Rheumatoid arthritis (RA) is characterized by hyperplasia of synovial tissue and by the accumulation of large numbers of leukocytes within the inflamed synovium (1, 2). Despite considerable efforts, the contribution of individual leukocyte subsets and stromal elements to the pathogenesis of the disease remains elusive (3). T lymphocytes have been proposed to play an important role, but there is now ample evidence that macrophages and activated synovial fibroblasts also contribute significantly to the destructive nature of the disease (4). For example, synovial macrophages produce important proinflammatory cytokines, such as tumor necrosis factor α and interleukin-1β (IL-1β), while synovial fibroblasts are the principal cells mediating joint destruction, through their production of proinflammatory chemokines, cytokines, and matrix metalloproteinases (5).
It has been assumed that the predominant interaction of T lymphocytes in the synovial microenvironment is with antigen-presenting cells such as monocyte/macrophages, dendritic cells, and B cells. However, interactions between infiltrating bone marrow–derived hematopoietic cells (such as lymphocytes) and endogenous stromal cells (such as fibroblasts) have been shown to directly contribute to the intensity and persistence of chronic inflammation (6–9). For example, T cell–fibroblast interactions within the synovium induce the expression of adhesion molecules, cytokines, and chemokines by synovial fibroblasts (3, 10). They also lead to the survival and active, chemokine-mediated retention of T cells within the synovium (6, 7, 11). Whether these cellular interactions also regulate the formation of lymphoid aggregates within the synovium, as occurs in lymphoid neogenesis, has not been analyzed in detail.
Chronically inflamed tissues such as the rheumatoid joint often contain lymphoid aggregates that share many of the structural and functional features of secondary lymphoid tissue (12–14). We and others have recently shown that many of the chemokines required for effective lymphoid organogenesis are also expressed by stromal cells in the rheumatoid synovium (11, 12, 15). In addition, rheumatoid fibroblast-like synoviocytes (FLS) are able to support B cell survival (16), induce osteoclastogenesis, and regulate bone erosion (17). Rheumatoid FLS also display features of “nurse-like” stromal cells in their ability to support the spontaneous migration of B cells beneath them in vitro, a process called pseudoemperipolesis (18–20). These findings suggest that, as occurs in lymphoid neogenesis, the interactions between stromal cells of the synovial membrane and infiltrating lymphocytes might dictate the distribution of leukocyte subsets within the synovial microenvironment.
A striking feature of the rheumatoid synovium is the distribution of T cell subsets within the rheumatoid synovial compartment. CD4 T cells preferentially accumulate in a perivascular distribution, whereas CD8 T cells are sparsely distributed throughout the synovial tissue (21). Furthermore, the ratio of CD8 to CD4 T cells within the synovial tissue is much lower than that within the synovial fluid. The molecular basis for this high degree of cellular organization within distinct microdomains remains unclear.
In this study, we set out to determine whether differential chemokine-dependent interactions between T cell subsets and synovial fibroblasts might explain this longstanding conundrum of distinct T cell subset distribution within distinct microdomains in rheumatoid synovitis. We found that FLS derived from the rheumatoid synovium are able to support high levels of T cell migration beneath them (pseudoemperipolesis). The ability to support pseudoemperipolesis was dependent on the expression of the chemokine stromal cell–derived factor 1 (SDF-1; CXCL12), which was constitutively overexpressed by rheumatoid FLS. Unlike the case for B cells, neither CD4 nor CD8 T cells required CD49d–vascular cell adhesion molecule 1 (VCAM-1) interactions for efficient pseudoemperipolesis. CD8 T cells migrated more efficiently and with a higher velocity than CD4 T cells when underneath rheumatoid FLS. Finally, studies using synovial T cells ex vivo confirmed the in vivo relevance of the CXCR4–SDF-1 interaction. These results support the concept that rheumatoid FLS directly affect the behavior of infiltrating T lymphocytes, and that T cell–fibroblast interactions contribute to the distinctive architectural features that define the rheumatoid microenvironment.
