Regeneration of B cell subsets after transient B cell depletion using anti-CD20 antibodies in rheumatoid arthritis




Transient B cell depletion with the monoclonal anti-CD20 antibody rituximab has resulted in favorable clinical responses in patients with rheumatoid arthritis (RA). However, little is known about the regeneration profile of different peripheral B cell subpopulations. The aim of this study was to delineate the regeneration profile of different B cell subsets in the peripheral blood after selective anti-CD20–mediated B cell depletion.


Seventeen patients with RA refractory to standard therapy were treated with rituximab. Patients 1–6 received 4 weekly infusions of rituximab at a dose of 375 mg/m2, and patients 7–17 received 2 infusions of rituximab (1,000 mg), 2 weeks apart. Four-color staining was performed at several time points, using CD38, IgD, and CD27 in addition to other cell surface markers. In one patient, the mutational status of the immunoglobulin receptor was examined.


The analysis revealed a distinct pattern of B cell regeneration. The first wave of repopulating B cells were immature B cells (CD38high,IgD+,CD10+,CD24high), the immunoglobulin receptors of which were not yet somatically mutated. In parallel, a recirculation of plasma cells was observed. Later, the number of naive B cells increased, and these cells predominated in the peripheral blood B cell pool. CD27+ memory B cells showed a slow and delayed repopulation, and the level of these cells stayed significantly reduced (<50%) compared with baseline values, for more than 2 years.


Our findings provide evidence for a characteristic regeneration pattern of B cell subpopulations, with long-lasting modulation of B cell subset composition, after selective anti-CD20–mediated B cell depletion.

Rheumatoid arthritis (RA) is a chronic inflammatory disease that targets primarily the synovial membranes of multiple joints, leading to progressive cartilage and joint destruction. For more than 2 decades, RA has been considered to be a predominantly T cell–mediated disease (1). In the past few years, however, a growing body of evidence has emerged suggesting that B lymphocytes are critical in the pathogenesis of RA.

B cells are involved in the inflammation cascade by releasing chemokines and cytokines, such as lymphotoxin α/β, that have essential effects on synovial follicular dendritic cells and the development of tertiary lymphoid tissues (2–4). Within ectopic lymphoid tissue, activated B cells take part in an antigen-driven specific immune response, undergoing affinity maturation and differentiation into plasma cells (3). Moreover, B cells are the precursors of a diverse group of autoantibody-secreting plasma cells. B cells with specificity for self immunoglobulins can bind and internalize immunoglobulin–antigen complexes, leading to an enhanced antigen-presenting function (3, 5). B cells also support the activation of autoreactive T cells by expressing costimulatory molecules such as CD80/86 and CD40 (6).

Additionally, data from clinical trials indicate that B cell depletion with the anti-CD20 antibody rituximab is highly effective in RA and is well tolerated by patients (7–9). This was recently confirmed by a randomized phase II trial (10), providing direct evidence for the contribution of B cells in the pathogenesis of RA. Available evidence suggests that clinical benefits depend on effective B cell depletion. Although complete peripheral B cell depletion is regularly seen in RA (10, 11), in other autoimmune disease, especially systemic lupus erythematosus (SLE), incomplete depletion has been reported in a subset of patients, even after full dosing with rituximab (12).

Rituximab is a chimeric anti-CD20 monoclonal antibody containing murine light- and heavy-chain variable region sequences and human constant region sequences. In 1997, rituximab was approved for the treatment of relapsed or refractory low-grade or follicular non-Hodgkin's lymphoma. The B cell–specific antigen CD20 is found during B cell development, starting at the pre–B cell level. It is stably expressed during further B cell differentiation and is lost during terminal differentiation to plasma cells. It is not present on stem cells, early pre-B cells, or plasma cells.

Rituximab is known to deplete B cells in vivo by several mechanisms, including effector cell–mediated cytotoxicity, complement-mediated cell lysis, activation of intracellular signaling, and apoptosis (13, 14). Although the relative importance of these different functions in RA is still unclear, they all may contribute to the mode of action of rituximab. Peripheral B cell depletion is usually observed for 6–9 months. Currently, only limited data are available regarding the effect of rituximab treatment on B cell regeneration (rebuilding of the B cell compartment).

