Increased Generation of Fibrocytes in Thyroid-Associated Ophthalmopathy

Abstract

Context: The pathogenic basis for Graves’ disease (GD) continues to elude our understanding. Specifically why activating antibodies are generated against self-antigens remains uncertain as does the identity of the antigen(s) that provokes orbital involvement in GD, a process known as thyroid-associated ophthalmopathy (TAO).

Objective: The aim of the study was to determine whether CD34⁺ fibrocytes are generated more frequently in GD, whether they infiltrate orbital connective tissues in TAO, and whether they express the thyrotropin receptor (TSHR).

Design/Setting/Participants: Generation of fibrocytes from peripheral blood mononuclear cells was examined in samples from 70 patients with GD and 25 healthy control subjects. Fibrocytes were characterized by flow cytometry. Orbital tissues and fibroblast culture strains were examined for their presence.

Main Outcome Measures: The frequency of CD34⁺ fibrocyte generation from peripheral blood cells, characterization of their phenotype, cytokine production, and their presence in affected orbital tissues were analyzed.

Results: CD34⁺CXCR4⁺Col I⁺ fibrocytes expressing IGF-I receptor are far more frequently generated from cultured peripheral blood mononuclear cells of donors with GD compared with healthy subjects. They express TSHR at high levels and TSH induces fibrocytes to produce IL-6 and TNF-α. Numerous CD34⁺ fibrocytes were detected in orbital tissues in TAO but were absent in healthy orbits. Tissue-infiltrating fibrocytes express TSHR in situ and comprise a subpopulation of TAO-derived orbital fibroblasts.

Conclusions: Our findings suggest that fibrocytes may participate in the pathogenesis of TAO because they express relevant autoantigens such as IGF-I receptor and functional TSHR and differentially accumulate in orbital tissue in TAO.

CD34⁺ fibrocytes express functional thyrotropin receptor and accumulate in the orbit in thyroid-associated ophthalmopathy.

Neuroendocrine control of immune function comprises an integrated system of signaling pathways involving diverse cell types (1). Despite their complexities, these interrelationships must be understood if we are to devise therapeutic strategies for autoimmune diseases. Graves’ disease (GD) represents a prototypic antibody-driven autoimmune process affecting the thyroid and orbital connective tissue (2). Generation of activating antibodies against the TSH receptor (TSHR) leads to excessive thyroid hormone synthesis uniquely in GD (3). But the mechanisms underlying orbital tissue inflammation and remodeling in thyroid-associated ophthalmopathy (TAO) remain uncertain as do their relationship with the processes occurring in the thyroid. Several investigators have attributed the involvement of the orbit in GD to anatomically restricted TSHR expression (4). Indeed, TSHR has been detected in affected orbital tissue (5) and derivative fibroblasts, especially under culture conditions favoring adipogenic differentiation (6). But low-level TSHR expression can also be detected in many other fatty tissue depots (7,8,9,10,11). Positive correlations between thyroid-stimulating immunoglobulin levels and the clinical activity of TAO have been reported (12,13). However, no direct evidence yet links TSHR or TSI to the pathogenesis of TAO. In addition to TSHR, we reported that the IGF-I receptor (IGF-IR) is overexpressed by orbital fibroblasts from patients with GD (14). Its interactions with IgGs from these patients or with IGF-I results in the production of T cell chemoattractants (15) and hyaluronan (16). Accumulation of that glycosaminoglycan constitutes a cardinal feature of GD (17). Recently we demonstrated that TSHR and IGF-IR form physical and functional complexes (18). The intimate association of these proteins may contribute to the immune reactivity in GD.

