The Antigen Presenting Potential of CD21low B Cells

Human CD21low B cells are expanded in autoimmune (AI) diseases and display a unique phenotype with high expression of co-stimulatory molecules, compatible with a potential role as antigen-presenting cells (APCs). Thus, we addressed the co-stimulatory capacity of naïve-like, IgM-memory, switched memory and CD27negIgDneg memory CD21low B cells in allogenic co-cultures with CD4 T cells. CD21low B cells of patients with AI disorders expressed high levels of not only CD86, CD80, and HLA-DR (memory B cells) but also PD-L1 ex vivo and efficiently co-stimulated CD4 T cells of healthy donors (HD), as measured by upregulation of CD25, CD69, inducible co-stimulator (ICOS), and programmed cell death protein 1 (PD-1) and induction of cytokines. While the co-stimulatory capacity of the different CD21low B-cell populations was over all comparable to CD21pos counterparts of patients and HD, especially switched memory CD21low B cells lacked the increased capacity of CD21pos switched memory B-cells to induce high expression of ICOS, IL-2, IL-10, and IFN-γ. Acknowledging the limitation of the in vitro setting, CD21low B cells do not seem to preferentially support a specific Th effector response. In summary, our data implies that CD21low B cells of patients with AI diseases can become competent APCs and may, when enriched for autoreactive B-cell receptors (BCR), potentially contribute to AI reactions as cognate interaction partners of autoreactive T cells at sites of inflammation.


INTRODUCTION
Beside the production of antibodies, B cells contribute crucially to the adaptive immune response as potent antigen-presenting cells (APCs) (1). After endocytosis via the B-cell receptor (BCR) and processing of an antigen, B cells present antigenic peptides by MHC II molecules to CD4 T cells. This cognate interaction is strongly enhanced by activation-induced expression of co-stimulatory molecules like CD80 and CD86 on the surface of the antigenpresenting B cell binding to CD28 and other molecules involved in T-cell-B-cell (T-B) interaction. Therefore, memory B cells already expressing higher levels of these molecules are better APCs (2,3) and in the absence of both molecules the activation of T cells is strongly impaired (4).
In the last 20 years, an accumulation of a circulating CD21 low Bcell population has been described in the context of different disease entities associated with chronic immune stimulation as in viral [human immunodeficiency virus (HIV) (5) or hepatitis C virus (HCV) (6)] or parasite infection [malaria (7)], in patients with immune dysregulation in common variable immunodeficiency (CVID) (8,9), in graft versus host disease (10), or in autoimmune (AI) disorders like systemic lupus erythematosus (SLE) (11) or rheumatoid arthritis (RA) (12). We recently demonstrated that an accumulation of CD21 low B cells was most frequently observed in SLE patients, followed by RA and primary Sjögren Syndrome (pSS) but less frequently in undifferentiated or mixed connective tissue disease (UCTD/MCTD) or systemic sclerosis (SSc) (13). The different naïve-like (IgD pos CD27 neg ) or memory (IgD pos CD27 pos , IgD neg CD27 pos , and IgD neg CD27 neg ) CD21 low B-cell populations display a common core phenotype and share altered signaling characteristics independent of the underlying autoimmune disorder (13), some of which have been previously described as activated naïve (14), atypical memory (15), or tissue-like-memory (16) B cells. Their high expression of activation markers and costimulatory ligands for T-cell help, such as CD80 and CD86 (12,(16)(17)(18)(19), distinctly discriminates them from their CD21 pos counterparts as potentially potent APCs for T cells (20,21). This is of special interest since these CD21 low B-cell populations contain increased proportions of antigen-specific clones in chronic infection (22,23) and of autoreactive clones in AI diseases (12,14,17,24). Furthermore, several studies indicated that B cells play a prominent role as APCs in the induction of autoimmunity [ (25,26) and reviewed in (1)]. Thus, given the conceivable pathological impact of increased co-stimulatory capacities, we addressed the co-stimulatory potential of the different CD21 low B-cell subsets in the context of AI disease in an allogenic superantigen-driven lymphocyte reaction.

