TED human orbital tissue shows increased adipocytes. To delineate changes in the cellular composition within orbital tissue associated with TED, we utilized snRNA-Seq to analyze samples from patients diagnosed with TED in comparison with samples from control patients without TED pathology (Figure 1A and Supplemental Table 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.182352DS1). The patients with TED ranged in age from 22 to over 70 years, while control patients were mostly over 70 years of age. Smoking history was limited, with only 1 individual acting as a control reporting smoking more than 10 years ago and another reporting rare use (Supplemental Table 1). Our analysis identified 4 primary cell populations present in both groups: fibroblasts, adipocytes, endothelial cells, and a diverse array of immune cells (Figure 1B). Fibroblasts were characterized by the expression of genes such as DCN, CNTN4, and ADAMTSL1 (Supplemental Table 2). Adipocytes were identified through the expression of PDE3B, LPL, and PLIN1 (Supplemental Table 2). Endothelial and smooth cells exhibited VWF expression, and the immune cell population comprised multiple subsets, including neutrophils, mast cells, and cells indicative of either innate (immune cell cluster 2, F13A1 and MRC1) or adaptive (immune cell cluster 1, PTPRC and RIPOR2) responses (Supplemental Table 2).

Overview of the experimental approach and cell-type distribution in the control and TED groups. (A) Schematic of the experimental workflow. (B) UMAP plots illustrating the distribution of cell types in the control (left) and TED (right) groups. (C) Bar graphs comparing the distribution of cell clusters between the control and TED groups identified in snRNA-Seq.

Notably, the group of patients with TED exhibited a substantial increase in adipocyte numbers compared with the control group (Figure 1C and Supplemental Figure 1), consistent with the adipogenic shift previously documented in TED pathology (18). Increases in IGF1, IGF1R, and TSHR in TED fibroblasts compared with the controls were observed (Supplemental Figure 2), aligning with previous studies (19). While increases in IGF1 and TSHR were observed in TED adipocytes compared with controls, IGF1R levels did not differ between the groups (Supplemental Figure 2). IGF2 expression was absent in both groups (Supplemental Figure 2).

Furthermore, adipogenic markers such as PDE3B and PLIN1 were detected in TED fibroblasts but not in controls (Supplemental Figure 3), suggesting an ongoing adipogenic transition within TED fibroblasts. Additionally, our snRNA-Seq data highlighted an increased presence of immune cells facilitating adaptive responses in TED samples, contrasting with a diminished representation of cells associated with innate immunity (Figure 1C). These observations not only validate known histopathological features in TED but also reveal insights into the disease’s immune landscape.

Exploring variations in immune cell distribution between the control and TED groups. To further explore the variations in immune cell distribution between the control and TED groups, we focused our analysis on immune cell clusters 1 and 2. These clusters were characterized by gene expressions indicative of adaptive (cluster 1) and innate (cluster 2) immune responses (Supplemental Table 2). A Gene Ontology (GO) pathway analysis of TED immune cells (Figure 2, A and B, and Supplemental Table 3), revealed pathways such as “T cell differentiation in thymus” and “positive T cell selection.” This analysis also uncovered heightened expressions of CD74, LYZ, HLA-DBA, CD163, and IRAK3 in TED immune cells (Figure 2C and Supplemental Table 3) — genes generally enriched in adaptive immune cells. Such findings suggest a pronounced shift toward an adaptive immune response in TED, consistent with autoimmune etiology (20, 21).

Immune cell analysis and pathway enrichment in the control and TED groups. (A) UMAP plot of immune cells in the control and TED groups. (B) GO pathway analysis of genes enriched in TED immune cells. (C) Heatmap displaying the expression of control- or TED-enriched genes in immune cells. (D) UMAP plot showing identified subclusters of immune cells across all samples. (E) Bar graphs comparing the distribution of immune cell clusters between the control and TED groups identified in snRNA-Seq.