Lymphocytes and other bone marrow-derived cells recruited to the orbit appear to drive tissue activation in TAO (19,20,21). Orbital fibroblasts comprise a heterogeneous population of cells possessing divergent phenotypes and potential for differentiation (22,23,24). Moreover, they exhibit attributes differing from their nonorbital counterparts that may underlie anatomic-selective involvement of the orbit in GD (25). Whereas healthy orbital tissue derives from neural ectoderm (26), the identity of cells from other embryonic origins and that contribute to the tissue remodeling characteristic of TAO remains incomplete. Fibrocytes derive from monocytoid or B cell precursors circulating in peripheral blood mononuclear cells (PBMCs) (27). They infiltrate connective tissues in response to injury and retain markers characteristic of their bone marrow origin (27,28,29). Fibrocytes have been implicated in normal physiological process such as wound healing and in fibrotic diseases such as idiopathic pulmonary fibrosis, asthma, liver, and kidney fibrosis (30,31,32,33,34). They participate in inflammation and tissue remodeling (35,36,37). Fibrocytes synthesize collagen I (Col I), display cell surface CD34 and CXCR4, and traffic to peripheral tissue sites in response to CXCL12, the cognate ligand for CXCR4 (37,38). CD34 represents a marker of hematopoietic stem/progenitor cells. CD34 may regulate cell differentiation and mediate adhesion to bone marrow stroma. On endothelium, it participates in L-selectin-mediated leukocyte recruitment (39). Importantly, fibrocytes can differentiate into myofibroblasts and adipocytes when treated with TGF-β and peroxisomal proliferator-activated receptor-γ ligands, respectively (40). They have not been shown previously to express autoantigens.

In this report, we describe for the first time a dramatically increased abundance of fibrocytes generated from cultured PBMCs in patients with GD. These fibrocytes spontaneously express TSHR at levels comparable to those found on thyroid epithelial cells. When treated with TSH, fibrocytes produce high levels of proinflammatory cytokines. We also find that CD34⁺TSHR⁺ fibrocytes infiltrate orbital tissues in situ in TAO and comprise a large fraction of orbital fibroblasts cultured from those diseased tissues. Our current findings suggest a potential link between increased frequency of CD34⁺TSHR⁺ fibrocyte generation, their infiltration of the orbit, and their potential participation in the pathogenesis of TAO.

Subjects and Methods

Materials

Ficoll-Hypaque was purchased from Sigma Aldrich (St. Louis, MO). FacLyse buffer, Cytofix, anti-CD19, CXCR4, CD34, leukocyte-specific protein-1 (LSP-1), CD31, Col 1, anti-IGF-IRα PE (clone 1H7), anti-TSHR, isotype mouse IgG1 fluorescein isothiocyanate, phycoerythrin, allophycocyanin, and CyChrome were purchased from BD Biosciences (San Jose, CA). Fetal bovine serum was supplied by Life Technologies (Grand Island, NY). Bovine TSH (bTSH) was obtained from Sigma-Aldrich.

Subject samples

Subjects, aged 20–65 yr, were recruited from the patient populations of Jules Stein Eye Institute and Harbor-UCLA Medical Center. Informed consent was obtained as approved by the Institutional Review Boards of the Center for Health Sciences at UCLA and Harbor-UCLA Medical Center-Los Angeles Biomedical Institute. The blood donor cohort comprised patients evaluated for GD without or with clinically apparent TAO. Control subjects were healthy volunteers without known autoimmune disease who presented for esthetic or functional eyelid surgery or routine medical care. Individuals excluded from the study carried the diagnosis of other autoimmune diseases, asthma, chronic inflammatory processes, sinusitis, recent trauma, or HIV infection. Patients with GD comprised a clinically heterogeneous group and included six hyperthyroid, six hypothyroid, and 58 euthyroid patients at the time of participation. Fifty-one of 70 patients manifested TAO, whereas two exhibited pretibial dermopathy. A total of 16 patients with TAO were in the active inflammatory phase [clinical activity score (CAS) ≥ 3], whereas the remainder were stable (CAS < 3; n = 35). In addition, 25 healthy subjects without known thyroid disease served as controls. Fibroblast strains were initiated from waste tissue obtained during orbital decompression surgery or from healthy individuals undergoing cosmetic procedures. Thyrocytes were also generated from surgical waste of patients undergoing thyroidectomy for the treatment of GD or were obtained from the contralateral (normal) lobe of individuals with a thyroid neoplasm.

Fibroblast and thyrocyte cultures

Orbital fibroblasts were cultivated as described previously (41). Dermal fibroblasts were derived from normal-appearing tissues or were purchased from the American Type Culture Collection (Manassas, VA). Cell layers were covered with DMEM containing 10% fetal bovine serum, l-glutamine, and penicillin/streptomycin. Monolayers were disrupted by gentle treatment with trypsin/EDTA. All experiments were performed with fibroblasts between the second and 12th passages from culture initiation. We previously determined that these fibroblast strains are not contaminated with epithelial, endothelial, or smooth muscle cells and maintain a stable phenotype over this culture interval (22). Primary human thyroid epithelial cells were cultivated as described previously (42).

