Transient ER stress cell-autonomously promotes beta cell cycling in mice

Abstract

Aims/hypothesis

Regenerating endogenous pancreatic beta cells is a potentially curative yet currently elusive strategy for diabetes therapy. Mimicking the microenvironment of the developing pancreas and leveraging vascular signals that support pancreatic endocrinogenesis may promote beta cell regeneration. We aimed to investigate whether recovery from experimental hypovascularisation of the endocrine pancreas could trigger mouse beta cell proliferation.

Methods

A doxycycline (DOX)-inducible transgenic mouse model was used to induce conditional intra-islet hypovascularisation. In this model, vascular endothelial growth factor (VEGF)-A signalling within pancreatic islets is antagonised through beta cell-specific overexpression of a VEGF-A decoy receptor, soluble fms-like tyrosine kinase 1 (sFLT1). Cessation of sFLT1 overexpression was induced by DOX withdrawal. sFLT1 expression, vessel kinetics and beta cell proliferation upon DOX administration and withdrawal were analysed using quantitative RT-PCR and immunostaining. Single-cell RNA-seq was used to investigate the effects on the islet cells’ transcriptome and perform pathway enrichment analysis. RIP-rtTA;TetO-GFP mice were studied in parallel to assess the dependency of cell cycle induction on vessel manipulation. Additionally, in vitro experiments were conducted to further elucidate and validate our in vivo findings.

Results

Serendipitously, we discovered that sFLT1 overexpression in beta cells induces endoplasmic reticulum (ER) stress and activates proliferation-associated pathways. Upon cessation of sFLT1 overexpression, ER stress decreased and beta cell proliferation was promoted independently of vessel recovery, as shown by cumulative BrdU labelling over 7 days (mean ± SEM vs control: 14.3 ± 1.3% vs 5.2 ± 0.6%) during the DOX withdrawal period. Transient GFP overexpression also induced ER stress and a subsequent reduction thereof resulted in increased beta cell proliferation (mean ± SEM vs control: 7.2 ± 0.4% vs 5.1 ± 0.5%). Chemical, transient induction of ER stress in vitro by ER-stress-inducing compounds reproduced this beta cell cycling response, as assessed by cumulative EdU labelling during a 3 day washout period (mean ± SEM vs control: 2.6 ± 0.4% vs 0.8 ± 0.2% for thapsigargin and 3.8 ± 0.9% vs 1.0 ± 0.2% for tunicamycin), which further increased under high-glucose conditions when islets were exposed to thapsigargin (mean ± SEM vs control: 9.0 ± 1.2% vs 2.0 ± 0.4%).

Conclusions/interpretation

Our findings uncover a link between transgene (over)expression, ER stress, glucose and cell cycle activation in mouse beta cells.

Data and code availability

The single-cell RNA-seq data generated in this study are deposited at GEO (NCBI) with accession code GSE274443.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s00125-025-06649-3.

Introduction

Diabetes mellitus is a chronic disease caused by insulin deficiency. Current treatments focus on regulating blood glucose levels without addressing the underlying beta cell defect. To cure diabetes, restoring and preserving a functional beta cell mass sufficient for fine-tuned glucose control is necessary. The prospect of cell replenishment through endogenous beta cell regeneration is enticing and could be achieved by stimulating the proliferation of existing beta cells, inducing the redifferentiation of dedifferentiated beta cells, or generating new beta cells from non-beta cells [1]. However, the intricate processes involved in these beta cell regeneration modalities are not fully understood.

Most beta cells are generated during fetal development, with proliferation peaking in early postnatal life at an average of 2%, before declining to less than 0.1–0.5% in the adult human pancreas [2–6]. Self-duplication remains the primary mechanism for maintaining the postnatal beta cell mass in mice [7]. Stimulating the proliferation of remaining beta cells could restore functional mass in people with diabetes, making it crucial to uncover the signalling pathways involved, particularly in adulthood.

Beta cells rely on systemic and microenvironmental cues to adapt their function and numbers [8], with endothelial cells being one of the main sources of these cues. Pancreatic islets are highly vascularised mini-organs with extensive endothelial cell–beta cell crosstalk [9]. Vascular endothelial growth factor (VEGF)-A is a key signalling molecule in the endothelial cell–beta cell crosstalk [10]. Within islets, beta cells are the main source of VEGF-A, which, by binding to the VEGF receptor (VEGFR)2 on endothelial cells, stimulates their recruitment and proliferation [10]. Precise control of VEGF-A signalling is crucial for maintaining islet vascular homeostasis and functional beta cell mass [11]. In addition to microenvironmental signals, intrinsic signalling pathways have also been implicated in promoting beta cell proliferation. In particular, the unfolded protein response (UPR), which is activated in response to endoplasmic reticulum (ER) stress, has been linked to beta cell proliferation responses [1]. Previously, we used transgenic rat insulin promoter (RIP)-reverse tetracycline-dependent transactivator (rtTA)-tetracycline operator (tetO)-sFLT1 mice (RIP-rtTA;TetO-sFLT1 mice) to conditionally overexpress and secrete soluble VEGFR1 (soluble fms-like tyrosine kinase 1 [sFLT1]) in beta cells, thereby antagonising VEGF-A signalling. This model enabled us to investigate the impact of islet hypovascularisation and hypoxia on beta cell function and proliferation during adulthood, injury [12] and pregnancy [13]. Despite causing mildly impaired glucose regulation, intra-islet hypovascularisation did not affect baseline or stimulated beta cell proliferation rates. Here, we set out to examine whether recovery from experimental islet hypovascularisation promotes beta cell proliferation.

Methods

For detailed methods, please see electronic supplementary material (ESM) Methods.

Animal procedures

All experiments were approved by institutional and national authorities (licenses LA2230622, LA1230595). Mice were housed under controlled conditions with ad libitum access to chow and water. RIP-rtTA [14, 15] and TetO-sFLT1 [16, 17] mice on a CD1 background were intercrossed to generate double-transgenic (dTg) animals with doxycycline (DOX)-inducible, beta cell-specific expression of human sFLT1 [18]. RIP-rtTA;TetO-GFP,-LacZ mice were generated by crossing RIP-rtTA mice with TetO-GFP-LacZ mice (The Jackson Laboratory, Bar Harbor, ME, USA, https://www.jax.org/strain/018913). Genotyping was performed from ear biopsies by PCR using the Mouse Direct PCR Kit (B40015, Selleck Chemicals, Cologne, Germany; see ESM Table 1 for primer sequences). Experimental mice were 6–8 weeks old of both sexes. Adult male and female mice were randomly allocated to experimental groups. For in vitro studies, islets were isolated from 8-week-old male C57BL/6JRj mice (Janvier Labs, Le Genest-Saint-Isle, France). DOX (0.4 mg/ml, D9891, Sigma-Aldrich, St Louis, MO, USA) and BrdU (0.8 mg/ml, B5002, Sigma-Aldrich) were administered in drinking water for 14 days and 7 days, respectively. Blood glucose and body weight were measured after 2 h of fasting. Prior to being euthanised, mice were injected intravenously with biotinylated tomato lectin (B-1175-1, Lycopersicon esculentum, Vector Laboratories, Burlingame, CA, USA) to label perfused vessels.

