The precise role played by CD8⁺ T cells in the pathogenesis and inflammation of rheumatoid arthritis (RA) is unclear. In the synovial membrane, the most common IFN-γ-producing cell is the CD8⁺ T cell, suggesting that this population of T cells plays a major role in macrophage activation and perpetuation of the inflammatory response [1]. CD8⁺ T cells were recently associated with the presence of germinal centers in RA synovium [2], suggesting a role for CD8⁺ T cells in the formation or maintenance of those lymphoid structures in the synovium. Further studies indicated that CD8⁺ T cells exhibit oligoclonality in the peripheral blood [3, 4] and synovial fluid of RA patients [5], raising the question of whether this oligoclonality is antigen driven. However, recent studies have indicated that large numbers of virus-specific CD8⁺ T cells preferentially accumulate in the synovial fluid of RA patients and that these cells are also oligoclonal, suggesting that non-antigen-specific homing may be responsible for the observed oligoclonality of CD8⁺ T cells in the synovial fluid [6]. Because chemokines such as macrophage inflammatory protein-1α and RANTES (regulated upon activation, normal T-cell expressed and secreted) are expressed in RA synovial tissue [7, 8], subsets of CD8⁺ T cells may be preferentially recruited into the synovial tissue in a non-antigen-specific manner. If the expression of chemokines is also accompanied by a perturbation in CD8⁺ T-cell homeostasis in the periphery that favors differentiation into cell types that can be recruited into the synovium, then a vicious cycle may be set up in RA in which there is continuous generation of CD8⁺ T cells that can be recruited into the synovium, resulting in chronic inflammation and joint destruction.

Recently, memory CD8⁺ T cells were classified into three distinct populations, based on phenotype [9–11]: a central memory population, which is CD45RA^-CCR7⁺CD62L⁺CD28⁺IL-2⁺IFN-γ^-; and two effector memory populations, namely the CD45RA^-CD62L^-CCR7^- and the terminally differentiated CD45RA⁺CD62L^-CCR7^- populations. The two latter effector memory populations contain perforin, secrete IFN-γ and tumor necrosis factor-α, are cytotoxic, and are capable of rapid effector function after stimulation [9–11].

Although a linear model of differentiation has been suggested for these memory populations (i.e. central memory T cells CD45RA^-CCR7⁺CD62L⁺ → effector memory T cells CD45RA^-CD62L^-CCR7^- → effector memory T cells CD45RA⁺CD62L^-CCR7^- [10]), the exact relationship between these populations is not fully established. Indeed, Champagne et al. [12] suggested that the differentiation may not be linear at all. The central and effector memory phenotypes of CD4⁺ and CD8⁺ T cells in peripheral blood of RA patients are unknown. Determination of these phenotypes in RA may provide important insights into T-cell homeostasis, and we therefore examined the distribution of CD4⁺ and CD8⁺ T cells into these subpopulations because such a study may reveal differences in the differentiation of T cells in RA patients. Decreases in some of the subpopulations in peripheral blood may indicate that there is a selective migration of these cells out of the peripheral blood, decreased survival of these cells, or blockade in their differentiation. Perturbations in the homeostasis of memory T cells may play an important role in the pathogenesis of RA by generating effector cells that can contribute to the synovial inflammation of RA.

Naïve and memory subpopulations of CD4⁺ and CD8⁺ T cells from RA and SLE patients were compared with those in healthy control individuals to determine T-cell maturation differences between those groups.

As compared with the healthy control group, RA patients had fewer CD45RA⁺CD62L⁺ CD4⁺ naïve T cells (32 ± 4.8% in RA patients [n = 8] and 42 ± 6.5% in healthy controls [n = 8], respectively), although this difference was not statistically significant (Fig. 1a,1b). The CD45RA^-CD62L⁺ CD4⁺ central memory T-cell population was significantly increased in RA patients (50 ± 3.7% [n = 8]) as compared with the healthy control group (38 ± 4.4% [n = 8]; P < 0.05; Fig. 1a,1b). No differences were found in the CD45RA^-CD62L^-CD4⁺ effector memory population (15 ± 2.2% for RA patients and 18 ± 2.6% for healthy controls [n = 8 each]) or in the terminally differentiated CD45RA⁺CD62L^- CD4⁺ effector memory population (1.7 ± 0.5% for RA patients and 2.2 ± 0.6% for healthy controls [n = 8 each]; Fig. 1a,1b).

