BAFF-modulated repopulation of B lymphocytes in the blood and salivary glands of RITUXIMAB-TREATED patients with Sjögren's syndrome




Treatment with rituximab depletes B cells from the peripheral blood (PB) and salivary glands (SGs) of patients with primary Sjögren's syndrome (SS). The purpose of this study was to track the repopulation of B cell subsets in PB as well as their subsequent homing into SGs in patients with primary SS treated with rituximab.


A series of 4-color flow cytometry experiments delineated B cell subsets in 15 patients with primary SS. All were tested on days 8 and 15 of treatment. Nine of the patients were followed up monthly for 10 months, and the remaining 6 patients were followed up monthly for 24 months. Enzyme-linked immunosorbent assays were developed to measure serum levels of BAFF and rituximab. SGs were biopsied at the start of the study and 4 months after treatment in 15 patients, 12 months after treatment in 3 patients, and 24 months after treatment in 2 patients.


Baseline serum levels of BAFF correlated inversely (r = −0.92, P < 5 × 10−4) with the duration of B cell depletion: the higher the BAFF levels, the shorter the duration of B cell depletion. Four B cell subsets repopulated the PB: plasmablasts (CD19+, CD5−,IgD−,CD38++), transitional type 1 (T1) B cells (CD19+,CD5+,IgD+,CD38++), mature Bm2 cells (CD19+,CD5+/−,IgD+,CD38+/−), and memory B cells (CD19+,CD5−,IgD−,CD38−). Increased numbers of Bm2 cells and decreased memory B cells reappeared with time. Sequential SG biopsies revealed that B cells were absent in these glands for 12 months: they were detected 24 months after rituximab treatment. Memory and T1 B cells were the first B cells identified locally.


The timing of B cell repopulation is modulated by BAFF and is followed by reconstitution of the preexisting abnormalities.

Periductal aggregation of lymphocytes in the lacrimal and salivary glands (SGs) is promoted by autoimmune epitheliitis (1) and leads to xerophthalmia, xerostomia, and the production of antibodies to self (2). This disorder may occur alone as primary Sjögren's syndrome (SS) or against a background of other connective tissue diseases as a secondary symptom. Contrary to the long-held notion that primary SS is a T cell–mediated disorder, B cells appear to be key to its pathogenesis (3).

The paradigm based on B cell involvement, coupled with the ever-growing variety of B cell subsets, raises the issue as to which of these subsets underpins the autoimmune features characteristic of the disease. As soon as B cell antigen receptors are functional, B cells migrate from the bone marrow to secondary lymphoid organs, where they mature further. The emerging transitional B cells have long been described in rodents (4) and have recently been identified in the peripheral blood (PB) of humans (5). As transitional type 1 (T1) B cells, they present as CD20+,CD5+, CD10+/−,CD21+/−,CD23+/−,IgM+,IgD+/−,CD38++. Once they have evolved to type 2 (T2) B cells, they become CD20+,CD5+/−,CD21++,CD23+/−,IgM++, IgD++,CD38+/−. The latter subset differentiates either to noncirculating cells (6) that populate the marginal zone or to circulating cells that constitute germinal centers within lymphoid follicles (7). Developmental stages are then identified on the basis of the relative expression levels of IgD and CD38 on mature B (Bm) lymphocytes (8). As a consequence of their selection outside germinal centers in the presence of short-lived plasma cells, CD38−,IgD+ naive Bm1 cells progress to become CD38+,IgD+ activated Bm2 cells, some of which are selected as CD38++,IgD+ Bm2′ germinal center founder cells. Then, they differentiate into CD38++,IgD− Bm3 centroblasts and Bm4 centrocytes. Two types of B cells arise from germinal centers: CD38−,IgD− Bm5 memory B cells and CD38++, IgD− plasmablasts, which were first described by Odendahl et al (9). These return to the bone marrow, where they differentiate into long-lived plasma cells.

Patients with primary SS exhibit an increase in Bm2/Bm2′ cells (10) and a decrease in memory Bm5 cells (11). Cells of the second subset accumulate in the SGs, germinal centers of which include T2 B cells and B cells resembling marginal zone B cells (12). That the B cell–activating factor BAFF participates in the development of such abnormalities is suggested by the correlation between its high rate of production in primary SS and the increased number of circulating Bm2/Bm2′ cells (13). This view is reinforced by the presence of receptors for BAFF on B cells sequestered within the SGs (14).

The importance of B cells in autoimmune processes was the rationale behind the use of the anti-CD20 monoclonal antibody (mAb) rituximab in the treatment of primary SS. Furthermore, the pioneering work by Edwards and Cambridge (15) in rheumatoid arthritis (RA) has spawned a number of trials in autoimmune diseases (16–18) such as idiopathic thrombocytopenic purpura (ITP). The kinetics of B cell reconstitution have been examined in patients with systemic lupus erythematosus (SLE), RA, and more recently, lymphoma (19–21).