Immunostaining and image analysis

Pancreases or isolated islets were fixed in 4% (wt/vol.) formaldehyde (VWRK4078.9010, VWR, Radnor, PA, USA), paraffin-embedded and sectioned. Antigen retrieval, blocking and antibody staining were performed according to standard protocols. Primary antibodies were diluted in BOND Primary Antibody Diluent (AR9352, Leica, Wetzlar, Germany) and secondary antibodies in PBS + 0.1% (vol./vol.) Tween-20 (PBST) (see ESM Table 2 for primary antibodies and ESM Table 3 for secondary antibodies). Primary antibodies were validated by test immunostainings on appropriate positive and negative control tissues. TUNEL staining was performed after heat-mediated antigen retrieval. Nuclei were counterstained with Hoechst 33342 (B2261, Sigma-Aldrich). Imaging was performed on an Olympus BX61 (Olympus, Tokyo, Japan) or Zeiss LSM800 (Zeiss, Jena, Germany) microscope, and morphometric analysis of beta cell volume was carried out using Fiji (version 2.14.0) [19] as previously described [20].

Immunoblotting

Islets were lysed in RIPA Lysis and Extraction Buffer (89900, Thermo Fisher Scientific, Waltham, MA, USA) with 1% (vol./vol.) protease and 1% (vol./vol.) phosphatase inhibitors (P8849 & P5726, Sigma-Aldrich), proteins resolved by SDS-PAGE, transferred to nitrocellulose (1620215, Bio-Rad, Hercules, CA, USA) and stained with REVERT total protein stain (926-11016, Licor Bio, Bad Homburg, Germany). After blocking, membranes were probed with primary antibodies (see ESM Table 2 for primary antibodies), detected using IRDye-labelled secondary antibodies (see ESM Table 3 for secondary antibodies) with an Odyssey Fc imager, and quantified with Image Studio (Licor Bio). Primary antibodies were validated by test immunoblots on appropriate negative and positive control tissues.

Transmission electron microscopy

Pancreases were fixed in 2.5% (vol./vol.) glutaraldehyde, post-fixed in 1% (wt/vol.) osmium tetroxide and 2% (wt/vol.) uranyl acetate, dehydrated and embedded in epoxy resin. Ultrathin sections were stained with uranyl acetate and lead citrate and examined on a Tecnai 10 TEM equipped with a CCD camera.

In vitro experiments

Pancreatic islets were isolated from 8-week-old male C57BL/6JRj mice (Janvier Labs) by collagenase/protease (CIzyme RI, 005-1030, PELOBiotech, VitaCyte, Indianapolis, IN, USA) digestion and handpicked under a stereomicroscope. The isolated islets were randomly allocated to different in vitro treatments. Islets were cultured in RPMI 1640 containing 11.1 mmol/l glucose (61870-010, Thermo Fisher Scientific) with 10% (vol./vol.) FBS (F9665, Sigma-Aldrich) and 1% (vol./vol.) penicillin/streptomycin (15140122, Thermo Fisher Scientific). For the in vitro puromycin assay, isolated islets from adult dTg RIP-rtTA;TetO-sFLT1 mice from both sexes were randomly assigned to the different treatment groups. sFLT1 expression was induced with DOX (250 mg/ml, 48 h) and puromycin (A1113803, Thermo Fisher Scientific) was added 20 min before collection for immunoblotting [21]. For experiments under different glucose concentrations, media contained either 3.9 or 16.7 mmol/l glucose. ER stress was induced with thapsigargin (750 nmol/l, 6 h; T9033; Sigma-Aldrich) or tunicamycin (5 µg/ml, 8 h; T7765; Sigma-Aldrich). Cycling cells were labelled with 10 µmol/l EdU (Click-iT EdU Imaging Kit, C10340, Thermo Fisher Scientific).

RNA extraction and quantitative RT-PCR

RNA was isolated from islets using the RNeasy Micro Kit (74004, Qiagen, Venlo, the Netherlands). Reverse transcription was done as previously described [12]. Quantitative PCR was performed with TaqMan Gene Expression Assays (see ESM Table 4). Human sFLT1 expression was quantified using human-specific primers. Reference genes were selected by geNorm [22] and expression normalised to their geometric mean.

Single-cell RNA-seq

For single-cell RNA-seq (scRNA-seq) analysis, islets were dissociated into single cells and processed using the 10x Genomics Chromium platform. Libraries were sequenced on an Illumina NovaSeq 6000. Reads were aligned to a custom mouse reference genome including human sFLT1. Data preprocessing included cell filtering [23–25], doublet removal [26], normalisation and batch correction (Seurat, Harmony) [27, 28]. Clustering was performed with Uniform Manifold Approximation and Projection (UMAP) and Louvain algorithms. Differential gene expression was assessed by Wilcoxon rank-sum test with Bonferroni correction. Gene Ontology analysis was performed with Metascape and regulatory networks were inferred using SCENIC [29]. The results were visualised with Cytoscape [30]. Pseudotime trajectories for beta cells were analysed with Monocle3 [31].

Blinding/masking

Experiments were not blinded throughout the study; however, part of the analysis and the (re)counting of the proliferation samples was performed under blinded conditions to minimise potential bias.

Inclusion and exclusion criteria

We did not exclude any data, samples, or animals from the analysis. All collected experimental results are reported in the manuscript, and no outcomes that failed to support the main findings were omitted. All measured variables and conditions are presented in the results section.

Statistics

Data were analysed in GraphPad Prism v9.1.0 using unpaired t tests, one- or two-way ANOVA, or mixed-model analysis with Tukey’s post hoc test as appropriate. Results are shown as mean ± SEM, with significance set at p≤0.05.

