BCL6 in T cells promotes type 1 diabetes by redirecting fates of insulin-autoreactive B lymphocytes Authors: Landon M Clark, Jack C McAninch, Dudley H McNitt, Marguerite L Padgett, Tyler W Jenkins, Lindsay E Bass, Casey M Nichols, Jeffrey C Rathmell, Rachel H Bonami Published: 2026 ABSTRACT SummaryCurrently approved immunotherapies for type 1 diabetes (T1D) broadly target T cells and delay, but do not fully prevent, diabetes development, highlighting the need for more selective targets. Here, we show that T cell loss of the key germinal center protein, BCL6 prevents diabetes, despite the continued presence of anti-insulin B cells. This protection was associated with reduced anti-insulin B cell activation and proliferation. ΔBCL6 adoptively transferred CD4+ T cells showed reduced activation, relative to BCL6 sufficiency. Anti-insulin B cells primarily existed in extrafollicular niches and skewed toward phenotypically defined atypical and atypical memory B cell subsets in pancreas and pancreatic lymph nodes, which was reduced by T cell loss of BCL6. This work implicates CD4+ T cells in licensing extrafollicular insulin-binding B cells, a process which is restrained by T cell BCL6 deficiency. Overall, these findings support BCL6 inhibition as a promising T1D immunotherapy.Subject areas: Biological sciences Introduction Immunotherapies targeting T or B cells are used in rheumatic disease management and are beginning to be applied to type 1 diabetes (T1D) treatment (e.g., teplizumab), but no current treatment durably halts beta cell demise.1,2,3 A major challenge in many autoimmune diseases is that individuals typically present for disease diagnosis well after initial immune tolerance breach for autoantigens has occurred, as demonstrated by the predictive power of autoantibodies in diseases such as systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), and T1D.4,5,6 Thus, new T1D immunotherapies should ideally reduce disease burden even when initiated after autoreactive B cell tolerance breach has occurred (signaled by autoantibody seropositivity). Insulin is a major T1D autoantigen; insulin autoantibodies predict T1D in mice and humans and are associated with earlier age of T1D onset.7 Insulin autoantibodies are not directly pathogenic but are thought to signal pathologic B-T cell interactions.8,9 B cell-specific elimination of the diabetogenic MHC class II, IAg7, prevents diabetes in non-obese diabetic (NOD) mice.10 Overt diabetes depends on the presence of anti-insulin B cells in NOD mice but can occur even in the absence of detectable insulin autoantibody production.11 This is due to anti-insulin B cells acting as antigen-presenting cells (APCs) to drive CD4+ T cell activation.8,12 Insulin B:12–20-reactive T cell clones are found in mice (e.g., 8F10) and humans.13,14 Mutating a key residue in this insulin epitope prevents diabetes, further supporting a key role for immune recognition of insulin in driving T1D.15 Germinal centers (GCs) are microanatomical sites, where B cells interact with a subset of CD4+ T cells known as a T follicular helper cells (Tfhs) to support B cell affinity maturation, memory B cell formation, and antibody responses.16,17,18 The transcriptional repressor, BCL6, is necessary for Tfh and GC B cell formation.19,20,21,22 In T1D, anti-insulin B-T cell interactions influence one another to defy immune tolerance mechanisms that otherwise restrain their differentiation capacity.14 GC anti-insulin B cells, but not non-GC anti-insulin B cells, from VH125SD.NOD mice process and present the insulin B12:20 register recognized by the pathogenic anti-insulin 8F10 T cell clone.14 This unique APC programming, which supports 8F10 T cell proliferation, can also be conferred by CD40L + IL-4 stimulation of anti-insulin B cells,14 pointing to T cell-driven programming of the B cells as a switch that ultimately controls pathogenic T cell activation and differentiation. Anti-insulin VH125SD B cells and 8F10 T cells reciprocally enhance each other to adopt GC B cell and Tfh fates, respectively, to ultimately drive insulin autoantibody production.14 Tfh cells can transfer diabetes.23 Circulating Tfh cells are elevated in T1D patients, and abatacept response heterogeneity is linked to differences in Tfh-like populations.24,25,26,27,28 T cell expression of BCL6 also supports extrafollicular (EF) responses.29 We recently found that T cell loss of BCL6 blocks Tfh and GC B cell formation and prevents diabetes in NOD mice30 but it is unclear whether this protection was mediated by blocking GC or EF T/B interactions. Furthermore, atypical B cells are implicated in other autoimmune diseases such as SLE but have not been well characterized in T1D.31,32 These data collectively point to a need to examine the autoantigen-specific T/B lymphocyte interactions that GC and EF niches may support. Autoreactive B cell rarity complicates their study in a polyclonal repertoire. To circumvent this challenge, we deployed the VH125SD.NOD model, whereby an IgH locus-targeted B cell receptor (BCR) heavy chain transgene generates a trackable 1%–3% population of insulin autoreactive B cells.8 This model allowed us to address remaining knowledge gaps regarding how BCL6 loss in T cells impacts insulin autoreactive B cells responses and diabetes proclivity in a setting where B cell tolerance mechanisms have already been breached. Given recent advances in antigen-specific immunotherapy targeting insulin,33 we sought to gain a better understanding of how insulin-reactive B lymphocytes are activated and licensed to interact with T cells within complex in vivo niches that cannot fully be modeled in vitro. We show that despite transgene-induced expansion of anti-insulin B cells, T cell loss of Bcl6 led to near-complete prevention of diabetes in VH125SD.BCL6ΔCD4 NOD mice. Elimination of BCL6 from T cells led to a reduction in anti-insulin B cell activation and spontaneous proliferation in vivo. Proliferation was restored in VH125SD.BCL6ΔCD4 NOD mice by adoptive transfer of both BCL6+ and ΔBCL6 CD4+ T cells, suggesting CD4 T cells do not require BCL6 to drive anti-insulin B cell proliferation if they can override local immunosuppression that may be mediated by Tregs. We found that anti-insulin B cells chiefly reside in EF niches and skew toward atypical B cell (CD11b+ CD11c+) memory (CD80+ PDL2+) phenotypes compared to non-insulin-binding B cells, particularly in the pancreatic lymph nodes and pancreas. Multiple specific subsets of atypical memory and activated B cells identified via minimally supervised phenotypic analysis were reduced by BCL6 loss in T cells. These results highlight the importance of T cell BCL6 expression to activate and license anti-insulin B lymphocytes to upregulate T cell costimulatory molecules and support beta cell attack in T1D. Results BCL6 in T cells promotes spontaneous GC formation and diabetes in VH125SD.NOD mice T cell loss of Bcl6 prevents T1D in NOD mice,30 but it is unclear whether T cell expression of BCL6 supports the formation and expansion of anti-insulin B cells in the repertoire (initial B cell immune tolerance breach), anti-insulin B cell licensing by anti-insulin T cells, or both. Accelerated diabetes occurs in the VH125SD.NOD mouse model, which bypasses the key T1D pathogenic