Fibrocyte cultivation

PBMCs were subjected to culture under conditions similar to those described by Bucala et al. (27). Briefly, they were isolated from human blood by centrifugation over Histopaque-1077 (Sigma Aldrich), following the manufacturer’s protocol. Each culture well in a 24-well plate was inoculated with 5 × 10⁶ cells in 1 ml medium (DMEM with 5% fetal calf serum). After 12–14 d in culture, adherent cells (<5% of starting PBMC population) were washed and removed from the substratum by scrapping. Culture purity was verified to be greater than 90% fibrocytes by fluorescence-activated cell sorter analysis with anti-CD34 and anti-Col I antibodies. Cell viability was greater than 90% by trypan blue exclusion. Preparations were counted in a hemocytometer.

Induction of adipogenesis in orbital fibroblasts and fibrocytes

In some experiments, fibrocytes and fibroblasts were induced to differentiate into adipocytes with a differentiation medium containing rosiglitazone and insulin, following protocol B, as previously described (23). Adipocyte differentiation was confirmed using Oil Red O staining.

Immunofluorescence and confocal microscopy

Thin sections of orbital connective tissues were prepared as described previously (43,44). Frozen sections were fixed in 4% paraformaldehyde, blocked with goat serum (5%) for 30 min at room temperature, and incubated with primary mouse antihuman CD34, CD31, LSP-1, IGF-IRα, TSHR, Col I, or their isotype control monoclonal antibody overnight at 4 C. After washes in PBS containing 0.1% Tween 20, slides were incubated with secondary antibodies for 1 h at room temperature in the dark. For confocal microscopy, images were acquired and analyzed using a Eclipse 800 microscope (Nikon, Meliville, NY) interfaced with a Nikon PCM 2000 two-laser confocal system as described previously (18).

Flow cytometry

Dispersed fibroblasts and fibrocytes were washed in staining buffer containing PBS with 0.1% sodium azide and 1% BSA. They were incubated with 50 μl of heat-inactivated mouse serum for 5 min and then either PerCP-conjugated antihuman CD34, CD31, CXCR4, CD40, IGF-IR, TSHR, and Col I or PE conjugated anti-Thy-1 (CD90) MoAbs for 30 min on ice. They were then washed and resuspended in 400 μl of staining buffer at 4 C and subjected to fluorescence-activated cell sorter analysis in a Calibur flow-cytometer (BD Biosciences). Mean fluorescent intensity (MFI) was calculated as a ratio of mean fluorescence sample/isotype control fluorescence. Viable cells were gated on the basis of forward light scatter, and the data were analyzed with the FCS Express software program (De Novo Software, Los Angeles, CA). All studies were performed at least three times.

Cytokine measurements

Production of IL-6 and TNFα in cultured fibrocytes was determined by subjecting culture medium to ELISA kits from BD Biosciences (45).

Statistics

Unless otherwise stated, data values are reported as the mean ± sd. Statistical analysis was performed using a two-tail Student’s t test with a confidence level greater than 95%.

Results

Exaggerated fibrocyte generation in GD

Initial attempts to cultivate fibrocytes revealed that PBMCs from patients with GD generated substantially more numerous CD34⁺CXCR4⁺Col 1⁺ cells than did those derived from control donors. We therefore standardized the isolation procedure and culture conditions to quantify the fibrocytes generated from PBMCs. Despite heterogeneity among individual samples, those derived from GD yielded approximately 5-fold more fibrocytes than did controls [GD, 5268 ± 1260 fibrocytes per 10⁶ PBMCs (mean ± sd, n = 70) vs. control, 954 ± 329 fibrocytes per 10⁶ PBMCs (n = 25, P < 0.001 vs. GD)] (Fig. 1). Fibrocyte yields were increased among patients with GD, whether or not they manifested clinically apparent TAO (n = 51; P = 0.2) or dermopathy (n = 2). Fibrocyte yields were not statistically different in TAO patients with active disease (CAS > 3; 3,317 ± 1,746) compared to those with stable disease (CAS < 3; 6354 ± 1537; P = 0.4). Furthermore, the severity of exophthalmos of the more affected orbit failed to correlate with fibrocyte yields (P = 0.1). At the time of fibrocyte analysis, six patients with GD were hypothyroid, six were hyperthyroid, and 58 were euthyroid, but fibrocyte counts were not statistically different among these groups (P = 0.4; P = 0.5, P = 0.5, respectively). Moreover, the propensity for increased fibrocyte frequency appeared durable in that some patients had been diagnosed with GD 20 yr before their participation in the study.

