Sixty-one patients with active RA (Disease Activity Score in 28 joints (DAS28) >5.1), who fulfilled the revised classification criteria of the American College of Rheumatology for RA 20, were evaluated before and after infliximab therapy. Table 1 summarizes the characteristics of these patients.

Infliximab (Shering-Plough, Levallois-Perret, France) was given at a dose of 3 mg/kg intravenously at weeks 0, 2 and 6 and then every 8 weeks in combination with stable doses of methotrexate 7.5 to 15 mg/week orally or intramuscularly. Only patients on stable prednisone doses ≤ 10 mg/day and nonsteroidal anti-inflammatory drug treatment were included. According to EULAR response criteria 21, a positive clinical response to infliximab therapy was defined as a drop in the DAS28 from baseline by >1.2 or as a DAS28 <3.2 at week 14.

In addition, 30 healthy blood donors were included in the study. These donors were matched with patients for sex and age. Synovial fluid was obtained from 11 patients suffering from osteoarthritis.

The study was approved by the local Ethics Committee, and all patients gave informed consent.

Whole blood samples were analyzed on a FACSCalibur flow cytometer (BD Biosciences, Pont-de-Claix, France) with 10⁶ white blood cells acquired per analysis. DC subsets were measured using a DC kit from BD Biosciences. Peripheral blood mDC and pDC subsets were defined by the concomitant lack of lineage markers, HLA-DR expression, and mutually exclusive membrane expression of CD11c or CD123, respectively. Absolute numbers of blood DC precursors were calculated as the percentage of white blood cells expressed per milliliter of peripheral blood. Enumeration of blood DC was evaluated as published elsewhere 22. Plasma cells were analyzed by gating on CD19⁺ cells and by calculating the percentage of CD20^neg/CD38^high cells.

Synovial fluid was obtained at the initial time point from patients with RA (n = 9) and from patients with osteoarthritis (n = 11), with knee effusions. This synovial fluid was diluted appropriately with PBS in order to avoid clot formation. Synovial mDC and pDC subsets were defined by the concomitant lack of lineage markers (CD3^-, CD14^-, CD16^-, CD56^-, CD8^- and CD19^-), HLA-DR expression, and mutually exclusive membrane expression of CD11c or CD123, respectively. Results were expressed as the percentage of mDCs or pDCs among cells without the following lineage markers: CD3, CD14, CD16, CD56, CD8 and CD19.

Serum samples were collected and were stored at -80°C. IFNα levels were quantified with a human IFNα ELISA kit (BioSource International, Camarillo, CA, USA), according to the manufacturer's instructions. The detection limit of this IFNα ELISA is 25 pg/ml. This assay has been used previously by others groups for measurement of IFNα in the serum 323.

Peripheral blood mononuclear cells (PBMCs) of adult donors were isolated using Ficoll-Paque Plus (Amersham Biosciences, Saclay, France) gradient centrifugation. PBMCs (1 × 10⁶ cells/well) were cultured in RPMI supplemented with 10% FCS, and were stimulated in vitro with live influenza virus (10⁴ particles; Charles River Laboratories, Wilmington, MA, USA) with or without TNFα (10 μg/ml; R&D Systems, Lille, France) or TNFα blockers (Infliximab 20 μg/ml; Shering-Plough) in 96-well U-bottom plates. The infliximab dose used in vitro is comparable with the infliximab serum concentration found in vivo during the first weeks after the infusion 24. After 24 hours incubation, supernatants were collected. Depending on the conditions, cells were further incubated in fresh RPMI with live influenza virus (10⁴ particles; Charles River Laboratories). After 24 hours, the supernatants were again collected for IFNα quantification by ELISA.

PBMCs were isolated by Ficoll-Paque Plus (Amersham Biosciences, Saclay, France) gradient centrifugation – from RA patients treated by infliximab who had developed significant ANA titers, from healthy donors and from SLE patients. PBMCs (1 × 10⁶/well) were then cultured with 10⁴ influenza virus particles (Charles Rivers, Wilmington, MA, USA) with or without TNFα (10 μg/ml; R&D Systems) or TNFα blockers (Infliximab 20 μg/ml; Schering-Plough, Levallois-Perret, France) in a 48-well plate in 10% FCS RPMI supplemented with rhIL-2 (50 U/ml; R&D Systems, Lille, France). At day 15, supernatants were collected and tested for ANAs. The resulting B cells were analyzed using flow cytometry after gating on CD19⁺ cells and by calculating the percentage of CD20^low/CD38^high cells.

