Proteomic Analysis of Human Corneal Keratocytes Reveals Mechanical Strain–Dependent Changes in Cellular Function

Abstract

Purpose

This study aimed to determine how different strain intensities—including normal, moderately increased, and high strain—influence protein expression profiles and related biological processes in human corneal stromal keratocytes.

Methods

A well-established in vitro model using the Flexcell FX-5000 Tension System, which replicates the natural corneal curvature and enables precise strain application to keratocytes, was used. Keratocytes were exposed to three strain levels: 3% (normal), 6% (moderately increased), and 12% (high). Following strain application, cells were collected for liquid chromatography–tandem mass spectrometry–based proteomic analysis to generate protein expression profiles. Differentially expressed proteins (DEPs) among the three groups were identified and subjected to biological pathway enrichment to reveal strain-dependent biological processes. Western blot analysis was performed to validate the expression of selected DEPs.

Results

Keratocytes exhibited strain intensity-dependent responses. Three percent strain maintained keratocytes in a quiescent phenotype, consistent with our previous findings. Six percent strain activated protective and adaptive programs to preserve tissue homeostasis under stress. In contrast, 12% strain suppressed immune-related processes and induced extracellular matrix (ECM) remodeling. Notably, procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2 (PLOD2) and cathepsin L (CTSL)—two ECM remodeling-related proteins implicated in fibrotic responses—were significantly upregulated under 12% strain, highlighting a potential link between excessive mechanical stress and stromal fibrosis.

Conclusions

These findings demonstrate that corneal strain regulates keratocyte behavior in an intensity-dependent manner and suggest that high mechanical stress may drive pathologic stromal remodeling and fibrotic responses, offering mechanistic insights that may inspire future therapeutic strategies.

Globally, an estimated 11.7 million individuals experience severe vision impairment or blindness caused by corneal diseases.¹ Lesions to the cornea—the transparent and protective outermost layer of the eye—may result in profound visual loss and consequently in substantial limitations to daily functioning and overall quality of life.

The cornea is a unique viscoelastic tissue with specialized biomechanical features. It can deform under stress and recover its original shape, ensuring normal physiological function while preserving structural integrity.^2,3 Corneal strain, a central biomechanical parameter, is the internal deformation of the corneal microstructure relative to a theoretically unloaded state. This deformation arises naturally from mechanical forces acting constantly on the cornea, including IOP and intrinsic tension within the collagen lamellae, but can also be affected by forces introduced by external impact or fibrotic stiffening.^4,5 It is essential for maintaining corneal structure and optical performance. A range of high-incidence ocular disorders—including glaucoma, ocular hypertension, keratoconus, and trauma-related injuries—often lead to secondary corneal pathology and are accompanied by direct or indirect alterations in corneal strain.^6–9 We recently demonstrated that in two rat models of trauma-related injury, IOP—the primary factor contributing to corneal strain—significantly increased following injury, approximately doubling compared to controls.¹⁰

The human cornea consists of five main layers: the epithelium, Bowman's layer, the stroma, Descemet's membrane, and the endothelium, with the stroma accounting for about 90% of its overall thickness.¹¹ Keratocytes are the resident corneal stromal cells critical for corneal maintenance and repair. Under healthy conditions, keratocytes typically remain quiescent, showing low metabolic activity and limited production of stromal extracellular matrix (ECM) components such as collagen and proteoglycans.^12,13 When the cornea is injured, quiescent keratocytes are activated, undergoing apoptosis, proliferation, and migration, as well as transforming into fibroblasts or myofibroblasts that contribute to corneal repair.^14,15

Although the importance of corneal strain in regulating keratocyte function and maintaining corneal physiology has been recognized, a comprehensive understanding of its overall effects on keratocytes remains limited. It is still unclear whether different strain intensities produce distinct effects or whether increased strain—associated with pathologic conditions such as glaucoma subtypes, keratoconus, and corneal injury^9,10,16–19—alters keratocyte responses in fundamentally different ways.

Our previous work established an effective in vitro model using the Flexcell FX-5000 Tension System,²⁰ which mimics the natural curvature of the human cornea and enables the application of quantitatively adjustable strain to keratocytes. We found that a 3% strain applied using the Flexcell system preserved the keratocytes’ phenotype and maintained keratocytes in a relatively quiescent state,^10,20,21 indicating that 3% strain represents a physiologically normal strain condition for keratocytes in this model. In this study, to systematically examine how keratocytes respond to increased strain, higher strain intensities (6% and 12%) were applied in addition to the 3% strain condition. After strain application, liquid chromatography–tandem mass spectrometry (LC-MS/MS)–based proteomic analysis was performed to profile protein expression in keratocytes. Differentially expressed proteins (DEPs) associated with each strain intensity were identified and analyzed for pathway enrichment to reveal varying strain-related biological processes.