In the CD8⁺ T-cell population, 39 ± 6.2% were CD45RA⁺CD62L⁺ naïve cells for the RA patients and 28 ± 3.4% for the healthy control group (Fig. 1a,1b). The central memory CD45RA^-CD62L⁺ CD8⁺ T-cell population was significantly increased in RA patients (17 ± 3.5% [n = 8]) as compared with the healthy control group (9 ± 1.8% [n = 8]; P < 0.05; Fig. 1a,1b). No difference was found between patients and healthy control group in the CD45RA^-CD62L^- CD8⁺ effector memory populations (18 ± 3.2% for RA patients and 25 ± 4.5% for healthy controls [n = 8]), whereas the CD45RA⁺CD62L^- CD8⁺ terminally differentiated effector memory population was significantly decreased in RA patients (26 ± 2.4%) as compared with healthy controls (38 ± 4.8% [n = 8]; P < 0.05; Fig. 1a,1b).

No significant differences were found when CD4⁺ and CD8⁺ T cells of SLE patients were compared with the CD4⁺ and CD8⁺ T cells of matched healthy control individuals (Fig. 1c). In the CD4⁺ T-cell population, 35 ± 4.6% of cells from SLE patients and 45 ± 4.7% in the healthy controls exhibited a naïve phenotype; the central memory phenotype was expressed by 42 ± 3.8% of the CD4⁺ T cells from SLE patients (n = 12) and in 37 ± 3.1% of the CD4⁺ T cells from healthy controls (n = 12). Of the CD4⁺ T cells, 20 ± 3.6% and 16 ± 2.0% were effector memory cells in the SLE and healthy control groups (n = 12 in each), respectively, and only a very small population of the cells were terminally differentiated effector memory CD4⁺ T cells in SLE patients (2.4 ± 0.9%) and healthy controls (1.7 ± 0.5%; Fig. 1c). The CD8⁺ T-cell compartment of SLE patients consisted of 42 ± 5.6% CD45RA⁺CD62L⁺ naïve cells, 14 ± 2.9% CD45RA^-CD62L⁺ central memory, 20 ± 4.1% CD45RA^-CD62L^- effector memory, and 24 ± 4.9% CD45RA⁺CD62L^- terminally differentiated effector memory CD8⁺ T cells (n = 12; Fig. 1c). In the healthy control group, 39 ± 5.8% CD45RA⁺CD62L⁺ naïve cells, 9 ± 1.3% CD45RA^-CD62L⁺ central memory, 23 ± 3.4% CD45RA^-CD62L^- effector memory, and 29 ± 5.2% CD45RA⁺CD62L^- terminally differentiated effector memory CD8⁺ T cells were found (n = 12; Fig. 1c).

A positive correlation was found between the age and the percentage of CD45RA⁺CD62L^- terminally differentiated effector memory CD8⁺ T cells in the healthy control group (r² = 0.64 [n = 13]; P < 0.001; Fig. 1d), indicating that this effector population increases with age. However, no such correlation was detected in RA and SLE patients (Fig. 1d). Finally, the frequency of CD45RA⁺CD62L^- CD8⁺ T cells did not correlate with disease duration or treatment in either RA or SLE patients (data not shown).

The present study shows that the differentiation of peripheral blood CD8⁺ T cells is skewed in patients with RA and results in an increase in central memory CD45RA^-CD62L⁺ CD8⁺ T cells, with a concomitant decrease in terminally differentiated effector memory CD45RA⁺CD62L^-CD8⁺ T cells. The increase in central memory CD45RA^-CD62L⁺ T cells was also found in the CD4⁺ T-cell population in RA patients. This skewed differentiation was not observed in healthy age-matched control individuals and in SLE patients, indicating that this perturbation in homeostasis of T cells is a specific feature of RA.