Statistical analysis was performed using the GraphPad InStat software (version 3.0a for Macintosh; GraphPad Software, San Diego, CA, USA). Mann–Whitney tests were used for mean comparisons between groups. Wilcoxon's signed-rank test was used for the analyses of matched pairs. Correlation between DCs and activity markers were assessed using linear regression, given with the r² correlation coefficient. P < 0.05 was considered statistically significant.

To better delineate the involvement of known DC subsets in RA pathogenesis, we compared the number of circulating CD11c⁺HLA-DR⁺CD123^- mDCs and CD11c^-HLA-DR⁺CD123⁺ pDCs in peripheral blood from 61 active RA patients (free of TNFα-blocker treatment) and from 30 healthy volunteers. Interestingly, RA peripheral blood was characterized by a decreased number of both pDC and mDC subsets (mean ± standard deviation): mDC count = 10,214 ± 7,576 cells/ml in the RA group versus 16,228 ± 4,057 cells/ml in the healthy control group (P = 0.0002), and pDC count = 6,098 ± 4,710 cells/ml in the RA group versus 10,313 ± 4,201 cells/ml in the healthy control group (P < 0.0001) (Figure 1). We concluded that RA patients are characterized by a quantitative deficit in their peripheral circulating DCs.

We then looked for a correlation between absolute counts of blood DCs and the clinical status or laboratory tests known to reflect disease activity (DAS28, Health Assessment Questionnaire score, and C-reactive protein level). In RA patients, mDC counts were inversely correlated with each of these markers (P < 0.05, r² = 0.07, P < 0.02, r² = 0.11 and P < 0.05, r² = 0.11, respectively, for DAS28, Health Assessment Questionnaire score and C-reactive protein level). We did not find any statistical correlation between the pDC counts and DAS28, Health Assessment Questionnaire score or C-reactive protein level (Figure 2a,b,c).

The levels of both DC subsets are therefore decreased in the blood of RA patients with active disease, but only mDCs correlate inversely with disease activity – suggesting that this mDC decrease could reflect a migration to inflamed tissues. Accordingly, we found a higher percentage of mDCs in synovial fluid from active RA patients compared with that from patients with osteoarthritis (percentage ± standard deviation: mDC = 52.5 ± 13.7% in the RA group vs. 17.4 ± 18.3% in the osteoarthritis control group; P = 0.0005). In contrast, the percentage of pDCs in synovial fluid was not different between the RA and the osteoarthritis groups (percentage ± standard deviation: pDC = 8.4 ± 10.9% in the RA group vs. 2 ± 3.9% in the osteoarthritis control group, P = 0.1119) (Figure 2d). The preferential migration of mDCs to inflamed joints was also suggested by the increase of the mDC:pDC ratio in synovial fluid compared with that found in peripheral blood (median, 3.8:1; P < 0.01, Wilcoxon matched-pairs test) (Figure 2e).

Our initial results suggest that mDCs migrate from the blood to the inflamed synovial compartment. If this is the case, it seemed likely that effective therapy might block this migration and increase the blood mDC level.

Responders to the infliximab regimen (n = 46) were defined by a DAS28 decrease >1.2 after 14 weeks of infliximab therapy, whereas nonresponders (n = 13) were patients defined by a DAS28 variation <1.2 at week 14. Responders showed a substantial increase in their numbers of circulating mDCs (mean ± standard deviation = 11,915 ± 8,630 cells/ml at day 0 vs. 15,868 ± 11,467 cells/ml at week 14, P < 0.05 using Wilcoxon matched-pairs test) (Figure 3a), whereas the blood pDC level did not change significantly (5,632 ± 3,035 cells/ml at day 0 vs. 6,555 ± 4,656 cells/ml at week 14, P = 0.23) (Figure 3b). In contrast, nonresponders did not show statistically significant changes in mDC and pDC counts, and some patients even showing a decrease in both DC subsets during the course of treatment (mean ± standard deviation: mDCs = 7,991 ± 4,275 cells/ml at day 0 vs. 8,386 ± 3,689 cells/ml at week 14, P = 0.41; and pDCs= 5,542 ± 3,525 cells/ml at day 0 vs. 4,649 ± 2,032 cells/ml at week 14, P = 0.27) (Figure 3a,b). These data suggest the existence of a relationship between the fluctuations of the mDCs present in the blood and the variations of disease activity.