Patients
All experiments were performed with ethical approval by local authorities (Freiburg 239/1999 and 121/11 and Freiburg 66/13) according to the declaration of Helsinki. All patients and healthy donors (HD) had signed the informed consent.
In total 29 patients were included in the study at the outpatient clinic of the Department of Rheumatology and Clinical Immunology, University Medical Center Freiburg. 13 patients were diagnosed with RA, 8 with SLE, 2 with psoriatic arthritis, and 1 each with eosinophilic granulomatosis with polyangiitis (EGPA), pSS, antisynthetase-syndrome, sarcoidosis, spondyloarthropathy, and antiphospholipid syndrome. The patient cohort included 7 male and 22 female individuals with a mean age of 61.4 years (SD: +/-13.8 years). Patient and HD cohort were not matched for age and gender. Disease duration ranged from 2 to 56 years. Creactive protein (CRP) of patients ranged from <3 to 25.5 mg/l, with one outlier of 96.2 mg/l. Disease Activity Score 28 (DAS-28) of patients with RA ranged from 1.8 to 4, including 10 patients in remission (DAS-28 < 2.6) and 2 patients with low disease activity (DAS-28 < 3.2). Systemic lupus erythematosus disease activity index (SLEDAI) of patients with SLE ranged from 0 to 4, with the exception of one patient with active disease (SLEDAI of 8). Treatment of patients included no more than 8.75 mg prednisolone/day. Some patients were currently treated with monoclonal antibodies against TNF-a (n = 4) or IL-6R (n = 1). Other immunosuppressive therapies included methotrexate, hydroxychloroquine, leflunomide, azathioprine, and mycophenolatemofetil. Patients' characteristics and percentage of CD21 low B cells are summarized in Table 1.

Cell Staining and Isolation
Peripheral blood mononuclear cells (PBMCs) were isolated from EDTA blood by Ficoll density centrifugation following standard protocols. Surface staining on T-B-co-cultures, fresh or frozen PBMCs was performed with optimal amounts of the respective antibodies at 4°C for 15 min. Sorting of B-cell subpopulations of patients and HD was performed on freshly isolated PBMCs within 4 h after blood withdrawal. As indicated in Figure 1A CD19 pos B cells were divided into naïve (CD27 neg IgD pos ), IgM memory (CD27 pos IgD pos ), switched memory (CD27 pos IgD neg ), and CD27 neg IgD neg memory B cells. CD21 pos and CD21 low subsets were distinguished from each of these subpopulations by subsequent gating on CD21 low CD38 low and CD21 pos CD38 dim B cells. B-cell sorting of four to six of the eight potential subpopulations within each sample was conducted with the MoFlo Astrios (Beckman Coulter) or the Aria Cellsorter (BD). The sorted subsets were chosen according to the proportion of the respective populations. For information on sorted subpopulations please refer to Supplementary Table 1.
CD4 T cells were isolated from HD by negative selection using the Magnetic Cell-Sorting (MACS) CD4+ T-cell isolation (C) Graphs show the MFI of CD80, CD86 and HLA-DR in CD21 pos B-cell subpopulations of healthy donors (HD) and CD21 pos and CD21 low B-cell subsets of patients with autoimmune (AI) diseases in naïve, IgM mem, sw mem and CD27 neg IgD neg mem B cells. Statistical differences were considered significant at *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. RA, SLE, sarcoidosis, psoriatic arthritis, and spondyloarthropathy were marked by different symbols as indicated in the legend. ns, not significant. kit of Miltenyi Biotec following the manufacturer's instructions. The purity of isolated T cells was over 95% as determined by flow cytometry. For each experiment, B-cell subpopulations of one patient and one HD (B-cell HD) were matched with CD4 T cells of another HD (T-cell HD). DAPI was added before acquisition of data for exclusion of dead cells when necessary. To determine intracellular cytokines, T and B cells were co-cultured as described below. 10 µg/ml BFA (Sigma-Aldrich) was added 6 h prior to harvest. Cells were fixed and permeabilized with Cytofix/ Cytoperm and Perm/Wash Buffer (both BD Biosciences) following the manufacturer's instructions, and samples were stained with the respective antibodies at optimal concentrations. Flow cytometric data was acquired at the LSR Fortessa (BD Biosciences).

