Ethical approval for the project was obtained from the Leeds Teaching Hospitals National Health Service Trust Ethics Committee, and informed consent was obtained from each participant. Healthy control individuals were recruited from among local blood donors (n = 34). RA (n = 28) and osteoarthritis (OA; n = 12) patients were recruited through routine clinics at the Leeds General Infirmary (Table 1). They included patients with early, drug naïve (n = 7) and long-lasting, refractory (n = 21) RA (CRP range 5–155 mg/l) and patients with established, long-lasting OA (n = 12; CRP below detection range).

For the reconstitution studies we analyzed three RA patient cohorts (n = 31) and a cohort of non-RA patients with solid tumours (n = 7; Table 2). Each RA patient received high-dose cytotoxic therapy followed by autologous haematological transplants 263031. Each had disease that had proved resistant to multiple conventional antirheumatic drugs. Cohort 1 received an unmanipulated graft; cohort 2 received a graft that had undergone selection for CD34⁺ cells; and cohort 3 received a graft that had been CD34⁺ cell selected and T-cell depleted. The clinical progress of these patients was previously described elsewhere 263031. Control patients (Table 2) included five individuals with lung carcinoma, one with breast carcinoma and one with melanoma. They received unmanipulated autologous grafts following high-dose chemotherapy, as previously documented 32. For the IL-7 longitudinal studies, we analyzed four lymphoma and three sarcoma patients. All received intensive chemotherapy followed by reinfusion of unmanipulated autologous stem cells (Table 2). In addition, we studied three patients with systemic vasculitis who received the lymphocytotoxic monoclonal antibody CAMPATH 1H 33.

For our work on RA patients in clinical remission (Table 1), we recruited consecutive patients (n = 36) attending the rheumatology outpatient clinics with stable RA. They possessed no clinically significant synovitis and were deemed to be in 'remission' by the assessing consultant rheumatologist. Patients satisfied all of the following inclusion criteria: previous certified diagnosis of RA; over 18 years of age; disease duration of at least 12 months before remission; no disease flare within preceding 6 months; stable treatment within preceding 6 months; nil or minimal clinical evidence of active inflammatory disease and CRP below 15 mg/l within preceding 6 months; and no clinical indication to change treatment. We further refined this cohort by separating patients into those who satisfied the American College of Rheumatology (ACR) remission criteria and those who did not (Table 3).

IL-7, transforming growth factor-β₁, IL-6, tumour necrosis factor (TNF)-α and oncostatin M levels in sera and in tissue culture supernatants were measured using enzyme-linked immunosorbent assay (ELISA; R&D, Abingdon, UK), in accordance with the manufacturer's instructions. The sensitivities of the assay were <0.1 pg/ml for IL-7, 0.2 pg/ml for IL-6, 0.5 pg/ml for TNF-α, and 20 pg/ml for oncostatin M.

Peripheral blood mononuclear cells (PBMCs) were recovered as described previously 24, and CD4⁺ and CD8⁺ T cells were separated by negative selection (Metachem, Meylan, France). Purified CD4⁺ and CD8⁺ T cells (>92% pure for CD4⁺ and 89% pure for CD8⁺ T cells) were stained for CD45RB (FITC; Dako, Ely, UK), CD45RA (PE; Serotec, Oxford, UK), CD45RO (PE-CY5; Serotec) and CD62L (ECD Coulter, High Wycombe, UK) using conventional methods. Naïve T-cells were further sorted according to their CD45RB^bright, CD45RA⁺ and CD62L⁺ phenotype, using a FACS-Vantage cell sorter (Becton Dickinson, Oxford, UK). Memory cells and other subsets were identified based on their expression of CD45RB^bright/dull, CD45RA^±, CD45RO^bright/dull, and CD62L^± 24.

DNA was extracted from the different lymphocyte populations using standard proteinase K digestion followed by a phenol/chloroform extraction, either from total CD4⁺ and CD8⁺ populations after magnetic separation or from naïve cells after further cell sorting. TRECs were quantified using a real-time polymerase chain reaction based assay, as described previously 24. Briefly, TREC primers were F (d-CAC CTC TGG GCT ACG TGC TAG) and R (d-GAA CAC ATG CTG AGG TTT AAA GAG AAT); and glyceraldehyde-3-phosphate dehydrogenase primers were F (D-AAC AGC GAC ACC CAT CCT C) and R (d-CAT ACC AGG AAA TGA GCT TGA CAA). This analysis provided a final value that represented TREC DNA as a proportion of glyceraldehyde-3-phosphate dehydrogenase DNA, which is equivalent to the percentage of cells containing a TREC. Following the release of the entire T-cell receptor locus sequence late in 2002, we validated our assay utilizing an alternative set of TREC primers designed to minimize background signal when using PBMC DNA.

