Cutting Edge: A Threshold of B Cell Costimulatory Signals Is Required for Spontaneous Germinal Center Formation in Autoimmunity

Emerging data indicate that autoreactive B cell activation, epitope spreading, and pathogenic antinuclear Ab generation in systemic lupus erythematosus (SLE) occurs within spontaneous germinal centers (GCs) (1). Thus, delineating the cellular mechanisms underlying initial breaks in B cell tolerance and spontaneous GC formation is of significant importance. During lupus pathogenesis, autoreactive B cells engage cognate T cell help and initiate breaks in T cell tolerance via B cell–intrinsic, MHC class II–dependent Ag presentation and proinflammatory cytokine production (2–4). Complete CD4⁺ T cell activation requires costimulatory signals, including CD28 engagement by B7 receptors CD80 (B7.1) and CD86 (B7.2) on APCs. Animal studies have confirmed the importance of B7:CD28 signals in GC formation because both Cd28^−/− and Cd80^−/−/Cd86^−/− mice fail to generate GCs (5, 6). Conversely, the CD28 homolog CTLA-4 inhibits pathogenic T cell activation via cell-intrinsic and cell-extrinsic mechanisms (7).

Importantly, CD4⁺ T cell costimulatory signals could conceivably be provided by distinct APC populations, dendritic cells (DCs), and B cells. Based on our earlier identification of B cells as critical APCs in SLE (3), we predicted that B cell–derived costimulatory signals likely facilitate breaks in T cell immune tolerance during autoimmunity (1). However, Watanabe et al. (8) reported that B cell CD80/CD86 expression is redundant for T cell activation and GC formation after NP-OVA/Alum immunization. An important caveat is that candidate Ag immunization models may not accurately inform the biology of autoimmune GCs given the differences in (auto)antigen abundance and affinity, adjuvant load, and the impact of B cell/T cell tolerance mechanisms. For these reasons, we hypothesized that costimulatory signals might exert a more nuanced impact on GC formation in autoimmunity.

Using a well-characterized model of murine SLE, in this study we show that whereas autoreactive CD4⁺ T cell priming is B cell independent, B cell–derived costimulatory signals are critical for complete T follicular helper (Tfh) cell differentiation and autoimmune GC formation. Strikingly, even heterozygous deletion of B cell CD80/CD86 recapitulated this phenotype, indicating that during initial interactions between Ag-primed autoreactive T and B cells, a threshold of B cell costimulatory signals is required for CD4⁺ T cell activation and spontaneous GC formation.

Murine studies performed in a specific pathogen–free environment in accordance with Institutional Animal Care and Use Committee–approved protocols. To establish chimeras, donor bone marrow (BM; from femora and tibiae) was depleted of CD138⁺ plasma cells (130-098-257; Miltenyi Biotec) and mixed with µMT BM (20:80 ratio, 6 × 10⁶ total cells) and injected retro-orbitally into lethally irradiated (450 cGy times two doses) µMT recipients. For CD28-null chimeras, Was^−/−.Cd28^−/− and Cd28^−/−.µMT BM (20:80 ratio, 6 × 10⁶ total cells) was transferred into Cd28^−/−.µMT recipients. Wiskott–Aldrich syndrome (WAS) chimeras were sacrificed at 24 wk and virus-like particle (VLP) models at 8 wk posttransplant. For VLP experiments, CellTrace Violet–labeled 2 × 10⁶ TCR transgenic OT-II CD4⁺ T cells, negatively enriched using magnetic microbeads (≥95% purity), were transferred into either wild-type (WT) or Cd80^−/−.Cd86^−/− recipients (global model) or B cell–intrinsic WT and Cd80^−/−.Cd86^−/− chimeras (B cell–intrinsic model). Recipient animals were immunized i.p. with 10 μg of Qβ-OVA VLP and splenocytes analyzed at day 3 postimmunization (9).

Abs used were B220 (RA3-6B2), CD4 (RM4-5), CD138 (281-2), CXCR5 (2G8), CD86 (GL1), and ICOS (7E.17G9) from BD Biosciences; CD62L (MEL-14), CD11c (N418), CD11b (M1/70), PD-1 (J43), T-bet (4B10), and BCL-6 (BCL-DWN) from eBioscience; PD-1 (J43) from Life Technologies; CD19 (ID3), CD44 (IM7), CD4 (RM4-4), and CCR7 (4B12) from BioLegend; PNA (Fl-1071) from Vector Laboratories; and Fas (Jo2) from BD Pharmingen. Qβ-OVA and VLP-APC were from Dr. B. Hou. Flow cytometry of splenocyte suspensions was performed as described (10).