Mixed Lymphocyte Culture and Cell Activation
For T-B-cell co-cultures, 12,500 sorted B cells of patients and HD were cultivated either with or without 100 ng/ml of Staphylococcus enterotoxin B (SEB) (Sigma-Aldrich) in IMDM (Life technologies GmbH) containing 10% FCS (Biochrom) and 100 U/ml of penicillin/streptomycin (Life technologies GmbH). 18 h later allogenic CD4 T cells of a third-party T-cell HD (37,500 cells/well) were added and co-cultured for 24 h. After 24 h, supernatants were collected and stored at -80°C for subsequent analysis. As a positive control for activation, B cells were cultured with 2.5 µg/ml CpG alone for 42 h. Cells were washed and stained for flow cytometric analysis.

Measurement of Cytokine Production
Cytokines were measured from the supernatants of co-cultures and SEB-stimulated T-cell cultures as described above. Measurements were conducted by a bead-based immunoassay (LEGENDplex ™ Multi-Analyte Flow Assay Kit, Human T h Cytokine Mix and Match Subpanel, BioLegend) according to manufacturer's instructions and analyzed using the LSR Fortessa and the LEGENDplex software.

Statistical Analysis
The data was analyzed using GraphPad Prism 7.0. Data were tested for normal distribution. For comparison of three groups, statistical differences were determined by parametric one-way ANOVA or non-parametric Kruskal-Wallis testing (both unpaired) using Tukey's or Dunn's post-hoc test for multiple comparison. In case of two groups, unpaired t-test was applied. Differences were considered statistically significant at *p < 0.05, **p < 0.01, ***p < 0.001, or ****p < 0.0001. Data are shown as means with standard deviation (SD).

CD21 low B Cells Express High Levels of Co-Stimulatory Molecules Ex Vivo
CD21 low B cells defined by low expression of CD21 and CD38 constitute a common B-cell subset, comprising four stages of B-cell differentiation of naïve-like, non-switched, switched and CD27 neg IgD neg memory in analogy to their CD21 pos counterparts ( Figure 1A) (13). According to literature (6, 12, 13, 15-19, 24, 27, 28), all four subsets express common markers of CD21 low B cells as increased expression of CD11c, FCRL5, CD95 and ICAM-1, reduced expression of CD40, and are negative for CXCR5 ( Figure 1B and Supplementary Figure  1A). Concerning the co-stimulatory potential of CD21 low B cells, flow cytometric analysis revealed that the expression of the costimulatory molecules CD80 and CD86 was increased on all CD21 low B-cell populations when compared to the respective CD21 pos population of HD ( Figure 1C). Comparing between the four CD21 low B-cell populations, CD80 and CD86 were highest on the CD27 pos B-cell compartments, and the lowest expression was observed on naïve-like CD21 low B cells, which was similar to the pattern observed on CD21 pos B-cell populations. In contrast, HLA-DR expression was higher on naïve vs. memory CD21 pos B cells, while this relation was reversed among CD21 low B-cell populations, displaying significantly increased expression on both CD21 low switched memory B-cell compartments when compared to the respective CD21 pos counterparts ( Figure 1C). Expression of the co-stimulatory molecules ICOS-L and OX40-L was equally expressed on CD21 low and respective CD21 pos B-cell subsets except for significantly increased expression of ICOS-L on CD27 neg IgD neg memory CD21 low B cells (Supplementary Figure 1B). Additionally, we observed an increased expression of inhibitory programmed cell death ligand 1 (PD-L1) on CD21 low B cells reaching significance for CD27 neg B-cell subsets and comparable or slightly increased levels of PD-L2, which was only significant for switched memory CD21 low B cells (Supplementary Figure 1C). Thus, all CD21 low B-cell subsets share a distinct, activated phenotype prone for co-stimulation of T cells at sites of inflammation.