Further subdivision of immune cells into 6 distinct clusters (Figure 2D) revealed diverse contributions from each cluster between the groups (Figure 2E). Although the specific immune cell subtypes within each cluster remained unidentified (Supplemental Table 4), the markers noted suggest a composition enriched in monocytes/macrophages (clusters 0 and 2), NK cells and T cells (cluster 1), dendritic cells (cluster 2), and B cells (clusters 3 and 4). Cluster 1 cells from patients with TED showed elevated expression of CD4+ T cell markers, aligning with findings from a previous study that identified CD4+ T cells associated with hyperthyroidism and TED (22). These markers include PRF1, GZMA, GNLY, and GZMH (Supplemental Figure 4).

Differential expression of the IGF signaling pathway among TED orbital fibroblasts. Further investigation of orbital fibroblast populations from TED and control samples identified multiple subclusters of orbital fibroblasts, each distinguished by unique molecular markers (Figure 3, A and B, and Supplemental Table 5). A GO pathway analysis of genes enriched within these clusters identified varied biological pathways (Supplemental Figure 5). While some clusters showed no unique pathway alterations, the clusters predominantly found in TED samples exhibited distinctive pathway modifications (Supplemental Figure 5). Specifically, clusters labeled as “NFATC2,” “TSHR & PDE10A,” and “XACT” were markedly prevalent in TED samples (Figure 3C), and these clusters displayed changes in pathways related to “response to steroid hormone” and “cellular response.” These findings suggest biological shifts associated with TED or effects of therapeutic interventions like radiotherapy or drug treatments on mitigating TED symptoms (Supplemental Figure 5).

Analysis of orbital fibroblast subtypes and adipogenic gene expression in the control and TED groups. (A) Violin plots showing genes enriched in specific clusters across orbital fibroblast subtypes in all samples. (B) UMAP plot showing the distribution of orbital fibroblast subclusters in control (left) and TED (right) samples. (C) Bar graphs comparing the distribution of orbital fibroblast subclusters between the control and TED groups. (D) Heatmap displaying the expression of adipogenic and IGF-related genes across groups.

Despite the variability in molecular markers across fibroblast subclusters, a commonality in the expression of adipocyte-enriched genes (18) was observed between TED and control groups (Supplemental Figure 6). TED orbital fibroblasts, irrespective of the cluster, demonstrated higher levels of adipocyte-enriched genes compared with controls (Supplemental Figure 6). Additionally, an examination of genes known to regulate adipogenesis, such as CEBPB and CEBPD (23), FABP4 (24), FABP5 (25), PLIN2 (26), APOC1 (27), APOE (28), and RASD1 (29), revealed a comprehensive upregulation in adipogenic genes, notably CEBPB and CEBPD in all TED-associated orbital fibroblast subclusters (Figure 3D and Supplemental Table 6). This suggests an inherent adipogenic predisposition within TED fibroblasts. However, other markers like APOE, RASD1, and FABP4 were selectively upregulated in specific TED fibroblast subclusters (Figure 3D and Supplemental Table 6). Interestingly, the presence of fibroblast subclusters with a higher representation in TED samples did not directly correlate with increased adipogenic gene expression, and no differences were found in the expression of FABP5, PLIN2, and APOC1 between the groups.

Our study also identified disparities in the expression of genes related to the IGF signaling pathway across the fibroblast subclusters. Notably, TED orbital fibroblasts exhibited a significant reduction in IGFBP6 expression alongside an increase in IGFBP5 relative to controls (Figure 3D and Supplemental Table 6). This differential expression indicates a distinctive modulation of the IGF signaling pathway in TED fibroblasts, suggesting a unique pathophysiological mechanism at play.