Autoantibodies were measured as described (3) using the following reagents: calf thymus dsDNA (D3664-5 × 2 mg; Sigma-Aldrich); Sm/RNP (ATR01-10; AROTEC Diagnostics); and goat anti-mouse IgG–, IgG2c–, and IgG3–HRP-conjugated Ab (SouthernBiotech). Autoantigen microarrays were performed at The University of Texas Southwestern Microarray Core (11).

Immunofluorescence staining of frozen splenic sections was performed as described (3).

Single-cell BCR sequencing was performed as described (12). Briefly, single B220^lowCD138⁺ splenic plasma cells were sorted from representative WAS and B cell–intrinsic Cd80^−/−.Cd86^−/− WAS chimeras, and BCR κ L chain–specific transcripts were amplified in separate nested PCRs prior to sequencing. Processed sequences were submitted to ImMunoGeneTics V-QUEry and STandardization for alignment and somatic hypermutation (SHM) counts determined by using the ImMunoGeneTics mutation table output.

Using a chimeric murine lupus model, termed the WAS model, we have shown that B cell–intrinsic TLR signals, B cell Ag presentation to cognate CD4⁺ T cells, and cell-intrinsic cytokine signals coordinate to orchestrate autoimmune GC formation (3, 4, 10, 13). To address the relative importance of B cell– versus non–B cell (myeloid)–derived costimulatory signals in SLE, we used a similar strategy and generated WAS chimeras either globally Cd28 deficient or with B cell–intrinsic Cd80/Cd86 deletion. In keeping with established roles for B7:CD28 engagement in T cell activation, CD44^hiCD62L^lo/hi effector/memory (EM) CD4⁺ T cells’ formation was abrogated by global Cd28 deletion. Moreover, lack of CD28 costimulation prevented CXCR5 expression on developing Tfh cells. In contrast, B cell–intrinsic Cd80/Cd86 deletion exerted a limited impact on CD4⁺ T cell EM differentiation or CXCR5 upregulation, indicating that initial priming of autoreactive CD4⁺ T cells in this model is B cell independent (Fig. 1A, 1B, Supplemental Fig. 1).

CXCR5 upregulation and the corresponding downregulation of the T cell zone receptor CCR7 result in CXCL13-dependent migration of activated CD4⁺ T cells to the T–B border, where cognate interactions between primed CD4⁺ T cells and B cells facilitate Tfh maturation and GC formation (14). Surprisingly, whereas the proportion of CD44⁺CXCR5⁺ CD4⁺ T cells was similar, B cell–intrinsic CD80/CD86 deletion resulted in absent upregulation of Tfh surface receptors PD-1 and ICOS, preserved CCR7 expression, and reduced expression of the Tfh-defining transcription factor BCL-6 (Fig. 1C). Consistent with a critical role for B cells in completing Tfh differentiation, we observed a marked reduction in stringently gated CD44⁺CXCR5^hiPD-1^hiICOS^hi GC Tfh cells in the B cell–intrinsic Cd80^−/−.Cd86^−/− model (Fig. 1D).

The role for B cell costimulatory signals in Tfh maturation prompted us to test whether these events were unique to autoimmune settings. B cells can function as APCs following Qβ-VLP immunization, suggesting a potential role for B cell–derived costimulation in antiviral responses (9). To test this idea, we transferred CellTrace Violet–labeled CD4⁺ TCR transgenic (OT-II) T cells into either WT and Cd80^−/−.Cd86^−/− recipients (global model) or B cell WT and Cd80^−/−.Cd86^−/− chimeras (B cell–intrinsic model), which were subsequently immunized with VLP-comprising Qβ protein and OVA-derived peptide. As predicted, OT-II CD4⁺ T cells in global Cd80^−/−.Cd86^−/− mice failed to proliferate, upregulate CD44, or express BCL-6 in response to Qβ protein and OVA-derived peptide. In contrast, OT-II CD4⁺ T cells in B cell Cd80^−/−.Cd86^−/− chimeras divided and upregulated CD44, although BCL-6 expression was significantly reduced relative to WT chimera controls (Supplemental Fig. 1). Thus, whereas initial CD4⁺ T cell priming is B cell independent in the setting of both humoral autoimmunity and antiviral responses, B cell–derived costimulatory signals provide important contributions to Tfh maturation, including enhancing the expression of the Tfh master regulator BCL-6.

