henhaobio
Sign in

bioRxivImmunologyPreprint

A surviving beta cell subpopulation enriched in patients with T1D

Maxwell SpurrelliD, Zuzana TobiásováiD, Jordan S. PoberiD, John S. TsangiD, Kevan C. HeroldiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Type 1 diabetes (T1D) is characterized by the autoimmune destruction of pancreatic beta cells. While most beta cells are lost, a subset of beta cells persists years and even decades after disease onset. Studying these surviving cells is challenging, and thus how they escape immune killing remains poorly understood. Here, we applied a gene regulatory network inference-based clustering approach on existing islet scRNAseq data from cadaveric donors with T1D, autoantibody positive donors at risk for T1D, and non-diabetic donors to analyze beta cells from patients with established T1D. This approach identified a novel beta cell subtype enriched in T1D donors defined by the activity of several transcription factors which have well-characterized roles in beta cell survival, most notably IRF1. We found increased expression of immunomodulatory genes (e.g. SOCS1/3, HLA-E) as well as decreased expression of autoantigens and secretory genes, suggesting loss of cellular identity. We identified inflammatory cytokines as a driver of this phenotype by reanalyzing public data from primary human beta cells stimulated with inflammatory cytokines in vitro. We find a similar transcriptional program active in a subset of alpha cells, consistent with cell-extrinsic inflammatory cytokine signaling in vivo. We additionally use immunofluorescence to validate these findings. Overall, this study defines the features of surviving beta cells in patients with T1D, revealing several genes and pathways active in these cells as potential targets for T1D prevention and reversal.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (10 Oct 2026) · CC BY 4.0

Regulon-based clustering of islet scRNA-seq finds an IRF1-high, cytokine-driven beta cell state with reduced identity and autoantigen genes, enriched in type 1 diabetes donors.

Some beta cells persist for years to decades in type 1 diabetes (T1D), yet their human state is poorly understood because pancreas tissue comes only from deceased donors, and prior resilience work used non-obese diabetic mice. Regulon-based clustering of public Human Pancreas Analysis Program (HPAP) scRNA-seq defines a human surviving beta cell state and nominates IRF1, SOCS1/3 and HLA-E as candidate targets for protection.

Key results

  1. ShownClustering beta cells on SCENIC regulon activity rather than integrated expression revealed cluster 3 (C3), a subpopulation enriched in T1D donors and marked by higher IRF1, BCL6, ETV6, CEBPD, ETS2 and JUNB regulon activity with lower MAFA, MAFB and XBP1.

    Median C3 frequency 0.153 in T1D vs 0.037 in ND and 0.027 in AAb+ donors (FDR = 0.002 each, Wilcoxon); regulon PC4 higher in T1D vs ND (p = 0.002) and AAb+ (p = 0.001); C3 stable across clustering resolutions, and top C3 regulons recurred in most of 9 extra RegDiffusion retrainings; established vs recent-onset T1D 0.203 vs 0.109 (p = 0.067, not significant). · Identification of a T1D-enriched beta cell…

  2. ShownC3 beta cells upregulate cytokine-response and immunomodulatory genes but lose some identity, secretory and autoantigen transcripts, suggesting partial loss of cellular identity; class I HLA gain in T1D spans many clusters, with only HLA-E C3-specific.

    GSEA in C3 vs other beta cells: Interferon Gamma Response FDR = 5.12E-7, TNFa Signaling Via NFKB FDR = 3.97E-5, Pancreas Beta Cell depleted FDR = 1.49E-5; SOCS1/3, STAT1, CD274 up; IAPP, G6PC2, SLC30A8, PTPRN, PCSK1, CPE, NKX6-1 down; INS unchanged. Pseudobulk T1D vs ND beta cells: 434 DEGs (FDR < 0.05), incl. class I/II MHC. · Transcriptional profile of cluster 3 beta cells; Isolation of human beta cells from ND, AAb+ and…

  3. ShownReanalysis of public in vitro stressed primary human islet data showed that inflammatory cytokines, unlike ER stressors, reproduce the bidirectional C3 signature and most C3 transcription factor shifts (XBP1 was not reduced), with concordant chromatin motif accessibility.

    Maestas et al. (n = 5 ND donors, 48 hours): every cytokine condition shifted GSVA scores of the top 50 C3 up and down genes (FDR < 0.05 vs control); Brefeldin A and Thapsigargin shifted one direction only. Multiome (single donor): RNA vs chromVAR Pearson 0.72, p = 0.046. Benaglio et al. (n = 3): top HOMER de novo motifs IRF1, then JUNB. · Cluster 3-like phenotype is reproducible by…; Scoring cluster 3 gene sets in Maestas et al…; Identification and motif analysis of…

  4. SuggestedAn alpha cell cluster in the same donors shows nearly the same regulon and transcriptional changes, and its frequency tracks C3 beta cell frequency per donor, consistent with a shared cell-extrinsic cytokine signal rather than a beta cell-specific program.

    Alpha C3 shows higher IRF1, BCL6, CEBPD, JUNB, ETV6 and lower MAFB regulon activity; Interferon Gamma Response FDR = 1.20E-15, TNFa Signaling Via NFKB FDR = 4.63E-31. Frequency T1D 0.12 vs ND 0.07 (p = 0.397) and AAb+ 0.06 (p = 0.413); per-donor C3 beta vs C3 alpha frequency Spearman ρ = 0.65. · Homologous alpha cell population supports…

  5. SuggestedImmunofluorescence on donor pancreas sections found INS+IRF1+ beta cells in most T1D donors examined, often next to one another and restricted to some islets, and none in non-diabetic controls.

