Orbital adipose and connective tissue specimens were harvested from 14 patients with confirmed GO undergoing therapeutic orbital decompression procedures. All patients exhibited biochemical euthyroidism at the time of surgical intervention. GO patients were stratified according to the 1991 Nunery classification system (3): Type I subjects predominantly manifested orbital adipose tissue expansion with mild to moderate extraocular muscle involvement, without restrictive myopathy or diplopia; Type II subjects demonstrated significant extraocular muscle enlargement with resulting restrictive myopathy and diplopia within the central 20° visual field. Control specimens were procured from seven non-GO patients undergoing enucleation for uveal melanoma (UM), with inclusion criteria stipulating absence of extraocular extension or metastatic disease. Comprehensive clinical and demographic characteristics of the study cohort are detailed in Table 1 .

Exclusion criteria encompassed administration of systemic or local glucocorticoid therapy within three months preceding tissue acquisition. Disease severity and inflammatory activity were assessed using the standardized NOSPECS classification system and seven-point clinical activity score (CAS) as established by the European Group on Graves’ Orbitopathy (EUGOGO) (18, 19). Written informed consent was obtained from all study participants. This investigation was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki and received approval from the Institutional Review Board of the First Affiliated Hospital of Chongqing Medical University (approval number 2023-30, January 11, 2023).

Dulbecco’s Modified Eagle’s Medium (DMEM, #C11965500BT), fetal bovine serum (FBS, #10270-106-1), penicillin and streptomycin (#15140122), 0.25% trypsin/EDTA (#25200072), and phosphate-buffered saline (PBS, #C10010500BT) (all reagents from Gibco Laboratories, New York, NY, USA). Dimethyl sulfoxide (DMSO, #WL064, Meiluncell), TGF-β1 (R&D Systems, Minneapolis, MN, USA, #240-B-010), MMP-14 inhibitor (NSC-405020, #HY-15827, MCE);

TaKaRa MiniBEST Universal RNA Extraction Kit (#9767), PrimeScript RT Master Mix (#RR036B) and TB Green Premix Ex Taq II (#RR820B) (all from TaKaRa, Dalian, China); Radioimmunoprecipitation Assay (RIPA, #P0013B, Beyotime), Protein-free rapid blocking solution (#G2052, Servicebio), Enhanced BCA Protein Assay kit (#P0010, Beyotime), 5×protein loading buffer (Boster). Primary antibody: α-SMA (#19245S), COL1A1 (#39952S), CTGF (#86641S), GAPDH (#5174S) (all from Cell Signaling Technology, Boston, MA, USA); MMP14 (#AF0212), PI3K p85 alpha (#AF6241), Phospho-PI3K p85 alpha (Tyr607, #AF3241), pan-AKT1/2/3 (#AF6261), Phospho-AKT1/2/3 (Ser473, #AF0016), β-actin (#AF7018) (all from Affinity Biosciences); DyLight 488 Goat anti-Rabbit IgG (H + L) Secondary Antibody (#A23220, Abbkine), Goat anti-Rabbit IgG HRP Conjugated Secondary Antibody (#CW0103S, CWBIO). HRP-conjugated anti-rabbit secondary antibody (#DM-001, ProteinSimple, San Jose, CA, USA).