Enhanced adipogenesis in TED adipocytes. In our analysis of adipocytes derived from both TED and control groups, we identified multiple subclusters within each group, each defined by distinct molecular markers (Figure 4A and Supplemental Table 7). Notably, the TED group exhibited a substantially increased number of adipocytes compared with that in the control group (Figure 4B). Additionally, unique adipocyte subclusters named “ANXA1,” “CTH,” and “PDZRN4” were predominantly found in TED samples (Figure 4C). A GO pathway analysis of these subclusters, particularly the CTH subcluster, indicated a distinct shift toward “fat cell differentiation” (Supplemental Figure 7), suggesting that these subclusters may represent recently differentiated adipocytes. To further validate the distinct adipocyte populations identified through snRNA-Seq, Western blot analysis was performed to validate the expression of ANXA1, PDZRN4, and CTH in isolated, undifferentiated orbital fibroblasts from both patients with TED and healthy control patients. PDZRN4 expression was significantly higher in TED fibroblasts compared with control fibroblasts (n = 3, P < 0.001) (Figure 4D). Additionally, ANXA1 and CTH exhibited a trend toward higher expression in TED fibroblasts, though these differences did not reach statistical significance.

Characterization of orbital adipocyte subtypes and IGF-related gene expression. (A) Violin plots showing genes enriched in specific clusters across orbital adipocyte subtypes in all samples. (B) UMAP plot showing the distribution of orbital adipocyte subclusters in control (left) and TED (right) groups. (C) Bar graphs comparing the distribution of orbital adipocyte subclusters between the control and TED groups. (D) Bar graphs showing the expression of ANXA1, PDZRN4, and CTH in undifferentiated control and TED fibroblasts. n = 3 cell lines each, performed in triplicate.*P < 0.001 by unpaired t test. (E) GO pathway analysis of genes enriched in control (left) and TED adipocytes (right). (F) Heatmap displaying the expression of IGF-related genes across groups.

Further comparative GO pathway analysis across all subclusters of control and TED adipocytes revealed notable differences (Figure 4E and Supplemental Figure 7). TED adipocytes were enriched in pathways associated with “response to peptide,” “response to steroid hormone/glucocorticoid,” and “proliferation.” These findings suggest a complex interaction of differentiation and proliferation processes contributing to the increased adipocyte population observed in TED (Figure 4E).

A detailed examination of gene expression within the IGF pathway between the two cohorts further uncovered differences. Despite variability within subclusters, TED adipocytes consistently displayed a decrease in IGFBP6 expression across all subclusters compared with controls (Figure 4F and Supplemental Table 8), alongside an increase in IGFBP5 expression (Supplemental Table 8). Additionally, reductions in IGFL4 and IGFBP2 expression were observed in TED adipocytes, in contrast to the mixed expression patterns of IGF pathway genes seen in orbital fibroblasts. This pattern indicates a more pronounced dampening of IGF signaling in TED adipocytes.

Adipogenesis trajectory of TED orbital fibroblasts and adipocytes. To investigate the adipogenic differentiation of orbital fibroblasts in TED, we conducted an integrated analysis of TED orbital fibroblasts with adipocytes. Due to the limited number of adipocytes in control samples, this analysis was specifically focused on identifying changes within the TED cohort.

Our initial objective was to map a direct adipogenic pathway from specific fibroblast subclusters to differentiated adipocytes. Although the “NFATC2” fibroblast cluster in TED exhibited a closer association with differentiated adipocytes compared with the “CFD” cluster, a definitive adipogenic lineage from any particular fibroblast subcluster to adipocytes could not be conclusively established (Figure 5A).

Combined analysis of orbital fibroblasts and adipocytes in TED groups and pathway analysis. (A) UMAP plot showing merged orbital fibroblasts and adipocytes in the TED group (top) and previously identified (see Figure 3B and Figure 4B) clusters (bottom). (B) UMAP plots showing the expression of cluster-specific/enriched genes. (C) Gene modules in TED fibroblast (left) and TED adipocytes (right). (D) GO pathway analysis of gene modules in TED fibroblasts (left) and TED adipocytes (right).

We observed marked downregulation of TSHR in adipocytes compared with orbital fibroblasts (Figure 5B), accompanied by a consistent decrease in the expression of IGF1R, IGFBP5, and IGFBP6 (Figure 5B). Conversely, adipocytes demonstrated an upregulation of adipogenic markers such as ADIPOQ, FABP4, PLIN2, and RASD1 (Figure 4B), indicating successful adipogenesis.

