HIF2A–LOX Pathway Promotes Fibrotic Tissue Remodeling in Thyroid-Associated Orbitopathy

Abstract

Thyroid-associated orbitopathy (TAO) is a disfiguring periocular connective tissue disease associated with autoimmune thyroid disorders. It is a potentially blinding condition, for which no effective pharmacological treatment has been established. Despite a suggested role played by autoimmune thyrotropin receptor activation in the pathogenesis of TAO, the cellular and molecular events contributing to the fibrotic and inflammatory disease process of TAO are not fully defined. By developing a three-dimensional organoid culture of human orbital fibroblasts (OFs), we sought to determine the molecular mechanism underlying the fibrotic disease process of TAO. In this ex vivo model, we have demonstrated that hypoxia-inducible factor (HIF) 2α (HIF2A), but not its paralog HIF1A, accelerates extracellular matrix (ECM) deposition by inducing a collagen–cross-linking enzyme, lysyl oxidase (LOX). Inhibiting HIF2A and LOX with short hairpin RNA or small molecular antagonists effectively ameliorated fibrotic disease process within TAO organoids. Conversely, the overexpression of a constitutively active HIF2A in mouse OFs was sufficient to initiate LOX-dependent fibrotic tissue remodeling in OF organoids. Consistent with these findings, HIF2A and LOX were highly expressed in human TAO tissues paralleling excess ECM deposition. We propose that the HIF2A–LOX pathway can be a potential therapeutic target for the prevention and treatment of TAO.

Thyroid-associated orbitopathy (TAO) is a disfiguring and potentially blinding eye condition observed in autoimmune thyroid diseases, that is, Graves’ disease and Hashimoto’s thyroiditis (1, 2). Autoimmune activation of TSH receptor (TSHR) contributes to the pathogenesis of TAO (3–6); however, the downstream molecular and cellular events responsible for fibrotic tissue remodeling in TAO are not well defined. Recent advances in mouse models of disease allow us to better understand TSHR-dependent inflammatory disease process in TAO (7). Although animal models in general are effective in reproducing disease phenotype at the organ level, the presence of modifying factors, such as genetic background and gut microbiota, as well as interspecies difference in the proteome, could pose challenges to therapeutic screening of molecular targets in vivo (8).

A three-dimensional (3D) tissue culture system has emerged as a novel ex vivo approach to modeling human diseases and screening small molecules for therapeutic potentials (9). The 3D culture technique allows homotypic and heterotypic cell–cell interactions within the network of extracellular matrix (ECM) molecules mimicking in vivo–like tissue context (10). In this study, we sought to reproduce fibrotic tissue remodeling observed in TAO by culturing human orbital fibroblasts (hOFs) as a 3D organoid in a high-throughput hanging droplet culture. By using this system, we observed the role of cell-derived ECM in defining the ex vivo phenotype of TAO-derived hOFs. TAO-derived hOF organoids uniquely recapitulated the excess deposition of ECM, increased tissue stiffness, and proinflammatory gene expression observed in TAO (11, 12).

Adipose tissues become fibrotic and proinflammatory under nutritional stress and in disease states (13). Adipose ECM remodeling is determined by a balance between ECM deposition and turnover (14). The master regulator of ECM dynamics, however, has not been well defined. Hypoxia-inducible factors (HIFs) are basic helix-loop-helix Per-Arnt-Sim transcription factors, which regulate cellular metabolism and ECM remodeling in hypoxic conditions and disease states (15). HIF1α (HIF1A) and HIF2α (HIF2A) are two major basic helix-loop-helix Per-Arnt-Sim transcription factors responsible for hypoxia-inducible gene regulation. Their downstream target genes are mostly shared, but some genes are unique thereby conferring divergent phenotypes following the activation of HIF1A vs HIF2A (15). Adipocyte-specific overexpression of HIF1A leads to adipose tissue fibrosis and insulin resistance (16), suggesting a role played by HIF family members in connective tissue remodeling. A recent study demonstrates that hOFs derived from Graves’ ophthalmopathy display augmented induction of HIF1A under a hypoxic condition (17).

In this study, we hypothesize that HIFs may contribute to the fibrotic tissue remodeling of orbital adipose tissues in TAO. Among the downstream targets of HIFs, lysyl oxidase (LOX), an enzyme that cross-links collagen fibrils, mediates HIF-dependent tissue fibrosis (16, 18, 19). Using a hanging-droplet organoid culture of hOFs and genetically modified mouse-derived orbital fibroblasts (mOFs), we have demonstrated that HIF2A, but not HIF1A, induces LOX to promote fibrotic ECM remodeling, and thus increases tissue stiffness. Consistent with ex vivo findings, the HIF2A–LOX pathway was found to be highly upregulated in human TAO tissues paralleling excess ECM deposition.

Materials and Methods

Materials

Materials included DMEM (no. 11965092; Gibco/Thermo Fisher Scientific, Waltham, MA), fetal bovine serum (no. 16-000-044; Gibco/Thermo Fisher Scientific), l-glutamine (no. 25030081; Gibco/Thermo Fisher Scientific), antibiotic/antimycotic (no. 15240062; Gibco/Thermo Fisher Scientific), penicillin/streptomycin (no. 15140122; Gibco/Thermo Fisher Scientific), Ficoll-Paque Plus (no. 17-1440-03; GE Healthcare, Piscataway, NJ), puromycin (no. P8833; Sigma-Aldrich, St. Louis, MO), protamine sulfate salt from salmon (no. P4020; Sigma-Aldrich), Methocel A4M (no. 94378; Sigma-Aldrich), dexamethasone (no. D1756; Sigma-Aldrich), triiodothyronine, T3 (no. T6397; Sigma-Aldrich), troglitazone (no. 71750; Cayman Chemical, Ann Arbor, MI), porcine insulin (no. I5523; Sigma-Aldrich), TSH from bovine pituitary (no. T8931; Sigma-Aldrich), HIF2 antagonist 2 (no. SML0883; Sigma-Aldrich), GM6001 (no. 364206; Calbiochem, San Diego, CA), 3-aminopropionitrile fumarate salt [β-aminopropionitrile (BAPN), no. A3134; Sigma-Aldrich], and M22 TSH-stimulating human monoclonal antibody (RSR Diagnostics for Autoimmunity, Cardiff, United Kingdom).

hOF isolation and culture

Orbital adipose tissues were obtained as surgical waste samples from de-identified euthyroid patients with Graves’ orbitopathy (GO) undergoing orbital decompression and from healthy subjects without inflammatory orbital disease, who underwent cosmetic eyelid surgery. hOFs were isolated and grown as previously described (20). Briefly, tissues were minced into small pieces, placed on 150-mm culture dishes, and submerged in growth medium (DMEM supplemented with 10% v/v fetal bovine serum, 1% v/v l-glutamine, 1% v/v antibiotic/antimycotic) at a sufficient depth to cover the tissue chunks. Explants were cultured in a humidified incubator (at 37°C with 5% CO₂), with growth medium changed every 2 to 3 days. Orbital fibroblasts (OFs) from five GO patients and five healthy patients were isolated and expanded for subsequent experiments. All experiments were performed using OFs of three to six passages after the initial cell isolation.