Formalin-fixed, paraffin-embedded tissue specimens were sectioned at 3-μm thickness and subjected to a standardized immunohistochemical protocol. Deparaffinization was performed using xylene, followed by gradient rehydration through absolute ethanol series. Antigen retrieval was accomplished using EDTA buffer (pH 9.0) under optimized conditions. Following triple rinses with phosphate-buffered saline (PBS), endogenous peroxidase activity was quenched by incubating sections in 3% hydrogen peroxide solution for 25 minutes at ambient temperature under light protection. After subsequent PBS washing, nonspecific binding sites were blocked using 3% bovine serum albumin (BSA) applied dropwise onto sections with 30-minute incubation at room temperature. Primary anti-MMP14 antibody was applied to tissue sections with overnight incubation at 4°C in a humidified chamber. Following thorough PBS washing, sections were incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibody (Servicebio, Wuhan, Hubei, China) for 50 minutes at room temperature. Immunoreactivity was visualized using diaminobenzidine (DAB) chromogen with precisely timed development (45 seconds), followed by hematoxylin counterstaining. Processed sections underwent dehydration through graded alcohols and were permanently mounted using neutral mounting medium. Quantitative immunohistochemical analysis was performed using ImageJ software (National Institutes of Health, Bethesda, MA, USA), with integrated optical density (IOD) normalized to measured area serving as the semiquantitative metric for MMP14 expression levels.

Total RNA was isolated from primary orbital fibroblast cultures utilizing the TaKaRa MiniBEST Universal RNA Extraction Kit following the manufacturer’s optimized protocol. RNA quality and integrity were assessed prior to downstream applications. Subsequently, high-quality RNA samples were processed for cDNA library construction using the NEBNext^® Ultra™ RNA Library Prep Kit for Illumina platform through a systematic workflow including mRNA isolation, fragmentation, and double-stranded cDNA synthesis according to the manufacturer’s specifications. Preliminary quality assessment and quantification of the resultant cDNA libraries were performed using Qubit 3.0 fluorometric quantitation, with a stringent quality control threshold requiring effective library concentrations exceeding 10 nM. Libraries meeting these rigorous quality parameters were sequenced on the Illumina NovaSeq 6000 platform employing paired-end 150 bp (PE150) sequencing chemistry to ensure comprehensive transcriptomic coverage and robust read depth for differential expression analysis.

Orbital adipose/connective tissue specimens were meticulously dissected to remove fascial elements and vascular structures prior to processing. The resulting tissue was finely minced into approximately 1-mm³ fragments and established as explant cultures in 10-cm tissue culture dishes containing high-glucose Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 20% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Following fibroblast outgrowth from tissue explants and attainment of 80-90% confluence, adherent cells were enzymatically dissociated using 0.25% trypsin-EDTA solution and subcultured. Established orbital fibroblast (OF) cultures were subsequently maintained in standard proliferation medium comprising DMEM supplemented with 10% FBS and 1% penicillin/streptomycin under standard culture conditions (37°C, 5% CO₂, humidified atmosphere). All experimental procedures were performed using OFs between passages 3 and 7, with a minimum of three biological replicates per experimental condition to ensure reproducibility.

Total RNA was isolated from cultured OFs using the TaKaRa MiniBEST Universal RNA Extraction Kit according to the manufacturer’s standardized protocol. RNA concentration and purity were spectrophotometrically determined prior to reverse transcription. Complementary DNA (cDNA) synthesis was performed using the PrimeScript RT Master Mix under optimized reaction conditions. Quantitative real-time PCR was conducted on a Roche LightCycler 480 system (Roche Diagnostics, Basel, Switzerland) employing TB Green Premix Ex Taq II reagent with gene-specific primers. Oligonucleotide primer sequences utilized for target gene amplification are comprehensively detailed in Table 2 . Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) expression served as the endogenous reference control for normalization of target gene expression.