Induction of Co-Stimulatory Molecules On CD21 low B Cells In Vitro
To evaluate the co-stimulatory capacity, the four different subsets of CD21 low B cells and the respective CD21 pos populations of patients with AI disorders and HD were co-cultured with allogenic CD4 T cells of third-party healthy individuals in the presence of SEB. In the absence of T cells, CD80 and CD86 expression on SEB-stimulated B cells was comparable to unstimulated B cells (Figure 2A). Similarly, co-cultures of B cells and allogenic T cells in the absence of SEB did not induce an increased expression. However, after addition of SEB the expression of CD80 and CD86 was comparable to B-cell activation with CpG ( Figure 2A).
Unlike ex vivo, in T-B co-cultures without SEB ("unstimulated co-cultures") CD80 and CD86 expression on CD21 low B cells were only higher on naïve-like and nonswitched memory B cells, but not anymore on switched memory B cells when compared to their CD21 pos counterparts (Supplementary Figure 2A). Upon SEB-stimulation, CD80 and CD86 were similarly expressed on most CD21 pos and CD21 low Bcell subpopulations and significantly increased expression of both markers was exclusively observed on naïve-like CD21 low B cells ( Figure 2B). This was reflected by low or even significantly reduced further upregulation of the already elevated CD80 and CD86 expression on all CD21 low B-cell subpopulations (Supplementary Figure 2B). Thus, in the context of SEB-activated co-cultures, only naïve-like CD21 low B cells showed still increased expression of CD80 and CD86 compared to the respective CD21 pos population, while the induced expression of the co-stimulatory molecules on memory CD21 pos and CD21 low B cells was comparable.

Proficient Co-Stimulatory Potential of CD21 low B Cell Subsets
To evaluate the co-stimulatory capacity of B cell subpopulations in allogenic co-cultures, the upregulation of activation-associated molecules on third-party CD4 T cells was measured after 24 h. The short culture time was chosen due to limited survival time of CD21 low B cells in vitro (12,29). For analysis, the response of naïve and memory T cells was differentiated by gating on CD45RA pos and CD45RA neg T cells.
In unstimulated co-cultures of allogenic T cells and the different isolated B-cell subpopulations, we could not detect a significant upregulation of activation markers as shown for CD69 on naïve ( Figure 3A) or memory (data not shown) CD4 T cells. In T-B co-cultures with SEB, we observed a significant upregulation of CD69 and CD25 on naïve CD4 T cells when compared to unstimulated or SEB stimulated T-cell cultures without B cells ( Figures 3B, C), with one exception of CD25upregulation in co-cultures with CD21 pos CD27 neg IgD neg memory B cells of AI patients. Co-cultures with CD21 low B-cell subpopulations induced a comparable upregulation of CD69 and CD25 on naïve CD4 T cells compared to co-cultures with the respective CD21 pos population of patients or HD ( Figures  3B, C). Despite differences in CD86 and CD80 expression between naïve and memory B-cell populations, there were no detectable differences in their capacity to activate T cells, independent of whether CD21 pos or CD21 low B cells were the APCs. This was also true for the induction of ICOS for most of the CD21 pos and CD21 low B-cell subpopulations ( Figure 3D), but not for switched CD21 low B cells, which induced significantly less ICOS compared to switched CD21 pos B cells of HD ( Figure 3D).
On memory CD4 T cells the induction of CD69 was reduced compared to the naïve CD4 subset (p < 0.01, Figures 3B, 4A) and no significant differences were observed compared to SEB alone. In contrast, the expression of CD25 was higher on memory CD4 T cells after co-cultures with all the different B-cell subpopulations (p < 0.01, Figures 3C, 4B). However unlike for naïve CD4 T cells, the expression of CD25 on memory CD4 T cells was lower in cultures containing naïve-like and IgM memory CD21 low B cells of patients compared to co-cultures  with the respective CD21 pos B cells of HD ( Figure 4B). This is especially noteworthy since expression of both CD80 and CD86 was higher on naïve-like CD21 low B cells in these co-cultures. The expression of ICOS was significantly increased after coculture of T cells with IgM memory and class switched CD21 pos B-cell populations in the presence of SEB compared to T cells stimulated with SEB alone, but no other sorted subpopulation ( Figure 4C). Similar to the finding in naïve CD4 T cells, the induction of ICOS on memory CD4 T cells by class switched CD21 1ow B cells was significantly lower compared to the stimulation by their CD21 pos counterpart in HD.