This study reveals that corneal strain regulates keratocyte behavior in an intensity-dependent manner. It provides the first systematic experimental assessment of how different strain intensities modulate human keratocyte function. Our findings highlight the mechanical stress as a critical regulator of corneal homeostasis and pathology, as well as offer new mechanobiological insights that may be beneficial in the development of new therapeutic strategies for corneal disorders associated with abnormal strain, such as keratoconus, glaucoma, and corneal injury.

Methods

Human Cornea Samples Collection

Healthy human corneas were obtained from the Tissue Establishment, Eye Bank Umeå, University Hospital of Umeå, Sweden. Donors had voluntarily provided consent during their lifetime for postmortem corneal donation for both transplantation and research. When corneas were used for transplantation, the remaining portion of the cornea was supplied to the laboratory for research purposes. The Regional Ethical Review Board in Umeå determined that this study was exempt from formal ethical approval (2010-373-31M).

All cornea samples used in this study were anonymized and marked solely with the donor's sex and age. The study followed the ethical guidelines outlined in the Declaration of Helsinki.

Primary Human Keratocyte Isolation and Culture

Primary human keratocytes were isolated and cultured following previously described protocols.²² Briefly, the corneal epithelium and endothelium were carefully removed using a sterile scalpel, with the cornea sample washed twice in Hanks’ balanced salt solution (14170-088; Gibco, Life Technologies, Paisley, UK) both before and after removal. The remaining portion, the corneal stroma, was then cut into small fragments and digested overnight at 37°C in 2 mL Dulbecco's modified Eagle’s medium/nutrient mixture F-12 (DMEM/F-12) (31330095; Gibco, Life Technologies) supplemented with 2% fetal bovine serum (FBS) (F9665; Sigma-Aldrich, St. Louis, MO, USA) and 2 mg/mL collagenase from Clostridium histolyticum (C0130; Sigma-Aldrich) in a humidified 5% CO₂ incubator. Following digestion, the suspension was centrifuged at 200g for 5 minutes at room temperature, and the supernatant was discarded. The resulting pellet with enriched keratocytes was resuspended in DMEM/F-12 containing 2% FBS and seeded into a 25-cm² culture flask. Medium was changed every other day, and confluent keratocytes were passaged at a 1:3 ratio. Keratocytes used in this study were primarily isolated from the peripheral corneal tissue of donor corneas remaining after transplantation. Keratocytes were maintained in medium containing 2% FBS. Our previous work^10,22 showed that peripheral keratocytes cultured under these conditions maintain a slender, dendritic morphology and express high levels of characteristic keratocyte markers, including keratocan, lumican, aldehyde dehydrogenase (ALDH), and hematopoietic progenitor cell antigen CD34 (CD34), supporting preservation of keratocyte-associated features in vitro.

Mechanical Loading

Keratocytes were seeded in BioFlex 6-well plates (BF-3001C; Flexcell, Burlington, NC, USA) with 1.5 × 10⁵ cells in each well and cultured overnight at 37°C in a humidified incubator containing 5% CO₂. The next day, the plate was transferred to the BioFlex Loading Stations (LS-3000B25.VJW; Flexcell) equipped with dome-shaped loading posts of 25 mm diameter, designed to simulate the natural curvature of the human cornea at a 1:2 ratio. Keratocytes were exposed to constant equibiaxial strain at intensities of 3%, 6%, or 12% for 48 hours using the Flexcell FX-5000 Tension System, with corresponding control groups maintained under the same conditions without mechanical loading.

Proteomics Sample Preparation

Cells were washed twice with cold PBS and collected with a cell scraper. Then, the harvested cells were processed for peptide digestion with a modified SP3 protocol.²³ Briefly, the harvested cells were resuspended in lysis buffer (2% SDS, 20 mM Tris(2-carboxyethyl)phosphine (TCEP)) and heated at 95°C for 10 minutes. A 1:1 (v/v) mixture of SpeedBeads magnetic carboxylate-modified particles (Sigma-Aldrich; hydrophilic beads A, GE45152105050250; hydrophobic beads B, GE65152105050250) was washed four times with LC-MS–grade water and added to each sample in a binding buffer of 50% ethanol and 2.5% formic acid. Samples were incubated for 15 minutes at room temperature with gentle shaking (500 rpm), then transferred to a 0.22-µm filter plate (Sigma-Aldrich, MSGVN2210) and centrifuged at 1000 g to remove unbound proteins. The retained beads on the filter were washed four times with 70% ethanol.

A trypsin-containing digestion solution was made by adding trypsin to a digestion solution containing 100 mM HEPES (pH 7.5), 5 mM chloroacetamide, and 1.2 mM TCEP. Digestion was performed by adding trypsin-containing digestion solution to the samples at a ratio of 1 µg trypsin per 25 µg protein. Samples were incubated overnight at room temperature with gentle shaking (500 rpm). Peptides released during digestion were collected by centrifugation at 1000 g, while those remaining on the beads were eluted with 10 µL of 2% DMSO and pooled with the flowthrough. The combined peptide fraction was desalted on an Oasis HLB plate (186001828BA; Waters, Milford, Massachusetts, United States) according to the manufacturer's instructions and dried in a speed vacuum concentrator.

