neovascularization in experimental wound-healing models.^34,63,64 These studies highlight the therapeutic potential of ROCK inhibitors as anti-fibrotic agents in ocular diseases characterized by pathological ECM accumulation. However, whether such effects extend to the abnormal ECM alterations characteristic of FECD and other corneal endothelial disorders has not been systematically investigated.

Addressing this question, the present study identified constitutive activation of ROCK signaling in corneal endothelial cells from FECD patients and provided experimental evidence that pharmacologic ROCK inhibition promotes remodeling of the pathological ECM, in part through suppression of TGF-β/Smad signaling. Consistent with previous reports, ROCK activity appeared to be regulated primarily through downstream phosphorylation events rather than changes in protein abundance.^20,21 Inhibition of ROCK activity by ripasudil suppressed expression of multiple fibrosis-associated ECM components aberrantly upregulated in FECD (e.g., COL1A1, COL3A1, FN1, AGRN, TGFBI, CLU, TNC, FBN1, SPARC, VCAN) at both transcript and protein levels, while simultaneously promoting ECM turnover through induction of MMP1, MMP3, MMP10, and MMP12 and reduction of TIMP1. MMP1 and MMP3 are key enzymes involved in the degradation of fibrillar collagens and fibronectin, whereas MMP10 and MMP12 contribute to turnover of elastin-rich and non-fibrillar matrix components.^65,66 This shift in the MMP/TIMP balance indicates that ROCK inhibition may promote normalization of ECM homeostasis by both suppressing abnormal ECM production and enhancing ECM degradation. Mechanistically, ROCK inhibition was associated with reduced Smad2 phosphorylation, supporting the concept that ROCK activity amplifies TGF-β/Smad-dependent transcription in fibrotic endothelial cells.⁶⁷ Whether this effect reflects direct cross-talk between ROCK and Smad signaling pathways or indirect modulation through ROCK-dependent cytoskeletal tension and mechanotransduction remains to be determined. In addition, broader transcriptomic changes following ROCK inhibition, including downregulation of stress- and inflammation-associated pathways and enrichment of collagen catabolic pathways, indicate reprogramming of endothelial cells away from a persistent stress-fibrotic phenotype toward restoration of ECM homeostasis. Notably, ripasudil induced comparable ECM suppression and MMP induction in PBK and PEX-K specimens, suggesting that ROCK inhibition promotes ECM remodeling across distinct forms of corneal endothelial pathology.

The translational implications of these findings may be relevant. ROCK inhibitors, particularly ripasudil and netarsudil, have already demonstrated therapeutic potential in ophthalmology for glaucoma treatment and endothelial regenerative strategies, where they promote corneal endothelial regeneration by enhancing cell adhesion, proliferation, migration, survival, and function.^{30–32,34,68,69} The present findings extend the therapeutic scope of ROCK inhibitors for corneal endothelial disorders by suggesting that these agents may additionally target the fibrotic ECM environment that contributes to disease progression. Notably, ripasudil exerted more pronounced effects than other ROCK inhibitors tested, consistent with previous studies highlighting its potent cellular functional effects in vitro and its favorable clinical safety profile in ophthalmic use.^35,36,70 However, because the different inhibitors were not compared using matched dose–response analyses, differences in their potency and dose-dependent effects may have contributed to the observed variation in efficacy. Nevertheless, the overall findings indicate that ROCK inhibition can influence both endothelial cell function³⁵ and ECM homeostasis. By modulating these complementary aspects of corneal endothelial pathology, ROCK inhibitors could potentially contribute to slowing progression of FECD and reducing recurrence of guttae following regenerative therapies such as DWEK/DSO by regulating ECM remodeling during endothelial regeneration.⁷¹ The observation that similar ECM-modulating effects were also observed in other corneal endothelial disorders further suggests that this approach may have broader relevance across diseases characterized by pathological DM alterations.

Several limitations of this study should be considered. Although the ex vivo EDM culture system preserves native human FECD tissue and enables direct investigation of disease-associated mechanisms, it cannot fully reproduce the physiological environment of the living cornea, including aqueous humor dynamics, systemic influences, and interactions with stromal and immune cell populations that may affect ECM homeostasis and therapeutic responses in vivo. Consequently, local drug exposure, pharmacokinetics, and the magnitude and duration of the observed ROCK inhibitor effects may differ under clinical conditions. In addition, the RNA-sequencing analysis was performed using pooled specimens, which reduced patient-specific resolution and precluded assessment of individual biological variability. However, no significant differences in age or sex distribution were observed between the RNA-sequencing pools, and key differentially expressed genes were independently confirmed in individual specimens, supporting the validity of the transcriptomic findings. Furthermore, only a limited number of normal donor EDM specimens were available for comparative analyses, whereby donor-related variability regarding age, storage time, etc., may have influenced the observed results. Finally, although the present study suggests substantial effects of ROCK inhibition on ECM remodeling, it did not assess whether these effects translate into prevention, regression, or structural remodeling of established guttae in appropriate in vitro and in vivo models.^72,73 Future studies using relevant models of FECD-associated ECM deposition, as well as longitudinal in vivo and clinical studies, will be required to determine the optimal therapeutic window and to establish whether ECM-targeted ROCK inhibition can modify disease progression or delay recurrent guttae formation following DWEK/DSO, eventually translating into a clinically meaningful benefit for the patients.