One hundred and twenty-eight patients with RA were enrolled in the study. All fulfilled the American College of Rheumatology criteria for RA and attended the Rheumatology Clinic at Karolinska University Hospital, Stockholm, Sweden for corticosteroid injections of inflamed joints 11. Before the corticosteroid injections, synovial fluids were acquired from the knee joints (n = 128), the elbow (n = 1) or the shoulder joints (n = 2). Eighty per cent of the patients were women, median age of 56 years (range, 25–82 years) and a median disease duration of 9 years (range, 0–45 years).

Assessment of erosive disease was performed by radiographic evaluations of the ankle joints or wrist joints by the same two rheumatologists. Radiographic changes in one or more joints were found in 51 out of 70 (73%) patients included in these analyses. The majority of the patients were treated either with nonsteroidal anti-inflammatory drugs, with systemic or local corticosteroid treatment, with methotrexate alone or in combination with corticosteroids (prednisolone), or with TNF blockers alone or in combination with methotrexate. Some patients were untreated.

This study was approved in compliance with the Helsinki Declaration by the Ethics Committee of the Karolinska University Hospital, and all patients and healthy subjects gave informed consent.

Knee joint synovial biopsies were acquired according to a previously described procedure 12. Biopsies were taken at the site of inflammation, either close to cartilage or not close to cartilage, defined as either less than 1.5 cm or more than 1.5 cm from cartilage, respectively.

Frozen unfixed synovial biopsy sections were fixed with acetone. Sections were incubated overnight with the cocktail of primary antibodies – CD244 (R&D Systems, Minneapolis, MN, USA), CD4 (Becton Dickinson, San Jose, CA, USA), CD3 (DakoCytomation, Glostrup, Denmark) – or the isotype control antibodies – goat IgG (Caltag Laboratories, Burlingame, CA, USA), mouse IgG₁ (DakoCytomation) and rabbit immunoglobulin (DakoCytomation). Excess of antibodies were washed away before incubation with the secondary antibodies – anti-sheep/goat immunoglobulin-biotin (The Bidning Site, Birmingham, UK), avidin-Oregon Green 488 (Molecular Probes, Eugene, OR, USA), anti-mouse IgG-Rhodamine RedTM-X (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and anti-rabbit IgG-AMCA (Jackson ImmunoResearch).

Stained tissue sections were examined with a Leica DM RXA2 microscope (Leica Microsystems, Wetzlar, Germany) equipped with a Leica DC 300F (Leica Microsystems DI, Cambridge, UK) digital colour video camera connected to a PC computer. Photographs were analysed with Leica IM500 software (Leica Microsystems, Heerbrugg, Switzerland).

CD4⁺CD28^null T cells were identified by morphologically cell-like structures with co-localized immunostainings of CD3, CD4 and CD244, and were manually quantified in independent analyses performed by two persons. CD4^dim macrophages/monocytes and NK cells 13, which also might express CD244, could be excluded using the combination of CD3 and CD4 in the three-colour stainings to identify CD4⁺ T cells. The density of CD4⁺CD28^null T cells were calculated by dividing the number of CD4⁺CD28^null T cells by the total area of infiltrating T cells, measured with Image J software version 1.34s (National Institutes of Health, Bethesda, MD, USA).

The frequency of CD4⁺CD28^null T cells in peripheral blood and synovial fluid was analysed by four-colour flow cytometry (FACSCalibur instrument; Becton Dickinson Immunocytometry Systems, San Jose, CA, USA) in peripheral blood mononuclear cells and synovial fluid mononuclear cells after Ficoll separation (Ficoll-Paque Plus; GE Healthcare Biosciences AB, Uppsala, Sweden). The antibodies used were CD3-FITC, CD28-APC (Pharmingen; Becton Dickinson, San Diego, CA, USA), CD4-PerCP (Becton Dickinson, San Jose) and CD244-PE (Immunotech, Marseille, France).

The TCR-Vβ usage was determined by the IOTest1 Beta Mark kit (Beckman Coulter, Marseille, France). The TCR-Vβ stainings were combined with antibodies to CD4 and CD28 (see above) to identify CD4⁺CD28^null T cells and CD4⁺CD28⁺ T cells.

