bioRxivImmunologyPreprint
Transcriptionally distinct atypical B cell populations contribute to insulin autoreactivity and show differential activation, isotype switching, and somatic hypermutation in early-stage type 1 diabetes individuals
Lindsay E BassiD, Landon M. ClarkiD, Bryan Han, Daniel J. MooreiD, Rachel H. BonamiiD
Abstract
Autoreactive B cells must escape immune tolerance mechanisms to drive autoimmune disease, potentially through alternative paths to memory and antibody-secreting cell formation. "Atypical" B cells represent one such alternative pathway and are expanded in several human autoimmune diseases. Atypical B cells display phenotypic heterogeneity, but how this connects with differences in transcriptional programming, isotype switching and somatic hypermutation is incompletely understood in type 1 diabetes (T1D). Screening efforts to detect T1D prior to clinical diabetes onset affords an opportunity to study B cells early in immunopathogenesis, in which they primarily function as antigen-presenting cells. Prior evidence points to atypical B cell populations as reservoirs for islet autoreactivity in human T1D. We therefore developed sorting strategies to enrich for atypical and other autoreactive-prone B cell populations from the peripheral blood of early-stage T1D individuals and performed single-cell RNA/BCR-sequencing. We identified four transcriptionally defined atypical B cell clusters, each showing differential upregulation of atypical B cell defining genes. Additional gene expression changes were noted in broadly transcriptionally similar, but phenotypically distinct atypical B cell populations, with differences noted in the average extent of isotype switching and somatic hypermutation. Some atypical B cell populations showed extensive clonal sharing with classical memory or plasmablast subsets, consistent with relationships amongst these subsets. B cells reactive against a key islet autoantigen, insulin, were enriched in select atypical B cell populations. These findings highlight specific atypical B cell populations and features with potential value as biomarkers of disease progression or response to immunotherapy in T1D, and potentially other autoimmune diseases.
- organism: human
- assay: scRNA-seq
- research focus: clustering
- research focus: differential expression
- research focus: cell type annotation
- research focus: gene regulatory network inference
- approach: experimental assay
- approach: mixed model
- approach: statistical model
In-depth summary
Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (2 Oct 2026) · CC BY 4.0
B cells from 24 early-stage type 1 diabetes individuals include four atypical clusters differing in switching and mutation; Cluster 10 had the highest insulin-binding antibody rate (5/16).
Atypical B cells are expanded in several autoimmune diseases and earlier work points to them as reservoirs of islet autoreactivity, but how their phenotypic variety relates to transcriptional programs, class switching and somatic hypermutation in type 1 diabetes (T1D) was poorly understood. Sorting enriches these cells from screened early-stage individuals for paired single-cell transcriptome and B cell receptor (BCR) sequencing, and expressed BCRs are tested for insulin binding.
Key results
ShownOf 18 Seurat clusters, four (Clusters 4, 6, 7, 10) were atypical B cells. Each upregulated a different mix of atypical-defining genes, and only FCRL5 was raised in all four relative to pooled classical memory B cells.
77,031 cells (71,566 B cells) from 24 participants. Versus classical memory (32,311 cells), FCRL5 was the only gene up in all four atypical clusters; CXCR4, TXNIP, ZFP36L1 and ZFP36L2 were down in all. Differentially expressed gene counts ran from 72 (Cluster 4) to 735 (Cluster 7). Cluster 6 had the highest levels of the six atypical genes; Cluster 10 exceeded classical memory in CD27, AIM2 and TNFRSF13B. · 3.1 CD27+ CD21lo atypical B cells are…; 3.2 Transcriptionally defined atypical B cell…; 3.3 Distinct atypical B cell clusters show…; Figure legends
SuggestedBoth atypical and classical memory B cells yielded insulin-binding antibodies. The rate was highest in Cluster 10, the atypical cluster with the most class switching and clonal expansion, and none of the Cluster 6 antibodies tested bound insulin.
160 recombinant IgG1 antibodies from 12 participants were screened by insulin ELISA (positive if optical density > 0.33, 3-fold over the negative control, then confirmed by titration). Rates: Cluster 10 5/16 (31.3%), Cluster 4 2/35 (5.7%), classical memory 5/58 (8.6%), plasmablasts 2/19 (10.5%). The 14 binders came from 6 individuals; 9/14 had the CD27+ CD21lo sort phenotype. · 3.10 Insulin-binding B cells are found at an…; 4 Discussion; Figure legends
ShownIsotype switching varied widely between atypical clusters and, in Cluster 6, between sort phenotypes within the same transcriptional cluster.
Mean switched fraction per participant: Cluster 10 71.5% (higher) and Clusters 4 20.1%, 7 37.0%, 6 38.7% (lower), vs 47.6% for classical memory. Cluster 10 had more IgG2 (11.2%), IgA1 (25.3%) and IgA2 (9.4%); Cluster 6 had more IgG3 (6.5%). Within Cluster 6, CD27+ CD21lo cells were 51.7% switched vs 18.3% for P9 cells (beta-binomial mixed model). · 3.7 Early-stage T1D individuals show variation…; Figure legends
ShownHeavy-chain V gene somatic hypermutation was lower than classical memory in Clusters 6 and 4 and highest in Cluster 10. CD27+ CD21lo cells carried more mutations than the other three sort phenotypes in Cluster 6, and more than the CD27- CD21lo (P9) cells in Clusters 4 and 10.
