Naive- and Memory-like CD21^low B Cell Subsets Share Core Phenotypic and Signaling Characteristics in Systemic Autoimmune Disorders

The acute stimulation of the immune system induces a variety of adaptive changes in the composition, activation, and differentiation state of immune cells to properly defend the host against invading pathogens and to provide long-lasting memory. Similarly, chronic stimulation as in the context of chronic infections or other inflammatory conditions drives alterations of the immune system. However, these changes often differ in character and may also contribute to secondary development of autoimmunity, inflammation, or malignant lymphoproliferation. In several diseases associated with chronic immune stimulation, the persistent expansion of a B cell subpopulation with a low expression of CD21 has been identified. This expansion was described in autoimmune disorders like systemic lupus erythematosus (SLE) (1, 2), rheumatoid arthritis (RA) (3), primary Sjögren syndrome (pSS) (4), anti-neutrophil cytoplasm antibodies–associated vasculitis (AAV) (5), hepatitis C virus (HCV)–associated cryoglobulinemia (6, 7), or common variable immunodeficiency (CVID) with autoimmune manifestations (3, 8) encompassing an increased percentage of B cells expressing autoreactive BCR entities. It was also found in chronic infections, including HIV (9), malaria (10), or CMV (11), associated with an increase in pathogen-specific B cells. Although the expansion of autoantigen-specific CD21^low B cells in autoimmune disorders is implying a direct role of these cells in autoimmune pathogenesis, the expansion of pathogen-specific CD21^low B cells in infectious diseases underlines their involvement in regular immune responses against certain pathogens.

The distinct expression profile of CD21^low B cells is associated with the expression of inflammatory-type homing receptors, which favor cell migration to inflamed tissues (2, 7–9, 11). Indeed, CD21^low B cells have been extracted from sites of inflammation, like the synovial fluid of patients with RA or the bronchoalveolar lavage of CVID patients with interstitial lung disease (8). CD21^low B cells show a preactivated phenotype with an increased baseline expression of CD86 and a high basal phosphorylation of proteins of the proximal BCR signaling pathway (2, 4, 7, 12, 13). Inhibitory proteins like FcRL4 and CD32 (FcγRIIB) are upregulated and have been linked to the restricted differentiation and proliferative and functional capacity with impaired Ca²⁺ signaling upon BCR stimulation (3, 4, 6, 8, 9, 14, 15). CD21^low B cells are furthermore characterized by a high expression of CD95 (Fas). Disease-related Ag specificities of BCR are enriched among the CD21^low subpopulation (2–4, 9), indicating an Ag-triggered selection and expansion into this pool. Lately, a high expression of the transcription factor T-bet has been demonstrated for CD21^low B cells in different disorders (16–19).

Previous work has investigated CD21^low B cells mainly in the context of one disease and analyzed the prevalent specific B cell differentiation state associated with this disorder as determined according to the surface expression of the Ig isotype and CD27. Thus, mostly CD27⁻IgG⁺ cells were found in HIV (9) and malaria (10), CD27⁺IgM⁺ in HCV-associated mixed cryoglobulinemia, and more naive-like CD27⁻IgM⁺IgD⁺ B cells in RA, CVID (3), and pSS (4), although there are reports that different CD21^low subpopulations may exist within the same patient (3, 5, 9).

Therefore, we asked the question whether certain autoimmune disorders are associated with the preferential expansion of one differentiation stage of CD21^low B cells and how much the altered phenotype, signaling, and activation overlap between the different CD21^low populations to gain further insights into the core phenotype of CD21^low B cells.

EDTA blood was obtained from 218 patients with RA, connective tissue disease (CTD), or AAV seen at the outpatient clinics of the Department of Rheumatology and Clinical Immunology, Medical Center, University of Freiburg (Freiburg, Germany). Exclusion criteria were a preceding therapy with rituximab, a total dose of >5 g cyclophosphamide, or a current therapy with cyclophosphamide or >10 mg/d prednisone equivalent. Control blood was obtained from 24 healthy donors (HD). The study was approved by the ethics committee of the University of Freiburg (FR66/13), and all study subjects provided written informed consent before inclusion.