Figure 1.

Increased frequency of CD34⁺ fibrocyte generation from the PBMCs of 70 patients with GD compared with 25 healthy control donors. PBMCs were prepared as described in Subjects and Methods, and multiwell culture arrays were inoculated at a density of 5 × 10⁶ cells/well. Cultures were incubated for 14 d. Adherent cells (<5% of starting cells) were collected and manually counted in a hemocytometer. Statistical significance (P < 0.001) was determined with a two-tailed Student’s t test.

The phenotypes of fibrocytes from patients with GD appear similar to those from control donors

Fibrocytes generated in vitro from the PBMCs of patients with GD exhibit morphologies consistent with those described previously (27). They are oblong, spindle-shaped cells in culture that adhere tightly to plastic substratum and are morphologically similar to orbital fibroblasts (Fig. 2A). Some fibrocytes project arborized processes, whereas others are more rounded with nuclear prominence. If cultured for more than 2 wk, they spontaneously accumulate intracellular lipid droplets, based on Oil Red O staining (data not shown). Figure 2B demonstrates fibrocyte expression of CD34, smooth muscle actin, Col I, and CXCR4, consistent with their previously reported phenotype (27,28). Moreover, they display IGF-IR and CD11b but not CD3, CD19, or CD14, confirming the absence of contaminating T cells, B cells, or monocytes. This profile of fibrocyte surface markers is indistinguishable from that of control donor-derived fibrocytes.

Figure 2.

A, Orbital fibroblasts, dermal fibroblasts, and fibrocytes share spindle-shaped morphologic features (hematoxylin and eosin, ×20). B, Phenotypic analysis of fibrocytes from patients with GD reveals cell surface display of CD34 [upper left, insets contain images of immunohistochemical (upper) and immunofluorescence (lower) staining with anti-CD34 antibody]. Flow cytometric analysis of cell surface Col I, CXCR4, and IGF-IRα. Upper right panel demonstrates strong immunohistochemical staining for α-smooth muscle actin (×40). Lower left panel demonstrates strong staining by immunofluorescence with anti-IGF-IRβ (red). Analyses were performed as described in Subjects and Methods. Results were identical in 10 separate studies involving fibrocytes from five patients with GD and five healthy donors. ISO, Isotype control.

Circulating fibrocytes express high levels of TSHR

A central hallmark of GD concerns the loss of peripheral tolerance to TSHR, and more recent evidence would suggest a similar immunoreactivity to IGF-IR (14). Because fibrocyte generation from PBMCs is increased in GD (Fig. 1), we investigated whether these cells might express TSHR. Unexpectedly, we found that the receptor was abundant on fibrocyte surfaces. Equivalent TSHR display was found on fibrocytes derive from patients with GD and those from control donors (Fig. 3, A and B). Addition of TSH to cultures for 24h before staining abrogated anti-TSHR antibody binding, demonstrating specificity of TSHR detection (Fig. 3A). Even more surprising were the high levels at which TSHR is expressed by fibrocytes, levels which were comparable with those found on cultured human thyrocytes (Fig. 3C) (MFI: fibrocytes, 4.4 ± 0.5, n = 7 vs. thyrocytes, 2.8 ± 0.4; n = 3). High-level TSHR expression appears durable in culture and continues after fibrocyte differentiation into adipocytes (Fig. 3, E and F) (18). In contrast, undifferentiated orbital fibroblasts, even those from patients with GD, fail to express detectable TSHR (Fig. 3D) (MFI: orbital fibroblasts, 1.0 ± 0.1, n = 4).

Figure 3.

