Insulin production is sustained during DNA damage-mediated senescence in adult human beta cells

Abstract

Aims/hypothesis

Residual pancreatic beta cells in type 1 diabetes show reduced insulin production but the mechanisms remain unclear. Beta cells undergo stress responses during type 1 diabetes, including endoplasmic reticulum (ER) stress and DNA damage-associated senescence, which may affect insulin production. ER stress reduces insulin production but whether senescence disrupts insulin production in human beta cells has not been investigated.

Methods

DNA damage-mediated senescence was induced using bleomycin in human donor islets. Relative levels of prohormone convertase 1/3 (PC1/3), prohormone convertase 2 (PC2), carboxypeptidase E (CPE) and the endogenous PC1/3 inhibitor, proprotein convertase subtilisin/kexin type 1 inhibitor (proSAAS), were quantified by western blot. Levels of proinsulin and insulin were measured by ELISA. Flow cytometry was used to measure insulin expression in islet cells. FACS was used to sort endogenous senescent beta cells from islets for analysis of insulin content. Proinsulin immunofluorescence staining was quantified in endogenous senescent vs non-senescent beta cells in pancreas tissue from control donors and donors with type 1 diabetes. Publicly available datasets were used to interrogate relationships between senescence effectors, proinsulin-processing genes and insulin content. DNA damage was induced with bleomycin in the non-proliferative female-fetus-derived EndoC-βH5 human beta cell model to study the impact of the DNA damage response on insulin production in clonal cells growth-arrested due to p16^INK4A expression.

Results

DNA damage-mediated senescence led to increased PC1/3 without changes in levels of PC2, CPE or proSAAS in human islets. Consistent with these changes, no significant differences in proinsulin or insulin content were observed, compared with control islets. Flow cytometry confirmed maintenance of insulin content in DNA damage-mediated senescent beta cells vs control cells and sorted endogenous senescent beta cells had similar insulin content to non-senescent beta cells. Proinsulin staining was similar in endogenous senescent vs non-senescent beta cells from a control donor and donor with type 1 diabetes. Analysis of proteomics datasets from Humanislets.com and single-cell RNA-seq datasets from the Human Pancreas Analysis Program corroborated these findings. In EndoC-βH5 beta cells, which are growth-arrested, DNA damage led to decreased levels of CPE and proSAAS, and reduced levels of insulin.

Conclusions/interpretation

Our findings suggest that the expression of proinsulin-processing enzymes and the production of insulin are sustained in both chemically induced DNA damage-related senescence and in endogenous senescent adult human beta cells. Collectively, these findings suggest that senescent beta cells may be a source of insulin production among residual beta cells in type 1 diabetes.

Graphical Abstract

Supplementary Information

The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-025-06603-3.

Introduction

Type 1 diabetes results from chronic autoimmune-mediated destruction of insulin-producing pancreatic beta cells. In recent years, the role of beta cell stress and dysfunction have been recognised as factors contributing to the progression of the disease [1]. A variety of defects in insulin production have been reported in residual beta cells in the natural history of type 1 diabetes, including reduced INS nascent RNA, reduced mature insulin content, persistent proinsulin secretion, lower proinsulin content, higher proinsulin:C-peptide ratio [2–5] and reduced levels of the proinsulin-processing enzymes prohormone convertase 1/3 (PC1/3) and carboxypeptidase E (CPE) [2, 5, 6]. Various programs of beta cell stress involving DNA damage have been observed in pancreatic tissue from donors with type 1 diabetes, including endoplasmic reticulum (ER) stress and unfolded protein response (UPR), type I IFN response and cellular senescence [7]. The senescence stress response is a programmed state of growth arrest and DNA damage triggered by activation of cyclin-dependent kinase inhibitors. Cellular senescence leads to apoptosis resistance and a proinflammatory secretome, termed the senescence-associated secretory phenotype (SASP) [8]. Activation of the UPR and IFN response both involve ER stress, which impairs proinsulin processing and significantly reduces insulin content in human beta cells [2, 9]; however, the effect of senescence on insulin production and processing in human beta cells remains unclear.

Both residual beta cells in new-onset type 1 diabetes donors and senescent beta cells in the NOD mouse model express markers of DNA damage and senescence [10, 11]. Since DNA damage is a classical stressor leading to senescence in other cell types, these observations suggest that it may also be a trigger for senescence in beta cells. Consistent with this notion, exposure of isolated human islets to sublethal DNA damage using the chemotherapeutic agent bleomycin induces the hallmark features of senescence, which is also observed in a subset of residual beta cells in donors with type 1 diabetes [12]. This response is characterised by elevated and stable expression levels of cyclin-dependent kinase inhibitor p21 (encoded by CDKN1A) without an effect on the cyclin-dependent kinase inhibitor p16^INK4A (encoded by CDKN2A), activation of the DNA damage response, SASP and pro-survival BCL-2 family gene expression [13]. In contrast to DNA damage-associated senescence, beta cells that develop senescence during ageing show different senescence features, including upregulated p16^INK4a and the lysosomal senescence-associated β-galactosidase (SA-βgal) activity, without high p21 expression, DNA damage, SASP or a BCL-2 family pro-survival phenotype [14]. Previously, we showed that static glucose-stimulated insulin secretion is not impaired in response to DNA damage-mediated senescence in human islets [12]. However, reduced insulin content was observed in some donors, despite lack of effect on the mRNA level of INS or the genes encoding prohormone-processing factors PCSK1 (encoding PC1/3), PCSK2 (encoding PC2), CPE or PCSK1N (encoding proprotein convertase subtilisin/kexin type 1 inhibitor [proSAAS]) [12]. In contrast, we also showed that DNA damage in a well-characterised human beta cell model, EndoC-βH5 [15], impaired insulin secretion [12], suggesting that this human beta cell line responds differently to DNA damage compared with adult donor islets. Thus, the impact of DNA damage-mediated senescence activation on the pancreatic beta cell insulin production remain unclear.