Rituximab has been administered to patients with primary SS and has been claimed to lead to B cell depletion in the PB (22) and SGs (23). However, studies of the sequential repopulation of B cells following treatment of this disease have not been performed. There are several postulated mechanisms for the action of rituximab in B cell depletion. Should it turn out that antibody-dependent cell-mediated cytotoxicity (ADCC) predominates, its effects in primary SS would be dependent on the V and the F alleles of the Fcγ receptor IIIa (FcγRIIIa) –158 polymorphism, because the V and F alleles display a high affinity and a low affinity for IgG1, respectively (24). In addition, the baseline level of serum BAFF might influence the subsequent B cell regeneration but, in turn, depletion of B cells might affect the production of BAFF (25).

This study provides a unique opportunity to gain insight into the developmental stage at which autoreactive B cells escape tolerance, since data concerning whether rituximab depletion of B cells extends to tissue B cells are lacking. Our study was therefore conducted to address 3 issues. What are the dynamics of B cell subset redistribution following rituximab treatment in primary SS patients as compared with ITP patients? Do FcγRIIIa polymorphism and the preexisting BAFF level influence B cell depletion and repopulation? What are the kinetics of the reappearance of B cell subsets in the PB and SGs of the treated patients? Our findings are presented below.


Patients and disease controls.

Fifteen patients with primary SS were included in the trial. All of them fulfilled the American–European Consensus Group criteria for SS (26) and displayed an SG focus score of at least 3. Five patients with ITP who fulfilled the American Society of Hematology criteria for the disease (27) were enrolled as disease controls. This study was approved by the Institutional Ethics Committee at Brest University Medical School, and written informed consent was provided by all participants.

The approved dosing schedule for rituximab, 4 doses of 375 mg/m2 given intravenously on consecutive weeks, is based on the original trial in non-Hodgkin's B cell lymphoma (18, 28, 29). It is likely, however, that a course of 2 infusions may become the standard dosing regimen for patients with autoimmune diseases (18, 29, 30). Our primary SS patients received only 2 infusions of rituximab on days 1 and 8. Given the total depletion of circulating B cells on day 15, additional infusions appeared to be unnecessary.

Minor SG biopsy and polymerase chain reaction (PCR) analysis.

Minor SGs were biopsied before and 4 months after rituximab treatment in all 15 patients with primary SS. As described in detail elsewhere (12), reverse transcription–PCR (RT-PCR) using CD79b transcripts revealed the presence of B cells in 8 of the 15 SG samples. Framework region 2 and framework region 3 nested RT-PCR analysis excluded B cell clonal expansion in these 8 CD79b+ samples (12). Three of these 8 patients with infiltrating B cells underwent a third SG biopsy 12 months after treatment (the remaining 5 patients refused a third biopsy), and 2 of those 3 patients had a fourth SG biopsy 24 months after treatment (the remaining patient refused a fourth biopsy).

Flow cytometry analyses.

The distribution of B cell subsets was examined using a fluorescence-activated cell sorter that was adapted to provide a comprehensive analysis of the repopulating B cells. Except where indicated otherwise, antibodies were purchased from Beckman Coulter (Hialeah, FL). As previously reported (23), the percentage of CD19+ B cells was assessed at the following time points: prior to the first infusion, on day 8 after the first infusion and prior to the second, on day 15 after the second infusion, and monthly thereafter (for 10 months in 9 primary SS patients and 24 months in the remaining 6, as well as for 6 months in 2 ITP patients and 15 months in the remaining 3). Absolute numbers of cells were inferred from the size of the lymphocyte subsets and the leukocyte count.

After their depletion, or once the frequency of B cells had reached a minimum of 2.5% of circulating lymphocytes, their new subset distribution could be determined by staining mononuclear cells that had been separated with Ficoll-Hypaque. These studies were performed using phycoerythrin (PE)–Cy7–conjugated anti-CD19 mAb, PE-conjugated anti-CD5 mAb, PE–Cy5–conjugated anti-CD38 mAb, and fluorescein isothiocyanate (FITC)–conjugated anti-IgD mAb (BD PharMingen, San Diego, CA). To further characterize these subsets, the PB-derived B cells were enriched by rosetting with sheep erythrocytes. We used 8 quadruple analyses, combining 3 mAb (PE-conjugated anti-IgD, PE–Cy5–conjugated anti-CD38, and PE–Cy7–conjugated anti-CD5) with a fourth, FITC-conjugated, mAb selected from the following 8: anti-CD20, anti-CD27, anti-CD10, anti-CD23, anti-CD34, anti-CD62L, anti–Ki-67, or anti-IgM (Dakopatts, Glostrup, Denmark). An additional 4-color staining with PE-conjugated anti-CD24, FITC-conjugated anti-IgD, PE–Cy5–conjugated anti-CD38, and PE–Cy7–conjugated CD19 confirmed the identity of transitional B cells.