Peripheral blood mononuclear cells and synovial fluid mononuclear cells stimulated with HCMV antigens (see below) were analysed by flow cytometry after immunostaining with IFNγ-FITC, CD28-PE, CD3-APC and CD14-APC-Cy7 (all from Becton Dickinson, San Diego, CA, USA) and CD4-PerCp (Becton Dickinson, San Jose, CA, USA).

Flow cytometric data were analysed with CellQuest software (Becton Dickinson, Franklin Lakes, NJ, USA) or FlowJo software (Tree Star Inc., Ashland, OR, USA). The frequency of CD4⁺CD28^null T cells was calculated as the percentage of CD28-negative cells in the gated CD3⁺CD4⁺ population.

The functional capacity of CD4⁺CD28^null T cells from the synovial fluid and peripheral blood were assessed by IFN-γ production. Two million peripheral blood mononuclear cells and synovial fluid mononuclear cells from eight patients were either stimulated with plate-bound anti-CD3 antibodies (OKT-3) at 0 or 0.1 μg/ml for 4 hours, or by 2 μg/ml pp65 and immediate early HCMV antigens (JPT Peptide Technologies GmbH, Berlin, Germany) for 8 hours.

Activated cells were either detected by secretion (MACS Secretion Assay; Miltenyi Biotec, Bergisch Gladbach, Germany) or by upregulation of intracellularly stored IFN-γ (Becton Dickinson, San Diego, CA, USA) according to the manufacturers' protocols. The frequency of IFN-γ secreting CD3⁺CD4⁺CD28^null T cells was analysed by flow cytometry (see above).

The anti-CCP2 test (Immunoscan RA, Mark 2; Euro-Diagnostica, Arnhem, The Netherlands) was used to determine the levels of anti-citrullinated peptide/protein antibodies (ACPA) in the serum and synovial fluid. A cutoff value of 25 U/ml was used according to the manufacturer's instructions. Serum and synovial fluid samples were diluted equally (1:50) and were analysed on the same plate.

The presence of anti-HCMV IgG and IgM antibodies in the serum and the synovial fluid, from the same time point as the screening of CD4⁺CD28^null T cells, was tested in an enzygnost anti-HCMV/IgG ELISA and an enzygnost anti-HCMV/IgM ELISA (Dade Behring, Marburg, Germany). Sera from HCMV-seronegative patients were further examined for detection of IgG against HCMV using antigens prepared from a HCMV clinical isolate (C6) using an ELISA as previously described by Rahbar and colleagues 14. Control antigen was prepared from uninfected fibroblasts.

Comparisons of nonparametrically distributed data in two independent groups or compartments were performed by the Mann–Whitney test. The Spearman test for correlation was used for analyses of covariation of two nonparametrically distributed data.

To investigate the presence of CD4⁺CD28^null T cells in the synovial membrane, we used three-colour immunofluorescence microscopy. This technique utilizes our findings of CD244 expression by CD4⁺ T cells as a specific marker for the CD28^null subset in peripheral blood of patients with RA and muscle tissue of patients with myositis (Figure 1) (Fasth et al., submitted).

CD244 is hitherto mainly described as a receptor regulating activation of NK cells after interaction with its ligand CD48 15. CD244 might have similar functions on T cells, but this is not yet fully understood 16. Instead of detecting T cells lacking CD28, CD4⁺CD28^null T cells in synovial membrane biopsies were identified by co-expression of CD244, CD3 and CD4 (Figure 1). Two patient groups were analysed: patients having less than 0.5% and patients having more than 7% of circulating CD4⁺CD28^null T cells (Table 1). In all 11 patients, as expected, CD3-positive cells were abundant in the synovial membrane, and a majority of the CD3-positive cells also expressed CD4 (Figure 1). Only small numbers of CD4⁺CD28^nullCD244⁺ T cells were found. They were evenly distributed and without correlation with the size of the T-cell infiltrates, the frequencies of CD4⁺CD28^null T cells in peripheral blood or with ongoing medication. From these results we conclude that the vast majority of CD4⁺CD28^null T cells do not home specifically to the synovial membrane despite significant frequencies in the circulation.