The development of ANA is one of the most common side effects of TNFα-blocker therapies 2526. We therefore looked for a correlation between ANA appearance and DC count evolution in RA patients treated with infliximab.

The ANA levels were determined at the same time as the peripheral DC levels. After 14 weeks of treatment, we separated infliximab-treated patients into two groups: patients with positive ANA (n = 30) and patients with negative ANA (n = 16). The ANA level was considered positive when the serum dilution giving a positive signal in the indirect immunofluorescence on Hep-2 cells was above 1:250 and negative at the beginning of the treatment, or if the dilution increment reached at least three times the dilution observed at treatment onset. All of the data were obtained on day 0 of treatment onset and at week 14.

At week 14, the pDC levels were statistically lower in the ANA-positive group when compared with the ANA-negative group (mean ± standard deviation: circulating pDCs = 5,509 ± 3,161 cells/ml vs. 9,324 ± 5,834 cells/ml, P < 0.01) (Figure 4a). Although no statistically significant difference was found in the mDC subset between the two groups (data not shown), the decrease of peripheral pDC counts correlated with the increase of ANA titers (P = 0.02, r² = 0.15) (Figure 4b).

Because IFNα and pDCs have been implicated in autoantibody production in SLE pathogenesis 3, we measured the IFNα level in the blood of both ANA-positive and ANA-negative RA patients treated by infliximab. We found that RA patients developing ANA were characterized by higher levels of IFNα (310 pg/ml vs. 47 pg/ml, P < 0.01), suggesting that infliximab influences pDC homeostasis and promotes the production of ANAs through the secretion of IFNα (Figure 4c).

The presence of higher amounts of IFNα in RA ANA-positive patients prompted us to analyze the effects of infliximab on pDCs' ability to secrete IFNα in vitro. PBMCs from control donors were exposed to influenza virus alone or in the presence of infliximab. Influenza virus was used as a well-known strong pDC-IFNα inducer. We did not find any increase in cellular apoptosis of the cells in any of the conditions tested (data not shown). In both conditions (virus alone or virus + infliximab), we detected high levels of IFNα in the supernatant collected after 24 hours culture, without any differences between the two conditions (Figure 5). Repeat exposure of PMBCs to influenza virus, however, was able to induce large IFNα production only in the presence of infliximab. Furthermore, PBMCs pretreated with TNFα were unable to secrete significant amounts of IFNα. Although these studies were performed with PBMCs, it is probable that pDCs were the major source of IFNα given that they are the major IFNα-producing cells in peripheral blood. These data strongly suggest that infliximab maintains pDCs in an IFNα secreting state by quenching TNFα.

Jego and colleagues showed that pDCs exposed to viral infection were able to activate the B-lymphocyte compartment and to promote the generation of plasma cells and/or plasmablasts in an IFNα-dependent and IL-6-dependent fashion 4. To delineate the consequences of the sustained IFNα secretion induced by infliximab, we compared the proportion of circulating CD19⁺CD20^-CD38⁺ plasma cells in RA ANA-positive patients (n = 10) and RA ANA-negative patients (n = 10). RA ANA-positive patients exhibited a significant increase (P < 0.001) of the proportion of circulating plasma cells compared with RA ANA-negative patients (Figure 6a). The percentage of plasma cells in RA ANA-positive patients was similar to that observed in SLE patients.

We then tested, in vitro, whether infliximab effects plasma cell generation from B lymphocytes. PBMCs were cultured with influenza virus with or without infliximab. After 15 days, we measured the proportion of CD19⁺CD20^-CD38^high+ plasma cells. PBMCs cultured with influenza virus + infliximab were characterized by a higher proportion of plasma cells. Interestingly, concomitant addition of TNFα with virus stimulation inhibited plasma cell generation (Figure 6b). To analyze the in vitro effects of infliximab on ANA secretion, we repeated the experiment with PBMCs from RA ANA-positive patients, SLE patients and healthy control individuals. These PBMCs were cultured for 15 days in the presence of influenza virus with or without infliximab. After 15 days the secretion of ANA was found only in supernatants from cells from RA ANA-positive patients or SLE patients, and was further increased in the presence of infliximab (Figure 6c).

Taken together, those results suggest that infliximab promotes pDCs in an IFNα secreting state and allows for the differentiation of B lymphocytes into ANA-secreting plasma cells.