For the characterization of peripheral B cell subpopulations, a wide range of cell surface markers have been used, including IgD, IgM, CD27, and CD38. It is now well accepted that CD27 is a key marker for mutated memory B cells and further promotes the differentiation of memory B cells into plasma cells (15). In humans, 4 different peripheral memory B cell subsets have been identified, comprising IgD− and IgD+ B cells, and representing 40% of all circulating B cells. These memory B cells include classic isotype-switched B cells (15%), a small population of IgD+ B cells (<1% IgD only), and a population of IgM+ memory B cells that can be divided into an IgM-only subset (10%) and an IgM+,IgD+ memory subset (15%), both of which carry somatically mutated immunoglobulin genes (16). The IgM+,IgD+,CD27− naive B cell compartment comprises 60% of all circulating B cells (16). Originally described as phenotypic markers in B cell subsets of human tonsil (17), CD38 and IgD were recently used to characterize B cell subpopulations in peripheral blood (18). Using this classification, 5 mature B cell (Bm) subpopulations were identified, corresponding to different development stages from naive to memory B cells (Bm1–5) (18).

The aim of the present study was to delineate the regeneration profile of different B cell subsets in the peripheral blood after selective anti-CD20–mediated B cell depletion. In addition to long-term changes in B cell subpopulations, we focused specifically on the first 6 months of B cell regeneration, when the largest dynamic changes were likely to occur.


Patient samples, patient characteristics, and study design.

Peripheral blood samples were obtained from 17 patients with RA, at the indicated time points. All patients met the American College of Rheumatology (formerly, the American Rheumatism Association) revised criteria for the classification of RA (19). In all patients, RA was refractory to standard treatment with disease-modifying antirheumatic drugs, including methotrexate (MTX) and/or tumor necrosis factor α antagonists. Informed consent was obtained from all patients before entering the study, in accordance with the protocol approved by the ethics committee of the University of Wuerzburg. More than 70% of patients (12 of 17) continued to receive concomitant MTX. Rituximab was administered as follows: patients 1–6 received 4 weekly infusions of rituximab at a dose of 375 mg/m2 (total mean ± SEM dose, 2.6 ± 0.18 gm [range 2.4–3.6 gm]); patients 7–17 received 2 infusions of rituximab (1,000-mg), 2 weeks apart. Two patients in group 1 were re-treated 12 and 17 months, respectively, after the first infusion, using the same protocol.

Table 1 summarizes the characteristics of the patients. Patients 1–6 were followed up for more than 2 years, and patients 7–17 were followed up for 1 year. Six months after treatment with rituximab, the mean Disease Activity Score in 28 joints (DAS28) (20) was significantly improved compared with the DAS28 at baseline (4.1 and 6.1, respectively; P < 0.05). Similarly, both the mean C-reactive protein (CRP) level and the mean erythrocyte sedimentation rate (ESR) (both of which are markers of inflammation) were decreased significantly at 6 months (for the CRP, from 3.93 mg/dl at baseline to 0.79 mg/dl; for the ESR, from 28 mm/hour at baseline to 15 mm/hour; both P < 0.001) and were increased slightly 1 year after therapy. Overall, serum IgG and IgA levels remained stable, except in 2 patients in whom the total IgG level dropped below normal range at month 6. One of these patients regained normal IgG levels 12 months after therapy. In both patients, no higher infection rate has been observed so far. The mean IgM level showed a significant decline at 6 and 12 months after treatment (P = 0.05) but did not drop below the normal range. The mean rheumatoid factor (RF) concentration declined in 15 of 17 seropositive patients, from 293 units/ml to 110 units/ml at 6 months (P = 0.03). In one patient who had low levels of RF at study entry, RF became negative after treatment and remained undetectable for up to 1 year posttreatment.

Table 1. Characteristics of the patients*
CharacteristicBaseline6 months12 months
  • *

    MTX = methotrexate; RF = rheumatoid factor; DAS28 = Disease Activity Score in 28 joints; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein.

Age, mean (range) years48 (28–77)  
Female sex, %88  
Disease duration, mean (range) years11.4 (1–28)  
Receiving MTX, %70.6  
RF positivity, %88  
DAS28 score, mean (range)6.1 (3.6–7.8)4.1 (3.0–6.9)4.9 (2.4–7.7)
RF, mean (range) IU/ml292.7 (17–1,180)110.5 (<12–341)155.8 (<12–617)
ESR, mean (range) mm/hour28 (12–38)15 (7–33)21 (8–46)
CRP, mean (range) mg/dl3.93 (0.6–10.4)0.79 (0.03–2.24)1.6 (0.09–5.9)
IgG, mean (range) mg/dl1,213 (742–1,770)1,033 (250–1,490)1,100 (656–1,440)
IgM, mean (range) mg/dl155 (73–335)105 (32–110)93.7 (42–118)
IgA, mean (range) mg/dl269.3 (134–680)219 (107–391)238 (93–448)

Cell preparation.

