A surviving beta cell subpopulation enriched in patients with T1D

Reviewer #1 (Public review):

Summary:

The authors have leveraged publicly available single-cell RNA sequencing datasets from isolated islets downloaded from the PANC-DB resource to study the transcriptional profile of insulin-producing beta and glucagon-producing alpha cells from pancreas donors with, or at-risk (islet autoantibody positive) of Type 1 diabetes and donors without diabetes. Their rationale is that any remaining beta cells in these donors with T1D have resisted the autoimmune attack and can therefore provide insights into the transcriptional pathways that mediate this protection. They have developed robust bioinformatic pipelines to address this hypothesis. Their analyses identify beta (and alpha) cells clustered by their differential transcriptional profiles and gene regulatory networks (GRNs), which are present in varying proportions in individuals with and without T1D. The Differentially expressed genes (DEGs) identified align with previously reported datasets. The use of the SCENIC tool, a pipeline for GRN inference using transcriptomic data, involves scoring transcription factor (TF) activity with a rank-based approach, which is considered robust to technical artefacts and adds a novel perspective to this study. Through GRN analysis and regulon score generation, the authors identify a specific cluster of beta cells, cluster 3 (C3), that is enriched in individuals with T1D. This cluster was also slightly enriched in individuals without diabetes (ND) who were > 35 years of age. Their data aligns, supports and extends upon many earlier studies identifying key protective genes, e.g. CD274 (PD-L1) and HLA-E. Together, this provides insights into the transcriptional profile of beta cells that have resisted immune-mediated destruction, which could help with the design of stem cell-derived islet therapies and guide targeted immunotherapy drug trials in the future.

Strengths:

This largely agrees with and extends previous studies from a range of groups using different tissue repositories. This strengthens the validity of the conclusions. The identification of key GRNs associated with preserved beta cells could also aid in the future design of cell and immunomodulatory-based therapies.

Weaknesses:

The regulon scores are hypothesis-generating, not proof of the mechanism by which beta cells are protected. The observation that C3 is enriched in ND >35y could indicate that it is a regulon associated with beta-cell senescence, for example. In the context of T1D, this regulon could reflect beta-cell senescence or stress, which incidentally co-occurs with survival and, as such, is not necessarily a true reflection of survival characteristics. The authors could perhaps expand upon this possibility in a revision.

The authors have leveraged valuable datasets to generate a detailed profile of residual beta cells in Type 1 diabetes and have successfully achieved their study aims. The findings are largely consistent with and extend the existing literature, highlighting key regulatory networks, some of which are supported at both the RNA and protein level (e.g., IRF1). However, a key interpretative consideration is that GRN-derived regulon activity does not distinguish between causal and reflective biological states. In particular, it remains unclear whether these networks represent mechanisms of immune protection or instead reflect underlying beta-cell states such as stress adaptation or senescence. Clarifying this distinction will be important for understanding the functional significance of these regulatory programs and their potential therapeutic relevance.

Reviewer #2 (Public review):

Summary:

This work identifies a novel beta cell population primarily present in the islets from individuals with Type 1 Diabetes (T1D). This population is defined by increased expression of previously described transcription factors, including IRF1, BCL6, JUNB, and CEBPD. The authors postulate that the activation of these genes in beta cells during immune infiltration could be protective against beta cell destruction. This hypothesis aligns with experiments in NOD mice identifying a protected beta cell population. Overall, this work provides a hypothesis for how some beta cell populations survive immune infiltration in T1D.

Strengths:

This work uses a clever analysis approach, defining regulons using SCENIC and using these to recluster the data. This approach identified a novel beta cell population enriched in islets from individuals with Type 1 Diabetes that was very stable to different clustering resolutions. The authors also took many potentially confounding technical factors into account, removing ambient RNA and doublets, and often controlling for batch effects using pseudobulk approaches.

In addition to identifying a novel cluster in one published single-cell dataset, the authors also downloaded additional single-cell datasets that included cytokine treatment of human beta cells to validate the presence of this population in other datasets. In these datasets, the authors were able to identify a similar population of cells, labeled by similar transcription factors.

Weaknesses:

While the authors use a sophisticated approach to identify a novel beta cell subpopulation, more analysis needs to be done to ensure this cluster is biologically meaningful. First, the authors did not take the duration of diabetes into account in this analysis. The duration of diabetes is important because there are different levels of immune infiltration at different stages of diabetes. It would also be important to consider age at diagnosis, as the progression of disease is very different in early vs late onset populations.

Additionally, more exploration of potential confounding factors should be done when looking at the novel population vs other populations in the dataset. This would be further strengthened by adding analysis from datasets that more directly measure transcription factor activity, like single-nucleus ATAC-seq from the different disease states.

Finally, these data can't distinguish the response to the environment (i.e., cytokines) and protective programs. Especially given the similar program in alpha cells, the response to the environment seems likely. More analysis should be done, looking for a similar signature in other populations in the data.

Reviewer #3 (Public review):

Summary:

The authors used a gene regulatory network inference-based clustering approach with existing scRNAseq data sets from cadaveric donors with T1D, auto-antibody positive, and non-diabetic donors and found a regulatory network associated with b-cell survival that is associated with increased expression of genes controlled by interferon regulatory factor 1.

Strengths:

Using established data sets of RNAseq previously performed, the authors identify an interesting population of surviving b-cells in T1D that express a key antiviral transcription factor (IRF1), antiviral genes such as GBPs and iFIT, and decreased expression of a limited number of genes that have been associated with the identity of b-cells.

Selective expression in T1D and not observed in islets from control or auto-antibody positive donors.

Expression changes, TFs identified are also identified in human islets treated with cytokines.

The lack of changes in genes associated with ER stress or the response of endocrine cells to ER stress.

Weaknesses:

The authors do an excellent job of identifying characteristics of the donors/islets in the methods; however, this needs to be addressed in the Figure Legends and Results. Specifically, the length of exposure to cytokines is critical in evaluating the comparisons made in this study.

Is it possible to evaluate sex as a variable in this analysis, and if yes, does one still observe similar changes in identity gene expression and IRF1-dependent gene expression?

Length of disease and evidence for the C3 populations? Does one observe the C3 population in alpha cells of islets with long-standing disease or in the samples that had too few b-cells to perform the analysis? Temporally, 24 h was used for ATACseq and 48 h for cytokine treatment. These are very late exposures, suggesting that secondary and tertiary effects are being compared.

Activation of stress response genes has been correlated with impaired cytokine signaling in islets (human and rodents), limiting the number of endocrine cells that are cytokine responsive. Was this observed in the authors' analysis?

Recent studies have identified induction of antiviral and antibacterial genes in islets in response to short exposures to IL-1, TNF, IFN's that are consistent with the C3 expression profile observed by the authors. While this work has mostly been performed in rodent islets, it has also been observed in human islets, and may be useful in comparing additional transcripts that may contribute to the observed profiles.