This observational prospective single cohort study was conducted at the expert center for Lupus-, Vasculitis-, and Complement-mediated systemic autoimmune diseases (LuVaCs) of the Leiden University Medical Center (LUMC) in the Netherlands. In this study, AAV patients treated with RTX were eligible and informed consent was required for study participation. The study was approved by the local medical ethics committee of the LUMC. Eleven unique AAV patients that received RTX were included in this study (Table 1). Seven patients received RTX as remission-induction treatment for active disease, of which 6 were included for flow cytometry studies which are shown in Figures 2–4. Additionally, four other patients and three patients from the previous group received up to 4 times maintenance treatment with 500 mg RTX every 6 months (Supplementary Tables 1 and 2), which were allowed to re-enter the study (Supplementary Figure 1). There was a total of 17 RTX maintenance treatments, of which 8 were included for flow cytometry studies (Supplementary Table 1). The flow cytometry data of these RTX maintenance patients were shown in the Supplementary Figures 5 and 7. Regarding the PBMC culture experiments, all available PBMC samples at week 0, 24, and 48 weeks after all RTX treatments in all patients were included, except one ANCA-negative patient (n = 23).

Treatment with RTX as remission-induction therapy was given as two times 1,000 mg RTX i.v. with a 2 week interval or as 500 mg RTX every 6 months as maintenance therapy. Directly before every RTX infusion, all patients received oral acetaminophen, clemastine i.v. and 100 mg methylprednisolone i.v. Additionally, patients were allowed to receive other immunosuppressive medication according to standardized local protocol including up to three times 1,000 mg methylprednisolone i.v. daily, followed by high dose oral corticosteroids (1 mg/kg/day, maximum 60 mg) with tapering over 3 months. Additionally, patients with severe renal and/or lung involvement were allowed to receive 500 mg cyclophosphamide i.v. and/or plasmapheresis. All patients also received prophylactic treatment with co-trimoxazole 480mg/day, proton-pump inhibition, vitamin D, calcium supplementation and, if indicated, bisphosphonates.

Clinical and laboratory data were collected at study visits just before RTX infusion and 2, 4, 12, 24, and 48 weeks after RTX treatment. Serum and PBMCs were stored for experimental studies at each study visit. Birmingham Vasculitis Activity Score (BVAS-3) was used to score disease activity during the study (27). The clinical diagnostics lab measured total immunoglobulin (Ig) levels in the sera of the patients. Standardized measurements of circulating CD19⁺ B-cells by routine flow cytometry protocols were performed at the laboratory of Hematology, with a detection limit of 1*10⁶ cells/L for CD19⁺ B-cells, further referred to as low-sensitive flow cytometry (LSFC).

Peripheral blood was collected in EDTA-coated tubes at each visit. Samples were processed within 4 h after collection and analyzed by flow cytometry after bulk-lyse standard operating procedure (www.EuroFlow.org), as described before (25, 26). According to these protocols, the membranes of 20*10⁶ nucleated cells per patient sample were stained with the EuroFlow 13-color IgH-isotype B-cell tube. Per sample, 10–20*10⁶ leucocytes were measured in LSR Fortessa X-20 flow cytometer (Becton Dickinson Biosciences, San Jose, Calif). Instrument set-up and calibration were performed according to EuroFlow standard operating procedures www.EuroFlow.org (28). For data analysis, Infinicyt software version 2.0.1 (Cytognos S.L., Salamanca, Spain). was used.

