Orbital connective tissue samples were collected from six GO patients who underwent orbital decompression and from six patients without GO who underwent eyeball enucleation. The baseline characteristics of all patients are listed in Table 1 . The severity and clinical activity of GO were graded according to the NOSPECS classification and the 7-item clinical activity score (CAS) scheme proposed by EUGOGO, respectively (15, 33). All patients signed informed consent forms. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Zhongshan Ophthalmic Center (2016KYPJ028).

Orbital connective tissues obtained from surgery were cut into small pieces and plated in 10-cm culture dishes in high glucose Dulbecco’s Modified Eagle’s Medium (DMEM) containing 20% fetal bovine serum (FBS) and 1% penicillin/streptomycin (all from Gibco Laboratories, New York, USA) at 37°C in a 5% CO₂ humidified incubator. After the cells had migrated out of the tissue pieces and reached confluence, they were passaged with 0.25% trypsin/EDTA (Gibco Laboratories, New York, USA). The cultured cell strains were used between the third and sixth passages. We replaced the culture medium with medium containing 1% FBS before stimulation with TGF-β1 (10 ng/mL) (R&D Systems, Minneapolis, MN, USA) or IL-1β (1ng/mL) (Cell Signaling Technology, Boston, MA, USA). In some experiments, cells were pretreated with DHA (Selleck Chemicals, Shanghai, China), U0126 (ERK inhibitor) (Cell Signaling Technology, Boston, MA, USA), and Stattic (STAT3 inhibitor) (MedChemExpress, Shanghai, China). Detailed experimental conditions are outlined in the main text or figure legends. Each experiment was repeated on at least three independent specimens.

Cellular proliferation was assessed using a Cell Counting Kit-8 (CCK-8) assay kit (Bimake Biotechnology, Houston, Texas, USA). Cells were seeded in 96-well plates and treated with different concentrations of DHA for the indicated times. CCK-8 reagent diluted at a 10:1 ratio was then added to the medium, and the cells were cultured for another 2 h before measuring well absorbance at 450 nm using a Synergy H1 microplate reader (Berten instrument company, Vermont, USA).

DNA synthesis was determined by the 5-ethynyl-2′-deoxyuridine (EdU) incorporation assay using a BeyoClick™ EdU Cell Proliferation Kit with Alexa Fluor 488 (Beyotime Biotechnology, Shanghai, China). Briefly, cells were seeded in 24-well plates and treated with different concentrations of DHA for 24 h. Cells were then exposed to EdU for 2 h, fixed in 4% paraformaldehyde, permeabilized in 0.3% Triton X-100, and stained according to the manufacturer’s instructions. Then counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (Servicebio, Wuhan, Hubei, China) for 5 min. Images were taken using a fluorescence microscope (Nikon Ts2FL, Tokyo, Japan) at 488 nm excitation.

Total RNA was extracted from OFs using an RNA Quick Purification kit (ESscience Biotechnology, Shanghai, China). RNA was reverse-transcribed into cDNA using a PrimeScript RT Reagent Kit (TaKaRa, Dalian, Liaoning, China). Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) was performed using a SYBR Fast qPCR kit (Kapa Biosystems, Boston, MA, USA) according to the manufacturer’s instructions. Glyceraldehyde phosphate dehydrogenase (GAPDH) was used as internal control. The primer pair sequences for qRT-PCR are listed in Table 2 . The data are presented as the means ± SD of triplicate values from three independent experiments.

A protein extraction kit (KeyGEN Biotech, Nanjing, Jiangsu, China) was used to lyse cultured OFs. Protein concentrations were measured with a BCA kit (Beyotime Biotechnology co., Shanghai, China). Equal amounts of protein samples were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), transferred to a polyvinylidene fluoride (PVDF) membrane, and incubated overnight at 4°C with the following primary antibodies: GAPDH (D16H11) XP^® Rabbit mAb, β-Tubulin (9F3) Rabbit mAb, anti-alpha smooth muscle actin antibody [1A4] (α-SMA), CTGF (D8Z8U) Rabbit mAb, Smad2 (D43B4) XP^® Rabbit mAb, phospho-Smad2 (Ser465/467) (138D4) Rabbit mAb, Smad3 (C67H9) Rabbit mAb, phospho-Smad3 (Ser423/425) (C25A9) Rabbit mAb, Stat3 (124H6) Mouse mAb, Phospho-Stat3 (Tyr705) (D3A7) XP^® Rabbit mAb, p44/42 MAPK (Erk1/2) (137F5) Rabbit mAb, Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (D13.14.4E) XP^® Rabbit mAb. All antibodies were purchased from Cell Signaling Technology (Boston, MA, USA), except for anti-α-SMA antibody, which was from Abcam (Cambridge, UK). HRP-conjugated goat anti-rabbit and horse anti-mouse secondary antibodies (Cell Signaling Technology, Boston, MA, USA) were used for detection. Blots were visualized using enhanced chemiluminescence (WBKLS0500, Merck Millipore, MA, USA). PageRuler™ Prestained Protein Ladders #26616, #26619, and #26625 from ThermoFisher Scientific (Darmstadt, Germany) were used to determine molecular weights. Bands were quantified using ImageJ Software (NIH, Bethesda, MA, USA).