The following Abs were used in this study: CD11c Alexa Fluor (AF) 700, CD19 allophycocyanin, CD19 Brilliant Violet (BV) 421, CD19 BV605, CD21 PE/Cy7, CD27 PerCP/Cy5.5, CD38 PerCP/Cy5.5, IgM BV421, λ FITC, and Siglec10 UNLB, all from BioLegend (San Diego, CA); CD19 allophycocyanin-H7, CD21 PE/Cy7, CD27 BV605, CD32 PE, CD38 PE-CF594, CD38 V450, CD69 FITC, CD86 PE, CD95 PE, IgG AF700, ICAM-1 PE, Akt (pS473) AF488, ERK1/2 (pThr202/Y204) AF647, IκBα PE, PLCγ2 (pY759) AF647, SYK (pY352) PE, and S6 (pS235/236) allophycocyanin, all obtained from BD Biosciences (San Jose, CA); CD22 PE (Beckman Coulter, Krefeld, Germany); CD62L FITC (ImmunoTech, Marseille, France); IgD FITC (Southern Biotech, Birmingham, AL); IgG FITC (Dako, Hamburg, Germany); IgM AMCA and IgM Cy-5 (Jackson ImmunoResearch Laboratories, West Grove, PA); λ PE (Leinco Technologies, Fenton, MO); FcRL4 UNLB, a kind gift from M. Cooper (Emory University School of Medicine); secondary anti-mouse IgG AF647 (Life Technologies, Carlsbad, CA); and anti-κ F(ab′)₂ for BCR stimulation was obtained from Southern Biotech.

PBMCs were isolated by density-gradient centrifugation using Ficoll-Paque (Pancoll Human; PAN-Biotech, Aidenbach, Germany), according to standard protocols.

Whole blood staining was performed on 50 μl of EDTA blood using OptiLyse C (Beckman Coulter), following the manufacturer’s instructions, after incubation with mAbs specific for CD19, CD21, CD38, and CD27. Flow cytometric data were acquired on a LSRFortessa equipped with an UV laser (BD Biosciences, Heidelberg, Germany). FACS data were analyzed using the FlowJo software package (version 7.6.5; Tree Star, Ashland, OR). CD21^low B cells were defined as CD19^highCD21^lowCD38^low lymphocytes.

Isolated PBMCs were incubated with unlabeled mouse mAbs against FcRL4 and Siglec10 and, in a second step, marked with fluorochrome-conjugated anti-mouse IgG. After washing fluorochrome-conjugated mAbs specific for CD11c, CD19, CD21, CD22, CD27, CD32, CD38, CD62L, CD95, IgD, IgM, and IgG were added. Cells were analyzed by flow cytometry. B cells were gated into CD21^low (CD19^hiCD21^lowCD38^low) and CD21^pos (CD19⁺CD21^posCD38^low/+) subsets, and for each subset, naive/naive-like (CD27⁻IgM⁺), IgM memory (CD27⁺IgM⁺), switched memory (CD27⁺IgM⁻), IgG memory (CD27⁺IgM⁻IgG⁺), CD27⁻ switched memory (CD27⁻IgM⁻), and CD27⁻ IgG memory (CD27⁻IgM^-IgG⁺) subpopulations were identified (for gating strategy, see Supplemental Fig. 1). For analysis, subsets with <50 events were excluded.

Upon isolation, PBMCs were rested for 2 h at 37°C to minimize nonspecific B cell activation. Cells were stimulated with 15 μg/ml anti-κ Ab at 37°C for 7, 10, and 30 min as indicated and subsequently fixed and permeabilized using Cytofix and Perm Buffer III (BD Biosciences). Cells were labeled with mAbs against CD19, CD21, CD38, CD27, IgM, IgG, λ, SYK(pY352), PLCγ2(pY759), ERK1/2(pThr202/Y204), AKT(pS473), S6(pS235/236), and IκBα. To define the stimulated and the unstimulated population, respectively, cells were differentiated in λ⁻ and λ⁺ cells. Upon gating on CD21^low and CD21^pos cells, naive, IgM memory, IgG memory, and CD27⁻ IgG memory subpopulations were identified by their expression of IgM, IgG, and CD27. For Ca²⁺ mobilization, loading of PBMCs with Indo-1 was performed as described before (15). Cells were stained with Abs against CD19, CD21, CD22, CD27, CD38, IgM, IgG, and λ. After baseline acquisition for 45 s, 8 μg anti-κ was added for BCR stimulation. The following changes in intracellular Ca²⁺ concentrations were recorded for 7 min.