We then analysed the presence of CD4⁺CD28^null T cells in the synovial fluid by screening synovial fluid samples from 128 patients by flow cytometry. CD4⁺CD28^null T cells could be detected in synovial fluid at a median frequency of 1.9% (range 0–27%). These percentages, however, were significantly lower than in paired peripheral blood samples (median, 4.4%; range, 0–50%; P < 0.001) (Figure 2a). Despite significant differences in the two compartments, patients with the largest population of CD4⁺CD28^null T cells in the synovial fluid still tended to have the highest frequencies in peripheral blood (r = 0.47, P < 0.0001) (Figure 2b).

Analysis of 14 patients with two simultaneous synovial fluid samples demonstrated that the presence of CD4⁺CD28^null T cells in one inflamed joint reflects the occurrence also in other affected joints (Figure 2c). Patient groups undergoing different medical treatments (untreated, nonsteroidal anti-inflammatory drug, methotrexate or TNF blockade) did not have different frequencies of CD4⁺CD28^null T cells in the peripheral blood or the synovial fluid (data not shown).

Increased frequencies of circulating CD4⁺CD28^null T cells in patients with RA have been associated with HCMV infection; whether this is also valid for synovial CD4⁺CD28^null T cells is not known. Strikingly, we found significant CD4⁺CD28^null T-cell populations only in the synovial fluid of HCMV-seropositive individuals (P < 0.01) (Figure 2d). Comparative analysis of the frequencies of CD4⁺CD28^null T cells in the synovial fluid and peripheral blood of HCMV IgG-seropositive patients clearly showed that CD4⁺CD28^null T cells were less frequent in synovial fluid compared with peripheral blood also in this subset of patients (P < 0.0001) (Figure 2f). The serum titres of anti-HCMV IgG correlated with the levels in the synovial fluid (r = 0.87, P < 0.0001; data not shown). Seven patients (15%) displayed both anti-HCMV IgM and anti-HCMV IgG antibodies, indicating a recent or ongoing infection, but the frequency of CD4⁺CD28^null T cells in the circulation and synovial fluid was similar to IgG-seropositive patients lacking IgM (data not shown).

These data demonstrate that CD4⁺CD28^null T cells are only present in the synovial fluid and peripheral blood of RA patients seropositive for HCMV, and demonstrate that, despite significant frequencies in peripheral blood, CD4⁺CD28^null T cells are infrequent in the synovial compartment.

We have previously reported that CD4⁺CD28^null T cells from the peripheral blood rapidly respond to low TCR stimulation mimicked by a low concentration of anti-CD3 antibodies 6. To investigate whether CD4⁺CD28^null T cells in the synovial fluid displayed the same hyperresponsiveness, we measured the frequency of IFN-γ-secreting CD4⁺CD28^null T cells in mononuclear cells from synovial fluid. Indeed, these cells showed a similar response as CD4⁺CD28^null T cells from the peripheral blood (Figure 3, upper panel), indicating that CD4⁺CD28^null T cells in the synovial compartment have a capacity to function as local effector T cells.

With the strong association between CD4⁺CD28^null T cells and HCMV seropositivity we also investigated whether IFN-γ production could be provoked from CD4⁺CD28^null T cells in peripheral blood and synovial fluid, after stimulation by the HCMV-derived pp65 and immediate early antigens. Interestingly, synovial fluid-derived CD4⁺CD28^null T cells from all investigated patients showed specificity for these antigens (Figure 3, lower panel). These results indicate that HCMV-derived peptides can activate at least a subset of CD4⁺CD28^null T cells to function as effector T cells.

Hitherto we have shown that CD4⁺CD28^null T cells are less frequent in the synovial fluid compared with peripheral blood. We next investigated whether there was a selective recruitment of certain subsets of CD4⁺CD28^null T cells to the joint. Using antibodies detecting 24 different TCR-Vβ chains, we could identify TCR-Vβ subsets of the CD4⁺CD28^null T-cell population in both peripheral blood and synovial fluid. Two distinct patterns of TCR-Vβ distribution were found in the joint.