Immunofluorescence staining of salivary glands.

Biopsy specimens embedded in OCT (Miles, Naperville, IL) were snap-frozen in isopentane, and sequential 4-μm–thick tissue sections were cut and mounted onto poly-L-lysine–coated slides. Serial sections were incubated with unconjugated mouse anti-CD3, anti-CD4, or anti-CD8 mAb, together with unconjugated goat anti-CD20 antibody (Neomarker, Fremont, CA), for 40 minutes. After 3 washes in phosphate buffered saline (PBS), slides were incubated for another 40 minutes with FITC-labeled donkey anti-mouse antibody (Jackson ImmunoResearch, West Grove, PA), together with tetramethylrhodamine isothiocyanate (TRITC)–labeled donkey anti-goat antibody (Jackson ImmunoResearch) in PBS supplemented with 2% donkey serum (Sigma, St Louis, MO). After another 5 rinses, the sections were fixed in 4% cold p-formaldehyde and then examined with the Leica TCS-NT confocal imaging microscope (Leica, Wetzlar, Germany).

Control experiments established that human IgG did not interfere with the staining. This was based on the fact that mouse anti-human IgG, which was added instead of the anti-CD3/CD19, could not be detected with the FITC-labeled donkey anti-mouse IgG antibody (both from Jackson ImmunoResearch). Furthermore, goat anti-human IgG (Rockland, Gilbertsville, PA) added instead of the anti-CD3/CD19 was not detectable with the TRITC-labeled donkey anti-goat IgG antibody.

Percentages of cells expressing the second marker were determined by 2 independent observers (J-OP and CD) in 50 cells expressing the first marker. The overlay of green and red was seen as yellow on the screen of the confocal microscope.

In further experiments, TRITC-labeled goat anti-human CD20 antibody was combined with FITC-labeled rabbit anti-IgD antibody or with biotinylated anti-IgM mAb (plus FITC-labeled streptavidin), and then with anti-CD35 or anti-CD38 mAb (plus FITC-labeled donkey anti-mouse antibody) or with 1 of the following 4 FITC-conjugated mAb: anti-CD5, anti-CD19, anti-CD21, or anti-CD27.

Serologic tests.

Rheumatoid factor (RF) was determined by class-specific enzyme-linked immunosorbent assays (ELISAs) based on the use of rabbit IgG (Sigma) as the antigen being recognized. These ELISAs were performed in our laboratory as previously described (31).

Another ELISA was developed to quantify BAFF, as described elsewhere (13, 32). Briefly, microtiter plate wells were coated with mouse anti-BAFF monoclonal IgG1 (Research Diagnostics, Flanders, NJ) or an irrelevant control mouse IgG1 mAb. Sera diluted 1:20 in PBS containing 0.1% Tween 20 and 0.1% bovine serum albumin were added to the wells and incubated for 4 hours, followed by sequential incubations with rabbit anti-BAFF antibody (Upstate Biotechnology, Lake Placid, NY), biotinylated goat anti-rabbit antibody, and horseradish peroxidase (HRP)–labeled streptavidin. Absorbance values in the IgG1 control wells were subtracted from those in the duplicate wells coated with anti-BAFF mAb. A standard curve was constructed using serial dilutions of recombinant BAFF (PeproTech, Rocky Hill, NY). Samples with BAFF concentrations <1 ng/ml, which is the lowest limit of detection, were assigned a value of 0.

BAFF levels were measured before rituximab treatment and at weeks 16, 28, and 40 after treatment in all patients with primary SS and before rituximab treatment in the 5 patients with ITP. As advised in the article by Zhang et al (33), we wanted to be sure that RF did not compete with BAFF for binding to the capture IgG1 and then to the revealing antibody. Therefore, 20 randomly selected sera were tested in the ELISA and then adsorbed with protein A–agarose (Sigma) to deplete immunoglobulins and tested again.

Rituximab concentrations in the 15 primary SS patients were determined at week 16, using a previously validated ELISA (34). Sera diluted 1:100 were allowed to react for 90 minutes with the MB2A4 antiidiotype mAb (Serotec, Oxford, UK). After 3 washes, the plate was incubated for 60 minutes with HRP-labeled anti-human Fcγ antibody (Sigma), further washed, and the ELISA substrate was added. A standard curve was prepared by diluting known quantities of rituximab in normal serum. Sera from 10 RA patients drawn 8 days after their first infusion of rituximab served as positive controls for the mAb, and sera from another 20 RA patients (stored at –80°C) served as negative controls in this ELISA.