Isolated PBMCs were incubated with 10 μg/ml anti-κ for 36 h. Surface molecules were labeled with fluorochrome-conjugated mAbs specific for CD19, CD21, CD38, CD27, λ, CD86, CD69, and ICAM. To account for potential internalization of BCR, cells were permeabilized (Beckman Coulter IntraPrep Permeabilization Reagent) before staining with anti-IgG and anti-IgM mAbs. Analysis and gating were performed as described above.

Data analysis was performed with GraphPad Prism version 7.0 for Windows (GraphPad, La Jolla, CA) and R (20) with libraries ggplot2 (21) and ggiraphExtra (22). Normal distribution was analyzed using D’Agostino–Pearson normality test followed by paired or unpaired t test, respectively, Wilcoxon matched-pairs test or Mann–Whitney U test for not normally distributed collectives, and Bonferroni–Holm post hoc test for multiple comparisons (Figs. 2–4). Mixed effect models followed by Holm–Sidak multiple comparisons test or Friedman test followed by Dunn multiple comparison test were used for datasets consisting of paired samples (Supplemental Figs. 2–4) and repeated measures–ANOVA followed by Holm–Sidak multiple comparisons test or Kruskal–Wallis test followed by Dunn multiple comparison test for unpaired samples (Fig. 1), respectively. Correlations were measured with Spearman rank correlation coefficient (Fig. 1), and a multiple linear regression model was used for analysis of more than one variable (Fig. 1). The p values <0.05 were considered significant and depicted as *p < 0.05, **p < 0.01, and ***p < 0.001.

We screened 218 patients with different autoimmune disorders for the presence of CD21^low B cells in peripheral blood. In 69 patients (31.7%), we found an increase in the proportion of CD21^low B cells of >7.3% of all B cells (internal reference values: 1.0–7.3%, referring to the 5th and 95th percentile of 53 healthy controls). These patients are referred to as “CD21^low patients.” Over 30% of CD21^low patients were identified among SLE patients (40.7%), seropositive (35.1%) and seronegative (33.3%) RA patients, patients with AAV (33.3%), and with pSS (31.3%) but less frequently in undifferentiated CTD (UCTD) or mixed CTD (MCTD) (21.2%) or systemic sclerosis (SSc) (10.5%) (Table I). There were no significant differences regarding the percentage of CD21^low B cells in CD21^low patients between the different disease entities (Fig. 1A).

Correlations between the proportion of CD21^low B cells and disease features were made for SLE and RA cohorts. We found an inverse correlation (r = −0.413 and −0.294) with the frequency of total B cells in patients with SLE and RA, respectively (Fig. 1B). In an additive multiple linear regression model including the variables age and duration of disease, we further identified an association between frequency of CD21^low B cells and disease duration (coefficient β₁ = 0.666, p = 0.016) in patients with SLE (Fig. 1C). No significant correlations were found between proportion of CD21^low B cells and age, sex, disease activity (SLEDAI and DAS28), inflammatory markers (C3d and CRP), type of autoantibodies (ANA, anti-dsDNA, anti-Ro/La, and rheumatoid factor, anti-CCP) in SLE and RA, respectively, or organ manifestations (arthritis, skin, neurologic manifestation, renal involvement, hematologic/immunologic manifestation, serositis, secondary Sjögren syndrome) in SLE (data not shown).