Although the naïve/memory phenotype of T cells has previously been investigated in RA in numerous studies using CD45RA and CD45RO expression as markers of naïve and memory cells, respectively, that approach has suffered from the limitation that large numbers of CD45RA⁺ CD8⁺ T cells are actually effector memory cells [10, 13]. The CD45RA/CD45RO oversimplification has also resulted in rather confusing conclusions regarding T-cell homeostasis, such as defects in primary T-cell homeostasis based on reduced T-cell receptor excision circle (TREC) levels in naïve CD4⁺ T cells (defined as CD45RO^-) in RA patients [14]. Our findings suggest that reduced TREC levels in the CD45RO^- CD4⁺ T-cell population may not be due to a reduction in TRECs in naïve cells but rather to reduced TRECs in the CD45RA⁺CD45RO^-CD62L^- effector memory CD4⁺ T cells. It should be noted that previous studies have reported 'false naïve' CD45RA⁺ populations of CD4⁺ and CD8⁺ T cells in peripheral blood of RA patients [15]; however, the nature of these cells, the exact phenotype, and the significance was not known at that time.

Our finding that peripheral blood CD8⁺ T cells exhibit increased central memory phenotype and decreased terminally differentiated effector memory phenotype suggests that the peripheral blood homeostasis of CD8⁺ T cells is perturbed in RA. Perturbations in CD8⁺ T-cell maturation have been shown for HIV-specific CD8⁺ T cells, in which there is an accumulation of preterminally differentiated CD45RA^-CD62L^- CD8⁺ T cells [12, 16], and such a lack of differentiation may result in functional or homing defects. In RA we found a decrease in terminally differentiated CD45RA⁺CD62L^- CD8⁺ T cells with a concomitant increase in the CD45RA^-CD62L⁺ central memory population. If one accepts the linear model of differentiation [10], which we note has been challenged [12], then our findings indicate that in RA there may be an accelerated differentiation of naïve cells into central memory CD4⁺ and CD8⁺ T cells. This accelerated differentiation may be due to a non-antigen-specific effect in RA that differentiates all peripheral T cells irrespective of their specificity, or it may actually reflect an antigen-specific expansion of T cells potentially driven by autoantigen.

The decrease in CD45RA⁺CD62L^- effector memory CD8⁺ T cells in peripheral blood we found in RA patients may reflect a decrease in the survival of these cells. It should be noted, however, that peripheral blood T cells from RA patients do not exhibit an increase in apoptosis in in vitro cultures, which is in contrast to synovial membrane T cells [17, 18]. This may suggest that the skewed phenotype of the CD45RA⁺CD62L^- effector memory CD8⁺ T cells is more likely due to an increase in the migration of these cells into sites of inflammation. However, a blockade of the differentiation of central memory CD45RA^-CD62L⁺ CD8⁺ T cells into effector memory CD8⁺ T cells would also result in an increase in the central memory population with a concomitant decrease in the effector T cells, as observed in the present study.

Studies of the phenotype of CD8⁺ T cells in the synovial membrane and fluid may shed light as to whether this skewed phenotype is also found in these sites or whether there is an enrichment for CD45RA⁺CD62L^-CD8⁺ T cells, indicating increased recruitment into the inflamed synovium in RA. Inflammation and production of chemokines such as macrophage inflammatory protein-1α and RANTES [7, 8] in the synovium may result in preferential recruitment of such effector memory CD8⁺ T cells (which are important contributors to IFN-γ production) and subsequent macrophage activation, because terminally differentiated CD45RA⁺CD62L^- CD8⁺ T cells have been shown to express higher levels of perforin and may be more potent effector cells [10]. The question arises of whether the observed skewed differentiation of CD8⁺ T cells in RA patients is due to medication, especially steroids. As shown in Table 1, 38% of the RA patients and 58% of the SLE patients were receiving steroid treatment. However, the skewed memory phenotype was only observed in the RA patients, suggesting that this treatment is not responsible for the differences in CD4⁺ and CD8⁺ T-cell phenotypes.

Findings from the present preliminary study show that peripheral blood CD8⁺ T cells in RA exhibit a skewed effector memory phenotype. This skewed phenotype was not found in CD4⁺ T cells in RA and was not seen in age-matched healthy control individuals or in SLE patients. The skewed phenotype may be a result of accelerated differentiation and migration into sites of inflammation. An understanding of the mechanisms that are involved in this skewed differentiation of effector memory CD8⁺ T cells may prove valuable in elucidating the pathogenesis of RA.