Reagents included Recombinant Human TGFβ1 (PeproTech, Cat# 100-21, Rocky Hill, NJ, USA), Human IL11 Recombinant Protein (Gibco, Cat# PHC0115, Grand Island, NY, USA), and PDK inhibitor dichloroacetic acid (DCA, Sigma-Aldrich, Cat# D54702, St. Louis, MO, USA). Commercial kits included Human IL11 DuoSet ELISA (R&D System, Cat# DY218, Minneapolis, MN, USA), Lactate ELISA kit (Abcam, Cat# ab65331, Cambridge, MA, USA), rtPCR kit (TAKARA, Cat# RR047, Tokyo, Japan), SYBR qPCR kit (TAKARA, Cat# RR820, Tokyo, Japan), RNAsimple Total RNA kit (TIANGEN, Cat# DP419, Beijing, China), and Capillary Western Immunoassay kit (ProteinSimple, 12-230 kDa, San Jose, CA, USA). Primary antibodies targeted αSMA (monoclonal, Abcam, Cat# ab7817, Cambridge, MA, USA), IL11Rα (monoclonal, Abcam, Cat# ab125015, Cambridge, MA, USA), IL11 (Abcam, Cat# ab187178, Cambridge, MA, USA), PDK2 (Abcam, Cat# ab68164, Cambridge, MA, USA), and GAPDH (CST, Cat# 2118, Danvers, MA, USA). Validations of the primary antibodies are provided on the manufacturer’s website.

Orbital fat/connective tissues and peripheral blood mononuclear cell (PBMC) were collected from 20 GO patients during orbital decompression surgery and from 10 control subjects during strabismus or orbital surgery. For GO subjects, the inclusion criteria were patients with extraocular muscle enlargement on CT/MRI scan and limited globe motility; the exclusion criteria were patients who received orbital radiotherapy or received effective systemic steroid therapy within 3 months. For control subjects, the inclusion criteria were patients with concomitant strabismus, cosmetic orbital decompression, enucleation due to trauma, or blepharoplasty; the exclusion criteria were patients with orbital inflammatory disease, thyroid dysfunction, orbital infection, or previous intraorbital malignant tumors. The baseline characteristics of the enrolled subjects are summarized in Table 1 . Written informed consent was obtained from each subject, and the study protocol was approved by the Institutional Review Board of Fudan Eye and ENT Hospital.

The connective tissue was fixed with neutral formalin, embedded in paraffin, and sectioned in the sagittal or horizontal plane. The slides were processed with hematoxylin and eosin (HE) staining, Masson trichrome staining, PDK2, IL11, and IL11Rα immunohistochemical staining. Brightfield images were taken with a microscope (Leica Microsystems). ImageJ (version 1.54k) was used for calculating the region of interest (ROI). Threshold was adjusted manually to accord with connective tissue region and staining positive region, and the integrated option density (IOD) and the area of target protein distribution (area) could be obtained. Average optical density (AOD)=IOD/area.

Orbital connective tissue was minced into 5-mm³ pieces and placed in six-well format containing high glucose Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, Cat# 11965, Grand Island, NY, USA) supplemented with 20% fetal bovine serum and 1% penicillin/streptomycin (Cytiva, Cat# SV30010, Wilmington, DE, USA). After reaching 90% confluence, the primary OFs were passaged with 0.25% trypsin (Gibco, Cat# 25200, Grand Island, NY, USA) and maintained in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin. The cells were grown and maintained at 37°C with 5% CO₂. The culture fluid was renewed every 2–3 days. All the experiments were carried out at low passage (P3–P8). Cells were serum starved for 16 h before stimulations.

Primary OFs were seeded on glass coverslips at 8×10³ per well. Cells were fixed with 4% paraformaldehyde for 20 min, permeated with 0.5%Triton for 30 min, and blocked with 3% BSA for 1 h at room temperature. Cells were incubated with primary antibodies (1:200 anti-αSMA) at 4°C for 12–16 h. After extensive rinsing, coverslips were incubated with the Alexa Fluor 555-labeled secondary antibodies (Invitrogen, Cat. No. A-21422) in the dark at room temperature for 1 h and stained with DAPI (1:1,000) for 5 min. A confocal microscope (Leica TCS SP2; Leica Microsystems) was used for imaging. Cell morphological changes were quantitatively described by cell length–width ratio.