Mouse OF culture

ROSA26-STOP-HIF2dPA (flox/flox) mice were provided by Dr. Ernestina Schipani (University of Michigan) and used for the isolation of orbital adipose tissue. Cells isolated from this depot were cultivated as previously reported (21) and cultured in a standard growth medium. A degradation-resistant HIF2A mutant was induced by adenoviral Cre [vector; Ad5 cytomegalovirus (CMV)-Cre] recombinase. Adenoviral GFP (vector; Ad5 CMV eGFP.dlE3) or empty vector (Ad5 CMV pLpA.dIE3) was used as control.

3D culture of organoids

A hanging-droplet spheroid culture system was used to generate 3D organoids. To facilitate stable morphology, methylcellulose (Methocel A4M) was added to growth medium. Prior to being seeded to a hanging drop culture plate (no. HDP1385; Sigma-Aldrich), cells were cultured in 100-mm or 150-mm dishes until reaching ∼90% confluence. After washing with HBSS, cells were detached using 0.25% trypsin/EDTA and resuspended in growth medium. After centrifugation for 5 minutes at 300g, the cell pellet was resuspended in growth medium containing 0.25% w/v Methocel A4M. Volume was adjusted such that 20,000 cells were contained in 25 μL of solution, and 25-μL drops were placed into each well of the drop culture plate (defined as day 0). Organoid medium (i.e., growth medium with 0.25% w/v Methocel A4M) was used throughout the duration of spheroid culture. Every day, 14 μL of culture medium was removed and 14 μL of fresh culture medium was added to each well.

Adipogenic differentiation of organoids and treatment with inhibitors

Adipogenic differentiation was induced by an adipogenic cocktail containing 250 nM dexamethasone, 10 nM T3, and 10 μM troglitazone. For the analysis of lipid droplet formation, organoids were transferred to ultra-low attachment six-well plates and stained by boron-dipyrromethene (BODIPY; no. D3922; Thermo Fisher Scientific) at a 1:500 ratio by volume for 1 hour. Fluorescence of BODIPY-stained lipid droplets was detected with a Nikon A1 confocal microscope (Nikon, Tokyo, Japan) and quantified with ImageJ software v1.51n (National Institutes of Health, Bethesda, MD).

Microindentation force measurement

The mechanical testing of organoids was performed using MicroSquisher (CellScale, Waterloo, ON, Canada) as recently reported (22). The device consists of a microscale compression system equipped with a 406-μm diameter cantilever. A single organoid was placed on a 3-mm × 3-mm plate for each measurement. Organoids were compressed to 50% deformation (as determined by microscopic camera) for 20 seconds. The force required to produce 50% strain was measured through the cantilever, and data were reported as force/displacement (μN/μm).

Sirius red staining

For Sirius red staining, sections of orbital adipose tissue were incubated in a solution consisting of 0.1% Direct Red 80 (Sigma-Aldrich) in picric acid for 1 hour at room temperature with agitation. They were next transferred to a 0.5% glacial acetic acid solution and incubated for 10 minutes at room temperature with agitation. Sections were then washed briefly in tap water, dehydrated through an ethanol series, briefly incubated in xylene, and coverslipped with Permount (Thermo Fisher Scientific, Carlsbad, CA) for microscopic analysis.

Immunostaining of organoids and tissues

Immunofluorescent staining of HIF2A, HIF1A, LOX, type I collagen (COL1), type IV collagen (COL4), type VI collagen (COL6), and fibronectin (FN) protein in orbital adipose tissues was performed on paraffin-embedded sections of orbital tissues from 4 to 5 non-GO subjects and 4 to 10 GO patients. Sections (7 μm thick) of paraffinized orbital adipose tissue were deparaffinized, rehydrated, and permeabilized with cold acetone for 30 seconds. After blocking with 5% normal goat serum/0.1% Triton X-100 in PBS (PBST) for 1 hour at room temperature, sections were incubated overnight at 4°C with rabbit anti-HIF2A monoclonal antibody at 1:200 dilution [no. A700-003, RRID: AB_2631884, Bethyl Laboratories, Montgomery, TX (23)], rabbit anti-HIF1A monoclonal antibody at 1:200 dilution [no. A700-001, RRID: AB_2631882, Bethyl Laboratories (24)], mouse anti-LOX antibody [no. sc-373995, RRID: AB_10915622, 1:200, Santa Cruz Biotechnology, Santa Cruz, CA (25)], rabbit anti–collagen I antibody [no. 600-401-103-0.5, RRID: AB_217595, 1:200, Rockland Immunochemicals, Limerick, PA (26)], rabbit anti–collagen IV antibody [no. 600-401-106-0.5, RRID: AB_217598, 1:200, Rockland Immunochemicals (27)], rabbit anti–collagen VI antibody [no. 600-401-108-0.1, RRID: AB_217576, 1:200, Rockland Immunochemicals (28)], or mouse anti-FN antibody [no. sc-8422, RRID: AB_627598, 1:200, Santa Cruz Biotechnology (29)]. After a subsequent wash in PBST, sections were incubated with secondary goat Alexa Fluor 488 anti-rabbit IgG [no. A-11070, RRID: AB_142134, 1:1000, Invitrogen, Carlsbad, CA (30)] or goat Alexa Fluor 594 anti-mouse IgG [no. A-11020, RRID: AB_141974, 1:1000, Invitrogen (31)] for 1 hour at room temperature. Slides were counterstained with 4′,6-diamidino-2-phenylindole [DAPI; no. D1306, RRID: AB_2629482, Invitrogen (32)] and mounted in ProLong Gold antifade reagent (no. P36931, Invitrogen).