Mean %VH mutation per participant: Cluster 6 3.2% and Cluster 4 3.4% (significantly lower) vs 4.8% for classical memory; Cluster 10 5.4%. In Cluster 6 the CD27+ CD21lo population exceeded the other three sort populations; in Clusters 4 and 10 it exceeded the P9 population (precision-weighted linear mixed model with a donor random intercept). · 3.8 BCR somatic hypermutation varies across and…; Figure legends
ShownClusters 4 and 10 were more clonally expanded than classical memory. Within participants, Cluster 4 and classical memory shared the most clones, and Cluster 7 shared many clones with plasmablasts.
Clonal expansion: Cluster 10 7.3%, closest to plasmablasts at 12%; classical memory clusters 0% to 2%. Shannon index ran from 3.8 (Cluster 7) to 7.1 (classical memory). Shared expanded clones: Cluster 4 with classical memory n = 81 (15/18 participants); Cluster 7 with plasmablasts n = 42 (12/15); Clusters 4 and 6 n = 23. · 3.9 Atypical B cell clusters exhibit increased…; Figure legends
Methods, data and reproducibility
Peripheral blood mononuclear cells from early-stage T1D individuals were sorted into four CD19+ populations by CD27, CD21, CD24 and CD38, then profiled by single-cell RNA/BCR-seq (BD Rhapsody). Seurat clusters were compared on expression, isotype, mutation and clonality. Selected, mostly clonally expanded BCRs were expressed as recombinant IgG1 antibodies and screened for insulin binding by ELISA.
- Blood mononuclear cells from early-stage T1D individuals (TrialNet Pathway to Prevention), sorted into four B cell populations for single-cell RNA/BCR-seq · human · n = 24 participants; 77,031 cells (71,566 B cells) · SRA BioProject PRJNA1489232
- Recombinant monoclonal antibodies built from single-cell BCR sequences and screened for insulin binding · human · n = 160 antibodies (143 expanded, 17 non-expanded clones) from 12 participants
- Anti-insulin monoclonal antibody sequences · human · n = 14 antibodies from 6 individuals · GenBank (accession numbers in Supplemental Table S11)
- Methods & tools
- BD FACSAria III, BD Rhapsody HT Express Single-Cell Analysis System, BD Rhapsody Sequencing Analysis Pipeline v2.2.1/v2.3, Seurat v5.1.0, Immcantation/Change-O, IMGT/HighV-QUEST, fgsea, msigdbr (MSigDB HALLMARK, GO BP), vegan, glmmTMB, lme4 / lmerTest, insulin ELISA
- Reagents & environment
- Seurat v5.1.0; BD Rhapsody Sequencing Analysis Pipeline v2.2.1 and v2.3 on Seven Bridges with GRCh38; R v4.5.3; GraphPad Prism v10.4.1.
- Reproducibility gaps
- The paper describes custom R scripts for the joint RNA/BCR analysis but gives no repository or availability statement. Raw sequencing data are deposited under SRA BioProject PRJNA1489232 (no URL printed); antibody sequences and GenBank accessions are only in Supplemental Tables S10 and S11.
Limitations (6)
- Only peripheral blood was profiled. How these populations relate to the pancreas or pancreatic lymph nodes, where insulin-binding DN2 B cells were reported before, is unknown. · stated by the authors
- Distinct transcriptional programs imply different functions, but only isotype switching and somatic hypermutation were assessed; other functions need direct testing. · stated by the authors
- Few insulin-binding antibodies were profiled, so finding none in Cluster 6 (despite its IGHV4-34 enrichment) may reflect sampling depth rather than true absence. · stated by the authors
- The DN2-like gate did not require IgD negativity, and not all markers defining the BND2 subset described in other T1D work were measured. · stated by the authors
- All participants were early-stage T1D, and the text describes no non-diabetic or later-stage comparison group. So whether these clusters or their insulin reactivity are specific to T1D cannot be judged. · noted in this summary
- Insulin-binding rates rest on small, preselected antibody sets (e.g. 5/16 for Cluster 10; clones chosen for expansion, mutation or switching), and no enrichment test is reported. The screening concentration also differs: 1 μg/ml in Methods vs 1 mg/ml in the Figure 8 legend. · noted in this summary
Read this if you study atypical B cells or islet autoreactivity and want paired single-cell transcriptome/BCR data linking atypical subsets in early type 1 diabetes to class switching, mutation, clonality and insulin binding.
- Citations
- 1
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Insulin-binding B cells can arise in early-stage type 1 diabetes individuals in the absence of somatic hypermutation or insulin autoantibody seropositivity
Lindsay E Bass, Tommi C Taylor, … Rachel H. Bonami·bioRxiv·Oct 22, 2026
- Islet-antigen reactive B cells display a unique phenotype and BCR repertoire in autoantibody positive and recent-onset type 1 diabetes patients
Catherine A. Nicholas, Fatima A. Tensun, … Mia J. Smith·Cell ReportsPeer-reviewed·Mar 19, 2025
- Gene expression profiling reveals B cells are highly educated by the pancreatic environment during autoimmune diabetes
J. Boldison, J. Hopkinson, … F. S. Wong·bioRxiv·May 19, 2022
- BCL6 in T cells promotes type 1 diabetes by redirecting fates of insulin-autoreactive B lymphocytes
Landon M. Clark, James McAninch, … Rachel H. Bonami·iSciencePeer-reviewed·Jun 1, 2026
- Imaging mass cytometry reveals early β-cell dysfunction and changes in immune signatures during type 1 diabetes progression in human pancreata
Nathan Steenbuck, Nicolas Damond, … Bernd Bodenmiller·bioRxiv·Mar 5, 2025
- An imbalance of naive and effector T-cell phenotypes in early type 1 diabetes across conventional and regulatory subsets
Veronika Niederlová, Aleš Neuwirth, … Ondřej Štěpánek·Nature CommunicationsPeer-reviewed·Dec 11, 2025