Culture supernatant was collected and assayed according to the manufacturer’s protocol. The R&D IL11 ELISA kit was adapted to measure IL11 concentrations.

Primary OFs were seeded on a 15-cm dish. OFs were deprived of serum for 12 h after reaching 95% confluence and incubated with siRNA for 12 h. TGFβ1 (10ng/ml) ± DCA (5mM) were given in 1% FBS and harvested for intracellular lactate measurement after 12 h according to the instruction of Lactate ELISA kit.

Protein samples were extracted from unfrozen cells with RIPA reagent (Beyotime, Cat. No. P00138), quantified using BCA protein assay kit (Beyotime, Cat. No. P0012), and immunodetected with a Western blot system (WES; ProteinSimple) according to the manufacturer’s protocol. The relative amount of each immunoreactive band was quantified by signal intensity and normalized to GAPDH in the same sample. Different loading concentration of protein samples was tested by a titration experiment. The suitable concentration was determined as 0.5 μg/μl for αSMA (1:10), IL11Rα (1:10), PDK2 (1:10), and GAPDH (1:50).

Total RNA was extracted using an RNAsimple Total RNA kit (Tiangen, Cat. No. DP419) and reverse transcribed to cDNA with a rtPCR kit (TAKARA, Cat. No. RR047).

Real-time quantitative PCR was performed on a CFX96 Real-Time System (C1000 Touch, Thermal Cycler) with the SYBR qPCR kit. Primer sequences used are shown in Table 2 . The amplification efficiency was evaluated by the standard curve method. The mRNA level was normalized to GAPDH by the −ΔΔCT method.

Primary OFs were seeded in six-well plates to reach 80% confluence and transfected with Lipofectamine 3000 according to the manufacturer’s instruction. Briefly, after 6-h starvation, the OFs were treated with Opti-MEM (Gibco, Cat. No. 3198508) containing PBS (blank group), Lipofectamin 3000 + scramble siRNA (RIBOBIO, siN0000001; negative control group), Lipofectamin 3000 + PDK2 siRNA (sense 5′→3′: GACCGAUGCUGUACUCUAUTT; antisense 5′→3′: AUAGAUGACAGCAUCGGUCTT), Lipofectamin 3000 + IL11 siRNA (RIBOBIO, stB0006813B), and Lipofectamin 3000 + GAPDH siRNA (RIBOBIO, siP0000001; positive control group) for 24 h. The transfected cells were treated with other reagents for further experiments.

Four- to six-week-old female NOD/ShiLtJGpt-Prkdcem26Cd52Il2rgem26Cd22/Gpt (NCG) mice purchased from GemPharmatech Laboratories were used in the xenograft experiments. NCG mice were preconditioned with irradiation (1.2 Gy). A total of 1×10⁷ human PBMC were transplanted intravenously into each mouse within 24 h after irradiation (30). Flow cytometry was performed to detect the proportion of human CD45⁺ cells in the peripheral blood after 3 weeks. Mice with over 25% hCD45⁺ cells were considered humanized NCG mice.

NCG mice were kept on a standard 12-h light–dark cycle. The human orbital tissue (fat and connective tissues, no eye muscle tissues) was obtained at orbital decompression surgery. These tissues were cut into small pieces of 1 × 1 × 1 mm³ and xenografted into orbit of mice at the same time with intravenous injection of PBMC. Mice were observed once a week and were sacrificed after 3 weeks. HE and Masson staining were performed on collected orbital connective tissues. Every 3 days, 500 mg/kg of DCA or PBS was administered by gavage.