Cell and tissue lysates were prepared using a protein extraction kit (KeyGEN, Nanjing, China; #KGP250, #KGP950), and protein concentrations were quantified using a BCA assay kit (Beyotime, Shanghai, China; #P0010). Due to equipment constraints, two distinct Western blotting methodologies were employed. A subset of samples was analyzed with an automated capillary electrophoresis system (Simple Western system with Compass software; ProteinSimple, San Jose, CA, USA; Version 5.0.0) using Wes Separation Capillary Cartridges (covering molecular weight ranges of 12–230 kDa and 66–440 kDa; #SM-W004 and #SM-W008, ProteinSimple). In parallel, other samples were prepared by boiling for 5–6 minutes after the addition of 5× protein loading buffer to achieve denaturation. These samples were then separated via SDS-PAGE (loading 30 μg of protein for cells and 60 μg for eyeball tissue) and transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, USA). The membranes were blocked at room temperature for 30 minutes using a protein-free rapid blocking solution, followed by overnight incubation at 4°C with the primary antibody. After washing with Tris-HCl-buffered saline containing Tween 20 (TBST), membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies for 1–2 hours at room temperature. Protein bands were visualized using a gel imaging system (Thermo Fisher, USA) and quantified with ImageJ software (National Institutes of Health, Bethesda, MA, USA).

Confluent orbital fibroblast monolayers in 6-well plates were mechanically disrupted using sterile pipette tips to create uniform linear wounds. Initial wound dimensions were measured immediately post-disruption to establish baseline values. Culture medium was then replaced with fresh medium containing MMP-14 inhibitor NSC-405020 (100 μM) and TGF-β1 (10 ng/mL), followed by incubation for 24 and 48 hours. Wound closure progression was documented via phase-contrast microscopy (Leica Microsystems GmbH, 4× magnification) with quantitative analysis of wound width reduction.

All experiments were conducted with biological triplicates using samples from distinct individuals, with technical duplicates for each condition. Data are expressed as mean ± standard deviation. Statistical analyses were performed using GraphPad Prism v10 (GraphPad Software) employing one-way ANOVA with post-hoc tests. Statistical significance was defined as p<0.05.

Comprehensive transcriptomic analysis was performed on orbital adipose/connective tissue specimens obtained from demographically matched normal control (NC) subjects and patients with distinct GO phenotypes (type I and type II). RNA sequencing identified 15,803 transcripts across all samples ( Figure 1A ). Principal component analysis revealed distinct transcriptional signatures among experimental groups, with the first two components (PC1 and PC2) accounting for 26.55% and 22% of total variance, respectively ( Figure 1B ). Differential expression analysis identified 229 significantly modulated transcripts between GO type I and NC specimens (106 upregulated, 123 downregulated; fold change ≥2.0, P<0.05) ( Figure 1C ). The GO type II versus NC comparison yielded 405 differentially expressed genes, comprising 279 upregulated transcripts—notably including matrix metalloproteinases MMP14, MMP9, and MMP2—and 126 downregulated transcripts (fold change ≥2.0, P<0.05) ( Figure 1D ). Based on these findings, subsequent analyses focused on GO type II versus NC comparisons. KEGG pathway enrichment analysis revealed significant involvement of 54 signaling cascades, predominantly including PI3K-Akt signaling, MAPK signaling, and ECM-receptor interaction pathways ( Figure 1E ). Complementary Gene Ontology analysis demonstrated significant enrichment in biological processes governing extracellular matrix remodeling, cell-matrix adhesion, collagen fibril organization, and cytoskeletal regulation ( Figure 1F ).

Immunohistochemical evaluation of orbital adipose/connective tissue specimens demonstrated distinct MMP14 expression patterns across experimental cohorts. Normal control (NC) and GO type I tissues exhibited minimal MMP14 immunoreactivity, whereas GO type II specimens displayed pronounced MMP14 upregulation (P<0.0001; Figures 2A, B ). Complementary Western blot analysis of protein extracts from orbital adipose tissues stratified by clinical severity (NC, GO grade IV, and GO grade VI) revealed progressive elevation of both α-smooth muscle actin (α-SMA) and MMP14 protein levels in GO specimens compared to controls. Notably, expression intensity of both markers demonstrated a significant positive correlation with histopathological fibrotic grade ( Figures 2C, D ), suggesting mechanistic involvement in disease progression.

