Class-switched B cells display response to therapeutic B-cell depletion in rheumatoid arthritis

Thirty-five patients with RA according to the American College of Rheumatology classification criteria [29] were included in this prospective observational study. All patients were 'RTX-naïve'. They had active disease according to a 28-joint disease activity score (DAS28) of greater than 3.2, which would qualify them for repetitive RTX treatment [17]. All patients had failed to at least one disease-modifying antirheumatic drug (DMARD) and had shown inappropriate response to at least one tumour necrosis factor (TNF)-blocking agent. Disease activity was reflected by a median of 6 swollen (interquartile range [IQR] 3 to 10) and a median of 5 tender (IQR 2 to 9) of 28 evaluated joints. Median DAS28 was 5.0 (IQR 4.3 to 5.9), median erythrocyte sedimentation rate (ESR) was 33 mm (IQR 27 to 46), and C-reactive protein (CRP) serum concentration was 11 mg/L (IQR 3 to 24). Other patient characteristics, including the number of previously used DMARDs and anti-TNF agents, are summarized in Table 1. Assessors for clinical parameters were blinded to the time-matched laboratory results. All patients gave their written informed consent to participate. The study was approved by the Cantonal Ethics Committee of Bern (ref. no. 254/07).

Clinical improvement was assessed every 3 months by DAS28 and graded as European League Against Rheumatism (EULAR) good, moderate, or non-response [30]. EULAR good response, which could have been achieved at any visit during the 12-month observation period, was used for group definition in retrospective B-cell and antibody analyses.

Freshly isolated PB (9 mL) was anticoagulated with EDTA (ethylenediaminetetraacetic acid) and immediately used for flow cytometry. B-cell analyses were also performed in anticoagulated synovial fluids (n = 3) or in tissue homogenates from another three patients undergoing urgent joint replacement surgery of the knee (n = 1), synovectomy in treatment-resistant synovitis of the knee (n = 1), or wrist joint synovectomy (n = 1). The study protocol for invasive procedures was approved by the Cantonal Ethics Committee of Bern (254/07). All study participants gave their written informed consent to participate.

Synovial tissue was immediately prepared, as previously described [31], by injecting 1 mg/mL collagenase (Sigma-Aldrich, Munich, Germany) into the tissue samples, followed by incubation for 20 minutes at ambient temperature. Digests subsequently were minced and incubated for an additional 50 minutes at 37°C in collagenase 1%. The cell suspension was strained by 70-μm nylon filters (Falcon, now part of BD Biosciences, San Jose, CA, USA), washed twice in phosphate-buffered saline, and recovered in RPMI 1640 medium (Invitrogen, Karlsruhe, Germany) containing 10% fetal calf serum plus kanamycin at 37°C in 5% CO₂ atmosphere overnight. Non-adherent synovial tissue cells were carefully obtained together with the supernatants, centrifuged, and thoroughly washed with phosphate-buffered saline before subsequent analyses.

Fixation of leukocytes and lysis of erythrocytes for flow cytometry were done for quantitative analyses in TruCOUNT™ tubes (Becton Dickinson, Basel, Switzerland). Cells were stained with BD Multitest™ reagent for CD3/CD16 + CD56/CD45/CD19 markers as well as phycoerythrin-conjugated anti-CD27 (clone L128) and fluorescein isothiocyanate-conjugated anti-IgD (clone IA6-2). All antibodies were purchased from BD Biosciences. Data were acquired by flow cytometry using the BD FACSCalibur Flow Cytometry System and CellQuest software (BD Biosciences Immunocytometry Systems). Analyses were performed after gating on anti-CD45 stained lymphocytes. The number and frequency of CD19⁺ PB B cells and subsets (percentage of CD19⁺ B cells) were determined in a minimum of 20,000 events in the CD45 gate. Data collection was continued to 1 × 10⁵ events in case there were fewer than 0.5% CD19⁺ cells. The number and frequency of CD27^-IgD⁺ 'naïve', CD27⁺IgD⁺ non-switched MemB, and CD27⁺IgD^- switched MemB were determined according to their IgD and CD27 surface expression. All measurements were performed under the standard operating procedure guidelines of our accredited laboratory for flow cytometry.