Organoids were fixed in 4% paraformaldehyde/PBS overnight with or without permeabilization in 0.5% Triton X-100 in PBS for 1 hour. To stain extracellular collagen and FN, no permeabilization was performed, and all the detergents were excluded from the subsequent procedures. After blocking in 3% BSA/0.1% PBST for 3 hours at room temperature, organoids were washed three times for 30 minutes with PBST. Samples were then incubated with primary antibody overnight at 4°C. The catalog numbers and the dilution of primary antibodies were as follows: rabbit anti-HIF2A monoclonal antibody (no. A700-003, 1:200) and rabbit anti-HIF1A monoclonal antibody (no. A700-001, 1:200), both from Bethyl Laboratories; rabbit anti–collagen I antibody (no. 600-401-103-0.5, 1:200), rabbit anti–collagen III antibody [no. 600-401-105-0.1, RRID: AB_217573, 1:200 (33)], rabbit anti–collagen IV antibody (no. 600-401-106-0.5, 1:200), rabbit anti–collagen V antibody (no. 600-401-107-0.1, 1:200), and rabbit anti–collagen VI antibody (no. 600-401-108-0.1, 1:200), all from Rockland Immunochemicals; rabbit anti–α smooth muscle actin antibody [no. 5694, RRID: AB_2223021, 1:100 (34)], and rabbit anti-Ki67 antibody [no. 15580, RRID: AB_443209, 1 μg/mL (35)], both from Abcam, Cambridge, UK; rabbit anti–cleaved caspase-3 antibody [no. 9664, RRID: AB_2070042, 1:400, Cell Signaling Technology, Danvers, MA (36)]; mouse anti-FN antibody (no. sc-8422, 1:200), mouse anti–connective tissue growth factor (CTGF) antibody [no. sc-365970, RRID: AB_10917259, 1:200 (37)], mouse anti-LOX antibody (no. sc-373995, 1:200), and mouse anti–Thy-1 antibody [no. sc-19614, RRID: AB_2116695, 1:200 (38)], all from Santa Cruz Biotechnology. After a subsequent wash in PBST, organoids were incubated with goat Alexa Fluor 488 anti-rabbit IgG (no. A-11070, 1:500, Invitrogen) or goat Alexa Fluor 594 anti-mouse IgG (no. A-11020, 1:500, Invitrogen) for 3 hours at room temperature. Alexa Fluor 594 phalloidin [no. A12381, RRID: AB_2315633, Invitrogen (39)] was used for F-actin staining and DAPI (no. D1306, Invitrogen) was used for nuclear staining. Samples were mounted in ProLong Gold as indicated above.

For two-dimensional (2D) cell culture, cells were grown using a four-well chamber slide (no. 154526PK, Thermo Fisher Scientific) until 80% to 90% confluence and fixed with 4% paraformaldehyde for 1 hour at room temperature. After repeated washes in PBST, cells were permeabilized with 0.3% Triton X-100 for 5 minutes. When staining extracellular collagen and FN, the permeabilization step was omitted and detergents were excluded in subsequent procedures. After blocking with 1% BSA for 1 hour at room temperature, cells were incubated with primary antibody overnight at 4°C. After repeated washes, samples were incubated with secondary antibody (1:1000) Alexa Fluor 594 phalloidin (no. A12381, Invitrogen) for F-actin. Cells were then incubated in DAPI for 1 hour at room temperature before mounting with ProLong Gold antifade reagent (no. P36931, Invitrogen).

Western blot analysis of HIF2A

Whole-cell lysate was extracted from 100 organoids in 250 µL of RIPA buffer (no. 89900, Thermo Fisher Scientific) with a proteinase inhibitor, cOmplete Mini (no. 11-836-170-001, Sigma-Aldrich). After centrifugation at 14,000g at 4°C for 30 minutes to remove cellular debris, supernatants were subjected to Western blotting (20 µg of protein per lane). Protein samples were then separated by standard SDS-PAGE in 7.5% acrylamide gel followed by transfer to nitrocellulose membrane (Bio-Rad Laboratories, Hercules, CA). Primary antibodies recognizing rabbit anti-HIF2A monoclonal antibody (no. A700-003, 1:500) and mouse anti–α-tubulin antibody [no. sc-32293, RRID: AB_628412, 1:2000 (40)] were used for immunoblotting. Membranes were treated with horseradish peroxidase–conjugated secondary antibodies: anti-rabbit IgG antibody [no. 170-6515, RRID: AB_11125142, 1:5000, Bio-Rad Laboratories (41)], anti-mouse IgG antibody [no. 170-6516, RRID: AB_11125547, 1:5000 (42)], and chemiluminescent signal was developed using ECL Western blotting substrate (Pierce/Thermo Fisher Scientific) for detection with FluorChem M (ProteinSimple, Santa Clara, CA). Quantification of Western blotting signals was conducted with ImageJ software.

Image acquisition and analysis

Bright-field images of each organoid were captured in ×4 or ×10 objective lenses using an inverted microscope (Olympus IX70). The largest cross-sectional area (CSA) was calculated using ImageJ software version 1.51n. Immunofluorescent images were obtained using a Nikon A1 confocal microscopy and NIS-Elements 4.0 software. Images in two dimensions were acquired with a ×20 air objective or ×100 oil objective with a resolution of 1024 × 1024 or 2048 × 2048 pixels. For images of organoids, serial z-axis imaging (2.2-μm interval) at a 65-μm range from a surface of organoids was conducted using a ×20 air objective with a resolution of 512 × 512, 1024 × 1024, or 2048 × 2048 pixels and was converted as Z-stack image using the maximum intensity projection feature of NIS-Elements 4.0 software. Intensity of immunofluorescent target proteins was quantified using ImageJ. Signal intensity of organoids was expressed as intensity/surface area measured at 65 μm from the top of the organoid in the z-plane. The surface area was calculated as follows: surface area = D × A/(A + π × H²), where D (μm) indicates organoid diameter, A (μm²) indicates area of sectioned organoid, and H (μm) indicates height (=65 μm).