Fibrocytes cultivated from PBMCs express high levels of TSHR, regardless of whether they derive from patients with GD (A) or healthy donors (B). C, These levels are comparable with those found on primary human thyroid epithelial cells. D, In contrast, undifferentiated orbital fibroblasts fail to express TSHR. E, Fibrocytes differentiated into adipocytes accumulate intracellular lipid droplets staining with Oil Red O. F, TSHR levels on fibrocytes remain elevated after differentiation. Fibrocytes, orbital fibroblasts, and thyrocytes were cultivated as described in Subjects and Methods. In A, fibrocytes were preincubated with bTSH (5 mU/ml) for 24h before staining with anti-TSHR antibodies. ISO, Isotype control.

Treatment of fibrocytes with TSH results in the induction of IL-6 and TNF-α

Display of TSHR on fibrocytes raises the possibility that this receptor might provoke cytokine production. When bTSH (5 mU/ml) was added to fibrocyte culture medium for 24 h, the agent dramatically increased levels of IL-6 and TNF-α (Fig. 4). With TSH stimulation, IL-6 levels increased 12.8 ± 0.5-fold and TNF-α levels increased 19.8 ± 0.6-fold over untreated controls (P < 0.001 for each vs. controls). Treatment with IL-1β (10 ng/ml), a concentration associated with near-maximal up-regulation of IL-6 in orbital fibroblasts (45), increased production of that cytokine but failed to appreciably up-regulate levels of TNF-α. Similar findings were obtained with fibrocytes from control donors (data not shown). Thus, it would appear that TSHR expressed by fibrocytes is functional and can up-regulate cytokine expression. This finding suggests that the TSHR pathway in fibrocytes might play a role in orbital inflammation.

Figure 4.

TSHR displayed on fibrocytes generated from PBMCs can function to initiate cytokine production. Cultured cells, in this case, from a patient with GD, were treated with bTSH (5 mU/ml) or IL-1β (10 ng/ml) for 48 h. The medium was subjected to ELISAs specific for IL-6 (left panel) or TNF-α (right panel). Data are expressed as the mean ± sem of three replicate culture wells from a representative experiment. *, P < 0.001.

A subset of orbital fibroblasts from patients with TAO may derive from circulating fibrocytes

Orbital fibroblasts from patients with TAO possess unique attributes (46). Given the putative role of fibrocytes in orchestrating tissue reactivity and remodeling, we next asked whether they might provide a link between systemic disease and the orbit in TAO. To test this possibility, we examined orbital fibroblasts and identified a subpopulation of cells resembling fibrocytes. CD34⁺Col I⁺ fibroblasts are plentiful among fibroblast cultures from donors with TAO (Fig. 5, A and B). These cells display a profile of membrane markers remarkably similar to that of fibrocytes (Fig. 2B). Importantly, orbital fibroblasts were uniformly CD31⁻ and CD14⁻ (data not shown). In contrast, CD34⁺Col I⁺ cells are absent in control orbital fibroblast cultures (Fig. 5, A and B). Thus, the phenotypic similarities of a subset of TAO orbital fibroblasts and circulating fibrocytes suggest that they might share a common derivation.

Figure 5.

Orbital fibroblasts from individuals with TAO exhibit phenotypic attributes that are similar to those of fibrocytes and display CD34. Flow cytometric analysis of orbital fibroblasts from donors without or with TAO were stained with anti-CD34 (A) or anti-Col I antibodies (B). C, Undifferentiated orbital fibroblasts fail to express TSHR (<2%; left panel). After differentiation into adipocytes, TSHR levels increase substantially and 22% the CD34⁺ cells display the receptor (right panel). These data are representative of results involving four different donors with TAO. ISO, Isotype control.

When TAO-derived orbital fibroblasts were subjected to culture conditions promoting differentiation into adipocytes, TSHR expression increased substantially in a subset to levels resembling those found on circulating fibrocytes. As shown in Fig. 5C, incubation in a differentiation medium resulted in increased TSHR display on 22% of the fibroblasts after 18 d. Of great potential importance is the finding that these TSHR⁺ fibroblasts were uniformly CD34⁺, strongly suggesting that they derive from fibrocytes. Thus, a major aspect of the phenotype peculiar to TAO-derived orbital fibroblasts might relate to their being comprised, at least in part, of infiltrating fibrocytes and their potential to differentiate into TSHR-expressing adipocytes.