IgG, IgA, and IgM serum concentrations as well as serum RF were determined by nephelometry (Dade Behring Nephelometer; Dade Behring, Marburg, Germany). IgG antibodies against cyclic citrullinated peptides (CCPs) were also determined with commercially available methods (INOVA QUANTA Lite™ CCP3 ELISA; INOVA Diagnostics, Inc., San Diego, CA, USA).

Results are presented as median and 25% to 75% IQR for data in non-Gaussian distribution. The Wilcoxon rank sum test was used for comparison of two-tailed groups. The exact significance in the Mann-Whitney U test was used for comparison of two independent groups. Statistical analyses were performed with SPSS software version 15 (SPSS Inc., Chicago, IL, USA). Results with a P value of less than 0.05 were considered significant.

Ten patients did not achieve a significant clinical improvement during the observation period. Six patients fulfilled the EULAR criteria for moderate clinical response, and another 19 for good clinical response. Best achieved clinical response, by definition, could have occurred at any time during the 12-month observation phase after RTX. Ten of our good-response patients fulfilled this EULAR definition for the first time after 3 months, another 7 patients after 6 months, and 2 patients only 9 months after RTX. Duration of good response was documented for a median of 3 months but in fact may have been substantially longer with regard to the 3-month visit intervals. The criterion of moderate response was documented continuously in these same 19 patients for a median of 9 months. Patients experiencing good response had significantly higher anti-CCP antibody titer (P = 0.046) at baseline but significantly lower CRP serum concentration (P = 0.010). They had significantly shorter disease duration (P = 0.008) and fewer DMARD treatment attempts (P = 0.010) but were comparable for ESR, RF, swollen and tender joint counts, corticosteroid dose, and DAS28 at baseline. Median CRP in the entire study population dropped significantly from 11 mg/L (IQR 3 to 24) at baseline to values below the detection limit of 3 mg/L, with an IQR of between less than 3 and 9 mg/L at month 3 (P = 0.019), less than 3 to 6 mg/L at month 6 (P < 0.001), and less than 3 to 7 mg/L at month 9 (P = 0.008). Good responders had significantly lower CRP serum concentrations than the other patients 6 months after RTX. ESR improved significantly from a median of 33 mm/hour (IQR 26 to 40) at baseline to a median of 11 mm/hour (IQR 7 to 25, P = 0.011) at month 3 and a median of 11 mm/hour (IQR 5 to 20, P = 0.006) at month 6. Good responders had significantly lower ESR than the other patients after 3, 6, and 12 months. After 12 months, median CRP concentrations as well as ESR were in the same range as baseline values.

Median total number of CD19⁺ B cells (Figure 1) in the entire cohort was 169/μL (IQR 77 to 281); this parameter did not significantly differ between good responders and patients with moderate or no response in the later course. A median of 69% of B cells (IQR 55% to 79%) at baseline were 'naïve' B cells. Their absolute number was significantly higher (P = 0.034) in subsequent non-responders or moderate responders (median 199, IQR 137 to 297) than in good responders (median 109, IQR 56 to 142). In contrast, the frequencies of non-switched MemB (median 8%, IQR 4% to 13%), of switched MemB (median 16%, IQR 12% to 22%), and of CD27^-IgD^- B cells (median 4%, IQR 3% to 7%), as well as their absolute numbers, were not significantly different between the two patient groups.

The first RTX infusion reduced the number (Figure 1) and frequency of PB CD19⁺ B cells to a median of 3/μL (IQR 2 to 7) and a median of 2% (IQR 1% to 4%) of CD45⁺ cells, respectively. The median of all B-cell fractions decreased significantly in number (P < 0.001), but increasing frequency of switched MemB and of CD27^-IgD^- B cells upon the first RTX infusion indicated relative resistance of these subsets. When a B-cell frequency of below 0.5% of lymphocytes [6] or the number of CD19⁺ cells below 5/μL was used as an arbitrary surrogate of complete depletion, these values were reached in 61.3% and 48.6% of the cases, respectively, 14 days after the first infusion. Neither the achievement of one of these limits upon first infusion nor the absolute number or frequency of any defined B-cell subset in the early depletion phase was indicative for subsequent clinical response.