Peripheral blood was collected from the patients, and peripheral blood mononuclear cells (PBMCs) were prepared by Ficoll-Paque Plus separation (Pharmacia Biotech, Freiburg, Germany).

Monoclonal antibodies.

For immunofluorescence staining, the following cell surface markers were used: CD19 (allophycocyanin, IgG1, SJ25C1), CD27 (phycoerythrin [PE], IgG1, MT271), anti-human IgD (fluorescein isothiocyanate [FITC], IA6-2), CD38 (peridinin chlorophyll protein [PerCP]–Cy5.5, IgG, HIT2), CD20 (PerCP–Cy5.5, 2H7), anti-human IgM (biotin-labeled, G20-127), and anti-human IgG (biotin-labeled, G18-145), followed by streptavidin–PerCP–Cy5.5, CD24 (PE, IgG2a, ML5), CD21 (PE, IgG1, BLy4), CD23 (PE, IgG1, ML233), CD5 (PE, IgG1, UCHT2), CD10 (PE, IgG1, HI10a), mouse G1/G2a (FITC/PE, X40), mouse IgG1 (PerCP–Cy5.5, X40), and mouse IgG1 (APC, X40). All antibodies were from Becton Dickinson (Heidelberg, Germany).

Flow cytometric analysis.

Immunofluorescence staining for flow cytometric analysis was performed by incubating PBMCs in phosphate buffered saline (PBS) with 10 μl of monoclonal antibodies for 20 minutes on ice. In each tube, 1 × 106 cells were suspended. The cells were then washed in PBS. Samples containing biotinylated anti-IgM and anti-IgG antibodies were additionally incubated with streptavidin–PerCP–Cy5.5 for 20 minutes on ice and washed again with PBS. Four-color staining was performed using the FACSCalibur system (Becton Dickinson, San Jose, CA). B cells were identified by forward versus side scatter gating on viable lymphocytes in combination with gating on CD19+ cells. A total of 5,000 to 15,000 events were collected for each analysis.

In all patients, flow cytometric analyses was performed before treatment. For evaluation of the long-term course, analyses were performed for 6 patients during the regeneration period and every 3 months thereafter, for up to 25 months. For evaluation of the short-term course, flow cytometric analyses were performed for 11 patients during the recovery phase, starting when CD19+ B cell counts were >0.5% and then every 4 weeks during the following 6 months of regeneration. In some of these 11 patients, the analyses were performed at later time points, up to 1 year after regeneration had started.

The frequencies of B cell populations were calculated using CellQuest software (Becton Dickinson). The total numbers of B cells of various phenotypes were calculated per milliliter of blood, based on the frequencies of these cells among lymphocytes and the white blood cell count. To investigate the long-term course, flow cytometric analyses were performed using a combination of CD19, CD27, IgD and IgM, or IgG. Phenotypic analyses for evaluation of the short-term course or during regeneration were performed using CD19, IgD, CD38, and/or CD27 in combination with the following monoclonal antibodies: IgM, IgG, CD10, CD5, CD21, CD23, CD20, and CD24.

Single-cell sorting and polymerase chain reaction (PCR) amplification.

The procedure for single-cell sorting and subsequent PCR amplification has been described previously (21). Individual CD19+,IgD+,CD10+ cells were sorted into 96-well plates using a FACSDiVa outfitted with an automated cell deposition unit (BD Biosciences, San Jose, CA), and complementary DNA synthesis was performed using the Titan One Tube RT-PCR System (Roche Diagnostics, Mannheim, Germany). The rearranged Vk1 genes were amplified by nested PCR, and the products were extracted using the Mini Elute Gel Extraction kit (Qiagen, Hilden, Germany) and directly sequenced using the BigDye Terminator Cycle Sequencing Ready Reaction kit (Perkin-Elmer Applied Biosystems, Warrington, UK). The sequences were compared with germline immunoglobulin genes using the Web-based JoinSolver program (online at to examine the frequency of light chain variable gene (VL) mutations.