Gating strategy for the identification of different major B-cell and plasma cell subsets was shown in previous publications (25, 26). B-cells were identified based on their positive CD19 expression and low forward (FSC) and sideward scatter (SSC) properties, which is typical for lymphocytes. PCs were identified by high expression for CD38 and CD27, with low expression of CD24 and CD21 in conjunction with low-to-intermediate forward light scatter and sideward light scatter. Both switched memory and PCs were subsequently subclassified based on their maturation stage and expression of distinct Ig isotypes and Ig subclasses, as previously described (25, 26). This strategy resulted in the following definitions of B-cell subsets:

Pre-germinal center B-cells, including:

Memory B-cells, including:

Plasma cells (PCs)

Switched memory B-cells were divided in further subsets based on CD27 expression (CD27⁺ and CD27⁻), the latter also referred to as “double negatives”. PCs were further subdivided in plasmablasts (PBs), prePCs and mPCs based on CD20 and CD138. Additionally, switched memory B-cells and PCs were further subclassified according to their Ig isotypes and Ig subclasses smIgA1⁺, smIgA2⁺, smIgG1⁺, smIgG2⁺, smIgG3⁺, and smIgG4⁺ memory B-cells and smIgM⁺-only, smIgD⁺-only, smIgA1⁺, smIgA2⁺, smIgG1⁺, smIgG2⁺, smIgG3⁺, smIgG4⁺, and smIg⁻ PCs, respectively. Absolute counts were calculated by using total leucocyte cell counts, which was measured by the clinical diagnostics lab of the LUMC. For HSFC, reliable interpretation was set at a minimum of 20 acquired events, according to the EuroFlow guidelines. Counts below 20 acquired events were considered less robust interpretation and therefore indicated in the graphs with gray areas. The absolute detection limit of HSFC (<1 acquired event was 1*10³ cells/ul (average of all patients and timepoints), which was indicated by the dotted lines in each graph.

In-house ELISA protocols were used to determine total IgG and IgM concentrations in the supernatants of the cultures and anti-MPO IgG and anti-PR3-IgG in sera of the patients and supernatants of the cultures. Nunc MaxiSorp™ flat-bottom ELISA plates (ThermoFisher Scientific, Waltham, MA, USA) were coated overnight at room temperature in coating buffer (0.1M Carbonate, pH 9.6) with goat anti human IgG Fc (Bethyl Laboratories, Inc, Montgomery, TX, USA), rabbit anti human IgM (Thermo Fisher Scientific, Waltham, MA, USA), human sputum myeloperoxidase (Elastin Products Company, Inc, Owensfille, MO, USA) or PR3 (isolated from neutrophils according to (32). Subsequently the plates were blocked with PBS/2% Casein for 1 h at room temperature. After washing, the plates were incubated for 1 h with sera, supernatants, and antibodies were diluted in PBS/0.05% Tween/2% Casein (Tween^® 20, Sigma-Aldrich, Darmstadt, Germany). All washing steps were performed with PBS/0.05% Tween. Goat-anti-Human IgG-HRP (Bethyl Laboratories, Inc, Montgomery, TX, USA) was used for the total IgG ELISA. Monoclonal anti human IgM biotin antibody (Sigma-Aldrich, Darmstadt, Germany) and poly streptavidin-HRP (Sanquin, Amsterdam, The Netherlands) was used for the IgM ELISA. ABTS containing 2 µl hydrogen peroxide 30% (Sigma-Aldrich, Darmstadt, Germany) was used as a substrate for the horseradish peroxidases (HRPs) used in the IgG and IgM ELISAs. Anti-human IgG-AP antibody (Sigma-Aldrich, Darmstadt, Germany) was used for anti-MPO and anti-PR3 IgG followed by substrate: 5 mg phosphatase substrate tablet (Sigma-Aldrich, Darmstadt, Germany) in 10 ml pNPP buffer (97 ml diethanolamine, 0.1g MgCL₂, and 0.2 NaN₃ in 1-L water, pH 9.8). Because MPO and PR3 are active enzymes HRP was not suitable to be used in this ELISA. ELISAs using ABTS were measured at 415 nm, while those using phosphatase substrate were measured at 405 nm.