LC-MS/MS Analysis

Dried peptides were dissolved in 0.1% formic acid, and 1 µg from each sample was loaded onto a Vanquish Neo LC system (Thermo Scientific, Rockford, IL, USA) coupled to a mass spectrometer. Separation was performed using a trapping column (PEPMAP NEO C18, 5 µm, 300 µm × 5 mm; Thermo Scientific, 174500) and an analytical column (nanoEase M/Z HSS C18 T3, 100 Å, 1.8 µm, 75 µm × 250 mm; Waters, 186008818). Peptide separation and elution were performed over a total of 120 minutes using a gradient of mobile phase A (water with 0.1% formic acid) and mobile phase B (80% acetonitrile with 0.1% formic acid). The gradient began with a transition from mobile phase A to 8% mobile phase B within 4 minutes, followed by a gradual rise to 27% mobile phase B over 87 minutes. This was followed by a rapid increase to 80% mobile phase B within 6 seconds, which was maintained for several minutes, and finally decreased to 2% mobile phase B within 30 seconds to reequilibrate the column.

Data acquisition was performed on an Exploris 480 (Thermo Scientific) in a data-dependent mode. Survey scans were collected from a mass range of 375 to 1500 at a resolution of 120,000 with a 40% radiofrequency lens and normalized automatic gain control (AGC) of 300%. A maximum cycle time of 2 seconds was used to select precursor ions for tandem-MS/MS (MS2) analysis. Precursors with charge states of 2 to 6 were included, with dynamic exclusion enabled for 35 seconds. MS2 scans were acquired at 15,000 resolution (at m/z 200) with an automatic AGC target value and a 1.4 m/z isolation window. Higher-energy collisional dissociation fragmentation at a normalized collision energy of 30 was induced. Isotopes were excluded from MS2 analysis.

Proteomics Data Analysis

Proteomic raw data were searched against the Homo sapiens UniProt FASTA (proteome ID UP000005640) using FragPipe (version 18) with the LFQ-MBR workflow for label-free quantification.²⁴ Statistical analysis was performed in FragPipe-Analyst (http://fragpipe-analyst.nesvilab.org/). Intensities were normalized using the variance-stabilizing normalization method²⁵ to minimize technical variation. Missing values were imputed with the minimum value in each sample. False discovery rate correction was conducted using the Benjamini–Hochberg method.

Bioinformatics Analysis

Visualization of Proteomics Data

DEPs were defined by a t-test (P < 0.05) and a fold change of ≥1.5 or ≤−1.5 compared to controls. Heatmap visualization of DEP expression was generated in RStudio (version 2025.09.1+401) using the pheatmap package (version 1.0.13). Protein intensity values were log₂-transformed prior to visualization. Clustering of rows was performed with Euclidean distance and complete linkage to display expression patterns across experimental groups, and a royal blue–white–red color scheme was applied for visualization.

Principal component analysis (PCA) plots were generated using FragPipe-Analyst (http://fragpipe-analyst.nesvilab.org/).

Biological Pathway Enrichment

The biological pathway enrichment of DEPs under different strain intensities was conducted using the Reactome database (www.reactome.org) with H. sapiens selected as the reference species and a significance cutoff of P < 0.05.

Cell Lysis Protein Extraction

Plates containing keratocytes were placed on ice and rinsed with chilled PBS (09-8912-100; Medicago, Vantaa, Finland). Ice-cold RIPA buffer (89901; Thermo Scientific) supplemented with Halt Protease and Phosphatase Inhibitor Cocktail (100×) (78446; Thermo Scientific) was added, and cells were detached using a precooled plastic cell scraper. The lysate was transferred to a precooled 1.5-mL tube and gently agitated for 30 minutes at 4°C. Samples were then centrifuged at 16,000g for 20 minutes at 4°C, and the resulting supernatant was collected and transferred into a new prechilled tube. The extracted cell lysis proteins were stored at −80°C until further analysis.