step involving the emergence of anti-insulin B cell clones within the overall B-cell repertoire.8 We used this model to test how T cell loss of BCL6 impacts anti-insulin B cell activation/pathogenic T cell formation, and ultimately T1D, by generating VH125SD.Bcl6fl/fl.NOD (termed VH125SD.NOD hereafter) and VH125SD Bcl6fl/fl.CreCD4.NOD (termed VH125SD.Bcl6ΔCD4) mice. Strikingly, despite the presence of anti-insulin B cell populations in peripheral organs, T cell loss of BCL6 in VH125SD.Bcl6ΔCD4 mice led to nearly complete diabetes protection relative to control VH125SD.NOD mice (Figures 1A and 1B). BCL6 in T cells promotes spontaneous GC formation, anti-insulin B cell infiltration of islets, and spontaneous diabetes development in VH125SD.NOD mice Cells from spleen, pLNs, mLNs, and pancreas were isolated from 8 to 12-week-old VH125SD.NOD and VH125SD.Bcl6ΔCD4.NOD mice (genotypes fully defined in STAR Methods). (A) A representative flow cytometry plot from spleen shows the frequency of insulin-binding B cells (insulin+) identified using biotinylated human insulin/streptavidin fluorochrome as in STAR Methods among total B cells (live singlet CD45+ CD19+ lymphocytes). (B) Diabetes was monitored in cohorts of female VH125SD.NOD mice (n = 14, black line) and VH125SD. Bcl6ΔCD4.NOD littermates (n = 14, purple line) from 10 to 35 weeks of age. Mice were considered diabetic after two consecutive blood glucose readings >250 mg/dL, p < 0.0001, log-rank test. (C–G) Representative flow cytometry plots of pancreatic draining lymph nodes and pancreata gating on anti-insulin B cells as in (A) and as described in STAR Methods for pancreas are shown (C). The frequency of anti-insulin B cells (among total B cells) in (D) spleen, (E) mesenteric lymph nodes (mLNs), (F) pancreatic draining lymph nodes, and (G) pancreata are plotted for individual mice of the indicated genotypes. (H–L) Representative flow cytometry plots from the pancreatic lymph nodes (pLNs) of (H) Tfh cells (live singlet CD45+ CD4+ PD-1hi CXCR5hi Foxp3− lymphocytes) are shown with (I and J) frequencies among total CD4+ Foxp3- CD45+ cells and (K and L) numbers of Tfh cells in pLNs and pancreata plotted for individual mice. (M–Q) Representative flow plots of (M) GC B cells from the pancreatic lymph nodes (live singlet CD45+ CD19+ Fas+ GL7+ lymphocytes) with (N and O) frequencies of GC B cells among total B cells and (P and Q) numbers shown for pLNs and pancreata. (C–Q) n = 6–8 mice per group, 5 independent experiments, Mann-Whitney U test, bars representative of mean ± standard deviation. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ns, not significant. The proportion of anti-insulin B cells showed no or limited reductions between the Bcl6-sufficient and deficient groups in spleen, mesenteric, or pancreatic lymph nodes (Figures 1C–1F), but a pronounced reduction in the pancreas (Figure 1G). Total B cell and CD4+ T cell frequencies were unchanged regardless of Bcl6 presence or absence (Figures S1B–S1G). Transitional (T1 or T2), follicular, pre-marginal zone, or marginal zone populations in the spleen were also unchanged (Figure S2). The Bcl6-deficient group showed reduced proportions and numbers of Tfh and GC B cells (gated as in Figures 1H and 1M, full gating scheme in Figure S1A) in pancreatic lymph nodes and pancreas relative to Bcl6-sufficient mice (Figures 1H–1Q), with the exception that pancreatic Tfh cell numbers trended down but were not significantly different (Figure 1L). T follicular regulatory cells (Tfrs), which regulate the GC response, were reduced in the Bcl6-deficient group in spleen and pLNs, with the exception of pancreas, in which Tfr are rare (Figures S3A and S3B).34 CD4+ T peripheral helper (Tph) populations, defined as CXCR5- PD-1hi ICOShi, showed no significant differences with T cell loss of BCL6 in spleen and pLNs, with a downward trend noted in pancreas (Figures S3C and S3D). Major differences in Foxp3+ Treg populations were not observed with BCL6 loss in any of the organs examined (Figure S3E). Thus, T cell loss of BCL6 impacts GC B and Tfh cell populations and prevents diabetes, despite retaining an expanded (1%–3%) pool of anti-insulin B cells that reach the periphery. These data suggest that T cell expression of BCL6 is critical in supporting pro-pathogenic functions in B cells and/or T cells necessary to cultivate islet attack. BCL6 loss in T cells reduces insulitis severity in VH125SD.NOD mice but does not impact T/B lymphocyte organization We next examined whether T cell expression of BCL6 was required for pancreatic islet infiltration by T and B lymphocytes. Insulitis is apparent at 8–12 weeks and hits a crescendo at 13–16 weeks in the VH125SD.NOD model, an interval at which 5%–10% of mice typically develop diabetes (Figure 2A).8 Average insulitis was reduced in 8–12-week-old VH125SD.Bcl6ΔCD4 mice, relative to Bcl6-sufficient controls, with a corresponding increase in the percentage of islets with no insulitis present in the Bcl6-deficient group (Figures 2B–2D). The magnitude of this decrease was less apparent in 13–16-week-old mice (Figures 2E–2H). T cell expression of BCL6 supports enhanced insulitis severity, but is dispensable for tertiary lymphoid structure organization in VH125SD.NOD mice Pancreata were harvested from female, pre-diabetic VH125SD.NOD and VH125SDBcl6ΔCD4.NOD mice at 8–12 and 13–16 weeks of age and were formalin fixed, and paraffin embedded. (A–H) About 10 μm pancreas sections were stained with hematoxylin and eosin (H&E) and blind scored, with individual mice plotted. (A and E) Representative H&E-stained sections are shown, with arrows pointing to islets. All islets were blind scored, with average insulitis scores shown for control VH125SD.NOD mice (black) and VH125SDBcl6ΔCD4.NOD mice (purple) at (B) 8–12 weeks and (F) 13–16 weeks. The percentage of islets, which had no lymphocytic infiltrate present were calculated for (C) 8–12 weeks and (G) 13–16 weeks. (D and H) The percentage of islets with each score for all pancreata for 8–12 and 13–16 weeks is shown. (I–L) Pancreas sections from the 8–16-week-old cohort that had the highest insulitis infiltrate (n = 6 mice per group) were obtained and stained with antibodies reactive against CD3 (T cells) or B220 (B cells). (I) Representative images show “disorganized” T cell-B cell infiltration (top), and “organized” TLS (bottom). Islets were scored separately for T cell (CD3) and B cell (B220) infiltrate as follows: 0 (no T/B infiltrate), 1 (>25% infiltrate), 2 (25%–50% infiltrate), 3 (50%–75% infiltrate), and 4 (>75% infiltrate). Average infiltration score for (J) CD3+ and (K) B220+ cells in islets is shown. (L) Infiltrated islets that scored 2 or above were blind scored as organized (blue) or disorganized (red) in both VH125SD.NOD mice and VH125SDBcl6ΔCD4.NOD, n = number of islets scored. (A–L) n = 6–8 mice per group, Bars represent mean ± standard deviation, ns, not significant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, (B, C, F, G, J, and K) Mann-Whitney U test or (L) chi-square test. Organized T and B cell zones, termed tertiary lymphoid structures (TLSs), can form in the pancreatic islets of NOD mice and contain GCs.35 To examine whether a pre-existing repertoire bias toward anti-insulin B cell formation impacted