- Top of page
- MATERIALS AND METHODS
The relative contribution of factors produced by T cells and fibroblasts to the accumulation of inflammatory cells within distinct microdomains in the rheumatoid synovium remains unclear. In this study, we set out to test the hypothesis that chemokines (such as SDF-1) produced by rheumatoid FLS are responsible for generating this distinctive microcompartmentalization. Using a coculture model of lymphocyte accumulation (pseudoemperipolesis), we found that FLS derived from 6 different patients with RA were able to support both CD4 and CD8 T cell pseudoemperipolesis at much higher levels than FLS from nonrheumatoid tissue. Pseudoemperipolesis was rapid (maximal within 2 hours) and depended on the expression of SDF-1 (but not VCAM-1) by the rheumatoid FLS as well as the expression of CXCR4 on the T cells. CD8 T cells were able to migrate much more efficiently and at higher velocity under the rheumatoid FLS monolayers than were CD4 T cells. This ability was an intrinsic property of CD8 T cells compared with CD4 T cells.
Only rheumatoid FLS were able to support high levels of T cell pseudoemperipolesis. It remains unclear whether this is a general property of rheumatoid FLS or whether it represents the activity of a small number of specialized “nurse-like” cells (NLCs) found only in the rheumatoid synovium (28). Previous studies have provided contradictory results for B cell pseudoemperipolesis in this regard. Shimaoka et al (19) have shown that only rheumatoid NLCs, but not conventional FLS, were able to support pseudoemperipolesis. In contrast, other investigators have found that conventional FLS from joints affected with either RA or osteoarthritis could support B cell pseudoemperipolesis (18). All observers have found that dermal fibroblasts do not support constitutive B cell pseudoemperipolesis. Burger et al (18) have suggested that these discrepancies are due to the way in which NLCs are prepared (via limiting dilution) and maintained (by the use of conditioned medium).
Our findings using an extended panel of primary FLS demonstrate that the ability to support basal levels of T cell pseudoemperipolesis is an intrinsic property of FLS, in agreement with findings of studies on B cells by Burger et al (18). However, we also observed higher levels of T cell pseudoemperipolesis with rheumatoid FLS compared with other nonrheumatoid FLS (resolved viral arthritis and skin). This is more in keeping with the findings of Shimaoka et al (19), who determined that only rheumatoid NLCs could support B cell pseudoemperipolesis. The additional, enhanced level of pseudoemperipolesis that we observed for rheumatoid FLS was seen for both CD4 and CD8 T cells, which further emphasizes that this feature is intrinsic to rheumatoid FLS.
We found that CD8 T cell pseudoemperipolesis was much more efficient than CD4 T cell pseudoemperipolesis. Since this was observed for all FLS as well as for fetal lung fibroblasts, this suggests that compared with CD4 T cells, CD8 T cells have an increased intrinsic ability to undergo migration beneath fibroblasts. Moreover, we found that the velocity of migration of CD8 T cells was much higher than that of CD4 T cells (Figure 2). It is tempting to speculate that this intrinsic ability of CD8 T cells might account for the differential localization of T cell subsets at sites of chronic inflammation and for the distinctive CD4:CD8 T cell ratio that is reversed between synovial tissue and fluid.
Recent studies have shown that stromal cells present within the inflamed synovium share many features with, and may even originate from, mesenchymal stem cell precursors found in peripheral blood (28, 29). A common feature of these mesenchymal stromal cells is their ability to produce high levels of the chemokine SDF-1 (CXCL12). We found that the high levels of T cell pseudoemperipolesis mediated by rheumatoid FLS was dependent on the ability of the rheumatoid FLS to express SDF-1. Dermal fibroblasts, which produce very little SDF-1, were unable to support pseudoemperipolesis. Rheumatoid FLS produced very high basal levels of SDF-1, as measured by real-time TaqMan PCR (40–2,240 copies). Other FLS and fetal lung and dermal fibroblasts expressed much lower levels of SDF-1 mRNA (range 0.75–25 copies). Despite the wide range in the number of copies, all 6 rheumatoid FLS lines were able to support pseudoemperipolesis, suggesting that a relatively low threshold level of SDF-1 exists for this function.