Statistical analysis.

Statistical analysis was performed using an unpaired 2-tailed t-test. Values are expressed as the mean ± SEM and were calculated using Excel software. P values less than 0.05 were considered significant.


Quantitative long-term B cell regeneration.

Total CD19+ B cells.

In 6 patients, we analyzed the peripheral B cell pool in greater detail, over a period of 2 years. Two of these patients were re-treated at 12 and 17 months, respectively. After treatment with rituximab, all patients showed complete depletion of CD19+ B lymphocytes (to <0.1%), as measured by flow cytometry. B cell repopulation occurred 6–9 months after therapy (mean ± SEM 7.6 ± 0.6 months). As shown in Figure 1A, the absolute numbers of CD19+ B lymphocytes returned to baseline levels 12 months after therapy. Only 1 patient did not recover normal B cell numbers, and in that patient the levels remained reduced at ∼11.2 cells/μl. The numerically low B cell levels were not accompanied by a special clinical course or a different B cell regeneration pattern.

Figure 1.

Levels of CD19+ B cells (A), CD27−,IgD+ B cells (B), CD27+ memory B cells (C), and CD27+,IgD+ B cells (D) after 1 course of rituximab (n = 6 patients). Two patients were re-treated at 12 and 17 months, respectively. Therefore, the data represent only 5 patients at 12–16 months and 4 patients at 23–25 months. Values are the mean and SEM. ∗ = P < 0.005 versus baseline.

Naive and memory B cell pools.

B cell subsets were defined according to the expression of CD27, IgD, and IgM. At baseline, the IgD+ B cell pool consisted of 53.4 ± 12.7 IgD+,IgM+,CD27− naive B cells/μl, and 5 ± 0.75 IgD+,CD27+ memory B cells/μl. B cell repopulation was carried out mainly by naive B cells, the number of which slowly increased and reached baseline values after 12–16 months. This subset increased continuously, reaching an elevation of 84.55 ± 23 cells/μl at 2 years after therapy (Figure 1B). Overall, CD27+ memory B cells did not recover numerically over 25 months, and the level was reduced to ∼50% of the baseline value (Figure 1C). In particular, the level of IgD+,CD27+ cells remained low continuously, to approximately one-third of baseline values, over a 2-year period (Figure 1D).

Early recovery period.

The long-term reduction in the number of CD27+ B cells contrasted with a previous report restricted to 2 patients, which described a high mutational frequency in the immunoglobulin VH receptor, particularly during early B cell recovery (22). It seemed likely that during the time of B cell regeneration, when total B cell numbers were still reduced, differences in repopulation kinetics of particular B cells would be best observed. Therefore, we studied the relative composition of the peripheral B cell compartment during the first 6 months of B cell regeneration, in a more narrow time frame.

Composition of B cell subsets before therapy.

Peripheral blood CD19+ B cells from 11 patients with RA were examined for the expression of IgD and CD38, according to the Bm classification (18), in addition, using CD27 as a marker for somatically mutated B cells (15). The IgD+ compartment comprised 78 ± 2.9% of all B cells, of which 67 ± 4.1% were CD27− and 11 ± 2.1% were CD27+. The IgD+ B cells could be divided into a larger subset of activated naive B cells expressing CD38 (62 ± 4%) (Figure 2B) and a minor subset of naive CD38− B cells (11 ± 1.6%) (Figure 2C), which partially overlapped with the IgD+,CD27+ memory compartment (data not shown). Additionally, we could identify some CD38high, IgD+,CD10+ B cells in peripheral blood (5 ± 0.6%) (Figure 2A). The IgD− B cell compartment could be divided into 3 different subpopulations. The majority expressed medium levels of CD38 (11 ± 1.1%) and another population that had already down-regulated CD38 (6.3 ± 1.2%) (data not shown). In addition, a small proportion of CD38high,IgD−,CD27high plasma cells in peripheral blood (4 ± 0.8%) could be identified (Figure 2D).

Figure 2.

Levels of different B cell subpopulations in peripheral blood after therapy with rituximab (n = 11 patients). A–C, IgD+ compartment. D, IgD− compartment. A, CD38high,IgD+,CD10+ immature subset. B, CD38+,IgD+ naive B cells. C, CD38−,IgD+ B cell subset. D, CD38+++,IgD− plasma cells. Bars show the mean and SEM. ∗ = P < 0.001 versus baseline; + = P < 0.05 versus baseline.