Evidence for fibrocyte infiltration of orbital tissue in GD

We next examined orbital tissues from individuals with TAO for evidence of cells displaying markers characteristic of fibrocytes. These were compared with tissues from healthy donors. As the images in Fig. 6 demonstrate, cells expressing both CD34 (denoted by green in Fig 6A) and LSP-1 (denoted as red in Fig. 6C) and therefore exhibiting a phenotype consistent with that of fibrocytes were abundant in the TAO-derived tissue (Fig. 6, A and C). Examination of serially sectioned tissue excludes these CD34⁺LSP⁺ cells as comprising vascular endothelium because they fail to express CD31 (Fig. 6E). In contrast, CD34⁺LSP⁺ fibrocytes are scarce in healthy tissue (Fig. 6, B and D). Given the high level of TSHR expression found on fibrocytes, we examined orbital tissue from patients with TAO and found abundant TSHR⁺ (green) cells that coexpressed CD34 (red) (Fig. 6G). The presence of TSHR⁺CD34⁺ fibrocytes in these affected tissues suggests that they may traffic to the orbit and mediate tissue reactivity and remodeling through local production of cytokines such as IL-6 and TNF-α.

Figure 6.

CD34⁺LSP-1⁺ TSHR⁺ fibrocytes can be identified in the orbital tissue of patients with TAO but are absent in tissues from healthy donors. A, CD34 expression (arrows, green fluorescein isothiocyanate) in TAO-derived tissue (inset, negative control staining). B, Absent CD34 expression in healthy tissue (inset, positive staining control). C, LSP-1 expression in TAO-derived tissue [red, arrows, nuclei counterstained with 4′,6′-diamino-2-phenylindole (DAPI) (blue)] (inset, negative control). D, Absence of LSP-1 expression in healthy tissue (inset, negative control). E, CD31 expression in disease-derived tissue is limited to vascular endothelium (red, arrows). F, Hematoxylin and eosin-stained consecutive thin sections of the same orbital tissue (×40). G, Fibrocytes present in orbital tissue from patients with TAO coexpress CD34 and TSHR. Thin-sectioned tissue from a donor with TAO was stained according to procedures described in Subjects and Methods with anti-CD34 (green) and anti-TSHR (red) antibodies. Nuclei were counterstained with DAPI (blue). Thin sections were then subjected to confocal microscopy. Inset contains a negative staining control.

Discussion

Fibrocytes constitute a population of cells possessing phenotypic attributes that both resemble and diverge from those of fibroblasts (27,28). They derive from monocyte or B cell bone marrow lineages, yet their functions within the marrow space and as circulating cells remain uncertain. Fibrocytes or their precursors migrate to sites of injury, in which they provide antigen-specific T cell stimulation, promote wound healing, and drive fibrosis (34,35,36,47). Because they exhibit phenotypic attributes similar to those associated with orbital fibroblasts cultured from individuals with TAO (46), we raised the question of whether fibrocytes might infiltrate orbital connective tissue in GD. Surprisingly the frequency of TSHR⁺ fibrocytes generated from the peripheral blood of these patients is markedly increased (Fig. 1). Moreover, among orbital fibroblasts cultivated from individuals with TAO, a substantial number exhibit the CD34⁺ phenotype. The subset of TAO orbital fibroblasts previously found to express TSHR, especially after incubation under culture conditions favoring their differentiation into adipocytes (6,18), appear to uniformly display CD34, suggesting strongly that they derive from fibrocytes (Fig. 5). In addition, these CD34⁺ fibrocytes can be detected in situ in orbital connective tissues from these same patients but appear to be absent in tissues from healthy donors (Fig. 6). Thus, recruitment from the peripheral blood and infiltration into tissues, coupled with high-level TSHR expression, could underlie, at least in part, the potential participation of fibrocytes in orbital GD.

