GO develops in approximately 50% to 60% of patients with Graves’ disease.
9 42 Graves’ disease is thought to be an autoimmune process involving autoantigens common to the thyroid and the orbit, so that autoimmune attack on the thyroid leads to hyperthyroidism and associated hypermetabolic consequences, and autoimmune attack on the cells of the orbit, particularly fibroblasts, leads to GO.
1 13 Differentiation of fibroblasts to adipocytes leads to accumulation of fatty tissue in the orbit, whereas proliferation of orbital fibroblasts causes accumulation of matrix components and connective tissue, and fibrosis of the ocular muscles. It has been shown that orbital tissue from patients with GO is infiltrated with B and T cells, and it has been reported that orbital fibroblasts from Graves’ patients can stimulate proliferation of autologous T cells.
23 24 25 26 34 Here, for the first time, we demonstrate that T cells from patients with GO can stimulate proliferation of autologous orbital fibroblasts, thus providing a mechanism by which infiltration of orbital tissue by autoimmune lymphocytes can drive the pathogenic features of GO.
We established an rAMCR assay in which irradiated T cells (enriched from patient PBMCs) are mixed with orbital fibroblasts cultured from the same patient. In all patient samples examined, autologous T cells drove a significant increase in fibroblast proliferation over orbital fibroblasts cultured alone
(Figs. 1A 1B) . It is worth noting that this phenomenon was not seen in fibroblasts derived from a nearby anatomic site in patients who were undergoing blepharoplasty and had undergone orbital decompression surgery
(Fig. 2) . This result agrees with earlier studies showing that deep orbital fibroblasts express a unique phenotype
14 19 20 and suggests that this phenotype makes them susceptible to autoimmune attack in Graves’ disease.
To determine whether direct cell–cell contact is necessary for stimulation of orbital fibroblast proliferation, T cells and fibroblasts were cultured separated by a semi-permeable membrane. Separating the orbital fibroblasts from direct T-cell contact resulted in a 65% decrease in proliferation
(Fig. 3) , indicating that whereas cell–cell contact is essential for maximum stimulation of proliferation, one or more T cell-derived diffusible factors contribute to the net increase in proliferation. To identify which ligand–receptor interactions are necessary, we performed rAMCR assays with blocking antibodies to MHC class II, CD40, and CD40L. Each antibody completely inhibited T-cell stimulation of increased orbital fibroblast proliferation
(Fig. 4) . This indicates that in the absence of signals from the fibroblasts, the T cells do not express proproliferative effector functions, and that the fibroblasts signal via the MHC–TCR and CD40–CD40L pathways. This suggests that the proproliferative signals are dependent on recognition of fibroblast autoantigens by the T cells. However, the antibody results also present an apparent paradox, in that blocking MHC class II and CD40–CD40L interactions completely inhibited fibroblast proliferation, whereas blocking cell contact via the cell-migration plate membrane only partially inhibited fibroblast proliferation. It should be noted that the enriched T cells were expanded in vitro for 8 days with PHA and IL-2 and were probably in a state of partial activation. We hypothesize that the IL-2-activated T cells produce some soluble factors (probably cytokines and chemokines) that stimulate fibroblast proliferation suboptimally, while also inducing expression on the fibroblasts of MHC class II
(Fig. 5) . With increased expression of class II, the fibroblasts can fully activate the T cells via an antigen-dependent mechanism requiring MHC–TCR (signal 1) and CD40–CD40L (signal 2) interactions.
30 43 The activated T cells then produce increased cytokines, driving fibroblast proliferation. In the cell-migration membrane experiments, the IL-2-activated T cells produced low levels of stimulatory factors but could not be further activated in the absence of cell contact. However, in the antibody blocking experiments, blockade of class II MHC resulted in signal 2 without signal 1, whereas blockade of CD40 or CD40L costimulation resulted in signal 1 without signal 2, both of which conditions result in T cell anergy.
29 44 Thus, the T cells no longer produce even low levels of proproliferative factors.
Previous reports have shown that orbital muscle and fat tissue of Graves’ disease patients is infiltrated with CD4
+ T cells.