Western Blot

Protein concentrations were measured using the Pierce BCA Protein Assay Kit (23227; Thermo Scientific) following the manufacturer's instructions. Equal amounts of protein (10 µg per sample) were heat-denatured for 5 minutes at 98°C and separated on Mini-PROTEAN TGX Precast Protein Gels (4561044; Bio-Rad, Hercules, CA, USA) at 110 V for 1 hour. Proteins were then transferred to PVDF membranes (88520; Thermo Scientific) at 110 V for 90 minutes. Membranes were blocked with 5% bovine serum albumin (A9647; Sigma-Aldrich, St. Louis, MO, USA) in TBS (1706435; Bio-Rad) containing 0.1% Tween-20 (9005-64-5; VWR, Fontenay-sous-Bois, France) at room temperature for 35 minutes, followed by overnight incubation at 4°C with one of the following primary antibodies, depending on the experimental requirement: cathepsin L (E3R3P) rabbit mAb (55914T; Cell Signaling, Danvers, MA, USA), lysyl hydroxylase 2 antibody (44709S; Cell Signaling), and β-actin antibody (4967; Cell Signaling). After washing with TBST, membranes were incubated for 90 minutes at room temperature with anti-rabbit IgG, horseradish peroxidase (HRP)–linked antibody (7074; Cell Signaling). Cathepsin L (E3R3P) rabbit mAb, lysyl hydroxylase 2 antibody, and β-actin antibody were used at a dilution of 1:1000. The HRP-linked anti-rabbit IgG antibody was used at a dilution of 1:2000. Membranes were then washed with TBST. Protein bands were visualized using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Scientific) and imaged with an Odyssey Fc Imaging System (LI-COR Biosciences, Lincoln, NE, USA). Band intensity was quantified using ImageJ software (version 1.53; National Institutes of Health, Bethesda, MD, USA).

Quantification and Statistical Analysis

Statistical analyses were performed using GraphPad Prism 10 (GraphPad Software, La Jolla, CA, USA). Each experiment was independently repeated at least three times with cells from different donors. Details of replicate numbers and statistical tests are provided in the figure legends. Significance levels were defined as *P < 0.05 and **P < 0.01.

Results

Keratocytes Exposed to Different Strain Intensities Exhibit Distinct Protein Expression Patterns

To study the impact of strain intensities on keratocyte protein expression patterns, cells from three independent donors were subjected to 3%, 6%, or 12% strain for 48 hours. An overview of the experimental workflow is illustrated in Figure 1. Heatmaps (Figs. 2A–C) showed an overview of the expression of DEPs in all samples. Proteins were identified as DEPs if their expression levels were statistically significant by t-test (P < 0.05) with a fold-change of ≥1.5 or ≤−1.5 compared to the control group. Heatmaps were arranged according to experimental groups and visualized the log2-transformed protein expression values across all strain intensities in keratocytes from each donor, with three technical replicates for each strain level. Rows were clustered using hierarchical clustering (Euclidean distance, complete linkage). A royal blue–white–red color scheme was used, where red represents higher and royal blue lower relative abundance compared to the protein mean.

Figure 1.

Experimental workflow of this study. (A) Keratocytes were isolated from human corneas, cultured in flasks, and exposed to varying mechanical strain intensities. (B) Following strain application, keratocytes were collected for LC-MS/MS–based proteomic analysis. (C) DEPs identified from the proteomic data were subjected to bioinformatics analyses. (D) Expression levels of selected DEPs were validated. Figures were created in BioRender. https://BioRender.com/mgat2ii.

Figure 2.

Keratocytes exposed to different strain intensities exhibit distinct protein expression patterns. (A–C) Heatmaps showing the expression of DEPs in keratocytes from donors 1, 2, and 3 under control conditions and after 3%, 6%, or 12% strain for 48 hours. Rows represent DEPs, and columns represent samples. Rows were clustered using hierarchical clustering (Euclidean distance, complete linkage). Colors indicate log₂-transformed protein intensities, with red for higher and royal blue for lower relative abundance. n = 3. (D–F) PCA plots showing the clustering of DEPs in keratocytes from donors 1, 2, and 3 under control conditions and after 3%, 6%, or 12% strain for 48 hours. The first principal component (PC1) represents the major source of variation, reflecting proteomic differences across strain levels, while PC2 accounts for secondary variation among replicates.

The heatmaps demonstrate that keratocytes exposed to different strain intensities exhibit distinct protein expression patterns. Replicates within each strain group displayed highly consistent clustering patterns, reflecting strong intragroup uniformity and homogeneity. This indicates high experimental reproducibility and reliability. The only exception was observed in the 3% strain group from donor 1, where two replicates clustered tightly together while one diverged (Supplementary Fig. S1A). In Supplementary Figure S1, both rows and columns were hierarchically clustered using Euclidean distance and complete linkage to provide a more detailed view of replicate similarity.

PCA plots (Figs. 2D–F) showed clear grouping of samples by strain level across all three donors, with replicates clustering closely together, indicating strong reproducibility. The dominance of PC1 over PC2 reflected distinct proteomic patterns among strain levels. A minor deviation was observed in the 3% strain group from donor 1, where one replicate diverged slightly from the others, consistent with the heatmap results.