During T-B co-cultures not only co-stimulatory but also inhibitory receptor-ligand pairs are upregulated. We therefore analyzed the expression of PD-1 which was absent after SEBstimulated co-cultures on naïve CD4 T cells (data not shown) but up-regulated on memory CD4 T cells ( Figure 4D). In order to determine the activation induced PD-1 expression we subtracted the baseline expression of PD-1 on unstimulated T cells. Significant induction of PD-1 compared to T cells stimulated with SEB alone was observed for all co-cultures with HD-derived CD21 pos populations except for IgD neg CD27 neg B cells, where the number of samples was too low ( Figure 4D). Similarly, all CD21 low B cells induced higher PD-1 expression but due to small sample size only the experiments with naive like CD21 low B cells reached significance. Although the experiments could not demonstrate a significant difference in the PD-1 induction between different B-cell populations, there was a trend to a higher induction by naïve CD21 pos B cells when compared to their CD21 low counterparts as well as to CD21 pos switched memory B cells.
T cells response to stimulation by proliferation. Due to the limitations in the co-culture period this could not be addressed by CFSE staining or thymidine incorporation assays and staining for Ki67 was not conclusive (data not shown).
In summary, all CD21 pos B cell subsets have similar capacities to induce the investigated activation markers on naïve and memory CD4 T cells with the exception of ICOS induction, which was much stronger after stimulation by switched memory T cells cultured alone and with or without stimulation with SEB were obtained as control. Subsets marked in green (*) were significantly different to the SEB-stimulated T-cell control (p < 0.05). All subsets were statistically significant to unstimulated T cells. Differences were considered significant at *p < 0.05 * and **p < 0.01. The different diseases were marked by different symbols as indicated in the legend.

Co-Stimulation by CD21 low B Cells Does Not Distinctively Bias the T Cell Cytokine Profile
T cells substantially shape the immune response by their cytokine profiles. We addressed if co-cultures with CD21 low B cells bias T cells toward a distinct cytokine profile. Thus, IFN-g, TNF-a, IL-2, IL-4, IL-10, IL-21, IL-17A, and IL-17F production was determined in supernatants of co-cultures after 24 h. Production of IL-17 and IL-21 was not detectable under these conditions (data not shown).
The production of TNF-a and IL-4 did not significantly increase after co-cultures with CD21 low and CD21 pos B-cell subpopulations over the stimulation of T cells with SEB alone ( Figure 5A). IL-2 production was significantly induced only by co-cultures with switched memory B cells which was, however, not the case for CD21 low B cells. For IL-10 all co-cultures with CD21 pos B cellpopulations of HD were able to induce higher levels than the culture of T cells with SEB alone, reaching the highest levels in cultures with switched memory B cells which was not the case for CD21 low B cells, again inducing significantly lower cytokine levels than their CD21 pos counterparts. IL-10 was produced by T cells and not B cells in the cultures as demonstrated by flow cytometric analysis ( Figure 5B). Otherwise we could not detect relevant differences between most CD21 pos and CD21 low populations,  although one has to remark that only few co-cultures reached cytokine levels above the respective SEB-control. In HD-derived samples, switched memory B cells generally induced significantly higher levels of IL-2 (p < 0.001), and IFN-g (p < 0.01) and tended to higher levels of IL-10 (p = 0.07) than co-cultures with naïve B cells. This pattern was also seen for CD21 pos B cells of patients [p < 0.001 (IL-2), p < 0.01 (IFN-g), and p < 0.01 (IL-10)], but not for CD21 low B-cell populations.