All descriptive clinical data were expressed as median with [range] for numerical data or given as percentage for nominal data. All flow cytometry data were expressed as median with [range] for absolute numbers or ratio as compared to baseline. Statistical difference between two groups was determined with Mann-Whitney U test. Correlations were tested with Pearson’s correlation test (flow cytometry methods) or spearman’s correlation statistical test (changes of ANCA/B-cell subsets). All data were analyzed using SPSS statistics (IBM, New York, USA), version 23 or Prism version 8.0 (GraphPad Software, La Jolla, CA, USA).

The characteristics of AAV patients treated with RTX were summarized in Table 1. Median [range] BVAS scores in patients with active disease (n = 6) that received remission-induction treatment with RTX decreased from 11.5 [6–30] to 0 [0–7] after 24 weeks and remained 0 [0–10] after 48 weeks (Supplementary Figure 3A). Patients that received maintenance therapy (n = 8) had a median [range] BVAS score at baseline of 4 [0–7] which decreased to 0 [0–4] after 24 weeks and to 0 [0–0] after 48 weeks (Supplementary Figure 3B).

Both standard clinical “low-sensitive” flow cytometry (LSFC) and EuroFlow-based HSFC (25, 26) were applied to detect B-cells in 14 AAV cases before and 2, 4, 12, 24, and 48 weeks after RTX, resulting in 68 samples. LSFC was unable to detect CD19⁺ B-cells in 43/54 samples after RTX, in contrast to HSFC which was able to detect CD19⁺ B-cells in all patients at all timepoints after RTX. Figure 1A illustrates the increased sensitivity of HSFC to detect B-cells up to 0.001*10⁶/L, while there is a strong positive and linear correlation between the methods (Pearson’s r = 0.99, CI [0.98–0.99]; p < 0.0001). Additionally, the number of CD19⁺ B-cells over time was shown for each individual patient after treatment with RTX as measured with LSFC (Figure 1B) and with HSFC (Figure 1C). These figures clearly emphasized the increased capability of HSFC to detect B-cells at all timepoints in contrast to LSFC.

With the ability to reliably detect low numbers of circulating B-cells with HSFC, kinetics of different B-cell and plasma cell subsets could now be assessed in AAV patients. At baseline, AAV patients with active disease (n = 6), had a median [range] of 125 [1.70–238]*10⁶ CD19⁺ B-cells/L while AAV patients in remission on maintenance RTX therapy (n = 8) had significantly less CD19⁺ B-cells/L with a median of 0.50 [0.14–17.1]*10⁶ (p = 0.003). Given the reduced B-cell numbers due to previous RTX maintenance treatments (Supplementary Figure 4A), the kinetics of B-cell and plasma cell subsets were studied separately in patients with active disease that received remission induction treatment with 2× 1,000 mg RTX (n = 6) (Figure 2). Of note, the effects of RTX as maintenance therapy (n = 8) are depicted in Supplementary Figure 5.

After remission-induction therapy with RTX, the nadir of circulating CD19⁺ B-cells was reached after 12 [4–24] weeks. At the nadir, concentration of CD19⁺ B-cells was 0.07 [0.01–1.05]*10⁶ B-cells/L (Figure 2A), corresponding to a significant decrease of −99.7% (p = 0.03) (Figure 2E). After the nadir, total circulating B-cells increased up to 0.58 [0.15–6.90]*10⁶ B-cells/L at 24 weeks and up to 0.65 [0.19–10.0]*10⁶ B-cells/L at 48 weeks after RTX, but remained strongly reduced as compared to baseline. As expected, RTX remission-induction treatment (2× 1,000 mg) decreased total B-cells significantly more than RTX maintenance treatment (500 mg) at all timepoints (Supplementary Figure 4B).