Lentiviral gene silencing

For HIF2A knockdown, lentiviruses carrying three unique HIF2A short hairpin RNA (shRNA) constructs in pLenti-GipZ-CMV-Puro (GE Healthcare) and pLenti-LKO.1-Puro (Sigma-Aldrich) were used with 50 μg/mL protamine for 16 hours. For HIF1A knockdown, lentivirus carrying two unique HIF1A shRNA in pLenti-LKO.1-Puro vectors (Sigma-Aldrich) were used. shRNA sequences are as follows: HIF2A knockdown: no. 1, Lenti-GipZ-HIF2A-VSVG, 5′-GCATTAAAGCAGCGTATC-3′; no. 2, Lenti-LKO-HIF2A-3805-VSVG, 5′-CCGGGCGCAAATGTACCCAATGATACTCGAGTATCATTGGGTACATTTGCGCTTTTT-3′; no. 3, Lenti-LKO-HIF2A-3806-VSVG, 5′-CCGGCAGTACCCAGACGGATTTCAACTCGAGTTGAAATCCGTCTGGGTACTGTTTTT-3′; HIF1A knockdown: no. 1, Lenti-LKO-HIF1A-3810-VSVG, 5′-CCGGGTGATGAAAGAATTACCGAATCTCGAGATTCGGTAATTCTTTCATCACTTTTT-3′; no. 2, Lenti-LKO-HIF1A-10819-VSVG, 5′-CCGGTGCTCTTTGTGGTTGGATCTACTCGAGTAGATCCAACCACAAAGAGCATTTTT-3′. After transduction, selection was performed using 1 μg/mL puromycin.

Gene expression analysis

Total RNA was extracted from 20 organoids using an RNeasy mini kit (Qiagen, Valencia, CA). Reverse transcription was performed with the SuperScript II kit (Invitrogen) as per the manufacturer’s instructions. Respective gene expression was quantified by real-time PCR with either Power SYBR Green or Universal TaqMan Master mix using a StepOnePlus machine (Applied Biosystems/Thermo Fisher Scientific). cDNA quantities were normalized to the expression of housekeeping gene 36B4 (Rplp0) and are shown as fold change relative to control. DNA sequences of primers and TaqMan probes are shown in Supplemental Table 1.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 7 (GraphPad Software, San Diego, CA). For comparison of two mean values, a two-tailed unpaired Student t test or nonparametric Mann-Whitney test was used to calculate statistical significance with a confidence level >95%. To analyze the difference among multiple groups with unmatched sample number, a nonparametric Kruskal-Wallis test followed by a post hoc Dunn multiple comparison test was used; for matched multiple group comparison, one-way ANOVA followed by a Tukey multiple comparison test was used. Data are presented as arithmetic means ± SEM.

Study approval

The use of de-identified human samples for cell isolation and histology was approved by the University of Michigan Institutional Review Board. Animal study and procedures were approved by University of Michigan Institutional Animal Care and Use Committee.

Results

3D organoids made of hOFs reproduce in vivo–like tissue remodeling and stiffness

We isolated primary hOFs from de-identified surgical waste of nasal superior orbital adipose tissues obtained through decompression surgeries for GO and blepharoplasties for healthy controls. Under standard 2D culture conditions, healthy OFs (N-OFs) and Graves’ OFs (G-OFs) displayed similar shape and proliferation (Supplemental Fig. 1a). These hOFs were equally positive for Thy-1 and α-smooth muscle actin, markers representing lipogenic and contractile characteristics previously reported (43) (Supplemental Fig. 1a). Consistent with equal expression of these markers, a comparable number of hOFs (∼20%) from each group differentiated into lipid-laden adipocytes in response to adipogenic stimulation (Supplemental Fig. 1a).

To reproduce an in vivo–like 3D tissue microenvironment, we employed a high-throughput hanging droplet culture (44). We used 20,000 hOFs per well to generate a spheroidal organoid and assessed organoid size, adipogenic potential, and tissue stiffness (Supplemental Fig. 1b; Fig. 1a). Under both proliferating and adipogenic conditions, N-OFs and G-OFs maintained uniformly shaped spheroidal organoids (Fig. 1a). Under an adipogenic condition, OF organoids became partially lipogenic as shown by the presence of lipid-laden adipocytes within a meshwork of adipocyte-associated ECM proteins, for example, COL6 (Fig. 1a). G-OF organoids showed a CSA larger than did that of N-OF organoids at 1 day in culture (Fig. 1b). CSA of N-OF organoids declined during 6 to 12 days through spheroid compaction (45) (Fig. 1b). Under adipogenic condition, N-OF CSA was maintained during 12 days (Fig. 1b). This could be partly due to the emergence of BODIPY-positive adipocytes as well as adipogenic ECM deposition within spheroids (Fig. 1b). G-OF organoids also demonstrated spheroid compaction in nonadipogenic standard medium (Fig. 1b). Adipogenic cocktail decelerated the compaction of G-OF organoids as well; however, the effect was relatively weaker than that observed with N-OF organoids (Fig. 1b). No differences in cell proliferation and apoptosis were observed between the two groups (Supplemental Fig. 1c). Adipocyte-related gene expression in N-OF and G-OF organoids was comparable (Supplemental Fig. 1d).

Figure 1.

3D hOF organoids emulate TAO tissue stiffness. (a) Representative micrographs of 3D organoids generated from hOFs from control subjects (N-OF) and GO-derived hOFs (G-OF) with (+) and without (−) adipogenic cocktail. Confocal micrographs (far right) show lipid accumulation in adipogenic conditions detected by BODIPY staining (red) and COL6 deposition (green). Scale bars, 100 μm. (b) Organoid CSA during the culture time course, with (+ Adip.) or without adipogenic mix. n = 18 to 24 for each group. (c) Microindentation-based measurement of tissue stiffness after 6- and 12-day culture (day 6 and day 12) in standard medium. n = 11 to 12. (d) Effect of adipogenesis on tissue stiffness. A 12-day culture adipogenic protocol was used. n = 8 to 12. Data are presented as arithmetic means ± SEM. Statistical analyses were performed by a nonparametric Kruskal-Wallis test followed by a Dunn multiple comparison test. *P < 0.05; **P < 0.01; ***P < 0.001.