TLS formation and organization, we performed immunohistochemistry staining to detect B and T lymphocytes in pancreas sections (as outlined in STAR Methods and shown in Figure 2I). CD3 T cell infiltration was reduced with loss of Bcl6 (Figure 2J), while B cell infiltration was largely similar between both genotypes (Figure 2K), with some organized TLS still noted in both groups (Figure 2L). Thus, T cell loss of BCL6 failed to fully eliminate TLS organization, despite diabetes protection. T cell expression of BCL6 promotes insulin-binding B cell proliferation, activation, and upregulation of the T cell costimulatory markers in T1D-associated organs Anti-insulin B and T cells can drive each other to proliferate in vitro and can push each other to adopt GC B cell and GC Tfh fates, respectively, in vivo.12,14 GC and EF responses depend on complex niches involving 3D structures, cell interactions, and cytokines that cannot be fully reproduced in vitro.17,36,37 We therefore sought to address whether T cell loss of BCL6 impacted spontaneous anti-insulin B cell proliferation in vivo. Given that GC B cells are highly proliferative,38 we examined proliferation of anti-insulin B cells via Ki67+ cells among non-GC B cells to eliminate this potential bias, as in Figures 3A–3C. Little to no insulin-binding B cell proliferation was seen in the spleen but was enhanced among insulin-binding B cells (relative to non-insulin-binding B cells) in the pancreatic lymph nodes, and especially pancreas, which was reduced in the BCL6ΔCD4 group (Figure 3D). BCL6 in T cells increases activation and proliferation markers of insulin-binding B cells relative to non-insulin-binding B cells Cells were isolated from 8 to 12-week-old, female, pre-diabetic VH125SD.NOD with and without Cd4-Cre Bcl6 deletion from (A–G) spleen, pancreatic lymph nodes (pLNs) and pancreata. (A–C) Representative flow cytometry plots show Ki67 staining overlays of insulin-binding (left) or non-insulin-binding (right) B cells (identified as in Figure 1) from each genotype within live singlet GL7- CD95- CD45+ CD19+ lymphocytes in spleen (A), pancreatic lymph nodes (B), and pancreata (C). (D) Non-GC (Fas− GL7−) B cells were further gated on insulin-binding (ins+) and non-insulin-binding (ins−) and the frequency of cells that were Ki67+ (a marker of proliferation). (E–G) Insulin+ or insulin− B cell expression of (E) CD86 (T cell co-stimulatory molecule), (F) CD44 (activation marker), and (G) CD69 (activation marker) is shown for spleen (left), pancreatic lymph nodes (middle), and pancreata (right), with n = 6–8 individual mice plotted per group. (D–G) Bars represent mean ± standard deviation. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ns = not significant, Kruskal-Wallis test with post hoc test of multiple comparisons. All other comparisons not shown are not significant. We found that levels of CD86, a T cell costimulatory molecule, were reduced in insulin-binding B cells in VH125SD.Bcl6ΔCD4 mice compared to control VH125SD mice in pancreatic lymph nodes and pancreas but not spleen (Figure 3E). Changes in activation markers CD44 and CD69 were less striking, with T cell loss of BCL6 only reducing CD44 in pancreatic lymph nodes and CD69 in the pancreata among insulin-binding B cells (Figures 3F and 3G). Few BCL6-dependent differences were seen in Ki67, CD86, CD44, or CD69 across non-insulin-binding B cell populations in either pancreatic lymph nodes or pancreas, except for CD86 levels in the pancreatic lymph nodes (Figures 3D–3G). Anti-insulin B cell activation and proliferation was less apparent in the spleen, as compared to pancreatic lymph nodes and pancreas (Figures 3D–3G). Together, these results show that T cell expression of BCL6 creates an environment that supports anti-insulin B cell activation, expression of the T cell co-stimulatory marker, CD86, and proliferation at sites that prime islet attack, all of which hold potential to impact their function as pathogenic APCs. IgG1 and IgG2b class switching is reduced among insulin-binding and non-insulin-binding B cells by T cell loss of BCL6 Although class switching occurs largely before GC B cell differentiation, BCL6 stabilizes B cell class switching to IgG1.39,40 T cell loss of BCL6 led to a reduced frequency of IgG1+ and IgG2b+ B cells among insulin-binding as well as non-insulin-binding B cells in pancreatic lymph nodes (Figures S4A–S4D). Although IgG-switched B cells are detected in VH125SD.NOD mice, circulating IgG anti-insulin Ab is almost completely suppressed.8,41 In line with this, we detected little insulin-specific IgG autoantibody regardless of Bcl6 genotype in VH125SD.NOD mice, with varying levels detected in NOD.BCL6fl/fl controls, as expected (Figure S4E). BCL6 supports CD4+ T cell activation following adoptive transfer into VH125SD.BCL6ΔCD4 recipients BCL6 deletion via Cd4-Cre impacts both CD4+ and CD8+ T cells. To evaluate the CD4-specific role of BCL6, we sorted CD4+ T cells from VH125SD.Bcl6ΔCD4 or VH125SD donors, CellTrace Violet (CTV) labeled them, and adoptively transferred them into T1D-protected VH125SD.Bcl6ΔCD4 mice, as depicted in Figure 4A, to investigate their BCL6-dependence in undergoing activation and driving anti-insulin B cell proliferation in vivo. Flow cytometry phenotyping 7 days later identified similar proportions of CTV-labelled cells in spleen, pancreatic lymph nodes, and pancreas regardless of donor T cell genotype, with a large enrichment in the percentage of CTV+ (transferred) cells noted among total CD4+ T cells in the pancreas, relative to spleen or pancreatic lymph nodes (Figure 4B). This suggests the proinflammatory environment that T cells needed for pancreatic infiltration was present, even in this T1D-protected backdrop, and that other mechanism(s) were mediating the diabetes protection observed. BCL6 loss in CD4+ T cells impairs CD4+ T cell activation and promotes Treg formation in some sites but does not affect anti-insulin B cell proliferation in the pancreas CD4+ T cells were isolated from 12 to 14 week-old female donors and ∼5 × 106 BCL6+ or ΔBCL6 CD4+ T cells were CellTrace Violet (CTV)-labeled and adoptively transferred into 8–12 week-old VH125SD.Bcl6ΔCD4.NOD recipients. Seven days after transfer, spleen, pancreatic lymph nodes, and pancreata were assayed as follows. (A) Experimental schematic. (B) (Left) Representative flow plots of CD4+ CTV+ labeled cells in FMO control, spleen, pancreatic lymph nodes, and pancreas. (Right) The proportions of CTV+ labeled cells in spleen, pancreatic lymph nodes, and pancreata are shown for individual recipients following BCL6+ or ΔBCL6 CD4+ transfers. (C–F) CTV+ (transferred) or CTV- (endogenous, non-transferred) CD4+ T cells were assessed regarding: (C) CD44 MFI levels, (D) % CD69+, (E) % Tfh (CXCR5+ PD-1+), and (F) % Treg (FoxP3+). (G–H) % GC B cells (CD95+ GL7+) is shown among total B cells (CD19+ B220+ CD45+ live cells) and (H) % Ki67+ among non-GC (CD95− GL7− CD19+ B220+ CD45+ live) B cells in either insulin-binding (INS+) or non-insulin binding (INS−) B cells. (B–H) n = 5–6 mice per group, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ns, not significant, Kruskal-Wallis test with post hoc test of multiple comparisons (B–F and H) or Mann-Whitney U test (G) was used for analyses. All