Alterations of B cell subsets during the early recovery phase.

All 11 patients exhibited a total depletion of B cells (<0.1%), as measured by flow cytometric analysis of CD19+ cells among lymphocytes. Peripheral B cell repopulation (>0.5%) occurred between 6 and 10 months (mean ± SEM 7.1 ± 0.4 months) after therapy.

IgD+ compartment.

At the first time point of regeneration, the IgD+ compartment showed a distinct IgD/CD38 expression profile dominated by B cells, with lower and broader IgD expression in particular, and high CD38 expression (44.7 ± 4%) (Figure 2A). Further analysis showed coexpression of CD10 by these cells. In comparison with the medium level of expression of CD38 by IgD+ cells that is usually observed in the peripheral blood, the CD10+ B cells were characterized by CD24high and IgMhigh, CD5+,CD21low, CD23, and CD27−, resembling a previously described peripheral B cell subpopulation of immature/transitional B cells (23, 24). Figure 3 illustrates a typical regeneration pattern for an individual patient. This relative increase in the number of CD38high,IgD+,CD10+ B cells compared with the level before therapy was highly significant in the entire group of patients (P < 0.001) (Figure 2A).

Figure 3.

Typical pattern of regeneration of the IgD+ compartment after treatment with rituximab. At the first time point of regeneration, the IgD+ compartment consisted mainly of CD38high,IgD+,CD10+,CD24+ B cells (red dots), the level of which continued to decrease. The lower quadrants are defined by the use of isotype control antibodies, and the upper quadrants are defined by the expression of CD10 (red dots). Boxes (from left to right) show the percentages of the immature CD38high,IgD+ subpopulation before therapy, and 1, 3, and 5 months, respectively, after regeneration had started. PerCP = peridin chlorophyll protein.

Unlike the CD38high,IgD+ subpopulation, the 2 naive subpopulations (Bm1 [CD38−,IgD+] and Bm2 [CD38+,IgD+]) showed different regeneration kinetics. Both subpopulations regenerated slowly, according to both relative and absolute numbers (Figures 2B and C). Whereas the CD38+,IgD+ subset continuously increased in the later regeneration period, the CD38−,IgD+ subpopulation remained low during the time period under review. At 6 months, the percentage of these 2 naive populations reached the baseline value (for Bm1 + Bm2, 74%), with a predominance of the activated naive B cell subset (for Bm1, 4.1 ± 0.5%; for Bm2, 71.8 ± 3.7%). By multicolor staining of IgD, CD38, and CD27, the CD27+,IgD+ memory B cells were mainly detected in the CD38−,IgD+ subpopulation (data not shown) and showed sustained suppression during the entire study period.

Mutational analysis of the CD19+, IgD+,CD10+ B cell subpopulation.

Because CD38high,IgD+,CD10+ B cells dominated during the early phase of recovery, we used single-cell analysis to investigate the mutational frequency of this subset early in the regeneration process. Of the 20 sequences analyzed, mutations were identified in only 2 (mutation frequency 0.52%), resulting in an overall mutation frequency of 0.05% (Table 2). This is consistent with an immature phenotype.

Table 2. Single-cell analysis of CD19+,CD10+,IgD+ B cells*
Sequence no.Vk1JLCDR3Mutation frequency, %
  • *

    The Vk1 family was amplified. In 20 sequence analyses, the overall mutation frequency was 0.05%. CDR3 = third complementary-determining region.


IgD− compartment.

For the IgD− compartment, the early regeneration period was characterized by a relative increase in the CD38high,IgD− B cell subpopulation (18.1 ± 3.8%; P < 0.001) compared with pretreatment values (Figure 2D). Figure 4 illustrates a typical regeneration pattern for an individual patient. These cells expressed high levels of CD27 (Figure 4) and were negative for CD20 (results not shown). Thus, their phenotype marks them as plasma cells. Despite their CD20− phenotype, these cells were not detectable during the time of peripheral blood B cell depletion. They reappeared in absolute numbers similar to those observed before therapy (Figure 2D). However, their relative values were high during the first months of B cell regeneration, and elevated relative levels were maintained in the entire group over the first 6 months.

Figure 4.