At baseline, the predominant B-cell population was pre-GC B-cells while plasma cells represented only a small subset of total B-cells (Figure 2F). The pre-germinal center (Pre-GC) B-cells, including immature, CD5⁺ and CD5⁻ naïve B-cell populations (Supplementary Figure 6) were rapidly reduced after RTX and 67% of the patients reached their nadir between 12 and 24 weeks, which was 0.01 (0.001–0.04) *10⁶ cells/L, corresponding to a significant decrease of 99.98% (Figures 2B, E). Additionally, the memory B-cell compartment was rapidly reduced (Figures 2C, E) and surprisingly also the plasma cell compartment (Figures 2D, E). However, the reduction of plasma cells was less profound than the memory B-cells, and the time to nadir was 12 [2–24] weeks for memory B-cells and 4 [4–12] weeks for plasma cells. The increasing numbers of total CD19⁺ B-cells 24 weeks after RTX was mainly due to an increased number of circulating plasma cells (Figure 2F). Because plasma cells were the least targeted population by RTX, the predominant B-cell population during RTX maintenance therapy was plasma cells (Supplementary Figure 5F). Of note, the absolute counts of Pre-GC B-cell subsets after maintenance therapy with RTX are shown in Supplementary Figure 7.

To further study the residual memory B-cells and PCs after RTX treatment, we subsequently investigated their phenotype by HSFC. All memory subsets, including “unswitched memory B-cells” (CD27⁺ with IgM⁺ and/or IgD⁺), “switched memory B-cells” (CD27⁺ with IgA⁺ or IgG⁺) and “double negative” (DN) B-cells (CD27⁻ with IgA⁺ or IgG⁺) were rapidly and strongly reduced after RTX (Figures 3A, B). No significant different responses among these subgroups nor among IgG⁺ or IgA⁺ memory B-cells were detected after RTX (Figure 3B). Noteworthy, increasing numbers of unswitched and switched memory B-cells could be reliably detected at 48 weeks which was not the case for DN B-cells (Figure 3A).

Subsequently we investigated the residual plasma cell subsets based on their maturation stage and their expression of Ig isotypes after RTX (Figure 3C). Not unexpectedly, CD20⁺CD138⁻ plasmablasts (PBs) were rapidly undetectable after RTX, while CD20⁻CD138⁻ “Pre-PCs” and CD20⁻CD138⁺ mature PCs (mPCs) were reduced but remained detectable at all timepoints (Figures 3C, D). The distribution of IgA⁺, IgG⁺, and IgM⁺ plasma cell subsets was shown in Supplementary Figure 8. At baseline and after RTX, the most prevalent Ig subtype on PBs, PrePCs and mPCs was IgA.

Because HSFC provided measurable, longitudinal changes of B- and plasma cell subsets in AAV patients after RTX, we investigated whether these changes reflected simultaneous changes of circulating Ig and ANCA levels. The absolute serum levels of IgM, IgG and IgA after RTX in patients that received RTX as remission-induction treatment are shown in Supplementary Figure 9. IgM levels decreased gradually without recovery after RTX in most patients to a median decrease of −54% [−61; +37] (p = 0.31) after 48 weeks (Figure 4A). Total serum IgG levels reached their nadir after 4 weeks, corresponding to a median decrease of −21% [−53; −13] (p = 0.03) (Figure 4A). Subsequently, IgG levels gradually recovered back to baseline levels at 48 weeks. Total serum IgA levels remained largely unaffected by RTX treatment, but decreased to a median change of −11% [−44; +40] as compared to baseline (Figure 4A). In AAV patients that received RTX maintenance treatment circulating Ig levels did not change much (Supplementary Figure 10).

Surprisingly, the changes in serum IgM, -G, and -A levels were not associated with the changes in the plasma cell compartment (r < 0.2) (Figure 4B), which was also the case for patients that received maintenance treatment with RTX (Supplementary Figure 10E). Moreover, changes in IgM levels were strongly associated with changes in unswitched memory B-cells (r = 0.61, p = 0.001) and naive B-cells (r = 0.56, p = 0.002) (Figure 4B). The changes in IgG levels were strongly associated with changes of the immature B-cell compartment (r = 0.64, p = 0.0001), CD5⁺ naive B-cells (r = 0.44, p = 0.02), and PBs (r = 0.51, p = 0.006). Changes in IgA levels were associated with changes of CD5⁺ naïve B-cells (r = 0.57, p = 0.002) and unswitched memory B-cells (r = 0.46, p = 0.01).