OFs were seeded on coverslips within 6-well plates (approximately 5 ×10⁴ cells per well). After the cells reached 50% to 70% confluence, they were treated with DHA (20 μM) with or without TGF-β1 (10 ng/mL). Unstimulated cells served as control. Following treatment for 48h, cells were washed twice with phosphate buffered saline (PBS), fixed with 4% paraformaldehyde for 20 min, permeabilized with 0.3% Triton X-100, blocked with 3% BSA for 30 min at room temperature, and incubated with primary antibodies against COL1A1 (E6A8E) Rabbit mAb (Cell Signaling Technology, Boston, MA, USA) and anti-α-SMA (Abcam, Cambridge, UK). Cells were then washed, incubated with the Alexa Fluor 488 or Alexa Fluor 546-labeled secondary antibody (both from Nanoprobes, New York, USA) for 1 h in the dark at room temperature, and counterstained with DAPI (Servicebio, Wuhan, Hubei, China) for 5 min. Finally, anti-fading agent was added to mount the coverslips. Confocal microscopy (Carl Zeiss 880, Oberkochen, Germany) was used to take immunofluorescence images.

OFs (1 × 10⁵ cells per well) were seeded into 6-well plates and grown to confluence. A straight-line scratch was created using a sterile 200 μL disposable micropipette tip. Cellular debris was removed by washing with PBS twice, and cultures were incubated in DMEM containing 1% FBS. The cells were then treated with DHA (0, 20 μM) with or without TGF-β1 (10 ng/mL) and defined microscopic scratch wound areas were photographed at the indicated times using an inverted phase-contrast microscope.

GO-derived OFs were treated with different concentrations of DHA (0, 10, 20 μM) and 1 ng/mL IL-1β (R&D Systems, Minneapolis, MN, USA) in DMEM supplemented with 1% FBS. After 24 h, cell culture supernatants were collected and centrifuged at 5,000 g for 10 min. HA concentrations in cell culture supernatants were detected with an ELISA kit (R&D Systems, Minneapolis, MN, USA) following the manufacturer’s instructions.

Each experiment was performed in triplicate, and data are expressed as the means ± SD. GraphPad Prism software was used to analyze the results. Statistical analyses were performed by one-way ANOVA or two-way ANOVA, followed by Tukey’s multiple comparison test or Sidak’s multiple comparisons test. P < 0.05 indicated a statistically significant difference.

Primary OF cultures were established from orbital connective tissues resected from GO and non-GO patients ( Table 1 ). Vimentin, which is present in cells of mesoderm origin such as fibroblasts, was primarily used to identify OFs ( Supplementary Figure 1 ). In addition, negative expression of S100, desmin, and cytokeratin served to confirm the exclusion of nerve cells, skin melanocytes, smooth muscle cells, cardiomyocytes, and epithelial cells in these cultures ( Supplementary Figure 1 ).

The effect of DHA on the proliferation of OFs was evaluated using both CCK-8 and EdU assays. CCK-8 assay results showed that at concentrations ≥ 40 μM, DHA significantly inhibited the proliferation of both GO OFs and non-GO OFs at 24, 48, and 72 h post-treatment ( Figure 1A ). Since DHA ≤ 20 μM did not affect the viability of GO and non-GO OFs, 10 and 20 μM DHA doses were applied in subsequent experiments. EdU assays showed that the basal OF proliferation rate was slightly higher in GO than in non-GO OF cultures (P < 0.01). Additionally, results showed that DHA suppressed proliferation and DNA synthesis in both GO and non-GO OFs in a dose dependent manner ( Figures 1B –E ). Moreover, we calculated the decrease in the proportion of EdU positive cells after DHA treatment in both GO OFs and non-GO OFs. Results indicated that the inhibitory effect of DHA on cell proliferation was significantly stronger in GO-OF compared with non-GO OFs (P < 0.05) ( Supplementary Figure 2 ).

To determine whether DHA exerts antifibrotic effects on OFs, the expression of fibrosis-related genes was next examined. As shown in Figures 2A, B , following stimulation with TGF-β1 (10 ng/mL) for 48 h, α-SMA, CTGF, COL1A1, and FN1 mRNA levels were markedly increased in both GO OFs and non-GO OFs. Notably, the mRNA expression of α-SMA and FN1 were significantly higher in GO TGF-β1-treated groups than in non-GO TGF-β1-treated groups (P < 0.0001). However, pre-treatment of cultures with DHA (10 or 20 μM) for 3 h resulted in significant, dose-dependent downregulation of these markers. Moreover, western blot analysis revealed that DHA exposure suppressed α-SMA and CTGF expression at the protein level in OFs subjected to TGF-β1 stimulation ( Figures 2C, D ). The antifibrotic effects of DHA were further assessed by double immunofluorescence staining of α-SMA and COL1A1. As shown in Figures 2E, F , TGF-β1-mediated upregulation of both α-SMA (localized to the cytoskeleton) and COL1A1 (distributed in the cytoplasm) was remarkably repressed by DHA exposure.