Results

Transient sFLT1 overexpression induces beta cell proliferation independent of vessel recovery

To study the effect of intra-islet blood vessel ablation and regrowth on beta cell proliferation, we used RIP-rtTA;TetO-sFLT1 mice. In this model, DOX administration induces sFLT1 expression and secretion by beta cells, antagonising vessel survival and angiogenesis by scavenging VEGF-A near beta cells (Fig. 1a). The effects of transient sFLT1 expression on intra-islet vascularisation and beta cell proliferation were examined by administering DOX for 14 days followed by its withdrawal (WD) for 1, 4, or 7 days during which BrdU was administered (Fig. 1b). Mice that received DOX for 14 days are hereafter referred to as +DOX mice. dTg mice not receiving DOX and single-transgenic (sTg) TetO-sFLT1 mice that received DOX, referred to as −DOX and sTg mice, respectively, were used as controls. Fasting blood glucose values were elevated in +DOX mice compared with sTg and −DOX mice but gradually normalised after DOX WD (Fig. 1c and ESM Fig. 1a). Islets isolated from +DOX mice contained high sFLT1 transcript levels, which decreased significantly by 4 days after DOX WD (WD4D) (Fig. 1d). This was accompanied by a 35% and 44% decrease in the number of sFLT1-positive beta cells at WD4D and 7 days after DOX WD (WD7D), respectively, compared with the +DOX condition (sFLT1⁺/insulin⁺ cells: +DOX 86.7 ± 4.9%; WD4D 55.9 ± 4.2%; WD7D 48.6 ± 4.8%) (Fig. 1e, f). To assess protein translation in RIP-rtTA;TetO-sFLT1 mouse islets following DOX treatment and WD, we performed an in vitro puromycin incorporation assay. Puromycin functions as an aminoacyl-tRNA analogue that incorporates into nascent polypeptide chains, enabling detection of active translation via immunoblotting [32]. The puromycin assay showed increased protein translation 48 h after sFLT1 overexpression; this returned to baseline 72 h following DOX WD (ESM Fig. 1b–e). The impact of transient sFLT1 expression on islet vascularisation was assessed by quantifying the area, size and number of islet vessels. The intra-islet blood vessel area was strongly reduced in +DOX mice compared with sTg and −DOX controls. Surprisingly, at WD7D islet vascularisation did not recover (Fig. 1g, h and ESM Fig. 1f, g). Even at 6 weeks after WD, a subset of beta cells remained sFLT1-positive and islet hypovascularisation persisted (ESM Fig. 1h–j). Intriguingly, cumulative BrdU labelling showed a nearly threefold increase in beta cell proliferation at WD7D (mean ± SEM 14.3 ± 1.3%) compared with −DOX (5.2 ± 0.6%) and sTg (3.6 ± 0.2%) mice (Fig. 1i, j). The fraction of Ki67- and phospho-histone H3-positive beta cells peaked at WD4D (Fig. 1k–m) and insulin volume at WD7D was increased compared with controls (Fig. 1n). No significant differences in beta cell apoptosis or DNA damage were observed across the experimental conditions, as assessed by TUNEL staining and immunostaining for the DNA damage marker γH2AX, respectively (ESM Fig. 1k–m). Given that transient islet hypervascularisation induced by beta cell-specific VEGF-A overexpression promotes beta cell proliferation through the active recruitment of bone marrow-derived macrophages [33], we quantified intra- and peri-islet macrophages in our model. However, no significant differences were observed between the experimental groups (ESM Fig. 1n–p). In summary, sFLT1 overexpression in beta cells significantly reduces intra-islet blood vessels, with the effect persisting for at least 6 weeks post-transgene expression. Beta cell cycling increases after transient sFLT1 overexpression without significant recovery of islet endothelial cells or changes in macrophage numbers.

Fig. 1.

Transient sFLT1 overexpression induces beta cell proliferation independently of vessel recovery. (a) Schematic of the RIP-rtTA;TetO-sFLT1 transgenic mouse model. (b) Experimental approach to study the effects of transient islet hypovascularisation. Mice received a vehicle solution (−DOX) or DOX (+DOX) for 14 days, followed by a 1, 4 or 7 day WD period during which BrdU was administered. (c) 2 h fasting blood glucose levels (sTg [n=15], −DOX [n=13], +DOX/WD [n=14]). The grey zone indicates the period during which DOX was administered. (d) sFLT1 expression (fold change relative to −DOX) in pancreatic islets isolated from RIP-rtTA;TetO-sFLT1 mice, normalised to the geometric mean of Actb and Ppia (−DOX [n=3], +DOX [n=3], WD4D [n=4]). (e) Pancreatic sections immunostained for sFLT1. (f) sFLT1/INS area per pancreas (−DOX [n=5], +DOX [n=5], WD1D [n=5], WD4D [n=5], WD7D [n=5]). (g) Pancreatic sections immunostained for INS and lectin (functional blood vessels). (h) Lectin/INS area per pancreas (sTg [n=6], −DOX [n=5], +DOX [n=5], WD4D [n=5], WD7D [n=6]). (i) Pancreatic sections immunostained for INS, NKX6.1, and BrdU. (j) Beta cell cycling based on cumulative BrdU labelling for 7 days (sTg [n=6], −DOX [n=7], WD7D [n=10]). (k) Beta cell proliferation based on Ki67 positivity (sTg [n=10], −DOX [n=8], +DOX [n=10], WD1D [n=6], WD4D [n=11], WD7D [n=12]). (l) Percentage of beta cells in mitosis based on pHH3-positivity (sTg [n=5], −DOX [n=5], +DOX [n=6], WD1D [n=4], WD4D [n=9], WD7D [n=9]). (m) Immunostaining for Ki67, NKX6.1 and pHH3 on pancreatic sections. The arrows indicate proliferating beta cells. (n) Total insulin volume per pancreas (sTg [n=4],−DOX [n=7], +DOX [n=5], WD7D [n=9]). Data are presented as mean ± SEM. *p≤0.05, **p≤0.01, ***p≤0.001 for −DOX vs +DOX/WD or as shown; †p≤0.05, ††p≤0.01 for sTg vs +DOX WD, by mixed-effects analysis followed by Tukey’s post hoc test (c) or one-way ANOVA followed by Tukey’s post hoc test (d, f, h, j, k, l, n). Scale bar, 100 µm. INS, insulin; NKX6.1, NK6 homeobox 1; pHH3, phospho-histone H3; WD1D, 1 day after DOX WD

sFLT1 overexpression causes ER stress and loss of end-stage maturation markers in beta cells