    IRF1+ beta cells in 3 of 4 nPOD T1D donors (the negative donor, 6551, had very few beta cells); none in the nPOD ND donor or the internal ND control; in donors 6520 and 6325 some insulin-containing islets lacked C3 cells entirely. · Visualization of cluster 3-like beta cells in…

Methods, data and reproducibility

Reprocessed public HPAP islet scRNA-seq, inferred beta and alpha cell regulons with pySCENIC (RegDiffusion adjacencies, AUCell scoring), clustered cells on regulon scores, and tested disease effects with covariate-adjusted donor pseudobulk models; validated with public in vitro cytokine scRNA/ATAC datasets and IRF1/insulin immunofluorescence.

  • HPAP islet scRNA-seq from PANC-DB: T1D (n = 12), autoantibody-positive AAb+ (n = 11) and non-diabetic ND (n = 33) cadaveric donors, 10X Chromium 3' v2, v3 or v3.1 · human · n = 56 donors; 35165 beta cells (2390 from T1D donors)
  • Maestas et al. scRNA-seq of ND primary islets treated 48 hours with IFNG, TNFA, IL1B, their combinations, Brefeldin A, Thapsigargin or DMSO · human · n = 5 ND donors · GSE237448
  • Maestas et al. single-nucleus multiome (RNA + ATAC) of islets treated with IFNG+TNFA+IL1B cytokine mix or control for 48 hours · human · n = 1 ND donor · GSE237448
  • Benaglio et al. snATAC-seq aggregated beta cell peak counts from islets stimulated 24 hours with IFNG, IL1B and TNFA or left unstimulated · human · n = 3 ND donors · GSE205853
  • FFPE pancreatic sections from nPOD (Network for Pancreatic Organ Donors with Diabetes) and an internal Yale control, stained for IRF1, insulin, CD45 and DAPI · human · n = 4 T1D, 1 nPOD ND, 1 internal ND control
Methods & tools
CellRanger 9.0.1 (GRCh38 2020-A), Seurat 5.1.0 (SCTransform v2, CCA integration), SoupX 1.6.2 and scDblFinder 1.18.0, UCell 2.8.0, pySCENIC 0.12.1 with RegDiffusion 0.1.1 and AUCell, cisTarget GRCh38 v10 motif databases, edgeR, limma 3.60.3 and emmeans 2.0.1, fgsea 1.30.0 and GSVA 1.52.3, Harmony 1.2.0, Signac 1.13.0, MACS2 2.2.9.1 and chromVAR (JASPAR2020), DESeq2 1.44.0 and HOMER 5.1, Immunofluorescence (IRF1, insulin, CD45) on Leica Stellaris 5 confocal
Software & environment
R 4.4.1, Seurat 5.1.0, Python 3.10.0, pySCENIC 0.12.1 with RegDiffusion 0.1.1, CellRanger 9.0.1 (GRCh38 2020-A), SoupX 1.6.2, scDblFinder 1.18.0, fgsea 1.30.0, GSVA 1.52.3, limma 3.60.3, Harmony 1.2.0, Signac 1.13.0, DESeq2 1.44.0, HOMER 5.1
Reproducibility gaps
No container or lockfile is described; the Zenodo record covers code and Supplementary Tables 1-20 only, and raw immunofluorescence images are not listed as deposited.
Limitations (6)
  • The eventual disease status of AAb+ donors is unknown, so C3 cells in their islets may mark progression toward T1D or a protective response in non-progressors. · stated by the authors
  • The analyses are cross-sectional, so they cannot show whether the C3 state promotes survival or only reflects the inflamed T1D islet environment; testing this needs direct assays of beta cell death. · stated by the authors
  • Only surviving T1D beta cells could be profiled, so the phenotype of destroyed cells, the death pathways survivors went through, and the cells producing the cytokines remain unknown. · stated by the authors
  • GRN inference depends on incomplete transcription factor and cis-regulatory annotations and infers factor activity rather than measuring binding (no ChIP-seq or equivalent assay). · stated by the authors
  • Some older ND donors had high C3 beta and alpha cell frequencies, which likely confounded the alpha cell disease comparison, and within-donor C3 versus non-C3 comparisons were too underpowered for significance testing. · stated by the authors
  • The validation is thin: immunofluorescence is qualitative across 4 T1D donors and both ND samples, the RNA-chromVAR concordance (p = 0.046) comes from a single donor, and C3 frequency did not differ significantly between established and recent-onset T1D (p = 0.067). · noted in this summary

Read this if you study beta cell survival or islet inflammation in type 1 diabetes, or want a worked example of regulon-based scRNA-seq subclustering that separates biological states from technical batch effects.

Citations
0

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

Similarity Map
SelectedSame fieldCross-field
THISBenaglio '22Anderson '26Rui '20Bode '24Yi '26Boldison '22
77% semantic matchGenomics
Paola Benaglio, Han Zhu, Mei-Lin Okino+20 more·Cell Genomics·Nov 11, 2022· 33 citations