FcγRIIIa sequencing.

FcγRIIIa genotypes were determined by PCR using 20 ng of DNA in 100 nM dNTPs, 1.5 mM MgCl2, and 2.5 units of Taq polymerase (Invitrogen, Cergy-Pontoise, France), with 200 nM 5′-ATATTTACAGAATGGCACAGG-3′ and 5′-GAGTGAATGACACCTCCTAGCTACC-3′ primers. The 386-bp PCR product was purified with High Pure PCR columns (Roche Diagnostics, Meylan, France) and analyzed using the BigDye Terminator Sequencing Reaction kit (Applied Biosystems, Foster City, CA) using the 5′-GATTGCAGGTTCCACACACAGGCGTC-3′ and the 5′-TCCAAAAGCCACACTCAAAGTC-3′ primers.

Statistical analysis.

Results are expressed as the mean ± SD. Correlations were established using Spearman's test, and comparisons were made with the Mann-Whitney U test for unpaired data and Wilcoxon's test for paired data.


Depletion and repopulation of B cells in blood and salivary glands.

B cell depletion was defined as a reduction in the proportion of CD19+ cells to <0.1% of circulating lymphocytes. In PB, B cells started to decline during the first week of treatment, as reported previously (23). On day 8, B cells dropped to 0.0% in 12 patients, 0.4% in 1 patient, 0.7% in 1 patient, and 4.2% in 1 patient. On day 15, all 15 treated patients had full depletion of B cells (Figure 1A).

Figure 1.

Repopulation of B lymphocytes in relation to serum levels of BAFF in 15 patients with primary Sjögren's syndrome (SS) treated with rituximab. A, Total numbers of CD19+ cells (B cells) before and after rituximab (RTX) treatment. Arrows indicate RTX infusions. Numbers to the right of the data points are patient numbers. B, Expression of CD24 cells versus CD38 cells, by flow cytometry. Results suggest that the repopulating B cells are immature B cells in primary SS (pSS) patient 1 at 5 and 10 months after RTX, but this was not the case before RTX treatment in either this patient or the patient with idiopathic thrombocytopenic purpura (ITP; patient 2). C, Inverse correlation of serum levels of BAFF with the duration of B cell regeneration: the Fcγ receptor IIIa –158 genotypes seem not to be related to the duration of B cell depletion. D, BAFF concentrations in 2 groups of primary SS patients identified before RTX treatment: group I had high levels of BAFF and group II had low levels. BAFF levels were determined before treatment (0 months) and at 4, 7, and 10 months after treatment. Values are the mean and SD. ∗ = P < 0.05 versus baseline.

Repopulation was considered to have been achieved when the proportion of CD19+ cells reached 0.1% of circulating lymphocytes, but at least 2.5% was needed in order to determine the B cell subset distribution (Figure 1B). One patient experienced repopulation from week 8 onward, but in the remaining 14 patients, B cells remained markedly suppressed for a minimum of 9 weeks and a maximum of up to 40 weeks (Figures 1A and B). (Photomicrographs of stained SG biopsy sections obtained before and 4, 12, and 24 months after rituximab treatment are available upon request from the corresponding author.)

In the 8 patients whose SGs contained significant numbers of B cells (Figure 2, first column, patient 15), biopsy specimens were completely devoid of B cells 4 months after rituximab treatment. Although CD20 is confined to B cells, the numbers of infiltrating T cells were modestly affected by rituximab, as shown by staining for CD3 (Figure 2, second column) and CD4 or CD8 (results not shown). (Histograms showing the repopulation of B cells in all 15 primary SS patients and in 3 ITP patients for up to 24 months after treatment are available upon request from the corresponding author.) B cell depletion lasted for an additional 8 months in the 3 patients sampled a third time, but returned 24 months after treatment in the 2 of these same 3 patients who were biopsied a fourth time (Figure 2).

Figure 2.

Photomicrographs of minor salivary gland (SG) biopsy sections from a patient with primary Sjögren's syndrome (SS) obtained before and at 4, 12, and 24 months after rituximab (RTX) treatment. Sections from primary SS (pSS) patient 15 were stained with tetramethylrhodamine isothiocyanate–conjugated anti-CD19 monoclonal antibody (mAb) for B cells and with fluorescein isothiocyanate–conjugated anti-CD3 mAb for T cells. At 4 months after RTX treatment, B cells had disappeared from all SGs; they then began to reexpand at 12 months, and reinfiltrated the SGs at 24 months (original magnification × 16).

Short-term repopulation of B cell subsets in peripheral blood.