PBMCs were separated as above from 5 ml blood from RA patients and healthy control individuals. An aliquot of PBMCs was stained with a combination of CD127 (FITC; Serotec), CD19 (PE; Serotec) and CD4 or CD8 (PE-CY5; Serotec) to quantify IL-7 receptor expression on different cell types by flow cytometry. Cells were resuspended in RPMI 1640 supplemented with penicillin and streptomycin, glutamine and 10% human AB⁺ serum (Sigma, Aldwich, UK) and proliferation was assessed in response to PHA (10 μg/ml, Sigma), IL-2 (20 units/ml; Sigma), IL-7 (1–100 ng/ml; Sigma) or anti-CD3 antibody (OKT3; 1 μg/ml) with or without anti-CD28 antibody (YTH913.12; 5 μg/ml) co-coated on plastic Proliferation was quantified by incorporation of ³H-thymidine (1 μCi/well) after 5 days of culture.

Bone marrow mononuclear cells were obtained from posterior iliac crest aspirates from RA patients and healthy control individuals after informed consent had been obtained (with local research ethics committee approval), following centrifugation on Lymphoprep (Nycomed Pharma AS, Oslo, Norway), as previously described 3435. Aspirates from RA patients were repeated after 6–8 months of therapy with infliximab (Remicade; Schering Plough, Kenilworth, NJ, USA). Long-term bone marrow cultures from 10⁷ bone marrow mononuclear cells were grown, in accordance with standard techniques 3435. By allowing the formation of an adherent layer consisting mainly of macrophages and cells of mesenchymal origin, this culture system has been considered appropriate for evaluating the regulatory role played by the bone marrow microenvironment in haematopoiesis 36. At weekly intervals, cultures were fed by demi-depopulation. The adherent layer was usually confluent after 3–4 weeks, and at that time point cell-free supernatants were harvested and stored at -70°C for cytokine quantification.

Nonparametric tests were used throughout. The Mann–Whitney U-test for two independent samples was used to compare healthy control individuals with RA patients. The Spearman rank correlation coefficient was used to determine correlations between two variables. A Wilcoxon sign rank test was used to compare pretherapy and post-therapy outcomes.

We measured serum levels of IL-7 in a cross-section of active RA patients (n = 28), healthy control individuals (n = 34) and OA patients (n = 12). There was no correlation between serum levels of IL-7 and age in healthy control individuals 3738, and sex did not make any difference. Circulating IL-7 levels (Fig. 1a) were significantly lower in RA patients than in healthy control individuals (P < 0.00001). In RA there was no association between levels of circulating IL-7 and disease duration, inflammation as measured by CRP (Fig. 1b; nonsignificant correlation [R = 0.201, P = 0.161]), presence of a shared epitope (n = 17), or antirheumatic therapy (nonsteroidal anti-inflammatory drugs, methotrexate, or steroids). OA patients exhibited slightly lower IL-7 levels than did control individuals (P = 0.035) but they had significantly higher IL-7 levels than did RA patients (P < 0.00001). After Bonferroni correction there was no longer a significant difference between control individuals and OA patients, but other results remained unaffected. Regression analysis did not reveal any further trends.

There are several sources of IL-7 production, including stromal cells in the bone marrow, dendritic cells and epithelial cells in the thymus, skin and gut 6. We compared the ability of bone marrow stromal cells, derived from RA patients (n = 9) and healthy control individuals (n = 15), to produce IL-7 spontaneously in long-term cultures (Fig. 1c). The production of IL-7 was significantly lower in RA patients than in control individuals (P = 0.001). Furthermore, production did not consistently change after clinical remission induced by therapeutic TNF-α blockade (n = 8; P = 0.725). We also examined the PBMC response to IL-7 in RA patients and healthy control individuals. Whereas RA PBMCs responded suboptimally to IL-2, mitogen (PHA) or antigen (anti-CD3/CD28), as previously documented 39, their response to IL-7 was similar to that in control individuals (Fig. 1d). Importantly, in a cross-sectional comparison of 10 RA patients and 10 healthy control individuals, we could not find a significant difference in the number of cells expressing the IL-7 receptor (CD127) or in its level of expression (data not shown). Altogether, these findings suggest a deficit in circulating levels of IL-7 in RA, possibly due to an inability to produce IL-7, at least in stromal cells of bone marrow origin.