Differences and Similarities in DEPs Induced by Varying Strain Intensities

Table 1 summarizes the number of DEPs induced by different strain intensities. In the 3% strain group, a total of 5355 DEPs were identified across all three donors, including 2257 upregulated (fold-change ≥1.5) and 3098 downregulated (fold-change ≤–1.5) proteins relative to their respective controls. Among these, 3180 were overlapping DEPs—defined as DEPs detected in at least two of the three donors—including 910 upregulated and 2270 downregulated proteins. In the 6% strain group, 1381 DEPs were identified across all three donors, comprising 393 upregulated and 988 downregulated, with 115 overlapping (10 upregulated and 105 downregulated). In the 12% strain group, 4128 DEPs were detected, including 1370 upregulated and 2758 downregulated proteins, with 1099 overlapping DEPs—213 upregulated and 886 downregulated. Figures 3A–F present Venn diagrams illustrating the DEPs in each strain group for each donor and their overlaps. Figure 3G shows the proportion of overlapping DEPs within the total DEPs of each strain group. The 3% strain group exhibited the highest proportion (59%), whereas the 6% strain group showed the lowest, as overlapping DEPs represented only 8% of the total.

Table 1.

Number of DEPs Induced by Different Strain Intensities

Strain Intensity	Total DEPs	Total Upregulated DEPs	Total Downregulated DEPs	Total Overlapping DEPs	Overlapping Upregulated DEPs	Overlapping Downregulated DEPs
3%	5355	2257	3098	3180	910	2270
6%	1381	393	988	115	10	105
12%	4128	1370	2758	1099	213	886

Figure 3.

DEPs induced by varying strain intensities in keratocytes. (A, B) Venn diagrams illustrating the upregulated and downregulated DEPs in keratocytes from donors 1 (blue), 2 (yellow), and 3 (green) under 3% strain. (C, D) Venn diagrams illustrating the upregulated and downregulated DEPs in keratocytes from three donors under 6% strain. (E, F) Venn diagrams illustrating the upregulated and downregulated DEPs in keratocytes from three donors under 12% strain. Numbers in overlapping regions indicate overlapping DEPs, while numbers in nonoverlapping regions denote DEPs unique to each donor. (G) Proportion of overlapping DEPs relative to the total DEPs in each strain group.

Subsequent analyses focused on the overlapping DEPs within each strain group. Table 2 summarizes the similarities and differences in DEPs among the three strain groups. The 3% strain group contained 766 unique upregulated and 1490 unique downregulated DEPs; the 6% strain group contained 4 and 45, respectively; and the 12% strain group contained 71 and 100. Notably, despite differences in strain intensity, the groups shared a considerable subset of DEPs. A total of 145 upregulated and 793 downregulated DEPs were shared across at least two strain intensity groups. This indicates that certain proteins exhibit robust and conserved expression responses to strain, independent of its intensity. Venn diagrams in Figure 4 display the distribution of DEPs, showing those unique to each strain group, as well as those shared across the three groups.

Table 2.

Differences and Similarities in DEPs Induced by Different Strain Intensities

Strain Intensities	Unique Upregulated DEPs	Unique Downregulated DEPs	Shared Upregulated DEPs Detected in at Least Two Strain Intensity Groups	Shared Downregulated DEPs Detected in at Least Two Strain Intensity Groups
3%	766	1490	145	793
6%	4	45
12%	71	100

Figure 4.

The distribution of DEPs across three strain groups. Venn diagrams of DEPs across three strain groups: 3% (blue), 6% (yellow), and 12% (green): (A) upregulated and (B) downregulated DEPs. Numbers in overlapping regions indicate overlapping DEPs, while numbers in nonoverlapping regions denote DEPs unique to each strain group.

DEPs Induced by Varying Strain Intensities Modulate Distinct Biological Pathways in Keratocytes

To determine the biological pathways in which DEPs induced by varying strain intensities are involved, pathway enrichment analysis was conducted using the Reactome database (www.reactome.org). Figure 5A shows the 10 most significantly enriched upregulated pathways induced by DEPs unique to the 3% strain group, indicating enhanced mitochondrial function with increased energy production and active clearance of defective or misfolded proteins. DEPs unique to the 6% strain group induced upregulation of five pathways (Fig. 5B), primarily involving proinflammatory, anti-inflammatory/proresolving processes and enhanced detoxifications. These findings suggest that keratocytes initiate an inflammatory response while simultaneously activating pathways that resolve inflammation and clear metabolic by-products. The 10 most significantly enriched upregulated pathways induced by DEPs unique to the 12% strain group (Fig. 5C) were largely related to connective tissue, particularly ECM dynamics, including increased synthesis, cross-linking, repair, and turnover, as well as broader processes of ECM maintenance, assembly, and remodeling. To identify biological pathways consistently upregulated by strain, we then analyzed the top 10 upregulated pathways induced by DEPs shared across at least two strain intensity groups (Fig. 5D). These pathways were primarily associated with mitochondrial quality control, including the removal of misfolded or damaged proteins and increased mitochondrial activity or biogenesis, as well as immune activation, inflammatory responses, and hematopoietic remodeling.

Figure 5.

Upregulated biological pathways induced by DEPs. (A) The 10 most significantly enriched upregulated pathways induced by DEPs unique to the 3% strain group. (B) Five upregulated pathways induced by DEPs unique to the 6% strain group. (C) The 10 most significantly enriched upregulated pathways induced by DEPs unique to the 12% strain group. (D) The top 10 upregulated pathways induced by DEPs shared across at least two strain intensity groups. The x-axis represents −log₁₀(P value). Pathway enrichment was performed using the Reactome database (reactome.org) with significance defined as P < 0.05.