A GO pathway analysis delineated distinct biological processes between the two cell types (Figure 5C). TED orbital fibroblasts were predominantly involved in pathways associated with “cytoskeleton organization/assembly” (Figure 5D). In contrast, TED adipocytes were linked to “response to fatty acid” and “insulin response” pathways (Figure 5D), underscoring the substantial role of insulin in adipogenic differentiation from fibroblast to adipocytes (18). These observations further reinforce the importance of the IGF pathway in the adipogenic process in TED, aligning with findings from earlier in vitro studies on patient-derived human orbital fibroblast cell lines (18).

Linsitinib treatment reduces adipogenesis in TED fibroblasts. The observed decrease in IGF-1 pathway gene expression in differentiated adipocytes relative to orbital fibroblasts aligns with previous research that IGF-1 signaling initially surges during the early stage of adipogenesis in vitro and then declines as adipocytes mature. This finding prompted us to investigate the effects of manipulating the IGF-1 pathway on adipogenesis in orbital fibroblasts. We utilized linsitinib, a small-molecule IGF-1R antagonist, which diminishes immune infiltration and fibrosis in orbit in TED animal models (30). Currently, linsitinib is undergoing phase II clinical trials as a potential oral therapy for TED.

In our established in vitro model using patient-derived human orbital fibroblast cell lines (18), we administered linsitinib in a dose-dependent manner (Figure 6, A–L). Consistent with prior observations, adipocyte differentiation was evident by day 5 in vitro, as indicated by Oil Red O+ adipocytes (Figure 6E). A notable dose-dependent reduction in the proportion of adipocytes was documented following linsitinib treatment on both days 5 and 9 (Figure 6M). Considering linsitinib’s known affinity for the insulin receptor (31), which also plays a role in adipogenesis, we conducted additional experiments in insulin-free media to show the insulin-independent effects of IGF-1R inhibition on adipogenesis. Despite the overall reduction in adipogenesis due to the absence of insulin, a critical factor in adipocyte differentiation, linsitinib treatment still resulted in a dose-dependent inhibition of adipogenesis (Figure 6N). To confirm that the decrease in adipogenesis associated with linsitinib treatment is due to inhibition of IGF-1R signaling, we performed Western blots to demonstrate that treatment with linsitinib is associated with a dose-dependent decrease in IGF-1R phosphorylation on days 5 and 9 (Figure 6O). These findings emphasize the critical role of IGF-1R signaling in the adipogenic process of TED orbital fibroblasts, highlighting how linsitinib’s influence on this pathway markedly affects adipogenesis, independent of insulin signaling.

Effect of linsitinib on adipogenesis and IGF1R expression in orbital fibroblasts. (A–L) Oil Red O staining of orbital fibroblasts treated with adipogenic media (A, E, and I) and 0.1 μM (B, F, and J), 1 μM (C, G, and K), or 10 μM (D, H, and L) linsitinib on days 0 (A–D), 5 (E–H), and 9 (I–L) after treatment. Arrows indicate Oil Red O+ adipocytes. Scale bar: 50 μm. (M) Bar plot showing the percentage of adipocytes after treatment with standard adipogenic medium and 0, 0.1, 1, or 10 μM linsitinib on days 0, 5, and 9. *P < 0.05; **P < 0.01; ****P < 0.0001 by Tukey’s multiple comparison test. (N) Bar plot showing the percentage of adipocytes after treatment with insulin-free adipogenic medium and 0, 0.1, 1, or 10 μM linsitinib on days 0, 5, and 9. *P < 0.05; **P < 0.01; ****P < 0.0001 by Tukey’s multiple comparison test. (O) Bar plot (top) and blots (bottom) showing the relative expression between phospho-IGF1R and total IGF1R after treatment with insulin-free adipogenic medium and 0, 0.1, 1, or 10 μM linsitinib on days 5 and 9. ****P < 0.0001 by Šidák’s multiple comparison test.