Patients receiving lymphocytotoxic therapy for conditions other than RA reconstitute more rapidly and completely than do RA patients. We previously studied three cohorts of RA patients who had received high-dose chemotherapy followed by stem cell reinfusion (Table 2). As previously reported 263031, CD4⁺ counts fell after treatment and subsequently remained low in all cohorts, with no significant differences due to graft manipulation (data not shown). In contrast, CD8⁺ T-cell counts initially rose before rapidly returning to basal levels.

In the present study we compared T-cell reconstitution in 12 RA patients (six from each of cohorts 2 and 3) and seven patients with solid tumours (Fig. 2 and Table 2). To avoid potential confounding effects of immunosuppressive drugs, RA patients were removed from the analysis if it subsequently became necessary to reinstitute antirheumatic therapies at times when disease activity resumed. The figure therefore represents 12 RA patients pretreatment and seven at 9 months.

The chemotherapy regimens differed between RA and non-RA patients, but the nadir lymphocyte counts were similar. Figure 2 illustrates the composition of the peripheral T-cell pool at baseline and at various times after treatment. The individual subsets were defined according to the lymphocyte differentiation pathway suggested by our previous work 24. The most naïve cells are represented in grey at the top of each bar chart. These cells progress to conventional memory cells and their precursors (striped bars) via post-naïve cells (white bars). Presumed 'central' memory cells are presented in black. Notably, at baseline RA patients possessed no CD4⁺ and CD8⁺ central memory cells in peripheral blood, as reported previously 24. After chemotherapy there was simultaneous accumulation of all subsets in cancer patients, resulting in rapid restitution of CD4⁺ T-cell counts within 3 months. The same was true of the CD8⁺ subsets except that there was an 'overshoot' of memory CD8⁺ T-cells. In contrast, there was no early expansion of any T-cell subset in RA, although some long-term restoration of naïve CD4⁺ subsets was observed. Naïve CD8⁺ T-cells also accumulated slowly, and there was a brief expansion of CD8⁺ memory cells. These marked differences between RA and cancer patients demonstrate that a limited early peripheral expansion after treatment may account, in part, for the lack of T-cell reconstitution in RA.

Although graft manipulation differed between cancer patients (un-manipulated) and RA (CD34 selected ± T-cell depletion) as mentioned above, we found that graft manipulation did not affect the rate of reconstitution in RA patients (data not shown). Other factors that differ between the RA and control group reflect the underlying disease. For example, RA patients may have been exposed to low-dose corticosteroid therapy as part of their prior treatment. It is not possible to exclude an effect of such a factor on our data.

In order to compare thymic activity after lymphodepletion, we measured TRECs longitudinally in CD4⁺ T-cells in the same RA and cancer patient cohorts. As a molecular marker of T-cell receptor rearrangement, TRECs provide a surrogate measure of recent thymic activity 4041. We measured TREC content in total CD4⁺-T cells as well as in naïve CD4⁺-T cells in isolation. Patterns of TREC variation were consistent between the seven cancer patients. TRECs rapidly accumulated after treatment but returned to baseline by 3 months (Fig. 3; open diamonds). Our data in cancer are therefore consistent with an early surge in thymic activity, followed by a slow return to baseline at a time when the T-cell counts have returned to baseline levels. Variation in the TREC content of naïve cells also followed that pattern. The reduction in TREC content of an individual subset such as naïve cells is better explained by proliferation within that subset 244243, therefore suggesting that naïve T-cells underwent peripheral expansion, resulting in TREC dilution in CD4⁺ cells (open diamonds). Similar results were observed for CD8⁺ T cells (data not shown).