B cell subsets repopulating the PB (all CD34−) were characterized according to the expression of IgD and CD5, and were gated based on CD38 expression. Three representative samples (patients 1, 8, and 14) are shown in Figure 3. First, the IgD−,CD5− cells were separated into CD38++ plasmablasts (CD20−,CD19+/−,IgM+/−,CD27+,CD62L+, CD10−,CD23−) and CD38+/− memory B cells (CD20+,CD19+,IgM+/−,CD27+,CD62L+,CD10−, CD23−). Then, the IgD+,CD5− cells were separated into CD38+,Bm2 and CD38−,Bm1 cells (CD20+, CD19+,IgM+,CD27−,CD62L+,CD10−,CD23+/−). Finally, the IgD+,CD5+ cells were separated into CD38++ immature T1 B cells (CD20+,CD19+, IgM+,CD27−,CD62L+/−,CD10+,CD23−) and CD38+/− mature Bm1 cells.

Figure 3.

Phenotype of blood B cells at the beginning of repopulation after rituximab (RTX) treatment in 3 patients with primary Sjögren's syndrome (SS). Enriched B cells from primary SS (pSS) patients 1, 8, and 14 were obtained at the indicated months after treatment with RTX, stained for IgD and CD5, and then divided into 3 subsets according to the expression levels of these 2 markers. The cells in the lower left quadrant (A) (IgD–,CD5–) were analyzed according to the expression pattern for CD38. Two populations are shown: plasmablasts (IgD–,CD5–,CD38++) and memory B cells (IgD–,CD5–,CD38–). Cells in the lower right quadrant (B) (IgD+,CD5–) were analyzed according to the expression pattern for CD38 and correspond to mature Bm1 and Bm2 cells (CD38+/–,CD19+,CD20+,CD27–,CD10–,IgM+, CD23+/–,CD62L+). The cell population in the upper right quadrant (C) (IgD+,CD5+) was divided into 2 subsets according to the expression pattern for CD38: transitional type 1 (T1) B cells (IgD+,CD5+,CD38++) and Bm1/Bm2 cells (IgD+,CD5+,CD38+/–). Solid histograms show cell staining; open histograms show isotype controls.

Staining with anti-CD24 confirmed the identity of T1 B cells (a representative example is shown in Figure 1B). Overall, this initial wave of B cells comprised 4 subsets: mean ± SD 36.4 ± 1.4% Bm1/Bm2 cells (22.0 ± 2.0 × 106/liter), 32.9 ± 2.2% plasmablasts (17.5 ± 1.5 × 106/liter), 16.9 ± 1.1% T1 B cells (9.4 ± 1.2 × 106/liter), and 12.3 ± 1.1% memory B cells (7.0 ± 1.2 × 106/liter) (Figure 4, patient 8). Of note, 63.5 ± 3.4% of Bm1/Bm2 cells expressed CD5. Similar patterns of reconstitution were seen in all patients with primary SS. (Histograms showing the repopulation of B cells in all 15 primary SS patients and in 3 ITP patients for up to 24 months after treatment are available upon request from the corresponding author.)

Figure 4.

Phenotype of B cells at the time of repopulation after rituximab treatment in a patient with primary Sjögren's syndrome (SS). Plasmablasts, memory B cells, CD5–,Bm1/Bm2 cells, transitional type 1 (T1) cells, and CD5+,Bm1/Bm2 cells obtained from primary SS patient 1 were analyzed against CD19, CD20, CD27, CD10, IgM, CD23, and CD62L. Immature B cells (see Figure 1B) were identified as CD10+ (arrow), CD62L– (arrow), and CD24++. Solid histograms show cell staining; open histograms show isotype controls.

The repopulating B cell subsets were distributed differently in ITP patients than in primary SS patients. There were more T1 and Bm2 cells in ITP patients than in primary SS patients (mean ± SD 28.9 ± 2.7 × 106/liter versus 9.4 ± 1.2 × 106/liter and 58.8 ± 3.2 × 106/liter versus 22.0 ± 2.0 × 106/liter, respectively; P < 0.05 for each comparison) and fewer plasmablasts and memory B cells (4.5 ± 2.0 × 106/liter versus 17.5 ± 1.5 × 106/liter and 3.2 ± 1.1 × 106/liter versus 7.0 ± 1.2 × 106/liter, respectively; P < 0.05 for each comparison). Remarkably, the number of Bm2 cells was increased in the primary SS and ITP patients at 12 months after treatment, regardless of their levels at the start (Figure 5A).

Figure 5.