Based on this unanticipated role for B cell costimulation on Tfh differentiation, we examined the phenotype of spontaneous GCs in B cell–intrinsic Cd80^−/−.Cd86^−/− WAS chimeras. Strikingly, expansion of PNA⁺FAS⁺ GC B cells was abrogated in both global Cd28^−/− and B cell–intrinsic Cd80^−/−.Cd86^−/− WAS chimeras (Fig. 2A–C). Immunofluorescence staining of splenic sections confirmed equivalent loss of autoimmune GCs in chimeras lacking global CD28 or B cell CD80/CD86 expression (Fig. 2D). Thus, in contrast with candidate Ag immunization (8), B cell–derived costimulatory signals are critical for spontaneous GCs during humoral autoimmunity.

Both Was^−/− mice and WAS patients develop diverse serum autoantibodies (10, 15), although in Was^−/− animals, these are predominantly a T cell–independent IgG3 subclass, with WT CD4⁺ T cell help required for IgG2c autoantibody generation (10). Consistent with these data, global Cd28 deletion limited class-switched autoantibody targeting dsDNA, although a subset of global Cd28^−/− WAS chimeras exhibited low-titer anti-dsDNA IgG autoantibodies (Fig. 3A). However, these were predominantly IgG3, with no IgG2c autoantibodies detected (Fig. 3B); these findings are consistent with our previous data in which interventions disrupting B cell:T cell cross-talk, including CD4⁺ T cell depletion and B cell–intrinsic MHC class II deletion, abrogate IgG2c autoantibodies (3, 10).

Thus, we predicted that absent GCs in B cell Cd80^−/−.Cd86^−/− WAS chimeras would be accompanied by loss of pathogenic IgG2c autoantibodies. However, significant complexity underlies plasma cell generation in SLE, with both GC and extrafollicular (EF) pathways contributing to the autoantibody repertoire (16). Strikingly, despite absent GCs, we observed widespread IgG2c self-reactivity by autoantigen microarray in B cell Cd80/Cd86^−/− chimeras that was enriched for autoantibody binding diverse connective tissue proteins such as α-actinin, vimentin, and collagen proteins (Fig. 3C). Moreover, B cell Cd80^−/−.Cd86^−/− WAS chimeras also developed high-titer anti-dsDNA autoantibodies, including the pathogenic IgG2c subclass, whereas T cell–independent anti-dsDNA IgG3 were absent (Fig. 3D). Surprisingly, a lack of autoimmune GCs exerted only a limited impact on the autoantibody repertoire. Whereas anti-dsDNA titers were unaffected, RNA-associated anti-Sm/RNP autoantibodies were lost in the absence of B cell costimulation, implicating spontaneous GCs as the likely source for this specificity (Fig. 3E). Anti-Sm/RNP autoantibodies were similarly reduced in Cd28^−/− WAS chimeras (Supplemental Fig. 2).

In keeping with EF B cell activation driving diverse autoantibodies, we observed GC-independent accumulation of IgG2c⁺ plasma cell foci in the EF red pulp of B cell–intrinsic Cd80^−/−.Cd86^−/− WAS chimeras, but not global Cd28^−/− mice (Fig. 3F). Splenic B220^loCD138⁺ plasma cells were also expanded in WAS and B cell Cd80^−/−.Cd86^−/− WAS chimeras, but reduced in the absence of T cell CD28 signals (Fig. 3G). Moreover, whereas the GC is the primary site for Ag-driven selection, we observed equivalent BCR mutational frequencies in sorted splenic plasma cells from B cell CD80/CD86–sufficient and –deficient chimeras (Fig. 3H), in keeping with prior evidence for B cell somatic mutation within EF foci in murine SLE (17). Finally, conflicting EF and GC-dependent models have also been proposed for the generation of CD11b⁺CD11c⁺ age/autoimmunity-associated B cells (ABCs) that act as precursors for pathogenic Ab-secreting cells in SLE (18–20). Interestingly, we observed preserved expansion of T-bet–expressing ABCs in B cell–intrinsic Cd80/Cd86^−/− WAS chimeras, but not global Cd28^−/− models, indicating that T cell–dependent EF activation can support development of this lupus-associated memory B cell subset (Fig. 3I, Supplemental Fig. 2).

Together, these data support a model in which both EF and GC activation pathways facilitate autoreactive B cell class-switch recombination, SHM, ABC formation, and differentiation into pathogenic plasma cells, with each pathway providing distinct contributions to the lupus autoantibody repertoire. Loss of B cell costimulatory signals uniquely dissociates these events by abrogating the formation of autoimmune GCs without preventing T cell–dependent EF B cell activation.