23 25 26 To investigate whether the stimulation of orbital fibroblast proliferation is dependent on CD4
+ T cells, we separated the enriched expanded T-cell population into CD4
+ and CD4
− fractions, by using magnetic bead separation
(Fig. 6A) . Neither fraction by itself stimulated orbital fibroblast proliferation. The enriched T-cell fractions (prepared from PBMCs on nylon wool columns followed by activation in vitro for 8 days) were typically >90% pure, as determined by flow cytometry. This raised the possibility that efficient stimulation of fibroblast proliferation requires both CD4
+ T cells and a CD4
− accessory cell, so that fractionating the enriched T cells on the basis of CD4 expression separated the accessory cells from the CD4
+ T cells. An alternative possibility is that the magnetic bead separation technique alters the T cells so that they no longer efficiently stimulate proliferation. To address this possibility, T cells were sorted into CD4
+ and CD4
− fractions by using a different technology, fluorescence activated cell sorting (FACS;
Fig. 6B ). Again, neither fraction stimulated the orbital fibroblasts to proliferate. Attempts to restore proliferation by combining the CD4
+ and CD4
− fractions were unsuccessful (data not shown), suggesting that this result could be an artifact caused by ligating CD4 with a specific antibody, a step common to both the magnetic bead method and FACS. Alternatively, a population of accessory cells may have been lost during purification, or the cells were not recombined in the optimal ratios.
To explore further the possibility that efficient stimulation of fibroblast proliferation requires an accessory cell, we used magnetic bead separation to remove potential accessory cells without affecting the CD4
+ T cells. We reasoned that the most likely accessory cells would be professional APCs such as DCs or monocyte-macrophages, which could present autoantigens to CD4
+ T cells resulting in T-cell activation. Magnetic beads coupled to antibodies for CD1a (DCs and DC precursors in peripheral blood), CD83 (DCs) and CD14 (monocyte-macrophages) were used to deplete the enriched T cells of antigen-expressing cells (<0.1%, data not shown) before rAMCR. None of these antibodies resulted in a decrease in orbital fibroblast proliferation
(Fig. 7) . This further supports the hypothesis that orbital fibroblast proliferation is dependent on CD4
+ T cells, but that current methods of purifying CD4
+ T cells alter their stimulatory capacity in vitro. However, it remains possible that one or more accessory cell types are involved in the rAMCR. It should be noted that the phenotypes and sorting characteristics of human APCs are much less well understood than in the mouse
45 and it is difficult to obtain sufficient blood cells for complex analysis from our surgical patients. Further analysis is necessary to determine whether CD4
+ T cells are sufficient for stimulation of orbital fibroblast proliferation or whether an accessory cell is required, and if so, what it is.
It should be noted that in the present study, fibroblasts were isolated from deep orbital fat obtained during decompression surgery, whereas the T cells were derived from circulating blood mononuclear cells. The ocular muscles are also an important site of disease, and have been shown to be infiltrated by T cells by immunohistochemistry.
23 26 However, biopsy of the extraocular muscles for research cannot be performed on ethical grounds, due to the risk of muscle impairment and increased patient symptoms. The number of T cells available from the fat biopsy tissue is too low to permit conventional analysis; however, it would be interesting to compare the response of fibroblasts to circulating and tissue infiltrating T cells by using more sensitive methods.
Previously, we and others have reported that Graves’-derived fibroblasts express MHC class II and CD40, and that ligation of CD40 on fibroblasts results in increased production of inflammatory cytokines.
19 22 28 In the present study, we extended these results by demonstrating that T cells from patients with Graves’ disease interacted with autologous orbital fibroblasts via the class II and the CD40–CD40L pathways to promote increased fibroblast proliferation. Although glycosaminoglycan synthesis was not measured in the present study, we have reported that CD40L stimulates orbital fibroblasts to produce hyaluronan,
15 and this study demonstrated that the interaction between orbital fibroblasts and autologous T cells was CD40–CD40L dependent. Thus, stimulation by T cells is responsible for many of the pathogenic properties of orbital fibroblasts in GO. This T-cell dependence supports previous indications that the pathogenesis of GO is antigen dependent. The exact antigens involved in GO have not been identified, although it has been hypothesized that they will be shared by thyroid and orbital tissue. Several candidates have been proposed, including TSHR,
46 47 48 49 and serum immunoglobulins from Graves’ patients can stimulate orbital fibroblasts to produce T-cell chemoattractants.
50 51
In addition to the role of T cells in stimulating autologous orbital fibroblasts reported herein, orbital fibroblasts can stimulate autologous T cells.
34 Taken together, these results describe a positive feedback loop in which thyroid-stimulating autoantibodies promote migration of T cells to the orbit, which stimulate fibroblasts in orbital tissue to upregulate expression of MHC class II and the presentation of autoantigens. This in turn further activates the T cells to produce surface and/or diffusible factors that drive activation and proliferation of orbital fibroblasts, leading to expression of fibroblast-based diseases including proliferation of fibroblasts and excess connective tissue, deposition of matrix glycosaminoglycans, intramuscular fibrosis, and differentiation and proliferation of adipocytes. New therapies that interrupt these feedback processes hold the potential to halt the progression of GO and may even allow for nonsurgical reduction of GO.
The authors thank Mike Kasim, MD, and Kimberly Cockerham, MD, for providing surgical specimens.