Next, we sought to determine the stiffness of hOF organoids using a compression-based force measurement. We employed a microscale indentation technique (11), which measures real-time force and displacement (Supplemental Video 1). Compression of organoids demonstrated a force-displacement relationship in agreement with a previously reported “viscoelastic model” (11, 46, 47). N-OF organoids cultured for 12 days required a higher force and energy to achieve 50% strain than did those cultured for 6 days, underscoring a time-dependent increase in tissue stiffness (Fig. 1c). G-OF organoids displayed increased tissue stiffness relative to N-OF organoids at both day 6 and day 12 (Fig. 1c). Adipogenic stimulation markedly increased tissue stiffness of both N-OF and G-OF organoids; however, G-OF stiffness remained twice as high as N-OF stiffness (Fig. 1d). Taken together, these mechanical force measurements suggest that G-OF organoids are significantly stiffer than N-OF organoids, and tissue stiffness increases with tissue maturation and adipogenic stimulation.

ECM accumulation within hOF organoids determines tissue stiffness

We examined the distribution of collagens within orbital adipose tissues from TAO and healthy subjects using Picrosirius red, a collagen-binding dye. TAO orbital adipose tissue showed higher overall collagen content relative to that of healthy controls (Fig. 2a). When the identity of deposited ECM molecules was probed by immunofluorescent staining, COL4, COL6, and FN were found to be highly deposited in TAO orbital adipose tissues (Fig. 2b). No difference in COL1 content was observed (Fig. 2b). The hOFs isolated from healthy and Graves’ subjects did not show significant differences in the expression of COL1, COL4, and COL6 in 2D culture (Supplemental Fig. 2), but type V collagen and FN expression were increased in G-OFs (Supplemental Fig. 2). Assembling the same cells into 3D organoids, however, reproduced TAO-specific deposition of COL4, COL6, and FN (Fig. 2b, right). When ECM gene expression was examined, only FN expression was found to be elevated in G-OF organoids, but no increase in COL1, COL4, and COL6 was observed (Supplemental Fig. 3a). These findings suggest that the deposition of COL4 and COL6 in G-OF organoids might be regulated through a posttranscriptional mechanism, for example, through a posttranslational collagen fibrillogenesis (16, 18).

Figure 2.

Unique ECM deposition in GO tissues and GO-derived OF organoids. (a) Representative Sirius red staining of periocular adipose tissues in control (normal) and GO (Graves). Scale bars, 100 μm. (b) Representative immunofluorescent staining of collagens (green) and FN (red or green) with DAPI (blue) in control and Graves orbital adipose tissues (left, n = 4 to 5) and 3D organoids reconstituted from hOFs isolated from respective tissues (right). Signal intensities were quantified per spheroid. n = 5 to 8 organoids. Scale bars, 50 µm (left) and 100 μm (right). G indicates G-OF; N indicates N-OF. *P < 0.05, unpaired Student t test (n = 4 in each group).

To test whether posttranslational ECM remodeling modifies ECM deposition and tissue stiffness in hOF organoids, we tested the role of proteinase-dependent ECM remodeling in the regulation of tissue stiffness. Matrix metalloproteinase (MMP) family members play a major role in collagen remodeling (14, 48). When N-OF and G-OF spheroids were treated with pan-MMP inhibitor GM6001, which inhibits MMP-dependent collagen turnover, we observed significant accumulation of COL1, COL4, and COL6 (data not shown) in parallel with increasing tissue stiffness (Supplemental Fig. 3b), suggesting that MMP-dependent ECM remodeling directly regulates hOF organoid tissue stiffness.

TSHR activation promotes fibrosis and stiffness of G-OF organoids

TSHR is expressed in G-OFs and considered to play a pathological role in GO as well as in hypothyroidism-associated orbitopathy (49, 50). We sought to determine the effects of hormonal milieu of Graves’ disease (hyperthyroidism and TSHR activation) on hOF organoid tissue stiffness and ECM deposition. We cultured hOF organoids in the presence of T3 (30 nM), TSH (5 mIU/mL), and both together (T3 plus TSH) during a 6-day period (Fig. 3a). TSH did not change the size of G-OF organoids, whereas T3 alone or T3 plus TSH promoted organoid compaction (Fig. 3b). In contrast, TSH alone or TSH plus T3 increased the tissue stiffness of G-OF organoids but not T3 alone (Fig. 3c), suggesting the dominant role played by TSH in regulating tissue stiffness. The response of G-OF organoids to TSH was dose-dependent (Fig. 3c, right). Whereas the expression of TSHR was similar between G-OF and N-OF organoids (data not shown), the effect of TSH on tissue stiffness was observed only with G-OF organoids but not N-OF organoids (Fig. 3d), suggesting that epigenetic alteration in G-OFs may underlie the unique response of G-OF organoids to TSH stimulation.

Figure 3.

TSHR activation promotes ECM deposition and increases tissue stiffness of G-OF organoids. (a) Experimental protocol. (b) Organoid size. n = 15 organoids. (c) Effect of 30 nM T3, 5 mIU/mL TSH, and both together on tissue stiffness of G-OF organoids (n = 12 to 16) and dose-dependent effect of TSH on G-OF organoid stiffness. n = 8 to 9. (d) The effect of TSH on tissue stiffness of N-OF organoids (n = 11). (e) TSH-dependent accumulation of COL6 and FN. Quantification is shown on the right. n = 5 organoids. Scale bars, 100 μm. (f) Effects of M22 thyrotropin stimulating antibody (5 µg/mL) on G-OF and N-OF organoid stiffness. n = 10 to 11. Data are presented as arithmetic means ± SEM. Statistical analyses were performed for unmatched multiple group comparison with a Kruskal-Wallis test followed by Dunn post hoc multiple comparisons for (b), (c), and (e) and unpaired Student t tests for two group comparisons (d and f). *P < 0.05; **P < 0.01; ***P < 0.001. N.S., not significant.

We next evaluated T3- and TSH-dependent ECM deposition in G-OF organoids. Whereas T3 showed no discernable effect, TSH and TSH plus T3 induced the robust accumulation of COL6 and FN (Fig. 3e). Taken together, these results suggest that although tissue stiffness of both N-OF and G-OF organoids are regulated by MMP and adipogenic cues, only G-OF organoid stiffness is uniquely regulated by TSHR activation. Supporting the role of TSHR activation in increasing TAO orbital tissue stiffness in the pathological milieu of Grave’s disease, TSHR-activating immunoglobulin (M22) increased the stiffness of G-OF organoids but not N-OF organoids (Fig. 3f).