Typical pattern of regeneration with recirculation of CD19low,CD38high,IgD−,CD27high,CD20− plasma cells (green dots) during the early recovery phase. The top row shows the distribution of IgD and CD38 on gated CD19+ B cells, and the bottom rows show the distribution of CD27/CD38. Multicolor flow cytometric analysis showed that the CD38high,IgD− subpopulations expressed high levels of CD27 (green dots, bottom row). A, Before therapy. B, First month of regeneration. C, Three months after regeneration. D, Six months after regeneration. Shown are the percentages of plasma cells.


Transient B cell depletion using anti-CD20 antibodies has been shown to be clinically efficacious in patients with autoimmune disorders such as SLE and RA (9, 10), providing direct evidence for the involvement of B lymphocytes in the development and propagation of T cell–mediated autoimmunity. Nevertheless, we still do not know the pathomechanisms that are important for the observed reduction in the level of inflammation. It can be speculated that by depleting B cells down to the pre-B cell level a new formation of the B cell pool can be achieved, which may be less predisposed to autoimmunity compared with the repertoire pretreatment.

Therefore, when using an anti-CD20 antibody such as rituximab, several questions must be answered. In a previous study (22), we showed that the B cell repertoire that reemerges after transient B cell depletion is polyclonal and has a diverse pattern of variable immunoglobulin genes. Particularly during the early recovery period, significant changes in the repertoire composition could be detected. Following that study, we were interested in determining whether there is a common B cell regeneration profile after anti-CD20–mediated B cell depletion, and therefore studied the regeneration of different B cell subpopulations in 17 patients with RA treated with rituximab.

The results of this analysis showed that the B cell memory compartment exhibited a delayed repopulation pattern, with a reduction in the level of CD27+ B cells to ∼50% of baseline values, for as long as 2 years after treatment. In particular, the IgD+,CD27+ subset remained continuously low (approximately one-third of the pretreatment level). Previously, it was suggested that IgD+,CD27+ B cells in peripheral blood belong to an independent functional subpopulation that may develop and mutate its immunoglobulin receptor outside a germinal center–dependent T cell–B cell interaction (25, 26). Furthermore, similarities between IgD+,CD27+ blood memory cells and splenic marginal zone B cells led to the conclusion that the IgD+ memory subset is a circulating counterpart of the splenic marginal zone B cell compartment (26, 27). Indeed, a reduction in the number of CD27+ memory B cells, including IgD+,CD27+ B cells, was observed in splenectomized patients (26, 28). Therefore, the spleen seems to be important for the production of this B cell subpopulation. We previously reported that rituximab treatment leads to effective B cell depletion in spleen as well as bone marrow (29). This could be important for a specific influence on the precursors of the IgD+,CD27+ pool.

Despite the long-term reduction in their CD27+ memory B cell pools, patients who are treated with rituximab are not prone to a significantly higher rate of infection (30). In particular, antimicrobial antibody titers did not decline significantly after treatment in patients with RA (9, 10). This is probably attributable to long-lived plasma cells that were not eliminated by anti-CD20 therapy. However, some effects on antibody production have been observed. We and other investigators observed a significant reduction (∼30%) in serum levels of polyclonal IgM, although the mean serum level of IgM remained within the normal range (11) (Table 1). In patients with low-grade lymphoma treated with rituximab, the humoral immune response to recall antigens was shown to be significantly decreased (30). However, those patients were heavily pretreated and were experiencing progressive disease. Therefore, they were already significantly immuncompromised before receiving rituximab treatment. The results may be different using only rituximab-mediated B cell depletion. In fact, we observed a good anti-tetanus recall response in an individual patient who was receiving rituximab (29).

Interestingly, the level of specific autoantibodies such as RF decreased significantly after B cell depletion. Therefore, part of the antibody production seems to occur in an anti-CD20–sensitive compartment, which may be important for the clinical effect observed in RA. In contrast to these observations in RA, a reduction in the level of antinuclear and anti–double-stranded DNA antibodies in SLE has not been regularly reported. This difference may be attributable to the variability of B cell depletion observed in numerous patients with SLE, even after receiving a full dose of rituximab (12). The phenomenon is correlated with the inheritance of the low-affinity Fcγ receptor type III allele (31), which is also reported to contribute to the pathogenesis of SLE (32). In addition, complement defects are likely involved. Alternatively, it can be speculated that autoantibody responses in lupus are frequently generated by B cells or mature plasma cells, which are less sensitive to anti-CD20 antibodies.