The median number of B cells increased continuously from months 3 to 12. This PB B-cell repletion was dominated by 'naïve' B cells. The time point of repopulation tended to be earlier in moderate or non-responders than in good responders; however, this difference was statistically not significant. In contrast, the absolute number of switched MemB at month 6 (P = 0.049), month 9 (P = 0.045), and month 12 (P = 0.003) was significantly higher in patients with no or moderate time-matched clinical response to RTX than in good responders (Figure 1). This result was the same when correlating switched MemB with best achieved response at any time of the treatment cycle. In contrast, the number and frequency of non-switched MemB, of CD27^-IgD^-, and of 'naïve' B cells were not correlated with the clinical response after RTX.

The B-cell composition in six simultaneously collected samples from the PB and synovium during B-cell repletion (Figure 2) gave further evidence for the importance of switched MemB: While 'naïve' B cells dominated quantitatively in PB during B-cell repletion, only the frequency of switched MemB was significantly higher in the synovium (P = 0.028) than in PB. A median of 50% (IQR 18% to 93%) of PB B cells, but only 8% (IQR 3% to 12%) of synovial B cells, were of 'naïve' phenotype, whereas 10% (IQR 10% to 41%) of PB B cells as compared with 67% (IQR 43% to 76%) of synovial B cells exhibited the CD27⁺IgD^- phenotype. (For histology, see Additional data file 2.)

Therapeutic intervention with RTX led to significant decreases in RF and CCP antibody serum concentrations from months 3 to 12 (P < 0.05) (Figure 3). In patients who were RF⁺ at baseline, the lowest RF concentration (median of 36% from baseline) was achieved after 6 months. Five of the 21 initially RF⁺ patients became RF^- upon B-cell depletion, three of them lasting until the end of the observation whereas two others started to have positive RF tests again after 6 months. All patients with persistent or transient RF conversion were good clinical responders. A decrease of RF serum concentrations in comparison with baseline levels was significantly stronger in good responders than in the other patients. These results were similar for month 3 (P = 0.022), month 6 (P = 0.002), month 9 (P = 0.002), and after 12 months (P = 0.001). A significant reduction in median RF concentrations was long-lasting in good responders but was limited to a maximum of 6 months in moderate and non-responders. RF serum concentrations dropped more steeply than expected from the kinetics of slowly reduced IgG, IgA, and IgM total Ig concentrations.

In contrast to RF, CCP antibody concentrations decreased more continuously and in parallel to Ig levels. They reached their minimum of a median of 60% from baseline levels after 12 months. Good responders started with a tendency toward higher anti-CCP serum concentrations than the other patients but showed a significantly stronger reduction of CCP-directed autoantibodies after 9 months (P = 0.041) and 12 months (P = 0.027).

Total IgG serum concentrations started to be significantly reduced (P < 0.05) already after the first infusion, whereas IgA and IgM serum concentrations were more stable. These isotypes came down somewhat later, with significantly reduced levels from months 3 to 12 (P < 0.001) when compared with baseline. Three different patients marginally underwent the lower limit for normal IgG (7 g/L) or IgA (0.7 g/L) serum concentrations during the observation phase. Two other patients developed IgM concentrations below the lower normal limit of 0.4 g/L. None of the patients with a drop of any Ig isotype below the lower normal limit developed clinical symptoms of immunodeficiency, but seven of these patients experienced good clinical response and one patient experienced moderate clinical response.

After 12 months, the IgG and IgA serum concentrations decreased to medians of 86% and 79% and IgM serum concentrations decreased to values between 75% and 68% from baseline. Median IgG, IgA, and IgM serum concentrations were somewhat lower in good responders at any follow-up visit, but at no time point were the relative decreases of these concentrations from baseline significantly different between good responders and the other patients.