Next, the composition of distinct subpopulations within the CD21^low population defined by the expression of CD27, IgM, and IgG was determined for 29 autoimmune CD21^low patients (15 patients diagnosed with RA, 7 with SLE, 3 with other CTDs, and 4 with AAV) as well as for 24 HD (for gating strategy, see Supplemental Fig. 1 and 2Materials and Methods). All patients had CD21^low B cells belonging to all four subpopulations, however, with a quite heterogeneous distribution (Fig. 1D). In most patients, CD21^low B cells consisted predominantly of naive-like or CD27⁻ switched memory B cells. Comparison of the mean percentages of the different subpopulations of the CD21^pos B cell population of patients with RA or SLE and HD demonstrated the previously reported reduction of naive B cells in SLE (23) and RA (Fig. 1D). In contrast, in the CD21^low population of SLE and RA patients, the CD27⁻ switched memory compartment was expanded, whereas the percentage of naive B cells was comparable between CD21^low and CD21^pos B cells (Fig. 1F, 1G). There was no detectable difference in the distribution of CD21^low B cell subpopulations between patients below and above 60 y of age (data not shown).

To compare characteristic phenotypic features of CD21^low B cells, the expression of CD19, CD11c, Siglec10, CD32, FcRL4, CD95, and CD62L was assessed for each CD21^low subpopulation and their CD21^pos counterparts in these 29 autoimmune CD21^low patients and 24 HD. The increased expression of CD19, CD11c, Siglec10, CD32, FcRL4, and CD95 as well as decreased expression of CD62L compared with the CD21^pos counterparts was common to all subpopulations in the analyzed disease conditions (Fig. 2). Interestingly, patient-derived CD21^pos B cells often show an intermediate phenotype between CD21^low B cells of patients and CD21^pos B cells of HD. The few CD21^low B cells of HD displayed similar phenotypic changes like patient-derived CD21^low B cells when compared with the respective CD21^pos population (Supplemental Fig. 2A and data not shown).

The mean fluorescence intensity (MFI) of the respective markers, however, differed significantly between the four CD21^low subpopulations (Supplemental Fig. 3). To determine if these differences just reflect the underlying differentiation stage of the B cell subpopulation, we further analyzed the expression patterns of surface markers within the respective CD21^low and CD21^pos B cell subsets of patients and HD (Supplemental Figs. 2, 3, as CD21^pos B cells from patients and HD showed largely similar patterns, we did not include patient’s CD21^pos B cells in these figures for reasons of clarity). The expression pattern of CD19 (high expression on IgM memory cells), CD11c (high expression on switched memory cells), CD95 (high expression on memory and especially CD27⁺ IgG memory cells), and CD62L (high expression on CD27⁺ IgG memory cells) was largely similar between patients’ CD21^low B cell subsets and CD21^pos subpopulations. In contrast, the inhibitory receptors FcRL4, Siglec10, and CD32 were differentially expressed (Supplemental Fig. 3). The increased expression of these markers on CD21^pos IgM memory cells when compared with the other CD21^pos subpopulations is not reflected among CD21^low subpopulations, and the pattern varies for the different receptors (Supplemental Fig. 3). CD32 and Siglec10 are expressed the highest on naive CD21^low B cells, whereas FcRL4 expression is higher on switched memory B cells. Very similar expression patterns were observed for CD21^low subpopulations of HD (Supplemental Fig. 2A and data not shown).

Several studies suggested elevated levels of basal phosphorylation of signaling molecules downstream of the BCR and impaired signaling following BCR engagement in CD21^low B cells (2, 13–15). To address key signaling pathways in the different CD21^low B cell subpopulations, we stimulated PBMCs of CD21^low patients and HD with anti-κ and stained for λ to distinguish between stimulated (λ⁻) and unstimulated (λ⁺) B cells (for gating strategy, see Fig. 3A).

Analysis of unstimulated (λ⁺) B cell subsets revealed elevated basal phosphorylation levels of the early signaling molecules SYK and PLCγ2 in all CD21^low B cell subpopulations compared with CD21^pos B cells of patients and HD (Fig. 3B, 3C). This was also true for ERK phosphorylation with the exception of the IgM memory B cells, whereas for AKT and S6, this was only seen in the naive populations (Fig. 3C). Protein levels of IκBα in CD21^low B cells were relatively comparable to CD21^pos B cells, with a slight tendency to lower expression (Fig. 3C). These changes were consistent for all investigated patient groups, and the observed differences between CD21^low and CD21^pos B cells were also seen in HD-derived cells (Supplemental Fig. 2B).