To investigate the islet cell response to hypovascularisation and determine whether the observed increase in beta cell proliferation after DOX WD arises from cell-autonomous mechanisms or signals from other islet cell types, we performed scRNA-seq on islet cells from −DOX (8699 cells), +DOX (19135 cells) and +DOX WD4D (10,528 cells) mice and compared the transcriptomes of beta cells, alpha cells, delta cells, endothelial cells and macrophages (Fig. 2a). We annotated the different islet cell types based on marker gene expression (Fig. 2b and ESM Fig. 2a). sFLT1 transgene RNA overexpression, identified by pFUSE-1 and hsFLT1 expression, was detected in +DOX beta cells (ESM Fig. 2b and ESM Methods). The scRNA-seq results confirmed the enrichment of proliferating beta cells and the absence of endothelial cell recovery or macrophage recruitment in the WD4D condition (Fig. 2c). A hypoxia response was evident in +DOX beta, alpha and delta cells, as indicated by the upregulation of Vegfa and glycolysis-related genes (ESM Fig. 2c). Next, we analysed gene expression differences between the +DOX and −DOX conditions (ESM Figs 2f–i, 3). Upregulated genes in +DOX beta cells encompassed ER stress and UPR genes, including Sdf2l1, Pdia6, Dnajc3, Fkbp11, Atf4, Ddit3, Hspa5, Atf5 and Nupr1 (Fig. 2d, e and ESM Fig. 2d). Cdkn1a (encoding p21) and Chgb (encoding chromogranin B, a protein involved in trafficking secretory proteins into budding insulin granules [34]) were also upregulated, alongside Cd81 and Aldh1a3, markers of immaturity and dysfunction [35, 36] (Fig. 2d). Additionally, the expression of several end-stage maturation markers, including Ins1, Mafa, Slc2a2, Ucn3, Glp1r, Sytl4, Ppp1r1a and Trpm5, were decreased in +DOX beta cells, [37] (Fig. 2d). We confirmed presence of ER stress in +DOX beta cells by electron microscopy, which showed an aberrantly swollen ER (ESM Fig. 2e). We also confirmed loss of beta cell end-stage maturation markers based on decreased MafA and urocortin 3 (UCN3) and increased CD81 immunostaining [35] in beta cells (Fig. 2f). Next, we investigated the effects of DOX on the transcriptome of other islet cell types. Alpha cells in +DOX mice retained normal expression levels of glucagon (encoded by Gcg) and other identity and maturation markers such as Mafb, Arx, Irx1 and Irx2 [37]. Genes expressed differentially between +DOX and −DOX alpha cells included those involved in stress responses and apoptosis, such as Upk3a [38], Rbm4b, Gadd45g and Calr, as well as genes associated with coagulation (wound healing), such as Klkb1 and Serpine2 (ESM Figs 2f, 3b). Delta cells in the +DOX condition similarly retained key identity markers such as somatostatin (Sst), Hhex and Rbp4 [37] and downregulated genes (Efna5, St8sia2, Dclk1, Cpne5 and Ramp1) involved in cell morphogenesis and central nervous system development (ESM Figs 2g, 3c). The +DOX endothelial cells upregulated genes involved in antigen processing and presentation via MHC class II (Cxcl12, Cd36, Cd74, H2-Aa and H2-Ab1), and downregulated genes involved in endothelial cell identity and function (Ace, Angpt2, Kdr, Eng, Esm1 and Tie1) (ESM Figs 2h, 3d). Lastly, +DOX macrophages upregulated genes involved in lysosomal degradation and apoptosis (Bcl2a1b, Ctsb, Ctsd and Lgals1), and downregulated genes involved in antigen processing and presentation via MHC class II (ESM Figs 2i, 3e). In summary, beta cell-specific sFLT1 overexpression increases ER stress and UPR in beta cells while reducing beta cell and endothelial cell maturation markers, and triggers a hypoxia response in all islet cells.

Fig. 2.

sFLT1 overexpression causes ER stress and loss of end-stage maturation markers in beta cells and cessation causes ER stress relief. (a) Schematic model of the potential pathways driving beta cell proliferation upon DOX WD in RIP-rtTA;TetO-sFLT1 mice. (b) Uniform Manifold Approximation and Projection (UMAP) plots of −DOX (8699 cells), +DOX (19,135 cells) and WD4D (10,528 cells) islet cells (n=2 per condition). (c) Bar plots showing the percentage of proliferating beta cells, endothelial cells and macrophages in islets isolated from −DOX, +DOX and WD4D mice. (d) Volcano plot displaying upregulated and downregulated genes in +DOX vs −DOX mouse beta cells. Genes with an absolute Log₂FC >0.3 and FDR <0.01 are marked in red. (e) Gene ontology analysis of the genes significantly upregulated between +DOX and −DOX beta cells. (f) Pancreatic sections immunostained for the beta cell (de)differentiation markers MafA, UCN3 and CD81. (g) Volcano plot displaying upregulated and downregulated genes in WD4D vs +DOX mouse beta cells. Genes with an absolute Log₂(FC) >0.3 and FDR <0.01 are marked in red. (h) Heatmap representing the expression of top marker genes for each beta cell cluster. (i) UMAP plots of 24,223 beta cells from the −DOX, +DOX and WD4D conditions. Pie charts indicate the percentages of each beta cell cluster. (j) Normalised BiP, p-eIF2α and p-IRE1α signals in islets isolated from −DOX, +DOX, WD4D RIP-rtTA;TetO-sFLT1 mice. Representative immunoblots for BiP (78 kDa), p-eIF2α (38 kDa) and p-IRE1α (110 kDa) are shown. β-Actin (43 kDa) served as a loading control. Levels were normalised to the total protein stain. Data are presented as mean ± SEM. Statistical analysis was done by one-way ANOVA followed by Tukey’s post hoc test (j). pFUSE-1 and hsFLT1 represent the sequences used to detect sFLT1 transgene expression. Scale bar, 100 µm. FC, fold change; FDR, false discovery rate; PP, pancreatic polypeptide

Cessation of sFLT1 overexpression results in ER stress relief and beta cell redifferentiation