Phenotype of B cells in a patient with primary Sjögren's syndrome (SS) and a patient with idiopathic thrombocytopenic purpura (ITP) before and after treatment with rituximab (RTX). A, Analyses of cells from primary SS (pSS) patient 8 and ITP patient 1 were performed before and 12 months after RTX treatment. The relative expression of CD38 and IgD on mature B cells (Bm), most particularly, activated Bm2 cells and germinal center Bm2′ cells, distinguishes the developmental stages (see ref.8). Values in the upper right corners of the scatterplots are the percentage of the cell type. B, Kinetics of the repopulation of B cells in the blood after RTX treatment in 6 patients with primary SS (patients 1, 2, 3, 8, 9, and 13) (●) and in the 5 patients with ITP (patients 1–5) (○). Shown are the absolute numbers of plasmablasts, memory B cells, transitional type 1 (T1) B cells, and Bm2 B cells. eBm5 = early Bm5.

Features of peripheral blood depletion and repopulation.

Although the patients displayed the 3 FcγRIIIa genotypes (5 had FF, 9 had VF, and 1 had VV), all experienced similar depletion of B cells to <0.1% of circulating lymphocytes (Figure 1C). This suggests, although it does not prove, that FcγRIIIa polymorphism is not predictive of the response to rituximab treatment in primary SS. This is at variance with previous reports in SLE (35), but is similar to the findings in B cell chronic lymphocytic leukemia (36). Repopulating cells were similar in all primary SS patients, inasmuch as the only difference between them related to the time of reappearance of B cells in the PB (see Figure 1A). (Photomicrographs of stained SG biopsy sections obtained before and 4, 12, and 24 months after rituximab treatment are available upon request from the corresponding author.)

This raises questions regarding the role of BAFF in the duration of B cell depletion. The preexisting levels of BAFF correlated inversely (r = –0.92, P < 5 × 10–4) with the number of months it took for B cells to repopulate the PB. The higher the serum level of BAFF, the shorter the elapsed time before B cell repopulation (Figure 1C). To confirm this impression, the patients were categorized according to their preexisting levels of serum BAFF (Figure 1D), with group I consisting of the 9 patients with high levels and group II consisting of the 6 patients with low levels. Intriguingly, there was a slight diminution in the levels of BAFF in group I sera after 4 months (P < 0.05), but the levels returned to their original values 3 months later. In contrast, BAFF remained below the limits of detection in group II sera until at least 10 months after treatment. Like the latter group of primary SS patients, the ITP patients, who did not have BAFF, displayed circulating B cells after 6.5 months.

Of note, serum levels of RF correlated with serum levels of BAFF, but not with the binding of serum RF to irrelevant mouse IgG1 coated on the BAFF ELISA plate. RF is therefore unlikely to bind to the capturing IgG1 or to the revealing antibody (Figure 6A). This was confirmed by preabsorption of immunoglobulins, which did not alter the level of BAFF, as determined by ELISA (Figure 6B).

Figure 6.

Absence of interference of BAFF with rheumatoid factor (RF) in patients with primary Sjögren's syndrome (SS), as determined by enzyme-linked immunosorbent assay (ELISA) performed in our laboratory. A, RF levels in the primary SS patients correlated with the BAFF levels (bottom), but not with the irrelevant IgG1 that had been coated onto the plate (top). OD = optical density. B, Changes in BAFF levels in the 15 patients with primary SS before and after IgG adsorption. To ensure that RF did not compete with BAFF for binding to the capture IgG1 and then to the revealing antibody (33), all 15 patient sera were tested in the BAFF ELISA before and after immunoglobulin depletion through a protein A–agarose column. No appreciable differences were seen.

Because of variability in the serum levels of BAFF, establishing a relationship between BAFF and B cell depletion required assessment of serum levels of rituximab. Unexpectedly, the level of rituximab was much lower in BAFF-containing group I sera than in BAFF-noncontaining group II sera (mean ± SD 1.6 ± 0.5 μg/ml versus 24.3 ± 15.2 μg/ml; P < 0.01). This finding suggests that the effects of rituximab were faster in the presence of BAFF than in its absence.

Long-term repopulation of peripheral blood.

Levels of the 4 repopulating B cell subsets were serially assessed for 24 months in the PB of all 15 primary SS patients as compared with the 5 ITP patients (Figure 5B). (Photomicrographs of stained SG biopsy sections obtained before and 4, 12, and 24 months after rituximab treatment are available upon request from the corresponding author.) Over the long term, memory B cells did not behave the same way in these 2 diseases. They decreased in primary SS patients, but increased in ITP patients. In contrast, T1 B cells did not change appreciably in primary SS patients, but diminished significantly in ITP patients, as compared with preexisting levels. In both diseases, the number of plasmablasts was consistent, while Bm2 cells continued to be extremely high, long after their initial depletion. At 14 months after treatment, these cells represented 92.2 ± 3.5% (mean ± SD) of B cells in the primary SS patients and 78.1 ± 6.7% of B cells in the ITP patients. However, 21.3 ± 1.9% of Bm2 cells were CD5+ in the primary SS patients. These cross-sectional results suggest that, unlike the ITP patients, the primary SS patients reproduced the original B cell abnormalities over time.