When comparing the phosphorylation pattern driven by B cell differentiation, in CD21^pos B cells, basal phosphorylation was significantly increased for all investigated molecules in IgM memory populations, whereas this pattern was not replicated in CD21^low B cells (Supplemental Fig. 4A).

To describe the activation capacity of the different B cell populations, we calculated the stimulation index (SI) as the ratio of the MFI of the respective phosphoproteins of λ⁻ and λ⁺ B cells after stimulation with anti-κ. Similarly, the degradation of IκBα was assessed as the ratio of the MFI of IκBα in λ⁻ and λ⁺ B cells. For Ca²⁺ signaling strength, the mean peak of the response curve was compared.

All CD21^low subpopulations independent of the underlying disease were capable of increasing the phosphorylation of signaling molecules SYK, PLCγ, ERK, and AKT (Fig. 4A, 4B), demonstrating their general BCR-induced signaling potential (Fig. 4A, 4B, Supplemental Fig. 4B). Because of the high basal phosphorylation levels in nonstimulated cells, the SI of SYK and PLCγ was significantly lower in CD21^low B cells compared with the CD21^pos compartment (Fig. 4B), indicating a preserved, however, reduced signaling capacity of CD21^low B cells. This disturbed activation of the proximal signaling cascade was commonly associated with an abrogated canonical NF-κB response (Fig. 4B) and poor increase of the intracellular Ca²⁺ concentration (Fig. 4C, 4D). Unlike the phosphorylation response of proximal signaling components, ERK phosphorylation after BCR stimulation was preserved in naive-like and IgM memory CD21^low B cells but not the IgG memory compartment. Although there was not a clearly altered pattern for AKT phosphorylation, S6 phosphorylation was severely decreased in all but the IgM memory CD21^low populations when compared with HD and NS to patient-derived CD21^pos B cells. Interestingly, for both signaling molecules of the AKT–PI3K–mTOR–S6 pathway, the CD21^pos compartments of patients showed a clearly impaired phosphorylation compared with HD, with an exception for IgM memory B cells (Fig. 4B). The alterations seen in CD21^low B cells of HD resembled the ones described for patient-derived CD21^low B cells (Supplemental Fig. 2C and data not shown).

Comparison among the respective CD21^pos and CD21^low B cell subpopulations revealed differentiation-dependent differences regarding the phosphorylation capacity (Supplemental Fig. 4B, 4C). In the CD21^pos compartment, naive and IgM memory B cells displayed significantly increased phosphorylation of the proximal signaling molecules SYK and PLCγ2 upon BCR stimulation compared with the other subpopulations, whereas phosphorylation of AKT and ERK was significantly higher in IgG memory B cells. IgM memory cells showed only a little increase of phosphorylation of S6 and degradation of IκBα. This signaling pattern between the different naive and memory B cell populations was similar in CD21^low B cells for SYK and PLCγ2 but different for all other molecules (Supplemental Fig. 4B, 4C). In general, in CD21^low B cells, the lowest response was detected in CD27⁻ IgG memory cells; the severe deficiency of S6 phosphorylation and degradation of IκBα, however, was common for all CD21^low B cell populations. This altered pattern was also seen in CD21^low B cells of HD (Supplemental Fig. 2C and data not shown).

Before the initiation of differentiation and induction of effector functions, an immediate consequence of the BCR-induced signaling cascade is the upregulation of activation markers, among these the costimulatory molecule CD86 and the adhesion molecule CD69.

Expression of CD86 on unstimulated (λ⁺) cells was increased in all CD21^low subpopulations compared with the respective CD21^pos subset of HD (Fig. 5). Upon stimulation, expression was increased on all subpopulations, and only CD21^pos and CD21^low IgM memory B cells failed to upregulate CD86 on their surface (Fig. 5, Supplemental Fig. 4D). The highest levels of CD86 were consistently detected on stimulated CD21^low B cells. In contrast, there was no difference in the expression of CD69 between unstimulated CD21^pos and CD21^low B cells. CD69 was upregulated on nearly all CD21^pos cells after BCR stimulation, but CD21^pos IgM memory B cells and all CD21^low subpopulations failed to upregulate CD69 (Fig. 5B, Supplemental Fig. 4D). Similar patterns for CD86 and CD69 were observed among CD21^low B cells of HD (Supplemental Fig. 2D and data not shown).