A 7-T Bruker Clinscan animal MRI scanner (Bruker BioSpin MRI GmbG, Germany) equipped with a four-channel phase-array surface coil was used to examine orbital connective tissues of mice. Isoflurane were used to anesthetize the mice. The mice were placed on a dedicated mice scan bed (Bruker, Ettlingen, Germany) and T2-weighted TurboRARE sequence scanning (TR/TEeff = 2,700/30 ms, RARE factor = 8, NA = 12, spatial resolution = 0.62 mm × 0.62 mm × 0.3 mm, slices = 7, no gaps, scan time 16 min) was applied to them. Two independent observers (observer 1 had 5 years of experience in orbital radiology; observer 2 had 3 years of experience in orbital radiology) manually delineate ROI. Two small circular ROI (0.09–0.10 mm²) were placed in extraocular muscle region or on the ipsilateral white matter of brain as the calibration function. Signal intensity of orbital fat and brain was measured, and the ratio was obtained as “OD_{Orbital Fat/Brain}/OS_{Orbital Fat/Brain}.” Take the ratio between the right eye and left eye and each mouse’s tissue-specific signal intensity ratio (Signal Intensity_{Orbital Fat/Brain}) was acquired.

Statistical software was GraphPad Prism 8.0.2 (version 26.0). All continuous variables with normal distribution were shown as mean ± standard error of mean (SEM). Shapiro–Wilk test was applied to assess normality of data. Student’s t-test or Welch’s t-test was applied to compare two independent experimental groups. Paired t-test was applied to compare data from GO or control OFs. Pearson correlation and linear regression were applied to analyze correlation between two continuous variables. Statistical significance was defined as p-value < 0.05.

Histological staining images were analyzed to compare collagen (Masson staining, blue), PDK2 (brown), IL11 (brown), and IL11Rα (brown) expression. As shown in Figure 1A , PDK2, IL11, and IL11Rα expression was augmented in the GO group. The average optical density (AOD) of PDK2 in GO group varied from 0.050 to 0.387. AOD in the control group varied from 0.091 to 0.290. AOD of IL11 in the GO group ranged from 0.055 to 0.188 while that in the control group ranged from 0.074 to 0.122. As for IL11Rα, AOD was 0.121–0.316 in the GO group and 0.117–0.187 in the control group. Consistent with previous views (31), collagen expression was elevated in GO extraocular muscles ( Figure 1A ), which contributes to the occurrence of GO myopathy. Correlation analysis among PDK2, IL11, and IL11Rα in the GO group indicated a significant positive correlation between PDK2 and IL11, same as PDK2 and IL11Rα ( Figures 1B, C ). However, analysis between IL11 and IL11Rα showed no significant relativity between IL11 and IL11Rα ( Figure 1D ). The expression levels of three markers in the GO group were significantly higher than those in the control group ( Figure 1E ). These results demonstrate that IL11 signaling may contribute to pathogenesis of GO fibrosis through PDK2-related glycolysis-facilitated pathway.

TGFβ1 is the most critical fibrotic cytokine in GO. We stimulated OFs with human recombinant TGFβ1, and αSMA expression was elevated in both GO and control groups ( Figure 2A ), consistent with our pervious study (9, 10). In the following experiments, DCA was adapted as a glycolysis inhibitor. After incubation with TGFβ1 (10 mg/ml) ± DCA (5 mM) for 48 h, αSMA (ACTA2) expression was measured. The results demonstrated a significantly decreased αSMA expression when DCA was added at the protein level and mRNA level in both GO groups and control groups compared with OFs stimulated with TGFβ1 alone ( Figures 2B ). When OFs differentiate into myofibroblasts, their length-to-width ratio decreases. Immunofluorescence staining and cell morphological change also showed the same result ( Figures 2C, D ). These results suggest that suppression of glycolysis can inhibit TGFβ1-induced fibrosis of OFs.