Wound healing assays were next performed to assess the effect of DHA on the migration ability of GO-derived OFs. As shown in Figure 3 , by 24 h post-treatment wound closure was slightly higher in TGF-β1-treated compared to control OFs. After 48 h, the wounds were nearly closed in both control and TGF-β1 cultures. In contrast, at both time points, wound closure was significantly delayed in DHA-treated OFs compared with non-DHA-treated cells.

To explore the molecular mechanism involved in the antifibrotic effect elicited by DHA on GO-derived OFs, RNA sequencing was first performed to identify DEGs. A total of 2589 DEGs were detected between cells treated with TGF-β1 alone and those pre-treated with DHA ( Figure 4A ). We further performed function enrichment analyses of these DEGs using the KEGG and GO databases. As shown in Figures 4B, C , TGF-β1-stimulated genes were mainly associated with cell differentiation, migration, and extracellular matrix organization. In turn, pre-treatment with DHA correlated with significant downregulation of genes associated with the MAPK signaling pathway, the ERK1 and ERK2 signaling pathways, as well as fibrosis-associated processes, including OF activation, differentiation, migration, and extracellular matrix organization ( Figure 4D ). Based on TRRUST analysis, the top 20 transcription factors differentially regulated by DHA treatment are shown in Figure 4E .

Next, western blotting was performed to further investigate whether the MAPK/ERK signaling pathway is involved in the antifibrotic effects of DHA on GO-derived OFs. As shown in Figures 4F, G , DHA treatment inhibited TGF-β1-induced phosphorylation of ERK1/2. Previous studies showed that TGF-β1 can promote fibroblast activation and tissue fibrosis by both canonical Smad and non-Smad pathways (41) and that STAT3 activation also plays an important role in TGF-β1 signaling (42). Therefore, we also examined the phosphorylation of Smad2, Smad3, and STAT3 by western blotting. Results demonstrated that DHA obviously reduced TGF-β1-induced phosphorylation of STAT3 ( Figures 4F, G ); however, DHA treatment had a non-significant effect on the phosphorylation of Smad2 and Smad3 (Data not shown).

To confirm whether ERK and STAT3 inactivation are responsible for the antifibrotic effect of DHA on GO-derived OFs, the ERK inhibitor U0126 and the STAT3 inhibitor Stattic were applied to cultured cells. As shown in Figures 5A–C , U0126 and Stattic effectively inhibited the phosphorylation of ERK1/2 and STAT3, respectively. Additionally, both inhibitors suppressed TGF-β1-induced α-SMA and CTGF protein expression ( Figures 5D–G ). Notably, DHA exposure mimicked the effects of U0126 and Stattic by blocking TGF-β1-induced ERK and STAT3 signaling in GO-derived OFs. These data indicate that DHA exerts antifibrotic effects on GO-derived OFs through inhibiting TGF-β1-induced MAPK/ERK and STAT3 signaling.

To investigate whether DHA also exerts anti-inflammatory effects on OFs, we applied qRT-PCR to detect the expression of inflammation-related genes. The changes observed in GO-derived OFs were more remarkable than those seen in non-GO OFs after stimulated by IL-1β (P < 0.0001). Results showed that DHA significantly suppressed IL-1β-induced mRNA expression of inflammatory markers, including IL-6, IL-8, CXCL-1, MCP-1, and ICAM-1, in OFs from both GO and non-GO patients ( Figures 6A, B ).

Hyaluronic acid (HA) is a hydrophilic glycosaminoglycan importantly involved in GO pathogenesis by promoting the enlargement of extraocular muscle bodies (43, 44). Therefore, we detected the mRNA expression of core genes associated with HA production, namely HAS1, HAS2, and HAS3, in cultured OFs. As shown in Figure 6C , HAS1 and HAS3 mRNA expression was upregulated in IL-1β-stimulated, GO-derived OFs, and this effect was inhibited by DHA pre-treatment. In contrast, we found no significant changes in the expression of these genes in non-GO OFs ( Figure 6D ). Furthermore, we detected HA production in culture supernatants from GO-derived OFs via ELISA. Results showed that IL-1β increased HA production in GO-derived OFs, whereas DHA pre-treatment decreased this effect in a dose-dependent manner ( Figure 6E ). These data demonstrated that DHA inhibits IL-1β-induced inflammation and attenuates HA production by GO-derived OFs.