The polarization of cultured naïve T cells could not be addressed due to the limited co-culture period. Therefore we analyzed the expansion of the different Th subpopulations according to their CXCR5 pos surface expression of CXCR5, CCR6 and CXCR3, allowing the differentiation of Th1, Th1/17, Th17, Th2 cells and circulating T follicular helper cells (cTfh) 1, cTfh1/17, cTfh17 and cTfh2 CD4 T cells (30) (Supplementary Figure 4A). In naïve CD4 T cells, non-cTfh cells (CXC5R neg CD45RA neg ) or cTfh cells (CXCR5posCD45RA neg ) the proportions of CXCR3 pos CCR6 neg , CXCR3 pos CCR6 pos , CXCR3 neg CCR6 pos , or CXCR3 neg CCR6 neg T cells were not significantly altered compared to unstimulated T cells after co-culture with either CD21 low B-cell subpopulations of patients or the CD21 pos counterparts of HD corroborating the findings of cytokine profiles (Supplementary Figure 4B).
In summary, in these short-term co-cultures, we could not detect a clearly altered cytokine profile or preferentially polarized Th-cell expansion when co-cultured with CD21 low B cells except switched memory CD21 low B cells, which seem to be less efficient to induce the often superior secretion of cytokines by T cells seen in cultures with switched memory CD21 pos B cells.

DISCUSSION
Because of the high base line surface expression of co-stimulatory markers it has been suggested that CD21 low B cells are potent APCs, contributing to the immune response by driving the activation and cytokine release of CD4 T cells (20,31). Thus, we addressed the co-stimulatory capacity of different human CD21 low B-cell subsets isolated from peripheral blood of patients with AI diseases by measuring activation of CD4 T cells in allogeneic co-cultures.
We chose SEB-enforced mixed lymphocyte reactions to answer the question about the co-stimulatory potential of distinct CD21 low B-cell populations from different autoimmune disorders in order to avoid effects of the heterogeneity of donor T cells and to overcome the lack of specific antigens which renders direct comparison to previously published data (21) not easy. Despite of the mentioned limitations, our data clearly revealed that primary human CD21 low B cells have the potential to serve as APCs: All memory CD21 low subsets independent of the underlying disease induced high and-in case of switched memory cells-even significantly higher amounts of HLA-DR when compared to the respective CD21 pos counterparts. Interestingly, the expression pattern of HLA-DR on CD21 low B cells is reversed to the pattern observed on CD21 pos B cells, of which naïve B cells have the highest expression. The high expression of HLA-DR on memory CD21 low B-cell subsets is most likely due to exposure to IFN-g (32, 33) but the cause of the relative reduction in naïve-like CD21 low B cells remains less clear. In addition, CD80 and CD86 are highly expressed on all four CD21 low B-cell subsets when compared to the respective CD21 pos subpopulation of HD ex vivo (12,(16)(17)(18)(19). This elevated expression was still seen after in vitro co-culture on naïve, but not on the different memory CD21 low B cells populations, since CD21 low memory B cells failed to further upregulate the expression as much as CD21 pos memory B cells, as previously reported (6,13,24,28,29,34,35). Other molecules involved in co-stimulation (36)(37)(38)(39)(40)(41) were similarly (OX40L) or variably (ICOS-L) expressed between CD21 low and CD21 pos subpopulations. Also, ICAM-1 known to contribute to co-stimulation by increasing cell-cell interaction (42) was rather increased on all subsets, although the differences were not significant for CD27 pos CD21 low subsets. On the other hand, CD21 low B cells tended also to express higher levels of the inhibitory ligands PD-L1 and PD-L2. Nevertheless, the ex vivo phenotype of CD21 low B cells marks these cells as activated cells well-equipped for sufficient co-stimulation of T cells at sites of inflammation.