other comparisons not shown are not significant. Bars represent mean ± standard deviation. BCL6+ transferred (CTV+) CD4+ T cells expressed higher levels of CD44 in spleen, pancreatic lymph nodes, and pancreas compared to transferred ΔBCL6 CD4 T cells (Figure 4C). BCL6+ transferred CD4+ T cells also showed elevated levels of CD44 relative to endogenous (CTV−) CD4+ T cells in in the spleens and pancreatic lymph nodes, but not pancreas of the same recipient mice (Figure 4C). The frequency of activated (CD69+) cells was only increased on transferred BCL6+ CD4+ T cells (relative to transferred ΔBCL6 CD4+ T cells) in the pancreas, but not other organs, with endogenous T cells in the pancreas showing a comparable level of activation (Figure 4D). In comparison to endogenous CD4+ T cells in recipient mice, a reduced frequency of transferred CD4+ T cells expressed the proliferative marker, Ki67 in all organs, regardless of BCL6 presence or absence in the donor CD4+ T cells, but it is plausible that highly proliferative transferred cells could have lost CTV labeling (Figure S5). An increased proportion of Tfh cells were observed among BCL6+ vs. ΔBCL6 transferred CD4+ T cells in the spleen and pancreatic lymph nodes (Figure 4E), with an opposite trend noted for Treg proportions with respect to BCL6 genotype (Figure 4F). The pancreas, by comparison, showed decreased Treg proportions among transferred cells, relative to endogenous CD4+ T cells in recipients, suggesting that despite the T1D-protected backdrop, 7 days was insufficient for normal frequencies of Tregs to develop among transferred cells, even when they expressed the same genotype as the recipient (Figure 4F). Loss of BCL6 in transferred CD4+ T cells does not block anti-insulin B cell proliferation in VH125SDBcl6ΔCD4 recipients We next evaluated whether CD4 T cells depend on BCL6 to drive anti-insulin B cell activation and proliferation in vivo. Of note, the transferred T cell repertoire is polyclonal and likely contains a low proportion of cognate anti-insulin T cells capable of responding to insulin presentation by anti-insulin B cells. Nonetheless, GC B cells were increased (albeit to a modest extent) in the pancreas with BCL6+ CD4 T cell transfer (relative to ΔBCL6 CD4 T cell transfer), with no significant difference observed in spleen or pancreatic lymph nodes (Figure 4G). Regardless of whether BCL6+ or ΔBCL6 CD4 T cells were transferred, roughly 20% of non-GC anti-insulin B cells in the pancreas expressed Ki67 (Figure 4H), which matched the proportion of Ki67+ non-GC anti-insulin B cells in VH125SD mice without CD4 adoptive transfers (Figure 3D). By comparison, non-insulin-binding B cells were relatively quiescent in both recipient groups (Figure 4H). Given that only ∼5% of non-GC anti-insulin B cells expressed Ki67 without adoptive CD4 transfer in the pancreas of VH125SDBcl6ΔCD4 mice (Figure 3D), this suggests that transferred CD4 T cells, regardless of genotype, preferentially elicited insulin-binding B cell proliferation that was otherwise suppressed in VH125SDBcl6ΔCD4 mice. The limited formation of GC B cells in recipients also points to non-GC sites as being competent to support anti-insulin B cell activation. Anti-insulin B cells primarily reside in Extrafollicular niches Follicular exclusion is a common feature among anergic B cells,42 thus anergy may be driving insulin-binding B cell relegation to the extrafollicular (EF) niche. Increasing evidence points to the likely importance or even centrality of atypical B cell populations and EF niches in chronic infections and auto-Ab-mediated diseases such as SLE.32,43,44 Given non-GC anti-insulin B cells show an activated and proliferative (Ki67+) phenotype in pancreas and pancreatic lymph nodes (Figures 3A and 3B), we hypothesized that anti-insulin B cells could be engaging T cells outside of GCs. We therefore used confocal microscopy analysis of splenic tissue sections from VH125SD.NOD mice to identify IgD, insulin-binding, CD3, and Ki67-positive cells. The majority of insulin-binding B cells co-stained with IgD (Figure 5A; Figures S6A and S6B), confirming insulin staining specificity. The vast majority of anti-insulin B cells predominantly localized to EF niches, but not B cell follicular ( BCZ), or GC niches (Figure 5A), consistent with their anergic state.8,45 The majority of T/B infiltrates in the pancreas did not organize into discrete T/B zones (Figure 2L), limiting our ability to accurately determine whether anti-insulin B cells exist in EF niches at this site. Insulin-binding B cells exist in extrafollicular niches and skew toward CD11c+ T-bet+ and CD11b+ CD11c+ atypical B cell subsets, some of which are reduced by loss of Bcl6 Spleen, pancreatic draining lymph nodes, and pancreata were harvested from pre-diabetic, 8–12-week-old VH125SD.NOD mice. (A) Around 10 μm sections of spleen form VH125SD.NOD underwent immunofluorescence staining with IgD, Ki67, insulin, and CD3. Representative merge image shown on left with identified borders of B cell zone (BCZ), T cell zone (TCZ), germinal center (GC) and extrafollicular area (EF). Circled cells are representative insulin-autoreactive B cells, which express both IgD and insulin. Quantification of % insulin+ B cells in predefined niches on the left graph. Quantification of insulin+ IgD + colocalization, or insulin+ IgD-. n = 4–6 individual splenic sections per group. (B) Representative flow plots show CD11c+/CD11b+ cells among B220+ CD19+ live singlet lymphocytes for both non-insulin-reactive (insulin−, left) and insulin-autoreactive (insulin+, middle, right) B cell populations in the pancreas. (C) Quantification of the proportion of CD11c+ CD11b+ B cells in pancreata for both insulin− and insulin+ B cells for VH125SD (black) and VH125SDBcl6ΔCD4 (purple) mice. (D) Representative flow plots of CD11c+/Tbet+ cells among B220+ CD19+ live singlet lymphocytes for both insulin− and insulin+ B cell populations in the pancreas. (E) Quantification of the proportion of CD11c+/Tbet+ cells among both insulin− and insulin+ B cells in the pancreas of VH125SD (black) and VH125SDBcl6ΔCD4 (purple) mice. (F and G) Quantification of the proportion of CD11c+/Tbet+ B cells and CD11b+/CD11c+ B cells in pancreatic draining lymph nodes (F) and spleen (G). (A–G) One-way ANOVA (A, left), Mann-Whitney U test (A, right) (A) or Kruskal-Wallis test (C–G) with post hoc multiple comparison test were used, n = 4–8 mice per group. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns, not significant. All other comparisons not shown are not significant. Bars represent mean ± standard deviation. BCL6 in T cells promotes anti-insulin B cell skewing toward atypical B cell subsets in T1D-associated organs Given that anti-insulin B cells resided primarily in EF niches (Figure 5A), we wanted to assess whether they skew toward atypical B cell phenotypes defined by CD11b, CD11c, and/or T-bet expression in T1D-prone mice.46 Representative gating (Figures 5B and 5D) and analysis of individual mice revealed that insulin-binding B cells skewed toward CD11b+ CD11c+ B cells in the pancreas and pancreatic lymph nodes, compared to non-insulin binding B cells, which was blocked by loss of BCL6 (Figures 5C and 5F). Insulin-binding B cells sporadically adopted a CD11c+ T-bet+ phenotype in the pancreas and pancreatic lymph nodes of some mice, with no or limited reduction by T cell loss of BCL6, respectively (Figures 5E and 5F). In contrast, we observed no differences in either of these atypical B cell populations in the spleen, regardless of insulin-reactivity or T cell BCL6 genotype (Figure 5G). Overall, these data show that insulin-binding B cells skew toward CD11b+ CD11c+ atypical B cell subsets in T1D-associated organs, which was reduced by T cell loss of BCL6. BCL6 in T cells supports insulin-binding B cell skewing toward an atypical memory-like population Given the heterogeneity of atypical B cells and to avoid user bias in identifying B cell changes when BCL6 is lost by T cells, we performed t-distributed Stochastic Neighbor Embedding (t-SNE)-based clustering analysis of anti-insulin B cells using a 27-marker panel, including atypical (CD11b, CD11c, and T-bet), plasma cell/plasmablast (CD138, TACI), memory (CD80, PDL2, and CD73), and activated (CD44, CD69, and CD86) B cell markers. We identified four clusters in the pancreas (Figure 6A). Of those, cluster 1 was significantly reduced in VH125SDBcl6ΔCD4 mice relative to VH125SD controls, while clusters 3 and 4 trended upwards (Figure 6B). Cluster 1 exhibited increased Ki67, CD11c, CD11b, PDL2, CD80, and CD86, consistent with a proliferative, atypical memory B cell cluster (Figures 6C and 6D). PDL2 and CD80 are not exclusively found on memory B cells, thus future studies are required to determine whether this population is memory. Pancreatic draining lymph nodes exhibited a similar reduction of this potential atypical-memory B cell population (which also expressed Ki67, CD11b, Tbet, PDL2, CD80, and the plasma cell marker, CD138) with T cell loss of BCL6 (Figures S7A–S7D). BCL6 loss in T cells promotes formation of a phenotypically defined atypical memory B cell population in the pancreas of VH125SD.NOD mice Pancreata were harvested from pre-diabetic, 8–12-week-old Bcl6-sufficient (VH125SD) and Bcl6-deficient (VH125SDBcl6ΔCD4.NOD) mice (n = 12–14 mice) and flow cytometry staining was performed using the panel of markers in Table S1 and (D). B220+ CD19+ CD45+ live singlets were concatenated and normalized via CyCombine and concatenated, with genotype metadata encoded. (A) t-SNE was used to perform dimensionality reduction based on phenotypic marker expression profiles. Clusters were manually defined as indicated. (B) The cluster frequency is shown for each genotype, with individual mice plotted and means indicated. ∗p < 0.05, Kruskal-Wallis test with post hoc multiple comparison test. Bars represent mean ± standard deviation. (C) A rainbow intensity scale indicates expression levels of the phenotypic markers indicated at the top of each t-SNE plot. Insulin-binding, CD45, B220, CD19, and viability dye markers were omitted from the t-SNE analysis given they were used in parent population gating upstream of t-SNE analysis. (D) Heatmap shows relative expression of the indicated markers for each cluster defined as in (A) using a log-like arcsinh scale and appropriate arcsinh factors. Discussion Overall, this study identified that BCL6-expressing T cells redirected insulin-binding B cells toward pro-pathogenic fates of atypical and atypical-memory-like populations, which were associated with diabetes development. Preferential proliferation of insulin-binding B cells persisted in adoptive transfer experiments in which neither transferred CD4 nor endogenous CD4/CD8 T cells expressed BCL6, suggesting that productive T/B interactions can occur independently of BCL6, provided whatever local suppressive environment is maintained in VH125SD.BCL6ΔCD4.NOD mice is disrupted. We previously observed T cell/BCL6-mediated diabetes protection in polyclonal NOD mice.30 Our data suggest this protection was not solely due to a block in the initial expansion of insulin-reactive B cells but also their acquisition of functional potential to foster pro-inflammatory anti-insulin T cell development. BCL6 loss in T cells provides robust protection against T1D despite the pre-determined insulin-reactive B cell tolerance breach in the VH125SD.NOD model. This suggests that BCL6 inhibition may succeed in controlling pro-pathogenic anti-insulin B cell activity, even after this critical APC population has breached immune tolerance mechanisms to expand in the repertoire. BCL6 inhibitors are currently in clinical trials for treatment of B cell non-Hodgkin’s lymphoma,47,48 streamlining the process for regulatory approval as a new potential T1D immunotherapy. Future studies will be required to determine the therapeutic potential of BCL6 inhibitors in T1D. We previously found that anti-insulin B cells can sporadically enter GCs.41 However, our results show that anti-insulin B cells primarily exist in the EF area and skew toward an atypical memory B cell phenotype, suggesting that EF responses, perhaps in addition to GC responses, may support T1D pathogenesis. We found that anti-insulin B cells selectively upregulated CD69 ∼2-fold in the pancreas of VH125SD.NOD mice, with an increased skewing toward a proliferative phenotype (Ki67+) relative to non-insulin-binding B cells, which was ablated by T cell loss of BCL6. Passive serum transfer studies show that islet autoantibodies are not directly pathogenic in T1D.9 Selective elimination of MHCII on B cells, but not other APCs, prevented T1D in NOD mice, pointing to antigen presentation as the chief pathogenic role of B cells in T1D.10 The VH125SD.NOD and VH125Tg.NOD models show limited or no anti-insulin IgG, respectively, yet accelerated diabetes development in both models,8,49 which is attributed to anti-insulin B cell antigen presentation function.12 We therefore favor changes in anti-insulin B cell APC function as the underlying B cell mechanism of protection here, as opposed to anti-insulin IgG suppression. Other B cell specificities could be supporting T1D in the VH125SD.NOD model, though we assume they play a minor role compared to anti-insulin B cells. Greater than 95% of the B cells express the anti-insulin heavy chain,8 presumably creating a semi-polyclonal repertoire that is less diverse than in non-BCR transgenic mice. Prior studies showed that mutating only two amino acids in VH125Tg to produce the VH281Tg model eliminated insulin-binding B cells from the BCR transgene-restricted/semi-polyclonal repertoire and conferred disease protection.49 Thus, while it’s possible that other islet reactive B cells are driving disease in VH125SD.NOD mice, we assume this is a rather insulin-centric model that, once immune tolerance breach to insulin occurs to initiate beta cell destruction, could undergo epitope spread to other islet autoantigens through T cell interactions with dendritic cells and other APCs. In the setting of adoptive transfer, both BCL6+ and ΔBCL6 CD4+ T cells enhanced insulin-binding B cell proliferation in the pancreas, suggesting a BCL6-independent mechanism of insulin-binding B cell proliferation at that site, which was negatively associated with Treg skewing by the transferred CD4 T cells. This suggests that cognate anti-insulin T cells form in both VH125SD and VH125SDBcl6ΔCD4 mice but are only restrained from supporting insulin-binding B cell proliferation in VH125SDBcl6ΔCD4 