We subsequently assessed differential gene expression between the WD4D and +DOX conditions. In WD4D beta cells, the expression of the sFLT1 transgene and stress-related genes including Nupr1, Ddit3, Atf5, Fkbp11 and Cdkn1a was significantly downregulated (Fig. 2g). This was accompanied by an upregulation of the beta cell maturation genes Mafa and Ucn3 (Fig. 2g), which was also confirmed by immunostaining (ESM Fig. 4a). In endothelial cells, macrophages, alpha cells or delta cells, differential expression of genes such as Igf1, Hgf, Pdgf, Ctgf or Tgfb1 [1], encoding potential beta cell mitogens, was not observed (ESM Fig. 2f–i). Based on the observed kinetics of ER stress gene expression and beta cell proliferation, and in light of the findings by Szabat et al [39] showing that alleviation of basal ER stress via insulin gene knockout promotes beta cell proliferation, we hypothesised that sFLT1 overexpression induces ER stress in beta cells, and that its relief upon DOX WD triggers proliferation. To test this hypothesis, we re-clustered the beta cells, identifying eight heterogenous beta cell subsets, and further used SCENIC to identify potential transcription factors and their targets within each subset (Fig. 2h, i and ESM Fig. 5). Cluster 0 was present in all conditions and characterised by high Ins1 expression and Maz and Junb regulon activity. In accordance with the loss of end-stage maturity markers (see above), cluster 1 (most abundant in +DOX and WD4D conditions) displayed relatively low Pdx1 regulon activity and enriched activity of regulons associated with ER stress, such as Xbp1. Furthermore, this cluster also showed higher expression of Aldh1a3. Cluster 2 was absent under −DOX conditions, emerged upon sFLT1 induction (+DOX) and was nearly undetectable again following DOX WD (WD4D) (Fig. 2i). These beta cells exhibited the highest levels of the sFLT1 transgene and ER stress genes Nupr1 and Ddit3; hence, cluster 2 is referred to as the ‘ER stress^high’ cluster (Fig. 2h, i). Cluster 2 had highly enriched activity of regulons Atf4, Cebpb, Cebpg and Ddit3, all associated with ER and cellular stress [40–42]. Nupr1, among the most robustly upregulated genes in +DOX beta cells and among the most significantly downregulated genes in the WD4D condition, is a target of Atf4, Cebpb, Cebpg and Ddit3 (ESM Fig. 5c), while Cdkn1a (another upregulated gene) is primarily regulated by Cebpg. Cluster 3, present in all conditions, had high Gpx3, Chgb and Chga expression and was enriched in Gata6, Sox4 and Foxp1 regulon activity; these transcription factors are interconnected through their roles in developmental processes, cell differentiation and disease states [43–45]. Clusters 4 and 5 were predominant in the −DOX condition and likely represent more mature beta cells, as indicated by the elevated expression of genes essential for beta cell function, such as Ppp1r1a and Slc30a8. These clusters were also characterised by high Mafa regulon activity. Cluster 6 exhibited enriched activity of Creb3l2 and Foxj2, multifunctional transcription factors involved in stress responses, cell cycle regulation and differentiation [46, 47]. Cluster 7 demarks proliferating beta cells based on the expression of cell cycle genes (Stmn1, Mki67, Cenpf, Top2a and Tubb5) and exhibited robust activity in cell cycle-related gene regulatory networks.

To validate the transient induction of ER stress at the protein level, we examined several markers of the UPR. We investigated the expression levels of BiP, an ER chaperone whose expression is generally induced under ER stress, along with the phosphorylated forms of eukaryotic translation initiation factor 2 subunit 1 (eIF2α) and endoplasmic reticulum to nucleus signalling 1 (inositol-requiring-enzyme 1α; IRE1α) in islets isolated from −DOX, +DOX and WD4D mice. p-eIF2α is part of the eIF2α–eukaryotic translation initiation factor 2α kinase 3 (PERK)–activating transcription factor 4 (ATF4) branch and p-IRE1α is part of the IRE1α–X-box binding protein 1 (XBP1) branch of the UPR. We observed elevated levels of BiP, phosphorylated eIF2α and phosphorylated IRE1α in the +DOX condition compared with −DOX and WD4D (Fig. 2j and ESM Fig. 4b), consistent with a transient ER stress response. Our scRNA-seq data also revealed upregulation of Cdkn1a in the +DOX condition compared with −DOX, followed by a return to lower levels in WD4D. Interestingly, protein expression of p21, a negative regulator of the cell cycle, closely mirrored this pattern, with a significant increase in +DOX beta cells and a decrease in WD4D beta cells (ESM Fig. 4c, d). Collectively, these findings demonstrate that alleviating sFLT1-induced ER stress restores beta cell identity and functional maturity and is associated with dynamic shifts in beta cell subpopulations driven by changes in stress signalling, transcriptional regulation and cell cycle activity.

ER stress relief precedes cell cycle activation in beta cells

To further dissect gene expression dynamics in the RIP-rtTA;TetO-sFLT1 mouse beta cells and their relationship to beta cell cycling, we performed pseudotime trajectory analysis starting from beta cells predominant in −DOX mice (as indicated in Fig. 2i). The top 78 differentially expressed genes along the trajectory are highlighted (Fig. 3a, b), including the transgene (pFUSE-1 and hsFLT1), key ER stress and UPR regulators (Nupr1, Ddit3, Pdia6 and Fkbp11) and proliferation-related genes (Mki67 and Cenpa) (Fig. 3c). Interestingly, the transient surge in ER stress and UPR activation preceded the upregulation of cell proliferation genes, corroborating our hypothesis that transient ER stress promotes beta cell cycling in our model of conditional sFLT1 overexpression. To investigate the signalling pathways potentially driving proliferation, we examined phosphorylated ERK1/2 (p-ERK1/2) activity. Analysis of p-ERK1/2 showed transient upregulation under +DOX conditions, followed by a decline at WD4D (ESM Fig. 4e). We next evaluated phosphorylated Akt (p-Akt) levels as a potential driver of beta cell proliferation. Immunoblot analysis revealed a reduction in p-Akt levels at WD4D (ESM Fig. 4f), suggesting that Akt signalling is unlikely to mediate proliferation in our model. In contrast, levels of phosphorylated S6 ribosomal protein (p-S6RP), a downstream effector of the mechanistic target of rapamycin complex 1 (mTORC1) pathway [48], were markedly elevated in beta cells at WD4D (ESM Fig. 4g, h). These findings suggest that transient ER stress not only precedes but may also actively prime beta cells for proliferation, potentially through mTORC1 pathway activation rather than canonical Akt signalling.

Fig. 3.

Pseudotime analysis shows high sFLT1 expression and ER stress occurring prior to increased beta cell proliferation. (a) UMAP showing the pseudotime trajectory that was constructed from the 24,223 beta cells derived from −DOX, +DOX and WD4D RIP-rtTA;TetO-sFLT1 mice (n=2/condition). −DOX beta cells were defined as the root of the trajectory. Cells are coloured by pseudotime. (b) Expression trends in pseudotime of the top 78 genes changing along the trajectory. Hierarchical clustering (using Euclidean distance) in three groups based on expression patterns along the pseudotime trajectory. The first group of genes are those mostly related to protein synthesis, cellular stress responses, apoptosis regulation and the transgene. The second group of genes are genes related to the regulation of the cell cycle and proliferation, and the third group of genes are involved in insulin production, non-coding sequences and Lrp4, which encodes LDL receptor-related protein 4. The z score for each gene’s expression value represents how much that value deviates from the mean expression of that gene across the cells. (c) Expression pattern in pseudotime ordering of sFLT1 (indicated as pFUSE-1 and hsFLT1), Nupr1, Ddit3, Pdia6, Fkbp11, Mki67 and Cenpa. The black dendrogram lines indicate hierarchical clustering of genes based on their z score expression profiles along pseudotime; genes with similar temporal expression patterns are grouped together