SEB containing T-B co-cultures clearly induced the expression of CD69 and CD25 on naïve and CD25 on memory T cells. CD69 expression on memory T cells was hardly affected by B cells after 24 h, which might be caused by different kinetics compared to naïve T cells. The co-stimulatory effect of the different B-cell subsets, whether naïve or memory, CD21 pos or CD21 low , was nearly identical on naïve T cells regarding CD69 and CD25 induction. Thus, the increased expression of CD80 and CD86 on naïve-like CD21 low B cells after co-culture did not translate into a measurably enhanced activation of naïve T cells compared to co-cultures with naïve CD21 pos B cells. On the contrary, CD25 induction, which was generally higher on memory than naïve T cells, was even reduced on memory T cells when cultured with naïve and IgM-memory CD21 low B cells instead of their CD21 pos counterparts. The reason remains elusive. Known signals inducing high CD25 expression as the TCR signal intensity (43) are unlikely to be different in this setting, and IL-2 (44) was even slightly higher in cultures with naïve-like CD21 low B cells. Nevertheless, lower CD25 expression may limit the responsiveness of T cells stimulated by interactions with CD21 low populations to IL-2 and thereby potentially influence the survival, differentiation, and function of conventional and regulatory T cells (T reg ) (45,46) induced in the context of co-stimulation by CD21 low B cells in vivo. The activation of PD-1 pos memory T cells by CD21 low B cell populations may additionally be reduced by the generally higher expression of PD-L1 on all four CD21 low B-cell subpopulations as reported previously for one of the CD21 lowlike populations (47), given the inhibitory signal transferred by PD-1 ligation (48).
In regard to ICOS expression, we could show for the first time that its expression on naïve and memory T cells was especially high in co-cultures with HD-derived switched-memory B cells when compared to naïve and non-switched memory B cells. Interestingly, this effect was lost for switched memory CD21 low B cells. The differential signals given by naïve and switched memory B cells causing the differences in ICOS expression are unknown, but expression of ICOS on T cells is required for the induction and maintenance of an efficient germinal center response and B-cell memory formation (49,50), implying that switched memory CD21 low B cells would be less potent inducers of ICOS-dependent steps during this T-B interaction. In addition, high ICOS expression is required for IL-10, but not the other cytokine production by effector T cells (51). This notion is reflected by our finding that the supernatants of cocultures with switched-memory B cells of HD contained the highest amount of IL-10, while IL-10 was significantly reduced in the co-culture with CD21 low switched memory B cells.
Previous reports in mice suggested an increased antigen presenting potential of age-related B cells (ABCs), the murine homolog of CD21 low B cells. In vitro and in vivo data revealed that ABCs located in the spleen have an enhanced ability to take up, process, and present antigens to T cells compared to follicular B cells resulting in increased activation of T cells as shown by IL-2 secretion and T-cell proliferation (20). We were not able to test antigen-specific APC function of human CD21 low B cells of patients with AI disease and comparing results from mice to humans is difficult given the discrepancies of ABCs and CD21 low B cells (52) and the different origin of B cells from lymphoid organs in the murine models and peripheral blood in humans (18,20). Local environment and origin of B-cell subpopulations might strongly contribute to different APC capacities. So far, only few and conflicting studies have been published addressing the co-stimulatory potential of CD21 low B cells in humans. In contrast to our findings, Shimabukuro-Vornhagen and colleagues described a CD21 low CD86 high B-cell population of HD to induce increased proliferation and upregulation of CD25 on autologous T cells in the presence of anti-CD3 when compared to CD21 pos CD86 low B cells (21). Comparing, however, the CD21 low CD86 high and CD21 pos CD86 high B-cell populations, the APC capacity was not significantly increased, which is consistent with our observation. Besides differences in experimental settings, as they looked at an autologous co-culture setting, co-stimulated with anti-CD3 instead of SEB and observed a culture time of seven days, the B-cell subsets were obtained from HD and contained an unidentified mixture of Bcell subpopulations, which impedes the comparison. On the other hand, very recently reduced co-stimulatory potential was shown for a CD20 high CD24 low/atypical memory B-cell population of SLE patients, largely corresponding to CD21 low B cells, compared to CD27 pos memory B cells (53). The reduced APC-capacity of total CD21 low B cells was shown in a setting with low dose CD3 stimulation after five days. Limited survival of CD21 low B cells in vitro due to increased apoptosis (12,15,17,29,53) may, however, strongly influence the results in all settings with incubation times of more than 2 days.