The strength of T cell costimulation needs to be tightly regulated to prevent systemic autoimmunity. Heterozygous mutations in CTLA4 or its regulator LPS-responsive and beige-like anchor protein (LRBA) promote spontaneous humoral autoimmunity in humans (21–23). Thus, we examined whether reduced surface CD80/CD86 expression on B cells is sufficient to limit autoimmune GC formation in SLE. To do this, we generated parallel cohorts of WAS chimeras in which B cells expressed zero to two copies of CD80/CD86. Strikingly, despite a relatively modest reduction in B cell costimulatory molecule expression (Fig. 4A), heterozygous CD80/CD86 deletion recapitulated the phenotype of the B cell–intrinsic Cd80^−/−.CD86^−/− model. Similar to biallelic CD80/CD86 deletion, no change was observed in CXCR5 upregulation on CD4⁺ T cells in B cell Cd80/Cd86–haploinsufficient chimeras (Fig. 4B). Rather, reduced B cell CD80/CD86 expression correlated with the failure of CXCR5⁺CD4⁺ T cells to upregulate ICOS, PD-1, and BCL-6 and downregulate CCR7 expression (Fig. 4C). This resulted in a corresponding lack of GC Tfh cells (Fig. 4D). In keeping with Tfh maturation being required for GC formation, we observed a parallel loss of splenic GC B cells in heterozygous B cell Cd80^+/−.CD86^+/− chimeras (Fig. 4E, 4F). Finally, heterozygous CD80/CD86 deletion exerted no impact on the generation of IgG/IgG2c anti-dsDNA titers but reduced anti-Sm/RNP autoantibodies (Fig. 4G–J).

In summary, to our knowledge, our work defines a new role for B cells during the pathogenesis of SLE, namely the provision of costimulatory signals above a threshold required for Tfh maturation and autoimmune GC formation. These data are consistent with lack of class-switched autoantibodies and systemic inflammation in Cd28^−/−.MRL/lpr mice (24) and with delayed disease onset in NZB/NZW F1 mice treated with either CD28 blockade or combined CD80/CD86 inhibition and cytotoxic immunosuppression (25, 26). Mechanistically, whereas both EF and GC pathways have been described in human SLE and murine lupus models, it has been technically challenging to define the source of individual autoreactive plasma cells. However, because plasma cells can be functionally divided into short-lived EF and long-lived GC-derived subsets, variable autoantibody persistence following B cell depletion provides indirect evidence for the cellular source for individual specificities. In this context, our data are strikingly consistent with the decline in anti-dsDNA IgG, but durable, persistence of RNA-associated autoantibodies following B cell ablation in SLE (16).

Our study also informs the understanding of autoimmunity in CTLA4 and LRBA haploinsufficiency (21–23). CTLA-4 regulates surface B7 ligand levels by capturing CD80/CD86 molecules from the surface of APCs via trans-endocytosis, thereby limiting available costimulatory signals for T cell activation (27). Because CTLA4 and LRBA haploinsufficiency is accompanied by histologic GC formation and the expansion of circulating CXCR5⁺ Tfh cells (28, 29), these data suggest that modest B7 elevation via reduced trans-endocytosis is sufficient to drive CD28-dependent Tfh differentiation. Based on this model, we performed the converse experiment and confirmed that even modest reductions in B cell CD80/CD86 surface expression is sufficient to abrogate autoimmune GCs.

The ubiquitous presence of self-antigens and stochastic nature of lupus pathogenesis have each complicated the characterization of underlying immune events. Self-reactive B cell transgenic models have demonstrated that initial dual BCR/TLR-dependent activation of autoreactive B cells occurs in a T cell–independent manner but that provision of CD4⁺ T cell at EF sites enhances B cell proliferation and differentiation into autoantibody-producing plasmablasts (30). Although DCs are classically considered the primary APCs driving adaptive T cell responses, T cell activation is preserved in DC-deficient MRL.Fas^lpr mice (31), and B cell Ag presentation promotes murine SLE (2, 10). Moreover, autoreactivity in SLE converges on a relatively limited subset of nucleic acid–containing autoantigens able to engage B cell endosomal TLRs. Thus, this B cell–centric model for lupus pathogenesis raises the important countervailing question as to whether myeloid cells are redundant for T cell activation during humoral autoimmunity. In this study, we show that the expansion of CD44^hiCD62L^lo/hi EM T cells and CD44⁺CXCR5⁺ Tfh precursors requires initial CD80/CD86 costimulatory signals provided by non–B cell lineages. In this context, we predict that B cell propensity for activation by dual BCR/TLR signals focuses antinuclear autoreactivity in SLE, both by enhancing B cell presentation of nuclear self-antigens to cognate T cells and, indirectly, via the generation of circulating immune complexes that enhance delivery of specific self-antigens to myeloid APCs. Taken together, our data highlight how myeloid lineages and B cells coordinate to drive autoreactive T cell activation and emphasize how tight regulation of costimulatory signals is required to maintain immune tolerance.

The authors have no financial conflicts of interest.