Reappearance of B cell infiltrates in the salivary glands.

Previously, we reported that a number of T2 B cells resided in the SGs of patients with primary SS (12). These CD19+ B cells express CD20, CD21, and IgM, but not CD35 or CD38 (Figure 7, left columns, patients 14 and 15). SGs from 3 of the 8 patients presenting with marked B cell infiltrates before treatment were analyzed 12 months after rituximab-induced B cell depletion, and SGs from 2 of these 3 patients were analyzed after 24 months. Whereas there were as few as 1–3% CD5+ among CD20+ B cells before treatment, CD5+,CD20+ B cells accounted for one-third of the B cells that repopulated the SGs in the same patients at 24 months after treatment. The yellow overlay of green-stained CD27 with red-stained CD20 (Figure 7, right columns) indicates that many B cells are memory B lymphocytes.

Figure 7.

Extensive phenotyping of B cells in salivary glands (SGs) from 2 patients with primary Sjögren's syndrome (SS) before and after treatment with rituximab (RTX). SG B cells from primary SS (pSS) patients 14 and 15 were analyzed before and 24 months after RTX treatment, using a number of double combinations of fluorescein isothiocyanate (FITC)–conjugated antibodies with tetramethylrhodamine isothiocyanate (TRITC)–conjugated antibodies. The overlay of green FITC fluorescence and red TRITC fluorescence is seen as yellow (original magnification × 63).

Other B lymphocytes expressed CD19, CD38, and IgM, but not IgD or CD21. Indeed, these cells were more similar to T1 B cells seen in the PB than to the aforementioned T2 B cells (12). We were unable to detect CD21+,CD35+ follicular dendritic cells, and staining with Ki-67 was unsuccessful. Thus, in summary, the majority of B cells in the SGs at 24 months after depletion consisted of memory and T1 B cells.


In this study, we show that 2 infusions of rituximab reduced B cells to <0.1% of circulating lymphocytes for >2 months in all but 1 of the primary SS patients. In addition, we provide evidence that all B cell subsets were equally sensitive to this treatment. Further B cell depletion was also achieved in solid tissues as early as 4 months after treatment, and this lasted for 12 months or more. Parenthetically, B cell reconstitution seems to begin earlier in primary SS than in SLE (19) and RA (20); however, differences in the treatment protocols mean that comparison of our results with previous reports should be viewed with caution. We also observed that a number of plasmablasts were sustained in the PB because of their protection from the effects of rituximab by their lack of CD20 expression as well as their BAFF-promoted survival (37). Similar plasma cell precursors have been detected after rituximab treatment in SLE (19) and RA (20).

The pattern of B cell repopulation was the same in all treated patients with primary SS. A high frequency of naive B cells carrying IgD and CD38 were up-regulated, and a few memory B cells were characteristically seen. The former B cells, IgD+,CD38++, are naive B cells exiting the bone marrow, and unlike germinal center founder cells, they express CD10, CD24, CD62L, and CD38. The latter cells distinguish T1 B cells, which are CD38++, from Bm1/Bm2 cells, which are CD38+/–. Thus, as hypothesized by Anolik et al (21), B cell repopulation after rituximab treatment recapitulates B cell ontogeny.

Interestingly, similar to immature B cells, mature Bm1/Bm2 cells carry CD5. Restricted progenitors of CD5+ cells, which have just been identified (38), could differentiate into circulating CD5+ B cells during adulthood. Thus, rituximab-induced B cell elimination appears to trigger a second round of ontogeny, with immature innate CD5+ B cells (39) and conventional B cells acquiring CD5 within the Bm2 subset (40).

The sole difference between the patients was in the timing of the reappearance of B cells. In this regard, we addressed the issue of which factors predict the duration of B cell depletion. Previous reports link the response to rituximab in SLE patients with FcγRIIIa polymorphism (35). Our limited data do not support a major role of FcγRIIIa polymorphism in patients with primary SS, at least in this group of study patients. Yet, whether or not ADCC is important in the therapeutic effectiveness of rituximab remains to be established. It is possible that FcγR polymorphisms other than FcγRIIIa-158 affect the killing function of macrophages and natural killer cells (41). Also related to dosing is the appearance in the recipient of an immune response to the infused rituximab (18). These were the findings that prompted us to launch a multicenter trial to compare a large number of primary SS patients. Correlation between rituximab levels and B cell depletion is currently being evaluated.