mice. This functional suppression, which may be indirectly mediated, is overcome with adoptive CD4 transfer. While Treg formation is not changed in the pancreas by loss of BCL6, their anatomical proximity to anti-insulin B and effector T cells could be. Thus, one possible interpretation of our findings is that local Treg suppression is a vital contributor to whether insulin-binding B cells and CD4+ effector T cells undergo activation and proliferation within a specific niche. Atypical B cells have been implicated in other autoimmune diseases.44,50 This is the first study to highlight their increased presence in an NOD model of T1D, in which we observed significant enrichment of insulin-binding B cells among CD11b+ CD11c+ populations in the pancreas and pancreatic lymph nodes, which was reduced by T cell loss of BCL6. The low numbers of insulin-reactive B cells within these populations preclude formal atypical B cell-focused antigen-presentation and transfer studies in this model. Future experiments using alternative models are thus needed to directly evaluate the potential of these subsets in supporting the activation and differentiation of pathogenic T cell clones as a key step in T1D pathogenesis. Overall, our study identifies BCL6 as a pivotal transcriptional regulator in T cells that supports anti-insulin B cell activation, proliferation, IgG class switching, GC B cell differentiation, and skewing into the activated and proliferative CD11b+ CD11c+ atypical and atypical memory B cell subsets in T1D. Our data point to T cell/BCL6 dependence of activating insulin-binding B cells within certain atypical B cell populations (chiefly residing in the EF, niche) as a potential explanation for diabetes protection conferred by T cell loss of BCL6. Limitations of the study Cd4-Cre eliminates Bcl6 in both CD4 and CD8 T cell populations due to the double-positive stage in the thymus.51 While other non-T cell populations can also express CD4 (e.g., CD4+ dendritic cells52), we referred to this model as T cell-specific loss of Bcl6 out of convenience but acknowledge the potential role of other CD4+ populations in mediating the observed diabetes protection. Additional experiments are therefore required to determine the individual contributions of BCL6 in specific CD4, CD8, and other immune populations in the context of T1D pathogenesis. We also did not examine how proinflammatory, antigen-specific T cells were modified by T cell loss of BCL6, though it is clear from the diabetes protection observed that they were adequately reduced and/or controlled. Loss of BCL6 also impairs Tfrs; Tfr cells have been shown to promote B cell APC function in other autoimmune diseases.53 This paper cannot delineate the role of Tfr cells in promoting T1D in this model, and future research will have to selectively target this population on whether this mediates disease. T cell loss of BCL6 blocks both GC and at least some atypical B cell populations (e.g., CD11c+ T-bet+ B cells29), thus we cannot discriminate whether the disease protection observed in VH125SD.Bcl6ΔCD4 mice is related to crippled EF vs. GC responses. Additionally, while populations of insulin-binding B cells were identified that expressed PDL2, CD80, and/or CD73 markers that are associated with memory B cell populations54 (sometimes in combination with atypical B cell markers), additional evidence will be required to confidently call any of the defined populations outlined here as memory. Further experiments will be required to formally evaluate the reliance of T cells on GC vs. EF B cells to drive T1D. Resource availability Lead contact Further information and requests resources should be directed to and will be fulfilled by the lead contact, Rachel H. Bonami (rachel.h.bonami@vumc.org). Materials availability Requests for new materials for this paper will be directed to and fulfilled by the lead contact upon request. Data and code availability •Data reported in the paper will be shared by the lead contact upon request Data reported in the paper will be shared by the lead contact upon request •The paper does not report original code The paper does not report original code •Any additional information to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments This work was supported by Breakthrough T1D grants 2-SRA-2023-1453-S-B (to R.H.B.), 2-SRA-2026-1859-S-B, and 3-PDF-2024-1495-A-N (to D.H.McN.), and NIH grants F30 DK145130 (to L.M.C.), F31 DK141224 (to L.E.B.), T32 GM007347 (to L.M.C. and C.M.N.), T32 AR059039 (to T.W.J.), TL1TR002244 (to C.M.N.), and R01DK105550 (to J.C.R.). We also acknowledge the Vanderbilt University Medical Center (VUMC) Flow Cytometry Shared Resource [supported by the Vanderbilt-Ingram Cancer Center (P30 CA068485) and the Vanderbilt Digestive Disease Research Center (P30 DK058404), the VUMC Tissue Pathology Shared Resource (supported by NCI/NIH Cancer Center support grant, P30 CA068485), the Vanderbilt Cell Imaging Shared Resource (supported by NIH grants CA68485, DK58404, and EY08126), and the Islet and Pancreas Analysis Core [supported by the Vanderbilt Diabetes Research and Training Center (NIH grant P30 DK20593)]. We would like to acknowledge Dr. Alexander Dent (Indiana University) for providing the original Bcl6fl/fl.C57BL/6/129Sv mice, which we backcrossed to the NOD strain and then introgressed the VH125SD transgene for these studies. We also acknowledge Dr. Mark Boothby (Vanderbilt University Medical Center) for critical manuscript review. Author contributions Conceptualization, L.M.C., D.H.M., and R.H.B.; methodology, L.M.C., J.C.M., D.H.M., M.L.P., T.W.J., L.E.B., and C.M.N.; investigation, L.M.C.; writing – original draft, L.M.C., J.C.M., D.H.M., and R.H.B.; writing – review and editing, L.M.C., J.C.M., D.H.M., J.C.R., and R.H.B.; funding acquisition, L.M.C., D.H.M., L.E.B., J.C.R., and R.H.B.; resources, J.C.R. and R.H.B. Declaration of interests R.H.B. has unrelated funding from the National Institutes of Health, Breakthrough T1D, argenx, and the Leona M. and Harry B. Helmsley Charitable Trust. J.C.R. is Founder and Scientific Advisor of Sitryx Therapeutics. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesB220BD Biosciences (RA3-6B2)756542; RRID: AB_3688772CD4Thermofisher (GK1.5)78-0041-82; RRID: AB_2722965CD8BioLegend (53–6.7)100731; RRID: AB_893427CD11bBioLegend (M1/70)101261; RRID: AB_2572121CD11cBioLegend (N418)117321; RRID: AB_755987CD19Thermofisher (1D3)64-0193-82; RRID: AB_2688119CD21/35BioLegend (7E9)123433; RRID: AB_2860651CD23BioLegend (B3B4)101623; RRID: AB_2563705CD38BD Biosciences (90/CD38)771651; RRID: AB_3693480CD3e (Hamster)BD Biosciences (500A2)553238; RRID: AB_394727CD44BioLegend (IM7)103085; RRID: AB_3106052CD45BioLegend (30-F11)103183; RRID: AB_2904274CD69Thermofisher (H1.2F3)63-0691-82; RRID: AB_2688097CD73BioLegend (TY/11.8)127213; RRID: AB_11219608CD80BD Biosciences (16-10A1)570553; RRID: AB_3685839CD86BD Biosciences (B7-2)565479; RRID: AB_2739258CD95 (Fas)BD Biosciences (Jo2)756863; RRID: AB_3668956CD138BioLegend (281-2)142505; RRID: AB_10960141Cell Trace VioletThermofisherC34557ACXCR5BioLegend (L138D7)145511; RRID: AB_2562127Foxp3Thermofisher (FJK-16s)45-5773-80; RRID: AB_914349GL7BD Biosciences (GL7)553666; RRID: AB_394981Human insulinSigma-AldrichI2643ICOSBioLegend (C398.4)313521; RRID: AB_10895762IgDBD Biosciences (217-170)747077; RRID: AB_2871833IgDBD Biosciences (A85-1)405718; RRID: AB_10730619IgG-Alkaline PhosphataseSouthern Biotech1030–04; RRID: AB_2794293IgG (H + L) Goat anti-hamster secondaryThermofisher ScientificA-21112; RRID: AB_2535761IgG1BD Biosciences (A85-1)746811; RRID: AB_2744066IgG2bBioLegend (RMG2b-1)406713; RRID: AB_2750279Igk light chainBD Biosciences (187.1)755937; RRID: AB_3688224IgMBioLegend (RMM-1)406543; RRID: AB_2861023Ki67Thermofisher (SolA15)407-5698-82Ki67BioLegend (11F6)151206; RRID: AB_2566801PD-1BioLegend (RMP1-30)109103; RRID: AB_313420PDL2BD Biosciences (MIH37)758326; RRID: AB_3690466StreptavidinBioLegend405247; RRID: AB_3697286TACI (CD267)BD Biosciences (8F10)742840; RRID: AB_2741091TbetBioLegend (4B10)644809; RRID: AB_2028583Viability DyeBeckman CoulterC36628Chemicals, peptides, and recombinant proteinsCollagenase PSigma11213857001CD4+ T cell isolation kit, mouseMiltenyi130-104-454OCT CompoundAgar ScientificAGR1180Tween 20, 0.5% (v/v)Sigma AldrichP1379Bovine Serum AlbuminJackson Immunoresearch001-000-161Normal Goat SerumSigma Aldrich566380Streptavidin Biotin Blocking KitVector LaboratoriesSP-2002Triton X-100Sigma AldrichT8787Prolong Gold Antifade MountantThermofisherP36934Critical commercial assaysFoxp3 Transcription Factor Staining Buffer SeteBioscience#00-5323-00Experimental models: Organisms/strainsVH125SD.NOD Bcl6fl/flBackcrossed from transgenic strains gifted by the James W. Thomas and Alexander Dent LaboratoriesSee also: The Jackson Laboratory, Strain 029288VH125SD.NOD Bcl6fl/fl CD4-Cre ±Backcrossed from transgenic strains gifted by the James W. Thomas and Alexander Dent LaboratoriesSee also: The Jackson Laboratory, Strains 029288 and 013234NOD Bcl6 fl/flThe Jackson Laboratory and Alexander Dent LaboratorySee also: The Jackson Laboratory, Strain 001976 and https://pubmed.ncbi.nlm.nih.gov/39556799/Software and algorithmsR v4.5.1The R Projecthttps://www.r-project.orgPrism v9.3.1GraphpadRRID: SCR_002798ImageJNIHhttps://imagej.netImageScope softwareAperioRRID: SCR_020993FlowJo v10BD Bioscienceshttps://www.flowjo.com/solutions/flowjoCytobank v9.2 R packageBeckmanhttps://www.beckman.com/flow-cytometry/software/cytobank-premiumDplyr_1.1.4 R packageOpen sourcehttps://dplyr.tidyverse.org/Stringr_1.5.1 R packageOpen sourcehttps://cran.r-project.org/web/packages/stringr/index.htmlReadr_2.1.5 R packageOpen sourcehttps://cran.r-project.org/web/packages/stringr/index.htmlcyCombine v0.2.19 R packagecyCombinehttps://github.com/biosurf/cyCombine Experimental model and study participant details Animals Bcl6fl/fl.C57BL/6/129Sv mice51 (kindly provided by Dr. Alexander Dent, Indiana University) were backcrossed at least 14 generations to the NOD background and intercrossed with Cd4-Cre.NOD mice (strain number: 013234; The Jackson Laboratory, Bar Harbor, ME55) to generate Bcl6fl/fl.Cd4-Cre+/−.NOD mice, as described in.30 This loss generates a non-functional BCL6 protein whereby the DNA-binding domain is lost, but the rest of the protein is still expressed. Bcl6fl/fl Cd4-Cre.NOD mice were then crossed with VH125SD.NOD mice,8 a gift from Dr. James W. Thomas (and available at The Jackson Laboratory, strain ID 029288), in which an anti-insulin IgH transgene is targeted to the IgH locus (site-directed, SD), to generate VH125SD+/−.Bcl6fl/fl.Cd4-Cre+/− (termed VH125SD.Bcl6ΔCD4) and control VH125SD+/−.Bcl6fl/fl (termed VH125SD) NOD lines. All studies compared littermates to limit microbiome-related differences across groups. All mice were housed under specific pathogen-free conditions and given autoclaved food and water. For all studies, mice were euthanized via isoflurane. Cervical dislocation was subsequently performed on deceased animals. All experiments utilized female mice in line with increased female NOD diabetes incidence, and unless otherwise stated, were 8–12 weeks old. These animal studies were fully approved by the AALAC-accredited Vanderbilt Institutional Animal Care and Use Committee (IACUC). Study approval All experiments involving animals were approved by Vanderbilt University IACUC and conform to regulatory standards, Animal protocol Ethics Approval Number #M1800152. Method details Diabetes monitoring Blood glucose was measured weekly from 10 to 35 weeks of age in female littermates by nicking tails. Diabetes was diagnosed after two consecutive blood glucose readings >250 mg/dL. Histological assessment of insulitis and tertiary lymphoid structure organization Mice were sacrificed and pancreata were dissected from non-diabetic female mice. Pancreata were incubated overnight in 10% formalin at room temperature and then dehydrated briefly in 70% ethanol and paraffin embedded. 10 mm sections were cut, deparaffinized, and subsequently stained with H&E by the Vanderbilt Tissue Pathology Shared Resource. Slide images were obtained using a bright field Aperio ScanScope CS. Visualization and analysis was performed by using ImageScope software (Leica Biosystems). Islets were individually blind scored for insulitis (lymphocyte invasion) as described in.56 Pancreas blocks which exhibited insulitis via H&E staining above were chosen from n = 6 mice per genotype for subsequent CD3 and B220 immunohistochemical staining using a Leica Bond-Max IHC stainer. Following heat-induced antigen retrieval with Bond Epitope Retrieval Solution 2 (Leica Biosystems), 10 mm pancreas sections were stained with antibodies reactive against either CD45R/B220 (RA3–6B2) or CD3 (M-20). Islets were scored on their average infiltration of CD3 (T cells) and B220 (B cells) as described.30 Islets with an insulitis severity score of 2 or higher were further blind scored as “organized” (defined by organized B/T cell zones) or “disorganized” (defined by B/T infiltration that was not organized into discrete tertiary lymphoid structure zones). Cell isolation from tissues Cells were isolated from spleen, pancreatic lymph nodes (n = 2 nodes per mouse), and mesenteric lymph nodes (n > 5 nodes per mouse) as previously described.30 Pancreata were promptly dissected and digested with 1 mg/mL of collagenase P (Sigma) in HBSS and incubated while shaking at 37°C for 30 min. The suspension was then disrupted with an 18G needle and resuspended in HBSS +10% BCS to inactivate collagenase and cells were pelleted and resuspended in flow cytometry staining buffer (detailed below). Cell counts were performed using a Bio-Rad Tc20 Automated Cell counter. Antibodies and spectral flow cytometry Cells were stained for flow cytometry utilizing the murine reactive antibodies listed in Table S1. We previously observed that endogenous (murine) insulin “masks” insulin-binding B cell detection with biotinylated human insulin in the pancreas, but not lymph nodes or spleen.57 This technical issue is overcome in VH125-derived B cells by staining cells with a second antibody (mAb123) that binds a different insulin epitope than mAb125.57 Thus, for pancreas staining, cells were first incubated with human insulin (Sigma #I2643) for 30 min in flow cytometry (FACS) staining buffer (1× PBS +5% BCS +0.1% azide +0.02% EDTA), followed by detection with anti-insulin mAb123-biotin/SA-fluorochrome, as described.57 Cells were first incubated with Fc Block (BD Biosciences) (and non-biotinylated human insulin for