We next examined downregulated biological pathways induced by DEPs. Figure 6A shows the 10 most significantly downregulated pathways unique to the 3% strain group. These pathways indicated suppressed degradation of cell cycle proteins, suggesting potential cell cycle dysregulation, such as arrest or reduced proliferation, along with decreased proteasome activity and assembly, and an overall reduction in protein synthesis, a pattern commonly observed in quiescent or stressed cells. Most of the top 10 downregulated pathways unique to the 6% strain group (Fig. 6B) were associated with growth suppression, including cell cycle arrest or quiescence in response to DNA damage or stress, restricted proliferation through differentiation, and metabolic downshifts reflecting reduced growth demands. In Figure 6C, the 10 most significantly downregulated pathways unique to the 12% strain group indicated immune suppression or dysfunction and decreased cell proliferation and transcriptional activity. In addition, a potentially increased vulnerability to DNA damage was observed, including reduced capacity for DNA repair and cellular maintenance. Meanwhile, altered cell–ECM interactions suggested that tissue structure and repair capacity may be affected. The top 10 downregulated pathways induced by DEPs, shared across at least two strain intensity groups, are displayed in Figure 6D, showing biological pathways consistently downregulated by strain. These pathways suggest a coordinated stress response characterized by globally suppressed protein synthesis, reduced proliferation and metabolic activity, altered cellular homeostasis, and increased risk of aberrant protein production.

Figure 6.

Downregulated biological pathways induced by DEPs. The 10 most significantly enriched downregulated pathways induced by DEPs unique to the (A) 3%, (B) 6%, and (C) 12% strain group. (D) The top 10 downregulated pathways induced by DEPs shared across at least two strain intensity groups. The x-axis represents −log₁₀(P value). Pathway enrichment was performed using the Reactome database (reactome.org) with significance defined as P < 0.05.

Validation of Selected DEPs in the 12% Strain Group

The 12% strain group represented an abnormally high strain level, similar to that observed in various ocular disorders such as glaucoma, keratoconus, and trauma-induced injuries. Some of these conditions are often characterized by pathologic changes in the corneal stroma, including fibrosis, which is closely linked to maladaptive ECM remodeling.

Two DEPs, procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2 (PLOD2) and cathepsin L (CTSL), previously reported to participate in ECM remodeling and tissue fibrogenesis,^26–29 were identified in ECM-related enriched pathways in the 12% strain group and were selected for validation. Table 3 presents the expression levels of PLOD2 and CTSL and their associated enriched pathways in the proteomics data. Western blot analysis (Fig. 7) revealed that the expression of PLOD2 and CTSL was significantly increased in keratocytes exposed to 12% strain for 48 hours.

Table 3.

Expression Fold Change and Associated Pathways of PLOD2 and CTSL in the 12% Strain Group

Protein	Sample ID	Expression Fold Change	Associated Enriched Pathways
PLOD2	Donor 1Donor 2Donor 3	1.311.751.66	• Collagen formation• Collagen biosynthesis and modifying enzymes• Extracellular matrix organization
CTSL	Donor 1Donor 2Donor 3	1.842.071.13	• Collagen formation• Extracellular matrix organization• Glycosaminoglycan metabolism

Figure 7.

Validation of the expression of PLOD2 and CTSL in keratocytes subjected to 12% strain. (A) Keratocytes were subjected to 12% strain for 48 hours. Protein expression level of PLOD2 and CTSL was assessed by Western blot. β-Actin served as a loading control for total protein. Representative image from one of three independent Western blot experiments conducted using samples from three different donors. (B, C) Densitometry analysis of the Western blot bands of PLOD2 and CTSL. n = 3. The densitometry of β-actin bands served as the normalization reference. Statistical analyses were conducted by an unpaired t-test. Data are presented as mean ± SD and normalized to the ctrl group. *P < 0.05, **P < 0.01.

Discussion

This study employs an effective in vitro model²⁰ that mimics the natural curvature of the human cornea and allows precise strain control. It provides the first comprehensive analysis of how different corneal strain intensities affect the behavior and function of resident corneal stromal keratocytes.

In this study, different strain intensities were applied to keratocytes using a previously established in vitro model²⁰ based on the Flexcell FX-5000 Tension System. Due to system limitations—such as membrane elongation capacity, rebound, and friction—the actual strain achieved will be lower than the set value. A previous study reported that cells experienced an average strain of approximately 37% ± 8% of the applied equibiaxial strain at the cellular level when using the Flexcell system.³⁰ Thus, in our model, 3%, 6%, and 12% strains correspond to actual strains that were experienced by keratocytes, which are 1.11% ± 0.24%, 2.22% ± 0.24%, and 4.44% ± 0.24%, respectively.