The early thymic response to lymphopenia did not occur in the 12 RA patients. In contrast, the TREC content of total CD4⁺ T-cells climbed gradually for several months after treatment (Fig. 3; closed diamonds). The TREC measurements in naïve cells also did not return to baseline, however, suggesting a lack of proliferation of CD4⁺ naïve cells. Therefore, a delay in achieving good release of newly developed T-cells also appeared to contribute to slow T-cell reconstitution in RA after high-dose chemotherapy. Similar results were observed for CD8⁺ T-cells (data not shown).

Figures 2 and 3 suggest that the development and expansion of CD4⁺ T-cells were compromised in lymphopenic RA patients. Both the development and expansion of T cells have been extensively documented in relation to IL-7 (for review 6). The relative deficiency in circulating IL-7 levels in RA patients identified in Fig. 1 therefore suggests a significant role for IL-7 in impaired T-cell reconstitution following high-dose chemotherapy. We measured serum levels of IL-7 longitudinally in four RA patients after lymphodepleting therapy (cohort 3, without relapse within 12 months) and seven non-RA patients (Table 2). Figure 4 clearly demonstrates a four- to fivefold rise and subsequent decrease in IL-7 levels, coincident with short-term lymphopenia in non-RA patients (triangles). In marked contrast, IL-7 levels did not change significantly in four RA patients over 12 months of follow up (squares).

In RA patients whose disease was controlled by in vivo TNF blockade, spontaneous release of IL-7 from bone marrow derived stromal cell cultures was variable, remaining reduced in some patients but returning to normal in others (Fig. 1). We therefore decided to investigate IL-7 levels in patients with well controlled disease and minimal levels of disease activity for at least 6 months before recruitment (n = 36; Table 1). Levels of IL-7 were heterogeneous and ranged from 2.47 to 16.25 pg/ml. No clinical parameter was significantly correlated with IL-7 levels (disease duration, remission duration, previous or current therapy, rheumatoid factor).

We measured TREC in total CD4⁺ T-cells in these patients in clinical remission in relation to age. The results were also heterogeneous (Fig. 5a; all triangles). Comparing these values with our previous results in healthy control individuals (small circles 24), there appeared to be two distinct patient groups. One of these groups had a CD4⁺ T-cell TREC content similar to or higher than that in age-matched healthy control individuals, and the other group exhibited lower TREC content. We used the median TREC content to distinguish two groups. Open and closed triangles relate to group 1 (above the median TREC value) and group 2 (below the median TREC value), respectively. The relationship between TREC content and age was present in group 1 (thick line; R = -0.738, P = 0.001) but not in group 2. No clinical parameter was able to predict TREC content (disease duration, remission duration, previous or current therapy, rheumatoid factor).

We reanalyzed the IL-7 data with respect to this dichotomy in TREC levels, and there was a significant difference in circulating levels of IL-7 between these two groups (group 1, n = 17: IL-7 12.71 ± 2.76 pg/ml, range 9.57–16.25 pg/ml; group 2, n = 19: IL-7 6.50 ± 1.88 pg/ml, range 2.47–9.30 pg/ml; P < 0.00001). Furthermore, there was a positive correlation between the levels of circulating IL-7 and the TREC content of total CD4⁺ T cells (Fig. 5b; n = 36, all diamonds; R = 0.777, P < 0.00001). No similar relationship was observed in healthy control individuals (n = 12; R = 0.219, P = 0.595).

We subsequently reanalyzed the data according to the ACR criteria for clinical remission 4445. Patients fulfilling or not fulfilling the ACR criteria (Table 3) are shown as open and black diamonds, respectively, in Fig. 5b. The two populations were undistinguishable in terms of TREC content (P = 0.807). There was no difference in their circulating levels of IL-7 (ACR positive: 9.07 ± 3.33 pg/ml, range 4.9–15.23 pg/ml; ACR negative: 9.31 ± 4.01 pg/ml, range 2.47–16.25 pg/ml; P = 0.838). Furthermore, the correlation between IL-7 and TREC content was maintained in both groups (ACR positive, n = 17: R = 0.680, P = 0.005; ACR negative, n = 19: R = 0.779, P = 0.001). These data suggest that, removing any influence of systemic inflammation, RA patients form two groups that are characterized by normal or low levels of thymic activity and IL-7. It is not possible to predict from these data whether these abnormalities are primary or, indeed, whether they have pathogenic significance. However, they may be important in the context of reconstitution capacity after lymphodepleting therapies.