pancreata) per the manufacturer’s instructions, followed by surface Ab staining (with human insulin for lymph nodes and spleen, mAb123 for pancreata) for 1 h at 4°C in FACS buffer. Cells were washed three times with FACS buffer and then stained with streptavidin-fluorochrome for 1 h at 4°C. Cells were then incubated with Foxp3/Transcription Factor Staining buffer (eBioscience, #00-5323-00) for 30 min at 4°C. Samples were washed twice with 1× permeabilization buffer from the kit and stained with the intracellular Ab mix in 1× permeabilization buffer overnight at 4°C. After overnight incubation, cells were washed three times with 1× permeabilization buffer, resuspended in FACS buffer, and acquired on the Aurora Spectral Cytometer (Cytek). Cells for single-color controls were prepared similarly as described. Flow cytometry data were analyzed using FlowJo v10, R v4.5.1, and Cytobank v9.2. The following R packages were used for analyses: dplyr_1.1.4, stringr_1.5.1 readr_2.1.5, magrittr_2.0.3, and cyCombine_0.2.19. Adoptive transfers of BCL6+ or ΔBCL6 CD4+ T cells Following splenic dissection and red blood cell lysis as described previously, splenocytes from 12 to 14 week-old female VH125SD mice or VH125SD.BCL6ΔCD4 or mice were labeled with CellTrace Violet (CTV) (Thermo Fisher C34557 A) per the manufacturer’s instructions, with staining performed in RPMI at 37°C for 5 min. Cells were then washed 2 times with RPMI. CD4+ T cells were then isolated via a CD4+ T cell isolation kit, mouse (Miltenyi, #130-104-454) per the manufacturer’s instructions. Approximately 5 × 106 BCL6+ or ΔBCL6 CD4 T cells were injected i.p. into 8-12 week-old female VH125SD.Bcl6ΔCD4 recipients. 7 days after transfer, spleen, pancreatic lymph nodes, and pancreata were assayed via flow cytometry to measure marker expression as indicated in figure legends. Confocal microscopy Confocal imaging was performed as described in.58 Briefly, tissues were fixed in periodate-lysine paraformaldehyde (PLP) solution for 4–6 h, incubated with 30% sucrose for 1–3 days, and cryopreserved utilizing OCT compound (Agar Scientific, AGR1180). Following this, tissues were sliced utilizing the Leica Cryostat CM1850 into 10 μm sections. Tissue sections were then rehydrated into a staining jar or dish with Wash Buffer (Tween 20, 0.5% (v/v), Sigma Aldrich Cat# P1379 and 1% PBS). Blocking Buffer (Bovine Serum Albumin 2% (w/v), Jackson ImmunoResearch Cat# 001-000-161, and Normal Goat Serum 10% (v/v), Sigma Aldrich Cat# 566380 and 1× PBS) was placed onto tissues for 60–120 min at room temperature to minimize non-specific binding of primary and secondary antibodies. Tissues were then washed and incubated with the Streptavidin/Biotin blocking kit (Vector Laboratories, Cat# SP-2002) Following this, cells were permeabilized using permeabilization buffer for 30 min at room temperature (Triton X-100 2%, Sigma-Aldrich, Cat# T8787, and 1× PBS). Cells were then washed and incubated with primary antibody cocktail as listed in Table S2 for 60–120 min. Tissues were then washed with wash buffer and incubated with secondary antibody cocktail as listed in Table S2. Slides were washed and mounted with mounting medium (ProLong Gold Antifade Mountant, ThermoFisher, Cat# P36934). Images were acquired on the Zeiss LSM780 via the 10× or 20 × 0.9 NA Plan Apo. Regions of B cell zone, T cell zone, germinal center, and extrafollicular sites were drawn by blinded experimenter, and the proportion of insulin+ B cells was quantified utilizing Imaris once borders were drawn for VH125SD-BCL6 sufficient mice. IgD+ HINS+ and IgD- HINS+ cells were quantified via manual counting. Minimally supervised analysis of flow cytometry data using t-SNE Flow cytometry (.fcs) files from samples stained with markers outlined Table S1 and Figure 6C were imported into R and underwent CyCombine normalization.59 Batch correction was performed using cyCombine v0.2.19 in R. Data were arcsinh-transformed (cofactor = 6000) and down-sampled to the minimum cell count across samples to ensure equal representation. Batch correction was then applied with an 8 × 8 self-organizing map grid, 'scale' normalization method, and 'condition' as covariate to preserve biological differences. Next, .fcs files were concatenated with key labeling conditions (Cre- and Cre+) and imported into Cytobank. Appropriate arcsinh cofactors were set for each marker as previously described60 and exported, and samples were gated on CD45+ B220+ CD19+ live singlet cells. Insulin-binding B cells were detected as outlined above and analyzed using viSNE on Cytobank with default parameters (iterations = 1000, perplexity = 30, theta = 0.5), with appropriate arcsinh cofactors set for each variable. Markers used to identify insulin-binding B cells were omitted from the subsequent marker expression and heatmap analyses for pancreas, and pancreatic draining lymph nodes. Density plots were created for each concatenated group (split by genotypes, density percent per contour 10%) and used to guide manually drawn clusters. Individual mouse percentage of each cluster were plotted, and heatmaps were generated to show marker expression for each of these clusters. ELISAs Competitive binding ELISA was used to detect insulin-specific antibodies, as outlined previously.30 Briefly, 384-well Maxisorp Nunc plates (Thermo Scientific) were coated with 1 μg/mL human insulin in borate-buffered saline overnight at 37°C. Sera were diluted 1:100 in 1× PBS+0.1% Tween 20 (PBS-T) and incubated in wells for 1 h at room temperature. To determine insulin-specific IgG, parallel samples were incubated in the presence of 100 μg/mL human insulin to enable subtraction of O.D. values from non-inhibited wells. After sera incubation, goat anti-mouse IgG secondary conjugated to alkaline phosphatase (catalog number 1030-04; Southern Biotech, diluted 1:250) was added. Plates were washed with PBS-T after each of the steps above. Wells were incubated with 10 mg/mL p-nitrophenyl phosphate substrate (Sigma-Aldrich) in a 50 mM potassium carbonate +1 mM magnesium chloride buffer. Optical density was read at 405 nm after 30 min using a Synergy LX Microplate Autoreader (Bio-Tek). Quantification and statistical analysis Statistical tests are indicated in the corresponding figure legends and significance values were calculated using GraphPad Prism v9.3.1 (GraphPad Software). For animal experiments, n refers to the number of mice per group unless otherwise stated in the figure legends. Unless otherwise stated, bars represent mean ± standard deviation. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns = not significant. Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.isci.2026.115990. Supplemental information Document S1. Figures S1–S7, Tables S1, and S2 References Associated Data Supplementary Materials Document S1. Figures S1–S7, Tables S1, and S2 Data Availability Statement •Data reported in the paper will be shared by the lead contact upon request Data reported in the paper will be shared by the lead contact upon request •The paper does not report original code The paper does not report original code •Any additional information to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information to reanalyze the data reported in this paper is available from the lead contact upon request.