Conditional overexpression of GFP in beta cells increases beta cell proliferation

To test whether conditional overexpression of a random transgene under control of the insulin gene promoter induces beta cell cycling, we developed RIP-rtTA;TetO-GFP transgenic mice in which DOX administration induced beta cell-specific GFP expression (Fig. 4a). A similar experimental design was used, with DOX administration for 14 days followed by DOX WD (Fig. 4b). Mice that did not receive DOX were used as controls. The 2 h fasting blood glucose values were measured weekly and showed no significant difference between groups (Fig. 4c). We validated GFP expression in +DOX islets and a significant decrease thereof at WD4D (Fig. 4d). The UPR genes Hspa5, Atf6, Atf4 and Ddit3 were significantly increased in +DOX mouse islets and decreased at WD4D (Fig. 4e), similar to the sFLT1 model (ESM Fig. 2d). To expand on this comparison, we also assessed Cdkn1a gene expression. Cdkn1a increased significantly in +DOX islets and decreased at WD4D (Fig. 4e). Protein levels of beta cell end-stage maturation markers MafA and UCN3 did not differ between -DOX, +DOX and WD4D mice expressing GFP (Fig. 4f). Notably, as in sFLT1 mice, transient GFP overexpression significantly increased beta cell proliferation at WD7D compared with −DOX (mean ± SEM 7.2 ± 0.4% vs 5.1 ± 0.5%) (Fig. 4g, h). The fraction of Ki67-positive beta cells peaked at WD7D (Fig. 4i, j). Of note, DOX administration per se did not promote beta cell cycling (ESM Fig. 6a–c). In summary, conditional GFP overexpression by beta cells transiently induces ER stress, with increased beta cell proliferation during the alleviation phase. These data suggest that any conditional transgene under control of the insulin gene promoter might evoke similar responses.

Fig. 4.

Conditional overexpression of GFP in beta cells increases beta cell proliferation. (a) Schematic of the RIP-rtTA;TetO-GFP transgenic mouse model. (b) Experimental approach to study the effects of transient GFP overexpression. Mice were untreated (−DOX) or received DOX for 14 days (+DOX), followed by a 4 or 7 day WD period during which BrdU was administered. (c) 2 h fasting blood glucose levels (−DOX [n=8], +DOX WD7D [n=9]). The grey zone indicates the period during which DOX was administered. (d) GFP expression (fold change relative to −DOX) in pancreatic islets isolated from RIP-rtTA;TetO-GFP mice, normalised to the geometric mean of Tbp and Hmbs (−DOX [n=6], +DOX [n=5], WD4D [n=4]). (e) Hspa5, Atf6, Atf4, Ddit3 and Cdkn1a expression (fold change relative to −DOX) in pancreatic islets, normalised to the geometric mean of Tbp and Hmbs (−DOX [n=5], +DOX [n=5], WD4D [n=4]). (f) Pancreatic sections immunostained for UCN3 and MafA. (g) Pancreatic sections immunostained for INS, NKX6.1 and BrdU. (h) Quantification of active beta cell cycling after 7 days cumulative BrdU labelling (−DOX [n=12], WD7D [n=24]). (i) Immunostaining for INS, NKX6.1 and Ki67 on pancreatic sections. (j) Beta cell proliferation based on Ki67-positivity (−DOX [n=6], +DOX [n=5], WD4D [n=5], WD7D [n=5]). Data are presented as mean ± SEM. Scale bar, 100 µm. *p≤0.05, **p≤0.01, ***p≤0.001, analysed by one-way ANOVA followed by Tukey’s post hoc test (d, e, j), mixed-effects analysis (c) or unpaired t test (h)

Transient transgene-independent ER stress promotes mouse beta cell proliferation in vitro

To assess whether transient, transgene-independent ER stress promotes beta cell proliferation, C57BL/6JRj mouse islets were exposed in vitro to thapsigargin (6 h) at 11 mmol/l glucose, followed by a 72 h washout period in EdU-supplemented medium (Fig. 5a). Thapsigargin exposure induced ER stress, as evidenced by upregulation of Hspa5, Atf6, Atf4 and Ddit3 compared with controls, which largely resolved after the washout (Fig. 5b). Beta cell proliferation, assessed by Ki67-positivity, was significantly increased following thapsigargin washout (mean ± SEM 0.8 ± 0.1% vs 0.3 ± 0.1% in controls) (Fig. 5c, d). Cumulative EdU labelling further confirmed enhanced beta cell proliferation during the recovery phase (mean ± SEM 2.6 ± 0.4% vs 0.8 ± 0.2% in controls) (Fig. 5c, e) but not during the initial 6 h exposure (ESM Fig. 6d, e).

Fig. 5.

Thapsigargin- and tunicamycin-induced transient ER stress promotes beta cell proliferation. (a) Experimental setup: C57BL/6JRj mouse islets were exposed to 0 or 750 nmol/l thapsigargin for 6 h in RPMI 1640 medium containing 11.1 mmol/l glucose, followed by a washout period of 0 (no washout) or 72 h. EdU was added during the washout phase to label proliferating cells. (b) RT-qPCR analysis of Ins1, Hspa5, Atf6, Atf4 and Ddit3 normalised to the geometric mean of Ubc and Ppia. Data are shown as fold change relative to the 6 h 0 nmol/l thapsigargin condition. (c) Representative immunostaining of islets for NKX6.1, Ki67 and EdU to assess beta cell identity and proliferation. (d) Quantification of beta cell proliferation index based on Ki67 staining. (e) Cumulative beta cell cycle entry measured by EdU incorporation at 72 h post-washout. (f) Experimental setup: C57BL/6JRj mouse islets were exposed to 0 or 5 µg/ml tunicamycin for 8 h in RPMI 1640 medium containing 11.1 mmol/l glucose, followed by a washout period of 0 (no washout) or 72 h. EdU was added during the washout phase to label proliferating cells. (g) RT-qPCR analysis of Ins1, Hspa5, Atf6, Atf4 and Ddit3 normalised to the geometric mean of Ubc and Ppia. Data are shown as fold change relative to the 8 h 0 µg/ml tunicamycin condition. (h) Representative immunostaining of islets for NKX6.1, Ki67 and EdU to assess beta cell identity and proliferation. (i) Quantification of beta cell proliferation index based on Ki67 staining. (j) Cumulative beta cell cycle entry measured by EdU incorporation at 72 h post-washout. Data are presented as mean ± SEM. Scale bar, 100 µm. *p≤0.05, **p≤0.01, ***p≤0.001, analysed by two-way ANOVA followed by Sidak’s post hoc test. Tg, thapsigargin; Tm, tunicamycin

To determine whether this effect was reproducible with another ER stress inducer, islets were exposed to tunicamycin (8 h) under similar conditions, followed by a 72 h washout (Fig. 5f). Tunicamycin similarly triggered ER stress, as shown by increased expression of Hspa5, Atf6, Atf4 and Ddit3, which subsided after the washout period (Fig. 5g). Beta cell proliferation was elevated post-washout (mean ± SEM 2.5 ± 0.5% vs 0.4 ± 0.1% in controls) (Fig. 5h, i) but not during the exposure itself (ESM Fig. 6f, g). EdU incorporation again confirmed increased proliferation during the recovery phase (3.8 ± 0.9% vs 1.0 ± 0.2% in controls) (Fig. 5j). These findings corroborate the notion that transient ER stress, followed by recovery, can act as a stimulus for beta cell proliferation, independent of transgene expression.