3D-specific LOX activity is responsible for the mechanical stiffness of G-OF organoids

Because no corresponding increases in the expression of collagen genes were observed in parallel with increased ECM deposition in G-OF organoids (Fig. 2b), we hypothesized that a nontranscriptional mechanism may underlie excess ECM deposition in G-OF organoids. To test our hypothesis, we first determined the expression of LOX, a key mediator of collagen cross-linking, as well as CTGF, a matricellular protein that promotes collagen fibrillogenesis (51, 52). Notably, the basal expression of LOX and CTGF were significantly higher in G-OF organoids relative to N-OF organoids (Fig. 4a). No difference in their expression was observed in 2D culture (Fig. 4a). Accordingly, quantitative immunocytochemistry confirmed increased LOX and CTGF content in G-OF organoids (Fig. 4b). Furthermore, LOX content in G-OF organoids increased in response to TSH stimulation (Fig. 4c). These data suggest that the increased LOX and CTGF expression may account for the accelerated ECM deposition and increased tissue stiffness in G-OF organoids.

Figure 4.

LOX regulates the tissue stiffness of G-OF organoids. (a) Increased expression of LOX and CTGF in G-OFs observed in 3D organoids but not in 2D cultured cells (real-time quantitative PCR; five independent experiments). G indicates G-OF; N indicates N-OF. (b) Immunofluorescent staining of LOX and CTGF in 3D hOF organoids. Quantification is shown on the right. n = 5. Scale bars, 100 μm. (c) TSH-dependent induction of LOX in 3D G-OF organoids. Quantification is shown in the bottom panel. n = 5. Scale bar, 100 μm. (d) Effect of the LOX inhibitor BAPN on ECM accumulation in G-OF organoids. Representative confocal micrographs and the quantification of signal intensities are shown. Scale bars, 100 μm. (e) Effect of BAPN on tissue stiffness of G-OF organoids (control and TSH stimulated). Without TSH (n = 16) organoids; with TSH (n = 10) organoids. Data are presented as arithmetic means ± SEM. Statistical analysis was performed with an unpaired Student t test for two-group comparison. *P < 0.05; **P < 0.01; ****P < 0.0001. N.S., not significant.

To assess the role played by LOX in ECM deposition and tissue stiffness, we examined the effect of BAPN, an irreversible inhibitor of LOX (53), on ECM deposition and organoid stiffness. BAPN treatment effectively suppressed the deposition of COL1 and COL4, as well as FN, but not COL6 (Fig. 4d). Inhibiting LOX activity was sufficient to reduce the tissue stiffness of G-OF organoids in the presence or absence of TSH stimulation (Fig. 4e). These results point to a major role played by LOX-dependent ECM cross-linking in promoting ECM deposition and increased tissue stiffness.

HIF2A drives fibrosis and tissue stiffness in GO

Regulatory HIF subunits HIF1A and HIF2A promote inflammation and tissue fibrosis (16, 54). HIF1A and HIF2A are also known as the upstream regulators of CTGF and LOX expression (16, 55, 56). We hypothesized that HIF family members are involved in the elevated expression of CTGF and LOX in G-OF organoids. When we quantified the expression of HIF1A and HIF2A in N-OF and G-OF organoids, increased expression of HIF2A but not HIF1A was observed in G-OF organoids (Fig. 5a). In parallel, only HIF2A protein content was found to be increased in G-OF organoids relative to control (Fig. 5b; Supplemental Fig. 4a).

Figure 5.

HIF2A-dependent LOX induction and tissue stiffness in G-OF organoids. (a) 3D-specific elevation of HIF2A expression in G-OFs. The expression of HIF1A and HIF2A in 3D OF organoids is shown (real-time quantitative PCR; n = 5 independent experiments). G indicates G-OF; N indicates N-OF. (b) Immunofluorescent staining of HIF2A and HIF1A within N-OF and G-OF 3D organoids [HIF2A, HIF1A (green), and DAPI (blue)]. n = 7 organoids. Scale bars, 20 μm. Nuclear HIF2A and HIF1A signals are quantified on the right. (c) Downregulation of HIF2A target genes in G-OF organoids treated with shRNA against HIF2A (real-time quantitative PCR; n = 5 to 7 independent experiments). (d) Effects of shRNA-mediated HIF2A suppression on HIF2A-dependent accumulation of LOX. Quantification is shown on the right. n = 5 spheroids. Scale bars, 100 μm. (e) Effects of shRNA-mediated HIF2A suppression on tissue stiffness; three independent shRNA clones examined against vehicle and empty lentivirus-treated groups are shown. n = 12 to 15 organoids per group. (f) Suppression of HIF2A–LOX-dependent accumulation of COL1, COL4, COL6, and FN. Representative confocal micrographs and the quantification of signal intensities are shown. n = 5 organoids. Scale bars, 100 μm. (g) Effect of a small molecular HIF2A antagonist (C₁₂H₆ClFN₄O₃) on LOX and FN expression in G-OF organoids. n = 4 organoids. Scale bars, 100 μm. (h) Suppression of tissue stiffness of G-OF organoids by an HIF2A antagonist. n = 9 to 10 organoids. Data are presented as arithmetic means ± SEM. Statistical analyses were performed with an unpaired Student t test for (a)–(d) and with a Kruskal-Wallis test followed by Dunn post hoc multiple comparisons for (e), (g), and (h). *P < 0.05; **P < 0.01; ***P < 0.001. N.S., not significant.

We postulated that increased HIF2A expression in G-OF organoids underlies increased collagen fibrillogenesis and tissue stiffness through the induction of LOX and CTGF. To test this, we examined the impact of shRNA-mediated HIF2A knockdown on ECM remodeling and tissue stiffness of G-OF organoids. HIF2A suppression using three independent lentiviral shRNA clones (average ∼40% transcript reduction, Supplemental Fig. 4b) decreased the expression of known HIF target genes LOX, IL1B, IL6, and CCND2 (cyclin D2) (55, 57) (Fig. 5c). In parallel, we observed reduction in the protein levels of HIF2A and LOX (Fig. 5d). Supporting our hypothesis, HIF2A knockdown significantly reduced tissue stiffness of G-OF organoids (Fig. 5e) while inhibiting the accumulation of COL4, COL6, and FN (Fig. 5f). Treating G-OF organoids with a HIF2A-specific allosteric inhibitor (C₁₂H₆ClFN₄O₃) (58) exerted a similar effect in reducing LOX and FN protein levels (Fig. 5g). As a result, treatment with HIF2A inhibitor normalized tissue stiffness of G-OF organoids (Fig. 5h).