Figure 5 illustrates the relative changes from baseline of serum anti-PR3-IgG and anti-MPO-IgG in patients treated with RTX as remission-induction and as maintenance treatment. Absolute values are shown in Supplementary Figure 11. All ANCA IgG, against PR3 or MPO significantly decreased for all patients that received remission-induction treatment with RTX (Figure 5A). ANCA IgG was less affected during maintenance treatment with RTX (Figure 5B). Altogether, changes in both anti-PR3 and anti-MPO IgG autoantibodies were strongly associated with changes of naive B-cells (resp. r = 0.76, p = 0.0001 and r = 0.73, p = 0.0001) and DN memory B-cells (resp. r = 0.72, p = 0.0001 and r = 0.58, p = 0.008) (Figure 5C). Surprisingly, but in line with the total Ig data, changes in ANCA did not associate with changes of plasma cell subsets except a weak association of changes in anti-PR3 IgG and mPCs (r = 0.43, p = 0.02) (Figure 5C). Of interest, changes in MPO-ANCA IgM and IgA autoantibody levels demonstrated similar dynamics and correlations as MPO-ANCA IgG, which were significant correlations with the changes in naive B-cells (resp. r = 0.76, p = 0.0001 and r = 0.75, p = 0.0001) and DN memory B-cells (resp. r = 0.58, p = 0.008 and r = 0.55, p = 0.01), and no association with any of the plasma cell subsets (data not shown).

To ultimately demonstrate whether the observed residual memory B-cells with HSFC harbored ANCA-specific B-cells, we investigated in vitro ANCA production before and after RTX treatment. To do so, total PBMCs from AAV patients were polyclonally stimulated to induce antibody-secreting cells (ASCs) and subsequently (ANCA) IgG was measured in their supernatants as a reflection of ANCA-specific memory B-cells (Figure 6A). At baseline of the PBMC culture, the number of CD19⁺ B-cells for HCs was 61[56-74]*10³/well out of 1*10⁶ PBMCs/well, corresponding to normal range references values of HCs (~6%) (33) (Figure 6B). Both PR3- and MPO-ANCA AAV patient samples had significantly lower starting numbers of B-cells in the culture as compared to HCs, possibly due to previous immunosuppressive treatment (Figure 6B).

Polyclonal stimulation of PBMCs from HCs resulted in a median [range] of 70 [45–170]*10³ CD27⁺⁺CD38⁺⁺ ASCs per well after 7 days (Figure 6C). Polyclonal stimulation of PBMCs from PR3-ANCA and MPO-ANCA AAV patient samples before RTX treatment resulted in 0.17 [0.001–100]*10³ and 2.6 [1–61]*10³ ASCs (Figure 6C). After 10 days of culturing, PBMCs from HCs produced a median of 8.3 [2.7–20.7] µg/ml IgG, while 0.03 [0–6.1] µg/ml IgG by PR3-ANCA patient samples and 0.7 [0–7.3] µg/ml IgG by MPO-ANCA patient samples was detected (Figure 6D). Importantly, relevant IgG levels could also be detected at 24 and 48 weeks after RTX. Most importantly, ANCA IgG production was detectable in 3/12 PR3-ANCA and 9/11 MPO-ANCA patient samples at baseline. At 24 weeks after RTX, ANCA IgG production was detectable in 4/12 PR3-ANCA and 7/9 MPO-ANCA patient samples and at 48 weeks after RTX in 0/2 PR3-ANCA and 3/4 MPO-ANCA patient samples (Figure 6E). Altogether, these PBMC cultures demonstrated that both PR3-ANCA and MPO-ANCA patients had residual ANCA-specific memory B-cells after RTX.