In six of the 13 patients analysed, the dominant TCR-Vβ within the CD4⁺CD28^null T-cell population in peripheral blood was also present in the joint. In one such representative patient, with total frequencies of 7%, 1% and 1.5% CD4⁺CD28^null T cells in the circulation, in the left knee synovial fluid and in the right knee synovial fluid respectively, 29–36% of the CD4⁺CD28^null T cells in all three compartments expressed TCR-Vβ13.1 (Figure 4a). In contrast, synovial fluid from five patients displayed a clear exclusion of the dominant CD4⁺CD28^null T cells from peripheral blood. In the most prominent case, a patient with total frequencies of 9% CD4⁺CD28^null T cells in peripheral blood and 4% in the synovial fluid, 71.5% of the CD4⁺CD28^null T cells in the circulation expressed Vβ20 but only 4.5% of the cells in the joint expressed Vβ20. In contrast, there was an enrichment of TCR-Vβ5.1-expressing CD4⁺CD28^null T cells in the joint (Figure 4b). Patients with the restricted pattern did not display lower overall frequencies of synovial fluid CD4⁺CD28^null T cells as compared with patients with the same TCR-Vβ dominance in peripheral blood and synovial fluid. Neither did different medical treatments co-segregate with either pattern of TCR-Vβ distribution (data not shown). Interestingly, in one patient displaying two TCR-Vβ subsets covering more than 75% of the circulating CD4⁺CD28^null population, one of the subsets had access to, and was even significantly enriched in, the synovial fluid, while the other dominant subset was clearly restricted (Figure 4c). Two patients did not display any of these distinct patterns of TCR-Vβ distribution.

In all 13 patients, the distribution of TCR-Vβ subsets in the conventional CD4⁺CD28⁺ T-cell population was not dramatically different in the synovial fluid from that in peripheral blood, as illustrated in Figure 4a–c. These results suggest that the access of CD4⁺CD28^null T cells to the joint is restricted to certain TCR-Vβ subsets, and that RA patients can be divided into two distinct groups with regard to different access of CD4⁺CD28^null T-cell subsets to the joint.

The presence of ACPA is RA specific and is an early predictor of an aggressive disease course 1718. We therefore analysed whether the frequencies of CD4⁺CD28^null T cells in the circulation and in the synovial fluid were associated with the presence of ACPA in these compartments. We found no differences in the frequencies of circulating or synovial CD4⁺CD28^null T cells in patients with or without ACPA in these locations (Figure 5a,b).

To investigate the association between CD4⁺CD28^null T cells and erosive disease, we allocated patients into two clearly distinguishable groups of either significant erosive disease (n = 21) or no signs of erosive disease (n = 16) according to radiographic evaluations. We had access to longitudinal follow-up data on patients without erosions up to 10 years after disease onset; therefore, patients allocated to the erosive group also had a disease duration of a maximum 10 years. Since subgroups of patients with different treatment did not display different frequencies of CD4⁺CD28^null T cells either in peripheral blood or in synovial fluid, treatment was not considered a parameter to stratify for in this analysis. The results indicated that CD4⁺CD28^null T cells were not specifically enriched in the joint or peripheral blood of patients with erosive disease (Figure 5c,d), which is in accordance with the lack of correlation with ACPA levels.

Increased frequencies of CD4⁺CD28^null T cells have been associated with extra-articular manifestations 34. Three cases of such manifestations were reported in our cohort. One patient, rheumatoid factor-positive and HLA-B27-positive, with iritis had no CD4⁺CD28^null T cells either in blood or the synovial fluid. In contrast, two patients with pleurisy had 31% and 33% CD4⁺CD28^null T cells in peripheral blood and 2.3% and 6.4% in the synovial fluid, respectively. Despite only two patients, these observations are in line with the notion of an association between CD4⁺CD28^null T cells and extra-articular manifestations in RA.

Taken together, our data from peripheral blood and the inflamed synovial joint show no association between the presence of CD4⁺CD28^null T cells and joint destruction.