We then examined the effects of shRNA-mediated inhibition of HIF1A on tissue stiffness of G-OF organoids. Even though HIF1A transcript was suppressed by 78% and 43% by two unique shRNA clones, respectively, neither tissue stiffness nor gene expression of LOX, CTGF, and IL6 was affected (Supplemental Fig. 5). Taken together, these data suggest that HIF2A, but not HIF1A, is uniquely expressed in G-OF organoids and induces LOX activity, which is causal for the accumulation of COL4, COL6, FN, and increased tissue stiffness.

HIF2A activation is sufficient to induce tissue fibrosis and rigidity

To further investigate the causal relationship between HIF2A activity and OF organoid tissue stiffness, we employed a mouse model wherein a constitutively active HIF2A, with proline to alanine substitution (HIF2dPA), can be induced by Cre recombinase (15). We isolated primary mOFs from orbital adipose tissues of ROSA26-HIF2dPA mice (15), treated them in vitro with adenoviral Cre, and generated 3D mOF organoids (Fig. 6a). We found that HIF2A transcript and protein were significantly increased in mOF organoids when treated with Cre-expressing adenovirus relative to control (Fig. 6b and 6c). In parallel, LOX expression was markedly induced in these organoids at gene and protein levels (Fig. 6b and 6c). In keeping with the finding, COL4, COL6, and FN were upregulated in organoids generated from adeno-Cre–treated mOFs. In this mouse model, COL1 expression was unexpectedly reduced upon HIF2A activation, suggesting the presence of a few differences between human and mouse phenotypes (Fig. 6d). Regardless, similar to human G-OF organoids, HIF2A-overexpressing mOF organoids became significantly stiffer than did controls (Fig. 6e). To determine whether HIF2A-dependent stiffness requires LOX activity in mouse organoids, we treated them with a LOX inhibitor, BAPN. BAPN treatment completely reversed the increased tissue stiffness conferred by mutant HIF2A expression, suggesting that the HIF2A-dependent increase in tissue stiffness is largely mediated by LOX enzyme activity (Fig. 6e). Our findings suggest that the activation of HIF2A is sufficient to increase mOF organoid stiffness by promoting LOX-dependent ECM deposition.

Figure 6.

Active HIF2A is sufficient for LOX induction and tissue stiffness. (a) Isolation of mouse OFs (upper); 3D mOF organoids treated with control vs Cre-expressing adenovirus to activate constitutively active HIF2A (HIF2dPA) (lower). Images of organoids captured on day 12 of culture are shown. Scale bars, 100 μm. (b) Expression of HIF2dPA and LOX in 3D mOF organoids. n = 4 to 5 independent experiments. (c) HIF2A-dependent induction of LOX. Representative confocal images and quantifications are shown. n = 4 organoids. Scale bars, 100 μm. (d) HIF2A-dependent regulation of COL1, COL4, COL6, and FN content within organoids. Representative confocal images and quantifications are shown. n = 4 organoids. Scale bars, 100 μm. (e) Cre-dependent induction of HIF2dPA increased tissue stiffness of mOF organoids. Treatment with BAPN reversed the HIF2dPA-dependent tissue stiffness. n = 8 to 10 organoids. Data are presented as arithmetic means ± SEM. Statistical analyses were performed with an unpaired Student t test for (b)–(d) and with a nonparametric Mann-Whitney U test for (e). *P < 0.05; **P < 0.01; ***P < 0.001. N.S., not significant.

HIF2A inhibitor is effective in reducing tissue stiffness induced by TSHR activation

To determine the efficacy of pharmacological inhibition of HIF2A in reducing the tissue stiffness of G-OF organoids in the hormonal milieu of Graves’ disease, an HIF2A allosteric inhibitor (C₁₂H₆ClFN₄O₃) was tested in G-OF organoids stimulated by TSH or M22. In either TSH- or M22-stimulated G-OF organoids, pharmacological inhibition of HIF2A reversed the tissue stiffness caused by TSHR activation (Fig. 7).

Figure 7.

Therapeutic potential of an allosteric HIF2A antagonist in reversing TSH- and M22-dependent tissue stiffness. G-OF organoids are stimulated with TSH or M22 with and without HIF2 antagonist (C₁₂H₆ClFN₄O₃; 5 µM). After 6 d of culture, tissue stiffness was determined by MicroSquisher. n = 8 to 10 organoids. Data are presented as arithmetic means ± SEM. Statistical analyses were performed with a nonparametric Kruskal-Wallis test followed by Dunn post hoc multiple comparisons. **P < 0.01, ***P < 0.001.

Augmented expression of HIF2A and LOX in TAO

To validate our ex vivo findings in vivo, we evaluated the expression of HIF2A and LOX in orbital adipose tissues isolated from human subjects with and without TAO. The expressions of HIF2A and LOX were relatively low in normal orbital adipose tissues but substantially elevated in tissues from TAO patients (Fig. 8a). The increased expression of HIF2A and LOX in TAO tissues corresponds with the degree of COL6 and FN accumulation (Fig. 8a). A positive correlation between HIF2A and LOX expression was observed (Fig. 8b). Unlike HIF2A, HIF1A expression was not different between TAO and non-TAO tissues (Fig. 8c).

Figure 8.

Augmented expression of HIF2A and LOX in GO tissues. (a) Overrepresentation of HIF2A and LOX and the accumulation of COL6 and FN in GO tissues. Non-GO (n = 5) and GO (n = 10) tissue slides were examined for the expression of HIF2A (green) and LOX (red). Non-GO (n = 4) and GO (n = 4) tissue slides were examined for the expression of COL6 (green) and FN (red). DAPI (blue). Quantification is shown at the bottom. Scale bars, 50 μm. G indicates GO; N indicates non-GO. Data are presented as arithmetic means ± SD. (b) Positive correlation between the levels of HIF2A and LOX. Pearson correlation scatter plot of signal intensities of HIF2A and LOX in non-GO (n = 5, blue triangle) and GO (n = 10, red circle) (R = 0.85) is shown. (c) Representative confocal image of HIF1A in human orbital tissues [HIF1A (green) and DAPI (blue)]. Quantification is shown in the right panel. n = 5 sections. Scale bar, 50 μm. Data are presented as arithmetic means ± SEM. Statistical analyses were performed by an unpaired Student t tests for (a) and (c) and by a Pearson correlation analysis for (b). *P < 0.05; **P < 0.01; ***P < 0.001. N.S., not significant.