The effects of PDK2 on glycolysis were explored in our previous study (9). Quantitative PCR and WES assay confirmed that 12-h transfection and translation with PDK2 siRNA significantly suppressed PDK2 transcription in both GO and control OFs ( Figures 3A, B ). Knockdown of PDK2 or IL11 or addition of DCA resulted in significantly decreased lactate production comparing with incubation with TGFβ1 alone in GO OFs. However, lactate production showed no significant change in control OFs ( Figure 3C ). Then, we compared fibrotic effect under a different treatment. Knockdown of PDK2 exerted inhibitive effects on the profibrotic effects of TGFβ1 in GO OFs at both mRNA and protein levels ( Figures 3D, E ). Immunofluorescence and length-to-width ratio revealed decreased αSMA expression and inhibited cell morphological change ( Figures 3F, G ). Collectively, the above results confirmed that PDK2-enhanced glycolysis can promote fibrosis in GO OFs.

IL11 has been proven to induce trans-differentiation of OFs to myofibroblasts (29). To explore whether IL11 is related to fibrotic process promoted by glycolysis, we compared the IL11 and IL11Rα mRNA levels of OFs by RT-qPCR between 48-h TGFβ1 (10 ng/ml) ± DCA (5 mM) treatment in the GO and control groups. The IL11 mRNA level was significantly lower in OFs treated with TGFβ1 + DCA than those treated with TGFβ1 alone ( Figure 4A ) in both groups. As for protein level, we detected the concentration of IL11 in the supernatant with ELISA, and the result corresponded with mRNA level ( Figure 4B ) in GO groups. While in control groups, cell strains presented inconsistency among each other. In addition, DCA decreased IL11Rα levels in both GO and control groups ( Figure 4C ). However, as for mRNA levels, no such tendency appeared ( Figure 4D ). These results suggest that enhanced glycolytic contributes to elevated IL11 expression.

To decipher the effect of IL11 on glycolysis-facilitated fibrosis, we compared the differentiation-promoting effect of IL11 to OFs. Immunofluorescence staining and length-to-width ratio revealed that the formation of SMA-positive fibers were increased and OFs’ differentiation with addition of IL11 in both GO and control groups ( Figures 5A, B ). Western blot and RT-qPCR analysis also confirmed that IL11 presented profibrotic effect, which was stronger under conditions of strong glycolysis ( Figures 5C, D ). These results indicate that the IL11 signaling elicits a fibrotic response in OFs and is more pronounced in GO.

The effects of IL11 on TGFβ1 induced myofibroblast differentiation were explored by siRNA transfection. Quantitative PCR of IL11 siRNA group verified a successful transduction ( Figure 6A ). After knockdown of IL11, we stimulated the OFs with TGFβ1 (10 ng/ml), DCA (5 mM), and TGFβ1 (10 ng/ml) + DCA (5mM) for 24 h to test mRNA level and 48 h to test protein level. Although αSMA (ACTA2) expression was still elevated, it was significantly reduced compared with that before knockdown in GO OFs ( Figures 6B, C ). The IL11 concentration in supernatant tested by ELISA showed a significant decrease when IL11 was knocked down ( Figure 6D ). The fluorescence immunostaining and cellular length-to-width ratio also exhibited similar results ( Figures 6E, F ). Interestingly, in several of the control groups, the TGFβ1-induced expression of αSMA was not significantly inhibited. This may be due to the weak glycolytic level in the control group, resulting to a less strong inhibition effect than in the GO group and indicating that IL11 signaling an important component of glycolytic-facilitated fibrosis. Taken together, IL11 increases sensitivity of GO OFs to TGFβ1, which may be a component therapeutic target for GO treatment.

We xenografted the GO orbital tissue into orbit of mice to build in vivo models. The fibrotic tissues exhibited lower signal intensity compared to control ones ( Figure 7A ), which was consistent with previous studies (32). The results demonstrated that relative signal intensity between the xenografted eye and control eye significantly decreased in the anti-glycolysis group ( Figures 7A, B ). HE staining confirmed the fibrotic effect of the GO orbital tissue orthotopic xenograft into the murine orbit ( Figure 7C ). We compared the ratio of fibrosis area and human orbital fat area in each group, and the result confirmed that inhibition of glycolysis by DCA could restrain orbital connective fibrosis in vivo ( Figures 7C, D ).