An increased proportion of CD21^low B cells has been identified in a variety of autoimmune disorders, including CTDs like SLE and pSS, RA, and AAV (3–5, 24), but data comparing number, phenotype, and function between the different autoimmune disorders are scarce. Culton et al. (5) described an expansion of CD19^hi cells, which are strongly overlapping with CD21^low B cells, in 34% of patients with SLE and 25% of patients with AAV. Our data reflect Culton et al. (5) finding with the higher percentage of CD21^low patients among patients with SLE when compared with AAV, RA, or pSS. In SSc, however, we observed a much lower prevalence of CD21^low patients, which is in contrast to data by Rubtsov et al. (24) reporting an increase of CD11c⁺ B cells in more than 30% of patients. Given the restriction of CD11c expression on B cells to the CD21^low B cell compartment, this inconsistent finding remains unexplained.

Expansion of CD21^low B cells correlates with B cell lymphopenia, which is a common finding in patients with SLE (1, 23) and also observed in patients with RA (25). This indicates that increased proportions of CD21^low B cells may partly reflect a reduction of CD21^pos B cells in some of the patients. A positive correlation of CD21^low B cells with duration of disease in SLE independent of age corroborates former data of our group (1) and is in line with the hypothesis of chronic immune stimulation causing an expansion of CD21^low B cells. The previously reported association of CD21^low B cells with disease activity (16, 26, 27) was not observed in our study. This is probably due to our patient selection excluding patients with therapies known to have a strong impact on the B cell compartment and thereby reducing the inclusion of active patients.

Although the first descriptions of CD21^low B cells did not differentiate between B cell developmental stages (1, 9, 28), later, typically only the most prominent CD21^low B cell subpopulations were described for the different diseases (e.g., naive-like in RA and CVID, nonswitched memory like in HCV-associated cryoglobulinemia, or switched memory like in malaria) (3, 6, 8–10). The co-occurrence of different developmental stages of CD21^low B cells was first recognized in HIV patients, in which CD27⁺CD21^low B cells “activated memory B cells” have been differentiated from CD27⁻CD21^low tissue-like memory B cells (9, 29), and more recently, also naive like and nonswitched memory CD21^low B cells were reported (30). In acute SLE, recent reports distinguished populations referring to our naive like and CD27⁻IgG/IgA CD21^low B cells (16, 31). Our data demonstrate for the first time, to our knowledge, that in healthy people and autoimmune diseases, the CD21^low compartment of every individual person comprises all four developmental stages and that the predominant populations differ between patients. Although naive and CD27⁻ switched memory B cells typically make up the largest portions of the CD21^low compartment, CD21^low CD27⁺IgM⁺ and class-switched memory B cells co-occur in the same patient.

Although a few reports have already demonstrated some shared features between different CD21^low populations in CMV and HIV (11, 30), we provide the first systematic report, to our knowledge, that all four CD21^low B cell subsets share a common core of phenotypic markers in HD and across different autoimmune diseases. This includes the elevated expression of CD19, by definition low expression of CD21 and CD38, the expression of CD11c, and increased expression of inhibitory receptors like Siglec10, FcRL4, and FcγRIIB (CD32). In SLE, a previous comparison of “activated naive B cells,” representing naive-like CD21^low B cells, and double-negative “DN2” cells, representing a subset strongly overlapping with CD21^lowCD27⁻IgM^- B cells, demonstrated, besides similarities, some differences (16). This finding is in agreement with our notion that some features of the different CD21^low subsets can be assigned to the underlying differentiation stage, like the expression of transcription factors TRAF5, ZEB2, and BACH2, which matched the expression pattern in their CD21^pos counterparts.