Baseline transcriptional analysis revealed significantly elevated MMP14 expression in primary OFs derived from GO patients compared to those isolated from normal control subjects ( Figure 3A ). To establish optimal experimental parameters, CCK-8 viability assays were employed to determine appropriate TGF-β1 concentrations and exposure times, while qPCR was used to assess fibrotic marker expression changes. Based on these preliminary studies, concentrations of 5, 10, and 20 ng/mL TGF-β1 were selected for subsequent experiments ( Supplementary Figures 1A, B ). Differential responsiveness to TGF-β1 stimulation was observed between GO and control-derived OFs. GO fibroblasts exhibited a concentration-dependent upregulation of MMP14 at 10 and 20 ng/mL TGF-β1, whereas control fibroblasts showed no significant expression changes under identical conditions ( Figure 3B ). Furthermore, extended exposure (48 hours) to TGF-β1 induced significant transcriptional activation of multiple fibrotic markers, including COL1A1, CTGF, α-SMA, and MMP14, specifically in GO-derived orbital fibroblasts ( Figures 3C, D ). These findings collectively suggest that TGF-β1-mediated MMP14 induction contributes significantly to the pathological fibrotic response observed in GO orbital fibroblasts.

To elucidate the molecular mechanisms underlying TGF-β1-mediated fibrosis in Graves’ ophthalmopathy, orbital fibroblasts (OFs) isolated from three independent GO patients were exposed to 10 ng/mL TGF-β1 for 24 hours prior to transcriptome analysis by high-throughput RNA sequencing. This systematic approach identified 7,229 significantly differentially expressed genes (DEGs) compared to untreated controls. Cross-referencing these DEGs with the MMP gene database yielded 1,758 MMP-associated transcripts ( Figure 4A ). Principal component analysis demonstrated distinct transcriptional profiles between TGF-β1-treated and control samples, with the first two principal components (PC1 and PC2) accounting for 91.32% and 3.64% of the observed variance, respectively ( Figure 4B ). Differential expression analysis visualized through volcano plotting revealed 412 significantly upregulated and 1,346 downregulated genes (fold change ≥2.0, P<0.05; Figure 4C ).

Protein-protein interaction (PPI) network mapping illuminated extensive functional connections between MMP14 and numerous proteins implicated in fibrotic processes and signal transduction pathways ( Figure 4D ). KEGG pathway enrichment analysis identified 54 significantly modulated signaling networks, with particularly strong enrichment in extracellular matrix-receptor interaction, PI3K-Akt signaling, MAPK signaling cascade, and cell adhesion molecule pathways ( Figure 4E ). Complementary Gene Ontology analysis revealed significant enrichment in biological processes critical to fibrogenesis, including cell adhesion, extracellular matrix organization, and collagen trimer formation ( Figure 4F ). To further characterize MMP14’s role in these regulatory networks, we constructed a correlation network integrating MMP14 with 255 selected genes derived from the enriched pathways, providing insights into potential functional relationships governing fibrotic transformation in GO ( Figure 4G ).

To investigate the functional significance of MMP14 in fibrotic pathogenesis, orbital fibroblasts derived from GO patients were treated with the selective MMP14 inhibitor NCS-405020 in the presence of TGF-β1. Immunoblot analysis revealed that pharmacological inhibition of MMP14 substantially suppressed TGF-β1-induced expression of both MMP14 and the myofibroblast marker α-SMA ( Figures 5A, B ), indicating attenuation of the fibrotic phenotype. The impact on cellular functionality was further assessed using scratch wound migration assays. While TGF-β1 stimulation significantly enhanced orbital fibroblast motility and wound closure capacity, concurrent treatment with NCS-405020 markedly impaired this migratory response. Temporal quantification of wound gap measurements confirmed significant dose-dependent inhibition of cellular migration following MMP14 inhibition ( Figures 5C, D ). These findings collectively demonstrate that MMP14 activity is essential for TGF-β1-mediated fibroblast activation and migration, key processes in the pathological tissue remodeling characteristic of GO.