Their behavior in vivo suggests that fibrocytes play important roles in host defense. For instance, they accumulate rapidly in thermal burns and are more frequently generated from the PBMCs of burned individuals (34). These findings appear analogous to ours. They efficiently present antigens to CD4⁺ T cells by virtue of their constitutive major histocompatibility complex-II expression (35), and thus, their accumulation in TAO might provide an explanation for how antigens are presented in tissues lacking secondary lymphoid structures (48,49,50). We found that fibrocytes express costimulatory molecules including IL-6, TNF-α (Fig. 4), and CD40 (our unpublished observations), which could provide T cell help and promote cell migration (51). A major source of morbidity associated with TAO relates to muscle fibrosis and reduced eye motility (52). Fibrocytes participate in fibrosis associated with chronic lung disease (53) and in animal models of bleomycin-induced lung fibrosis (31). Thus, it is possible that they mediate the debilitating aspects of TAO, including reduction in eye motility.

Multiple implications emanate from our current finding that tissue-infiltrating fibrocytes express high levels of TSHR. This observation offers potential insight into a link between systemic immune responses and anatomic site-specific manifestations including orbitopathy, acropachy, and dermopathy. High-level TSHR expression by these cells could provoke or amplify anti-TSHR immune responses by T and B cells. In addition, TSHR activation could underlie localized inflammation through fibrocyte production of cytokines such IL-6 and TNF-α. Alternatively, high-level display of autoantigens on fibrocytes could serve as a sink to dampen the biological impact of pathogenic antibodies, such as those directed against TSHR. Thus, our findings identify heretofore unrecognized potential disease mechanisms that could ultimately explain the development of TAO.

Mice harboring a disrupted TSHR gene lack central tolerance to that antigen, yet their persisting immune responses to TSHR vaccination remain comparable to those in wild-type animals (54,55). This result strongly suggests that loss of central tolerance may not play a critical role in the illicit immune responses against TSHR characterizing GD. Our current findings are congruent with the concept that peripherally expressed potential autoantigens, such as those displayed by fibrocytes, might overcome peripheral tolerance. This is especially true because fibrocytes can provide the necessary second signals required for robust lymphocyte activation. Given the apparent importance that the loss of peripheral tolerance plays in GD, fibrocytes might provide susceptible individuals with potential self-antigens that, in an appropriate context, might lead to disease. This concept formed the basis for earlier studies by Shimojo et al. (56,57) in which immunizing mice with fibroblasts overexpressing human leukocyte antigen-DR and TSHR overcame peripheral tolerance to the receptor. Thus, it is possible that excessive numbers of fibrocytes are released from the bone marrow, perhaps in response to nonspecific stress factors. This could result in host immunization with TSHR, IGF-IR, and any other potential self-antigens displayed on these cells. It has become clear that excessive fibrocyte generation can be associated with multiple forms of stress (58). Stress is also a common prelude to the onset of GD (59,60). Accelerated fibrocyte generation might result in loss of peripheral tolerance to antigen-displaying fibrocytes, leading to the development of GD in genetically susceptible individuals.

Others have reported previously that TSH and TSHR may play roles in tissues other than the thyroid, including fat (61,62). This possibility is proving particularly complex and tissue specific. Klein and Wang (63) demonstrated populations of mouse bone marrow cells that express either TSH or TSHR. Their findings suggest TSH production in monocyte-lineage CD45⁺CD11b⁺ cells (63). In contrast, CD11b⁻ cells, considered by the authors to represent lymphocyte precursors, display TSHR. When CD11b⁺ and CD11b⁻ cells were separated and maintained in vitro, exogenous TSH provoked selective production of TNF-α in CD11b⁻ cells (64). These observations in mice appear different from those we now report in human CD11b⁺ fibrocytes, which express TSHR (Fig. 3). Klein (65) also detected unanticipated TSH expression in thymus and intestine, whereas mouse bone marrow cells and thyroid produce a TSHβ splice variant (66). In aggregate, these findings implicate TSH and TSHR in normal immune function and suggest that they might play biological roles typically attributed to cytokines and their receptors (67). Abe et al. (68) reported that TSHR null mice exhibit osteopenia and found elevated levels of TNF-α in their bones. In that animal model, TSH directly inhibits TNF-α production and reduces the abundance of TNF-α-producing osteoclasts (69).

Our current findings suggest a potentially important role for the TSH pathway in regulating tissue reactivity. They identify a putative mechanism through which ligated TSHR could initiate inflammatory responses associated with GD by provoking cytokine production. Whether the magnitude of increased fibrocyte generation in GD changes with disease duration, treatment, or development of TAO is currently being assessed in longitudinal studies.