The similar expression patterns for CD19, CD11c, CD95, and CD62L when comparing the respective CD21^low B cell subsets of patients and CD21^pos B cell populations of HD suggest a prevalent differentiation-specific regulation of the expression of these surface molecules. In contrast, the increased expression of the inhibitory receptors FcRL4, CD32, and Siglec10 is mainly controlled by activation rather than differentiation of the cell because it is unique on IgM memory cells among CD21^pos B cells but shared between all CD21^low B cell populations. The activated status of CD21^low B cells, as well as IgM memory B cells, is corroborated by the increased basal phosphorylation of signaling molecules like SYK and PLCγ2 reported by us and others (2, 13, 14). This shared feature indicates constitutively active BCR signaling of CD21^low subpopulations independent of the differentiation state, possibly by chronic/repetitive engagement with (auto-)antigens or alternatively by constitutive rewiring of signaling networks.

This footprint is also associated with impaired signaling upon BCR restimulation in vitro as previously described by us and others for single CD21^low subpopulations in patients with CVID, SLE, HIV, malaria, and HCV-associated cryoglobulinemia (2, 6, 7, 14, 15, 32). In this study, we demonstrate that reduced BCR-dependent signaling concerning SYK, PLCγ2, Ca²⁺, and especially canonical NF-κB is also common to all CD21^low subpopulations when compared with CD21^pos cells of the same patient or HD. In contrast, ERK phosphorylation upon BCR stimulation is preserved and comparable between naive and CD27⁺ memory CD21^low B cells, although it is mainly seen in memory CD21^pos B cells (33). This is remarkable because in contrast to other BCR-induced signaling pathways, which are commonly downregulated by high Siglec expression, increased proapoptotic ERK signaling has been described after engagement of Siglec molecules on naive B cells (34), indicating that preserved ERK signaling in naive CD21^low B cells may contribute to the increased apoptosis of these cells in vitro (4, 8, 35). The increased expression of inhibitory receptors like Siglecs and FcRL4 is thought to act as a regulatory feedback mechanism on B cell activation in CD21^low B cells (36) because knockdown of certain inhibitory receptors partially restored B cell function in HIV-derived tissue-like memory B cells (37). This, however, may not be the complete explanation of the impaired BCR signaling of CD21^low B cell populations because CD21^pos IgM memory B cells also exhibit increased constitutive phosphorylation of key signaling molecules and higher levels of inhibitory receptors but respond to BCR stimulation. Unlike all other signaling pathways, phosphorylation of AKT and S6 is reduced in both CD21^pos and CD21^low populations from patients when compared with CD21^pos B cells from HD, indicating a disease-associated alteration of the PI3K signaling pathway. The general alteration of this pathway in CD21^low patients requires further investigations.

In summary, we show that a substantial proportion of patients with different autoimmune diseases present with an expansion of CD21^low B cells comprising various proportions of naive-like, IgM memory, CD27⁺, and CD27⁻ class-switched memory subpopulations. The unique core phenotype of all CD21^low B cell subsets in healthy controls and patients suggests either a common environmental imprint during their differentiation from naive and memory CD21^pos precursor cells as recently postulated for a Th1-driven inflammatory environment (17) or a common ancestor cell that subsequently differentiates into the four CD21^low subpopulations. This theory was supported by the discovery of related expanded BCR clones between activated naive and DN2 cells and the possibility to differentiate both in vitro from resting naive cells and DN2 from activated naive B cells (16, 38).

Given the increasing knowledge about the role of BCR, TLR, IL-21, and IFN-γ signaling in the development of CD21^low B cells in vitro and in vivo (16, 17, 19, 26, 39), their capacity to serve as Ag presentation cells because of preserved upregulation of CD86 expression, and the enriched autoreactive BCR repertoire among CD21^low B cells, it is tempting to speculate that the expansion of CD21^low B cells in patients with different autoimmune diseases may point toward shared, pathogenetically relevant mechanisms between these diseases, the clarification of which may allow for targeted treatment options.

We thank our patients and HD for collaboration and the nurses and physicians of the outpatient clinics of the Department of Rheumatology and Clinical Immunology for patient care. We thank Dr. Erika Graf and Dominik Stelzer from the Institute of Medical Biometry and Statistics, University of Freiburg for statistical advice.

The authors have no financial conflicts of interest.