The most dynamic changes were observed during the first months of peripheral B cell repopulation. Within the first weeks of B cell regeneration, the B cell compartment was dominated by a population of IgD/IgM− B cells with a phenotype resembling that of plasma cells. These cells expressed lower levels of CD19, high amounts of CD38 and CD27, and no IgD and CD20. The surface expression profile of this subset is consistent with that described for plasma cells (33). Even though these cells did not show elevated absolute numbers, they contributed a very high relative amount to the overall peripheral B cell population. In a recent study using mutational analysis of the rearranged B cell receptor, we observed an elevated level of highly mutated sequences during the first weeks of B cell recovery after rituximab treatment (22). Although we did not analyze the surface phenotype of the B cells carrying the highly mutated B cell receptors, it seems very likely that the plasma cells described here account for the majority of highly mutated sequences. The contribution of a CD27− B cell subpopulation that has been also described in SLE (34) is currently being studied.

Because plasma cells are CD20−, their numbers probably are not reduced by rituximab therapy. Nevertheless, they are not detectable during the peripheral B cell depletion phase of ∼7 months but recirculate at the time when B cell regeneration starts. Intense B cell production may change the local environment in the lymphoid tissues in a manner in which plasma cells, normally homed in the bone marrow or in survival niches in secondary lymphoid tissues, begin to recirculate to the periphery. Preliminary results show that most of these plasma cells express the IgA isotype and may therefore be released by gut-associated lymphoid tissue (data not shown). This mechanism may be similar to that involved in an active immune response after infections (35) or vaccination (36), when recirculation of specific plasma cells for establishment of B cell homeostasis has been described.

Apart from the recirculation of plasma cells, B cell recovery was characterized by a distinct pattern of newly developing B cells. It is not known from which B cell stage peripheral B cells regenerate after rituximab therapy. Theoretically, anti-CD20 antibodies can deplete B cells down to the pre-B stage, allowing the formation of a new B cell repertoire. Nevertheless, data from animal models treated with anti-CD20 antibodies support the view that B cells in lymphoid tissues are less affected than are circulating B cells. In a study by Gong et al, using a mouse model for anti-CD20–mediated immunotherapy, particularly germinal center and marginal zone B cells were more resistant to depletion in vivo (37). Therefore, it was interesting to observe high numbers of CD38high,IgD+ B cells coexpressing CD10 during the early phase of B cell recovery. According to the Bm classification, these cells were first assigned to be germinal center founder cells (18). Further analysis of these cells showed them to be CD27−. We analyzed the mutation status of their immunoglobulin receptors, which documented a very low mutation rate of 0.05%, thus providing evidence for an immature or naive phenotype. Recently, 2 reports provided strong evidence that CD38high,IgD+,CD10+ B cells in the periphery correspond to immature human transitional B cells recirculating through the periphery (23, 24). Analogous to those findings, the CD38high,IgD+,CD10+ cells dominating the B cell pool during the early recovery phase also expressed higher levels of CD24 and IgM, designating them as human transitional B cells.

During the later phases of B cell regeneration, a parallel increase in the naive CD38+,IgD+ population was seen, reflecting the transition from immature to mature B cells in the periphery. This regeneration profile was observed irrespective of the treatment regimen and repeated treatments in individual patients. Nevertheless, it should be pointed out that the data presented here concern alterations in only the peripheral B cell compartment. An important issue for further investigations is the effect of rituximab on tissue B cells, which currently is not known.

Our data provide evidence that by using rituximab, a complete depletion of the B cell compartment down to the pre-B stage is likely achieved. Repopulation of peripheral B cells appears to follow a characteristic pattern. First, immature B cells with a nonmutated B cell receptor and a CD38high,IgD+,CD10+,CD24high transitional phenotype regenerate. In parallel, a population of CD20− plasma cells is very prominent. Most likely, the latter do not resemble the depleted cells, which recirculate due to a changed environment when immature B cells expand in the secondary lymphoid tissues. In later phases, the B cell pool shifts to naive B cells of a phenotype regularly observed in healthy adults. In accordance with this model, CD27+ memory B cells recover later and remain numerically reduced for more than 2 years after transient B cell depletion.


We thank Anette Koss-Kinzinger and Kathrin Zehe for technical assistance.