Many studies have recognized the importance of corneal strain and have employed a wide variety of experimental methods to measure it.^31–34 Considering that the experimental model used in this study was based on the Flexcell system, in which keratocytes were subjected predominantly to shear and tangential strains, we refer to the work of Kwok et al.³¹ and Pan et al.³⁴ Pan et al.³⁴ demonstrated that, when IOP ranged from 10 to 21 mm Hg (the typical clinical IOP range),³⁵ the mean tangential and shear strains were approximately 0.20% to 0.50% and 2.80% to 4.50%, respectively. (The exact numerical values were not explicitly reported in the original study; the values presented here were estimated from the line plots provided in the study.) When IOP increased to 30 mm Hg—a pathologically elevated IOP level that can occur in ocular disorders such as glaucoma and following ocular trauma^36,37—the overall mean tangential and shear strains reached 0.50% ± 0.53% and 5.23% ± 2.38%, respectively. Kwok et al.³¹ reported that, at an IOP of 30 mm Hg, the overall mean tangential and shear strains in the central cornea were 0.25% ± 1.16% and 3.15% ± 2.14%, respectively, while those in the paracentral cornea were 1.82% ± 1.47% and 2.75% ± 1.11%, respectively.

The distinction between Flexcell-applied strain and previously reported corneal strain should be noted. Although the strain intensities used in this study differ numerically from tissue-level values reported in prior studies,^31,34 they fall within the ranges reported in the prior studies, providing experimental justification for the selected strain conditions and supporting their biological relevance within our model.

The role of strain in regulating corneal physiology and keratocyte function has been recognized previously, yet whether varying strain intensities will display different effects remains unclear. Altered corneal biomechanical properties have been reported in several common ocular disorders, including glaucoma, keratoconus, and trauma-related injury.^6–9 Corneas from clinical and subclinical keratoconus^16,17 and normal-tension glaucoma¹⁸ eyes exhibit significantly higher deformation amplitude ratios than healthy ones. Corneas in amyloidotic glaucoma eyes are also reported to be more deformable than those of controls.¹⁹ In addition, wounded corneas show reduced high-strain tangent modulus and tensile strength, along with elevated IOP, compared to unwounded corneas.^9,10 Collectively, these findings suggest that, in some ocular disorders, the cornea exhibits weakened biomechanical resistance to applied loading and is therefore expected to undergo greater deformation and have higher strain. Our previous studies demonstrated that 3% strain preserves the keratocyte phenotype and maintains it in a relatively quiescent state with reduced cell proliferation and migration,^10,20,21 which supports the rationale for selecting the 3% strain as the normal strain condition. On this basis, two higher strain conditions (6% and 12%), in addition to 3%, were included to examine strain intensity–dependent keratocyte responses. In this exploratory study, 6% and 12% were selected as stepwise increases relative to 3% and were within the operational limits of the Flexcell system. We did not assume a priori that these strain levels would necessarily elicit distinct biological responses; rather, this experimental design was intended to explore whether potential strain-dependent transitions in keratocyte programs emerge with increasing strain.

This study demonstrates that keratocytes respond differently and exhibit consistent intragroup, while distinct intergroup, protein expression patterns under different strain intensities. The 3% strain group showed the highest proportion of overlapping DEPs relative to its total DEPs (59%), followed by the 12% strain group (27%), whereas the 6% strain group showed the lowest proportion (8%). This indicates that under the normal strain condition (3% strain), keratocyte characteristics remain relatively consistent across different donors. By comparison, when strain increased to a higher level (12% strain), DEPs from different donors still maintained a certain degree of overlap, indicating that keratocytes may activate a relatively uniform physiological program to cope with excessive biomechanical stress. In contrast, 6% strain resulted in very few overlapping DEPs, suggesting that the keratocytes sensed a moderately elevated strain distinct from their normal condition (3% strain in this study), while keratocytes from different donors exhibited greater variability in their tolerance to this strain intensity. An interesting observation is that the 6% strain group clusters near the control group in PCA plots (Figs. 2D–F). Because the human cornea is continuously subjected to mechanical forces in vivo, keratocytes are always exposed to a baseline level of strain. A likely explanation is that control (unstrained) lacks strain and therefore represents a mild stressed state, similar to the 6% strain (the moderately increased strain compared to the 3% strain). Under such mild stress, keratocytes may activate compensatory or modest stress-responsive programs, resulting in similar protein expression patterns between these two groups.

Despite differences in strain intensity, the three groups still shared a subset of common DEPs, suggesting that some proteins were consistently regulated by strain, functioning like “strain-associated housekeeping” proteins that are highly sensitive to corneal strain and are consistently up- or downregulated to perform their significant physiological roles.

Biological pathway enrichment revealed that the shared DEPs across the three strain groups were primarily associated with active mitochondrial quality control and immune activation, along with a global reduction in protein synthesis and surveillance. These processes appear to serve as protective mechanisms during stress, reflecting the proactive response of keratocytes to stress or abnormal conditions such as pathologic alterations or disease.