Glucose availability modulates ER-stress-induced beta cell proliferation in vitro

To assess whether the in vitro proliferative response upon the transient exposure to ER stressors was affected by glucose concentration, we exposed C57BL/6JRj islets in vitro to thapsigargin (6 h) under low (3.9 mmol/l) and high (16.7 mmol/l) glucose concentrations (Fig. 6a). Under the low-glucose condition, thapsigargin reduced Ins1 expression but did not significantly induce canonical ER stress markers compared with controls (Hspa5, Atf4, Ddit3, Atf6), nor did it promote beta cell proliferation during exposure or after washout (Fig. 6b–e and ESM Fig. 6d, e). In contrast, under high-glucose conditions, thapsigargin robustly induced ER stress gene expression and significantly increased beta cell proliferation, as assessed by Ki67 staining (mean ± SEM 5.3 ± 0.7% vs 1.4 ± 0.2% in controls) and cumulative EdU labelling during a 3 day thapsigargin washout period (9.0 ± 1.2% vs 2.0 ± 0.4% in controls) (Fig. 6b–e), with a proliferative response exceeding that observed at 11 mmol/l glucose (Fig. 5c–e). Collectively, these findings demonstrate that glucose availability potentiates transient ER-stress-induced beta cell proliferation.

Fig. 6.

Glucose availability modulates transient ER stress-induced beta cell proliferation. (a) Primary islets from C57BL/6JRj mice were exposed to 0 or 750 nmol/l thapsigargin for 6 h in RPMI 1640 medium containing 3.9 or 16.7 mmol/l glucose, followed by a 0 (no washout) or 72 h washout period. EdU was added during washout to label proliferating cells. (b) Relative expression of Ins1, Hspa5, Atf6, Atf4 and Ddit3, normalised to the geometric mean of Ubc and Ppia. Data are shown as fold change relative to the 6 h 0 nmol/l thapsigargin condition. (c) Immunofluorescent staining for NKX6.1, Ki67 and EdU. (d) Quantification of beta cell proliferation based on Ki67 positivity. (e) Cumulative beta cell cycle entry after 72 h, assessed by EdU incorporation. Data are presented as mean ± SEM. Scale bar, 100 µm. ***p≤0.001, analysed by two-way ANOVA followed by Sidak’s post hoc test. Tg, thapsigargin

Discussion

Restoring the functional beta cell mass is a key objective in advancing diabetes therapies. In this study, we investigated whether recovery from transient islet hypovascularisation could stimulate beta cell regeneration. Transgenic overexpression of sFLT1 in beta cells induced islet hypovascularisation, and cessation of sFLT1 expression led to increased beta cell cycling, despite the absence of vascular recovery. Brissova and colleagues previously demonstrated that transient islet hypervascularisation promotes beta cell proliferation through interactions among intra-islet macrophages, endothelial cells and extracellular matrix components [33, 49]. However, in our model, we observed no significant differences in macrophage numbers or transcriptomic profiles across conditions.

Single-cell transcriptomic analysis revealed reduced expression of beta cell identity markers, alongside increased ER stress and elevated UPR gene expression in sFLT1-overexpressing beta cells. The loss of end-stage beta cell maturation markers is consistent with the requirement for endothelial cell–beta cell crosstalk and nutrient availability to maintain functional maturity [50]. ER stress, exacerbated by transgene production and islet hypoxia, likely amplified this dedifferentiation response [51–53]. Together, these findings suggest that transient hypovascularisation and ER stress can impair beta cell identity while simultaneously promoting proliferation, highlighting a complex balance between regeneration and functional maturity.

Transient GFP overexpression in beta cells also induced ER stress followed by an increase in beta cell proliferation, although the effect was less pronounced than in RIP-rtTA;TetO-sFLT1 mice. This difference in mitogenic response likely reflects the absence of a hypoxic microenvironment and elevated blood glucose levels in the GFP model that, in the sFLT1 context, amplify ER stress and drive beta cells toward a more immature, proliferative state. Although GFP is a non-secreted protein, its overexpression nonetheless triggered ER stress gene expression, which resolved upon cessation. Mechanistically, GFP overexpression likely induces ER stress indirectly by disrupting proteostasis through chaperone sequestration, overloading the ubiquitin-proteasome system and promoting protein misfolding or aggregation [54–56]. These disturbances may secondarily activate the UPR via elevated reactive oxygen species and impaired ER calcium homeostasis [56, 57]. Our findings align with previous reports in the mouse lens, where transgene overexpression alone was sufficient to initiate ER stress and UPR activation, independent of protein localisation or secretion [58]. Together, these observations underscore that transgene overexpression, even of non-secreted proteins like GFP, can perturb cellular homeostasis and activate ER stress pathways capable of influencing beta cell proliferation.

In vitro modelling of transient ER stress by exposing wild-type mouse beta cells to thapsigargin and tunicamycin successfully mimicked the mitogenic response observed in vivo. Our thapsigargin experiments further revealed that glucose potentiates beta cell proliferation when combined with an ER-stress-inducing compound, whereas low-glucose conditions fail to do so. In beta cells, glucose-driven protein synthesis and increased secretory demand can overwhelm the ER’s folding capacity, thereby intensifying ER stress and contributing to the overall stress burden. Previous studies have similarly shown that high-glucose levels activate the UPR in wild-type mouse islets [59]. This may also explain the stronger proliferative response observed in the sFLT1 model, where blood glucose values were elevated during DOX administration. Collectively, these findings highlight the synergistic role of glucose and transient ER stress in promoting beta cell proliferation and underscore the importance of metabolic context in shaping the cellular response to stress.