Discussion

Using the 3D organoid culture technique, we have demonstrated that HIF2A-dependent gene regulation and ECM remodeling are key molecular mechanisms responsible for orbital adipose tissue fibrosis in TAO.

Human 3D culture, recently employed to simulate a variety of human disease conditions (59), is often performed in synthetic scaffolds such as collagen hydrogels (60, 61). Li et al. (61) demonstrated that 3D collagen gels embedded with GO-derived hOFs display increased contraction and adipogenesis compared with those of normal hOFs. In this study, instead of 3D collagen gels, we used a liquid-based organoid culture, which is better suited for assessing de novo ECM synthesis and deposition. By evaluating human organoids derived from G-OFs and N-OFs, we could appreciate differences in the deposition and fibrillogenesis of collagens in the newly formed ECM network. In 3D hOF organoids, increased tissue stiffness and ECM architecture characteristic of TAO (11, 62) were modeled in a manner not attainable by 2D culture. ECM deposition observed specifically in 3D organoids underscores the necessity of cell–cell interaction, increased cell density, and cell-derived ECM deposition to recapitulate in vivo–like tissue architecture ex vivo (63, 64).

Despite excess collagen and FN deposition observed in G-OF organoids, no corresponding increases in gene expression of collagens were observed in our experiments. In contrast, we observed marked upregulation of LOX and CTGF at both gene and protein levels. LOX and CTGF are posttranslational modifiers that promote ECM fibrillogenesis; however, GO-specific upregulation of these molecules could not be detected in conventional 2D culture, suggesting that a 3D microenvironment is a necessary modifier for those genes. In a 2D setting, transcripts encoding LOX and CTGF were much more abundantly expressed than in 3D organoid settings (data not shown). The aberrantly high expression of LOX and CTGF in 2D condition makes it difficult to appreciate the potential role played by those molecules in regulating fibrillogenic process in vivo. In 2D, adherent cells are exposed to an artificially high mechanical tension, resulting in stretched cell shape and stress-fiber formation (65). In vivo, unlike in 2D culture, stress-fiber formation is limited to myofibroblast-like cell population observed in wound healing (66). Unlike 2D culture, 3D organoid culture provides a less rigid and physically amenable environment that allows us to observe in vivo–like ECM remodeling and gene regulation. By mimicking a relaxed, in vivo–like environment, the 3D organoid system helps us uncover a disease mechanism that is otherwise hidden from our experimental investigation.

LOX is a key regulator of collagen and ECM cross-linking (19, 67). Overexpression of constitutively active HIF1A under an adipocyte-specific promoter expanded adipose tissue and impaired glucose tolerance while increasing LOX expression and promoting tissue fibrosis (16). Our results suggest that HIF1A is not critical for LOX activation and fibrotic tissue stiffness in G-OF organoids or TAO tissues. In contrast, HIF2A was identified as a critical regulator of LOX-dependent ECM fibrillogenesis and tissue stiffness. GO-derived hOFs are known to be more responsive to hypoxia, displaying an adipogenic phenotype (68) and HIF1-dependent gene expression (17). Whereas Görtz et al. (17) demonstrated the preferential expression of HIF1A in GO-derived hOFs, we could not detect elevated expression of HIF1A in G-OF organoids or TAO tissues. We have not examined the effect of hypoxia on HIF2A-dependent gene expression and ECM remodeling in this study. LOX is induced preferentially by HIF2A relative to HIF1A (69, 70). Moreover, the HIF2A–LOX pathway underlies the progression of renal cell carcinoma (75) and mediates hypoxia-dependent metastasis (18). Similarly, increased activity of HIF2A may promote LOX-dependent ECM remodeling and tissue stiffness in TAO. Given the suspected association between cigarette smoking and aggravation of TAO (1, 2), hypoxia caused by smoking (71) may not only potentiate HIF1A-dependent biological effects (17), but it may engage HIF2A-dependent LOX activation and ECM deposition.

CTGF is a cysteine-rich matricellular protein that promotes fibrosis (52). It is induced by hypoxia in an HIF1A-dependent manner (16, 56). CTGF was highly expressed in G-OF organoids in our study; however, unlike LOX, CTGF expression was not altered by HIF2A suppression or activation, TSH stimulation, or HIF1A suppression (data not shown). Chronic tissue stiffness regulates CTGF expression through the activation of the YAP/TAZ pathway (72). Therefore, HIF2A–LOX-mediated tissue stiffness may indirectly increase CTGF expression by activating the YAP/TAZ pathway. As the effects of TGFβ and YAP/TAZ on CTGF synthesis are modified by HIFs (73), understanding the complex interplay among HIFs, TGFβ, and YAP/TAZ may further help us uncover the molecular basis of the tissue fibrotic process in TAO.

The molecular link between TSHR activation and augmented HIF2A activity, uniquely observed in G-OF organoids, remains unknown. Our results suggest that G-OFs might have acquired a gene regulatory network that leads to altered HIF2A expression, tissue remodeling, and responsiveness to TSHR stimulation. Defining the epigenetic molecular link between TSHR activation and HIF2A-dependent ECM remodeling should help us better understand the pathogenesis and fibrotic disease process of TAO.

In summary, using the experimental approach of 3D disease modeling, we identified a critical role played by the HIF2A–LOX pathway in promoting ECM accumulation and increasing tissue stiffness in TAO. We propose that the HIF2A–LOX pathway may serve as a novel therapeutic target in an effort to alleviate fibrotic tissue damage in TAO.

Glossary

Abbreviations:

2D

two-dimensional

3D

three-dimensional

BAPN

β-aminopropionitrile

BODIPY

boron-dipyrromethene

CMV

cytomegalovirus

COL1

type I collagen

COL4

type IV collagen

COL6

type VI collagen

CSA

cross-sectional area

CTGF

connective tissue growth factor

DAPI

4′,6-diamidino-2-phenylindole

FN

fibronectin

GO

Graves’ orbitopathy

G-OF

Graves’ orbital fibroblast

HIF

hypoxia-inducible factor

HIF1A

HIF1α

HIF2A

HIF2α

hOF

human orbital fibroblast

LOX

lysyl oxidase

MMP

matrix metalloproteinase

mOF

mouse orbital fibroblast

N-OF

non-Graves’ orbital fibroblast

OF

orbital fibroblast

PBST

Triton X-100 in PBS

shRNA

short hairpin RNA

TAO

thyroid-associated orbitopathy

TSHR

TSH receptor