Beyond the similarities in pathway regulation, each strain intensity exhibited distinct regulatory modules. Under 3% strain, the regulated biological processes were predominantly associated with quiescent cells, a hallmark of keratocytes under normal physiological conditions¹² and consistent with our previously published findings.¹⁰ Moreover, enhanced mitochondrial function and active clearance of defective or misfolded proteins indicate that, even in a quiescent state, keratocytes are not completely inactive but maintain a basal level of energy production and protein quality control to preserve intracellular homeostasis.

The influence of 6% strain on keratocytes was characterized by activation of defense and resolution programs that support tissue homeostasis following stress or injury, suggesting a potential adaptive survival mechanism. Cells showed slowed or arrested cell cycle progression while initiating responses to limit chronic inflammation and promote healing, indicating an active strategy to counteract stress-related damage. This suggests that when keratocytes sense a moderately elevated strain (6% strain) distinct from their normal condition (3% strain), they no longer maintain a quiescent state but instead activate protective programs to restore balance and preserve corneal stability.

Under 12% strain, keratocytes exhibited suppressed immune activity, weakened DNA repair and autophagy, and impaired ECM adhesion, while simultaneously engaging in active ECM remodeling. This pattern suggests that keratocytes under 12% strain shift from protective functions toward reshaping the stromal environment—either as part of a repair response or as maladaptive remodeling that may increase the risk of fibrosis. Within the context of our model, 12% strain represents a high-strain condition. Given reports of increased corneal deformation in ocular disorders such as glaucoma, keratoconus, and corneal injury, our findings highlight the detrimental impact of high corneal strain on keratocyte function and tissue homeostasis. These results further suggest that high strain contributes to corneal fibrotic responses, offering mechanistic insight into how excessive strain may influence corneal disease progression. Further studies are needed to define whether, how, and to what extent the corneal strain can be modulated to promote favorable clinical outcomes.

Two upregulated DEPs, PLOD2 and CTSL, were specifically identified in the 12% strain group and validated by Western blot analysis, which confirmed the reliability of the proteomic findings. PLOD2 and CTSL were enriched in pathways associated with ECM remodeling, indicating that abnormally high strain activates ECM remodeling-related biological processes. Previous studies reported that PLOD2 affects collagen stability and fibril diameter³⁸ and regulates the matrix fiber alignment and the topographic type of ECM alignment.^26,27 CTSL has been reported to exert opposing roles in fibrosis across different tissues. Elevated CTSL expression has been reported in liver fibrosis,²⁹ whereas decreased CTSL expression contributed to the development of lung fibrosis in systemic sclerosis.²⁸ Based on our findings, we propose that excessive strain may contribute to stromal disruption and subsequent fibrosis in certain ocular disorders. Notably, the dual roles of CTSL may suggest that under excessive strain, keratocytes may activate both profibrotic and antifibrotic processes. The eventual development of fibrosis may therefore depend on the balance between these opposing processes and the relative dominance of either effect.

Overall, our study provides the first systematic characterization of keratocyte responses to varying strain intensities, highlighting an intensity-dependent regulation of cellular behavior and function. The graded responses observed—from maintenance of quiescence under normal strain, to activation of protective programs at moderately increased strain, and finally to activated ECM remodeling under high strain—illustrate how mechanical stress shapes corneal homeostasis and dysregulation. These findings highlight the critical role of corneal strain in modulating keratocyte function and suggest that excessive mechanical stress may contribute to stromal disruption and fibrosis in certain ocular diseases.

We recognize that keratocyte mechanobiology in vivo is governed by multiple mechanical cues, including strain as well as factors such as ECM stiffness, substrate topography, and tissue elasticity, which may act together to shape cellular responses.^39–43 These mechanical factors are likely to act in concert rather than independently. Accordingly, the responses observed here likely reflect integrated mechanotransduction to time-varying mechanical cues within the local microenvironment. While the present study focused on strain as the primary mechanical input, delineating the relative contributions and potential interplay among these mechanical factors represents an important direction for future investigation.

It should also be noted that disease-related changes in corneal shape and biomechanics may evolve over longer time scales and involve tissue remodeling; therefore, ocular disorder–related in vivo strain may not be directly translated into short-term in vitro strain. At the same time, we do not exclude the possibility that, in some contexts, increased corneal deformation may be accompanied by keratocyte responses that partially overlap with those observed in our model. For example, we previously observed upregulated aldehyde dehydrogenase 3A1 (ALDH3A1) in trauma-related injury models with significantly elevated IOP, as well as in keratoconus-derived keratocytes, which is consistent with the increased strain-associated ALDH3A1 upregulation observed in our model.¹⁰

Further studies and a more comprehensive understanding of the thresholds and mechanisms by which keratocytes respond to mechanical stress may inspire strategies to prevent or mitigate corneal pathology associated with altered corneal biomechanics and guide the development of targeted therapeutic interventions.