Based on high T-bet expression and the preferential usage of IgG3 isotypes in CD21 low B cells (32,(54)(55)(56), the formation of CD21 low B cells has been linked to Th1 responses and IFN-g (54,57,58). Whether CD21 low B cells can vice-versa shape T-cell differentiation toward a Th1 profile is not clear. In mice, Hao et al. had reported an increased potential of ABCs to polarize naïve CD4 T cells toward a Th17 profile when performed under Th17 polarizing condition (31). Under Th1 and T reg polarizing conditions, no difference was observed between the co-culture with follicular B cells and ABCs. Due to the limited survival of CD21 low B cells in vitro, we performed short-term cultures, which don't allow for effector polarization from unprimed T cells. Acknowledging these limitations of the used in vitro conditions, we did not detect clear changes in the composition of naïve, non-Tfh and Tfh memory T cells according to their chemokine receptor profiles.
Accordingly, our data demonstrated no significant alterations for most cytokine concentrations in the supernatants of cocultures with CD21 low B cells compared to CD21 pos subsets with a few exceptions. Compatible to data of a murine model (20), the co-culture with naïve CD21 low B cells induced higher IL-2 levels when compared to naïve CD21 pos B cells, although much less than co-cultures with CD21 pos memory B cells. More importantly, however, the superior co-stimulatory capacity of switched memory CD21 pos B cells compared to all other CD21 pos B-cell subsets, which we could observe for ICOS induction and secretion of IL-2, IFN-g, and IL-10, is in line with previous observations (3) but was not observed for the respective CD21 low B-cell population, which was just within the range of the other B-cell subsets.
Given the well-equipped function as potential APCs, CD21 low B cells might present potential targets for therapeutic interventions in AI patients, since the high expression of costimulatory molecules on these cells is combined with an increased incidence of auto-reactive BCR entities among CD21 low B cells in RA, CVID, and SLE (12,14,17). The impact of CD80 and CD86 mediated co-stimulation on disease activity of AI diseases was emphasized by the efficacy of CD80/ CD86 inhibition (59,60). Abatacept, a fusion protein directly binding to CD80/CD86, approved for second line therapy in patients with RA, has been shown to directly decrease the expression of both receptors on B cells, resulting in reduced co-stimulation of T cells (61). How much of this in vivo effect is owed to a direct effect on the CD21 low B-cell compartment remains elusive. Nevertheless, either the elimination or the inhibition of the co-stimulatory potential of CD21 low B cells appears reasonable therapeutic approaches in patients with AI disorder and expansion of this B-cell population. On the other hand, the co-stimulatory potential of CD21 low B cells seems to be important for an efficient immune response against influenza (62). Further studies are currently conducted to address this issue and to elucidate the impact of Abatacept in other disease conditions linked to the occurrence of potentially autoreactive CD21 low B cells.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/ Supplementary Material.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Freiburg 239/1999 and 121/11 and Freiburg 66/13. The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
MR performed the experiments, analyzed and interpreted the data, and wrote the manuscript. KP, VS, and IH performed experiments. RV contributed to the acquisition and sample collection of patients and critically revised the manuscript. KW and BK designed and coordinated the study, interpreted data and wrote the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by the Deutsche Forschungsgemeinschaft (grant TRR130; P07 to KW and P12 to RV) and the German Federal Ministry of Education and Research (grant BMBF 01E01303).

ACKNOWLEDGMENTS
We thank our patients and healthy controls for their collaboration; the nurses and physicians of the outpatient clinics of the Department of Rheumatology and Clinical Immunology for patient care, Jan Bodinek-Wersing from the Lighthouse Fluorescence Technologies Core Facility (Universitätsklinikum Freiburg) for cell sorting and the Immunologisch-Rheumatologische (IR) Biobank Freiburg for the sample management and support.