Our findings are consistent with previous studies linking the UPR to beta cell proliferation. Szabat et al demonstrated that physiological insulin production induces baseline ER stress, which suppresses adult beta cell proliferation; notably, deletion of both insulin alleles stimulated proliferation prior to the onset of hyperglycaemia [39]. Sharma et al further showed that the UPR senses insulin demand and, through activation of the activating transcription factor 6 (ATF6) branch, can promote beta cell proliferation [59]. Xin et al, using scRNA-seq on human cadaveric donor islets, identified an INS^loUPR^high beta cell population enriched for proliferative markers [60]. Additionally, the salt-inducible kinase (SIK) inhibitor HG-9-91-01 was found to induce beta cell proliferation via transient ATF6-dependent UPR activation, although additional signalling pathways were required for a full mitogenic effect [61]. In our model, sFLT1 overexpression induced ER stress, as evidenced by upregulation of BiP and activation of both the PERK–eIF2α–ATF4 and IRE1–XBP1 branches of the UPR, preceding the onset of beta cell proliferation. Unlike other studies suggesting a role for Akt signalling [39], we did not observe increased p-Akt levels in the sFLT1 model. Instead, we found elevation of p-S6RP, a downstream effector of the mTORC1 pathway [48], coinciding with peak beta cell proliferation. These findings implicate mTORC1 signalling as a potential driver of ER-stress-mediated beta cell proliferation and align with previous reports linking ER stress and hypoxia to mTORC1 activation [62]. Importantly, mTORC1 can be activated independently of Akt signalling [63–65].

Regenerative responses to ER stress relief are of major clinical interest as ER stress is closely related to beta cell failure and the progression of type 1 and type 2 diabetes [66–68]. Glucagon-like peptide-1 (GLP-1) agonists are now established as first-line therapy for type 2 diabetes and their beneficial effects may be partly attributed to the modulation of ER stress and the UPR [69, 70]. In new-onset type 1 diabetes, initiating insulin administration after a variable period of increased metabolic workload due to decreased beta cell numbers may replicate an ER stress/UPR response with subsequent relief, thereby possibly contributing to the honeymoon phase observed in the weeks after starting insulin administration. Given the regenerative potential of alleviating ER stress, it is tempting to propose early intensive insulin therapy to reduce the strain on the remaining beta cells and promote their proliferation. However, intensive diabetes management did not prevent functional beta cell mass decline in children with newly diagnosed type 1 diabetes [71, 72], and parenteral insulin therapy did not slow disease progression in the DPT-1 trial (ClinTrials.gov registration no. NCT00004984). Nonetheless, it is worth exploring early insulin therapy to mitigate metabolic stress on remaining beta cells in combination with immune interventions that halt the concurrent inflammatory stress.

Our study highlights unexpected effects of ER stress in transgenic mouse strains commonly used for beta cell research. Specifically, we used the RIP-rtTA;TetO-X mouse, a frequently used model in other beta cell studies [12, 33, 73, 74]. One should carefully consider evaluating ER stress in these strains, as it may bias interpretations of beta cell behaviour. In summary, our findings that ER stress relief provides a regenerative impetus to beta cells should encourage further studies on alleviating the metabolic workload of beta cells in early-onset diabetes and could have potential implications for pharmacological intervention.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ATF4

Activating transcription factor 4

ATF6

Activating transcription factor 6

DOX

Doxycycline

dTg

Double-transgenic

eIF2α

Eukaryotic translation initiation factor 2 subunit 1

ER

Endoplasmic reticulum

IRE1α

Endoplasmic reticulum to nucleus signalling 1 (inositol-requiring-enzyme 1α)

mTORC1

Mechanistic target of rapamycin complex 1

PERK

Eukaryotic translation initiation factor 2α kinase 3

RIP

Rat insulin promoter

rtTA

Reverse tetracycline-dependent transactivator

S6RP

S6 ribosomal protein

scRNA-seq

Single-cell RNA-seq

sFLT1

Soluble fms-like tyrosine kinase 1

sTg

Single transgenic

TetO

Tetracycline operator

UCN3

Urocortin 3

UPR

Unfolded protein response

VEGF

Vascular endothelial growth factor

WD

Withdrawal

WD1D

1 Day after DOX WD

WD4D

4 Days after DOX WD

WD7D

7 Days after DOX WD

VEGFR

VEGF receptor

Acknowledgements

We are grateful to H. Azibou, N. Nasiri, M. Beya, A. Demarré, V. Laurysens and G. Stangé (Vrije Universiteit Brussel, Brussels, Belgium) for technical assistance, K. De Keersmaecker and M. Caruso (KU Leuven, Leuven, Belgium) for advice on protein translation assays, and R. Scharfmann and A. Fouque (INSERM, Paris, France) for sharing the puromycin incorporation protocol. Some of the data were presented as an abstract at the European Society for Paediatric Endocrinology meeting in 2024.

Data and code availability

The single-cell RNA-seq data generated in this study are deposited at GEO (NCBI) with accession code GSE274443. Other data that support the findings of this study are available from the corresponding author upon request. There are no restrictions on data availability.

Funding

This work was supported by grants from the following: the Research Foundation – Flanders (FWO) (G040719N; G099323N); the Fonds National de la Recherche Scientifique (FNRS); the FWO and FRS-FNRS under the Excellence of Science (EOS) programme (Pandarome project 40007487); and the Walloon Region strategic axis FRFS-WELBIO, Belgium (to MC). SB, SC, AVM and LW are doctoral fellows from the FWO with grant numbers 1S89821N, 11P3Z24N, 11I3123N and 1187425N, respectively. WS holds an FWO Senior Clinical investigator Grant (1806426N) and a BreakthroughT1D Career Development Award (5-CDA-2024-1491-S-B).

Authors’ relationships and activities

The authors declare that there are no relationships or activities that might bias, or be perceived to bias, their work.

Contribution statement

WS, NDL, HH and SB designed the study. SB wrote the initial draft of the manuscript. SB, AVM, LW, JP, SC, YH, LD, GL, CV and YT carried out the experiments, including mouse treatments and follow-up, microscopy imaging and analysis and in vitro experiments with ER-stress-inducing compounds. IS contributed to the design of the scRNA-seq experiments, performed the experiments and assisted in the interpretation of the results. DK and XY analysed the scRNA-seq data and assisted in figure generation. GL, AVM and SB performed histological analyses. MC, KM, EdK and FC provided resources, shared their expertise for data interpretation, and gave critical feedback on the manuscript. Writing, review and editing were performed by AVM, LW, SC, LD, JP, GL, DK, IS, KM, FC, EdK, XY, CV, WS, NDL, YH, MC and HH. Funding was acquired by HH, WS and NDL. All authors approved with the final version of the manuscript. WS and NDL are the guarantors of this work.