Relative to complement-mediated cellular cytotoxicity, caspase-dependent apoptosis, and sensitization to cytotoxic agents (18), the importance of ADCC as a mechanism for the activity of rituximab may vary from patient to patient. The efficacy of rituximab treatment is also diminished if the disease mechanism provides B cells with survival signals. In particular, repopulation of B cells might be modulated by BAFF through its effects on their lifespan (42). It is known, for example, that BAFF controls a steady-state number of circulating B cells that survive differentiation (43). This is consistent with equality in the length of time required for B cell repopulation in 3 of the 5 ITP patients and in the 6 group II primary SS patients, all of whom had low levels of BAFF. The influence of BAFF on B cell return is further supported by our previous finding that B cells infiltrating the SGs not only expressed BAFF receptors, but they also had the ability to synthesize this cytokine (14). As a result, rituximab-induced B cell depletion would decrease the synthesis of BAFF. Even so, BAFF could bind the newly formed B cells and thus be rendered undetectable in the ELISA.

Relevant to this issue is the suggestion that the beneficial effects of rituximab-based B cell depletion may be offset by a BAFF-mediated antiapoptotic effect on reemerging B cells (25). In this regard, the higher the level before treatment, the shorter the duration of the B cell lymphopenia. Similar bindings were reported in patients with B cell chronic lymphocytic leukemia (44). Thus, if rituximab is to be routinely applied as a treatment strategy for primary SS, advance knowledge of the level of BAFF may help in tailoring the doses of rituximab and the frequency of administration. Our results at 4 months posttreatment are consistent with the report that BAFF levels rose at 1–2 months (25) and decreased to pretreatment levels thereafter. Indeed, as stated by Cambridge et al (25), the relationship between the timing and speed of B cell repopulation and the levels of BAFF is likely to be complex.

One existing view is that treatment with rituximab may be less effective at purging B cells from sites of disease in affected tissues, although data on this topic are lacking. Hence, our study is novel in that sequential biopsies allowed us to compare the repopulating B cells in PB and SGs. The results indicate that at 24 months after rituximab treatment, B cells aggregating in SGs are T1 B cells. Since transfer experiments have established that T1 B cells give rise to T2 B cells, T2 B cells present in SGs at baseline should originate from T1 B cells. Accordingly, in primary SS patients, B cells should be initiating lymphoid neogenesis (45).

Memory B cells were detected early during repopulation, first in the PB and subsequently in SGs (compare Figure 5B with Figure 7). Later, they represented as few as 1% of circulating B cells. These findings are consistent with those reported by Anolik et al (19), Sanz and Anolik (46), and Rouzière et al (47). These are likely to be autoreactive memory B cells, and their depletion may have been less complete in solid tissues than in the PB. They might recirculate and home preferentially to afflicted organs. This sequence of events resembles the pathophysiology of memory B cell accumulation in the SGs, as proposed by Hansen and colleagues (11). Therefore, it is tempting to increase the number of infusions, rather than the dosage, of rituximab in an attempt to prolong the depletion of memory B cells and thereby delay their entrapment in SGs (48). Alternatively, due to the absence of resident memory B cells after treatment with rituximab (Weil JC: personal communication), new plasmablasts and plasma cells might have competed successfully in the bone marrow (49). Since the characteristic abnormalities in the proportions of PB and SG B cell subsets, and possibly in their functions (21), reappear 24 months after depletion, our data provide strong evidence that the microenvironment contributes to B cell abnormalities in this disease.

In summary, rituximab depletes B cells from the PB and SGs in patients with primary SS. A combination therapy that targets BAFF and CD20 can postpone B cell repopulation in patients with excessive BAFF because plasmablasts are protected from depletion by BAFF (30, 37). However, this picture is still far from complete (50). Rituximab should therefore be administered before irreversible atrophic changes occur in the SGs of patients with primary SS (51).


Drs. Youinou, Pers, and Saraux had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Pers, Youinou.

Acquisition of data. Daridon, Bendaoud, Le Berre, Bordron, Hutin, Renaudineau, Dueymes, Loisel, Berthou.

Analysis and interpretation of data. Pers, Devauchelle, Daridon, Youinou.

Manuscript preparation. Pers, Youinou.

Statistical analysis. Devauchelle, Saraux.


We acknowledge Mrs. Barbara Turpin (Honorary President of the Association Française du Gougerot-Sjögren et des Syndromes Secs) for permanent support. The secretarial help of Cindy Séné and Simone Forest is greatly appreciated. Grateful thanks are also due to Professor Haralampos M. Moutsopoulos (Department of Pathophysiology, National University Medical School, Athens, Greece) for critical reading of the manuscript and to Professor Rizgar A. Mageed (William Harvey Institute, Queen Mary School of Medicine and Dentistry, London, UK) for critical reading and editing of the manuscript.