Although we have not shown a direct correlation between the expression of SDF-1 mRNA and SDF-1 protein production, our findings are consistent with those of other investigators who have found high levels of SDF-1 within the rheumatoid synovium (11, 27, 30, 31). Our finding of high levels of VCAM-1 mRNA expression by rheumatoid fibroblasts (45–970 copies) compared with other FLS and fetal lung and dermal fibroblasts (0.03–5.1 copies) is also in keeping with previous studies which have shown that VCAM-1 protein is constitutively expressed on rheumatoid, but not skin, fibroblasts (32).
Inhibition studies of T cell pseudoemperipolesis using pertussis toxin and function-blocking antibodies confirmed the functional relevance of the overexpression of SDF-1 by rheumatoid FLS. We found that while blockade of SDF-1–CXCR4 interactions inhibited T cell pseudoemperipolesis mediated by rheumatoid FLS, there was no effect of blockade of chemokines involved in either the CCR5 (RANTES, MIP-1α) or CXCR3 (IP-10, Mig) system. Both CCR5 and CXCR3 have been implicated in the recruitment of T cells to the inflamed synovium (33). Therefore, taken together, these results suggest that while CCR5 and CXCR3 may be important for T cell recruitment to the synovium (endothelial selection), T cell retention depends more on SDF-1/CXCR4 (stromal selection).
We did not find any effect of α4 integrin (CD49d) blockade on either CD4 or CD8 T cell pseudoemperipolesis despite high levels of VCAM-1 on rheumatoid fibroblasts (Figures 5A and B). This is unlike the case for B cells, which require both SDF-1 and VCAM-1 to support B cell pseudoemperipolesis (Figure 5C) and survival (18, 19). In agreement with the findings of Burger et al (18), we found that B cell pseudoemperipolesis was relatively inefficient (∼5% of total input cells) compared with CD4 (25%) and CD8 (50%) T cell pseudoemperipolesis. It is therefore tempting to speculate that while the functional consequence of SDF-1–mediated T cell pseudoemperipolesis is to regulate T cell positioning within the synovium, B cell pseudoemperipolesis supports SDF-1/VCAM-1–dependent survival and activation. This is in keeping with findings of our previous studies, which have shown that T cell survival within the rheumatoid synovium depends on the production of type I interferon and does not require VCAM-1–α4β1 interactions (6, 7). Whether T cell pseudoemperipolesis regulates T cell activation and cytokine production is currently under investigation in our laboratory. In addition, recent studies have shown that T cells can directly activate rheumatoid fibroblasts in coculture (9).
We and others have recently suggested that SDF-1 produced by stromal cells and CXCR4 expressed on infiltrating cells play an important role in the accumulation of CD4 memory T cells in the rheumatoid synovium (11, 27). The close relationship between synovial stromal cells and infiltrating T cells suggests that there may be intrinsic features of T cells and fibroblasts that contribute to the pathology of RA. In this study, we found that a critical requirement for both CD4 and CD8 T cell pseudoemperipolesis was the ability of fibroblasts to make SDF-1 and for the migrating T cells to express CXCR4. Neither process on its own was sufficient (Figure 7). Intriguingly, even when SDF-1 protein was added to skin fibroblasts (which do not produce SDF-1 at high levels), T cell pseudoemperipolesis did not occur, suggesting that additional factors, such as an appropriate extracellular matrix to support SDF-1 presentation, are perhaps needed.
Our findings emphasize the critical role of interactions between leukocytes and the stromal environments in which they reside in driving the pattern of leukocyte accumulation in chronic inflammatory diseases (34). Moreover, they suggest that stromal cell–derived factors such as SDF-1 that guide lymphocyte positioning within tissues might be attractive therapeutic targets in chronic inflammatory joint disease.