Here, we investigated how DNA damage-mediated senescence affects proinsulin processing and mature insulin content in human donor islets and beta cells. We characterised the effect of DNA damage-mediated senescence on proinsulin-processing enzymes, proinsulin and mature insulin content in human islets and EndoC-βH5 cells. In addition, we corroborated these findings with analyses of publicly available datasets from Humanislets.com [16], the Human Pancreas Analysis Program (HPAP) [17, 18] and immunohistochemistry on pancreas sections from control donors and donors with type 1 diabetes. Our study has implications for understanding how different stress responses in residual beta cells in type 1 diabetes contribute to defective insulin production and suggest that senescent beta cells may be a previously unappreciated source of residual insulin production in type 1 diabetes.

Methods

Culture of human donor islets and EndoC-βH5 cells

Islets isolated from 21 human donors without diabetes were used in this study (electronic supplementary material [ESM] Table 1). Adult human donor islets for research were provided by the Alberta Diabetes Institute (ADI) IsletCore at the University of Alberta in Edmonton, Canada (http://www.bcell.org/adi-isletcore.html). Islet isolation was approved by the Human Research Ethics Board at the University of Alberta (Pro00013094). All donors’ families gave informed consent for the use of pancreatic tissue in research. Human donor pancreatic islets were also provided by the NIDDK-funded Integrated Islet Distribution Program (IIDP). Female-fetus-derived EndoC-βH5 cells [15] were purchased from Human Cell Design (Toulouse, France) and were validated by the supplier and confirmed mycoplasma free. Details of human islet and EndoC-βH5 culture and treatments are outlined in ESM Methods and the human islets checklist can be found at the end of the ESM file.

Western blot analysis and quantification

Islets or EndoC-βH5 cells were harvested and lysed in ice-cold radioimmunoprecipitation assay (RIPA) lysis and extraction buffer (ThermoFisher, Waltham, MA, USA) containing 1X Halt complete protease and phosphatase inhibitor cocktail, EDTA free (ThermoFisher), and protein concentration was determined by BCA assay (ThermoFisher). Western blotting was performed as described [12] using approximately 1–10 µg of total RIPA lysate protein. After transfer to nitrocellulose membranes and blocking with 5% wt/vol. non-fat milk in TBS-T (containing 0.1% vol./vol. Tween-20), membranes were probed with validated primary antibodies for p21, PC1/3, prohormone convertase 2 (PC2), CPE, proSAAS, β-actin and vinculin, as shown in ESM Table 2. Signals were detected with HRP-conjugated secondary antibodies (Jackson Immunoresearch) diluted 1:50,000 to 1:100,000 in 5% wt/vol. non-fat milk in TBS-T and ECL reagents (SuperSignal West pico plus, ThermoFisher) after exposure to x-ray film. See ESM Methods for further details.

ELISA

Human proinsulin or mature insulin were measured by specific ELISAs on islet protein lysates or EndoC cells extracted with acidified ethanol (0.15 mol/l HCl, 95% vol./vol. ethanol) overnight at 4°C. Samples were diluted 1:100–1:5000 prior to ELISA using commercial kits (Mercodia, Uppsala, Sweden). SASP factor GDF15 was detected by ELISA (R&D Systems, Minneapolis, MN, USA) in conditioned media of human islets or EndoC-βH5 cells collected on day 4 or 5 after drug washout, as previously described [12]. See the ESM Methods for information on ELISA specificity.

Flow cytometry, FACS cell sorting of islet cells and image cytometry

Intracellular flow cytometry of human donor islets for p21 and insulin was performed as previously described [19, 20]. Our published flow cytometry data on DNA damage-induced senescent and control islets from three different donors [20] were reanalysed to examine insulin median fluorescence intensity (MFI) in the live islet cell populations. FACS sorting of beta cell-enriched populations with the surface marker ectonucleoside triphosphate diphosphohydrolase 3 (ENTPD3) and the SA-βgal fluorogenic substrate 5-dodecanoylamino fluorescein di-β-d-galactopyranoside (C₁₂FDG) (ThermoFisher) was performed using modifications to previous methods [21, 22]. After resting islets in culture for 24 h, they were dissociated with 0.025% trypsin-EDTA and adjusted to a concentration of 1 × 10⁶ cells/ml in islet media. Islet cell suspensions were incubated with 16.5 μmol/l C₁₂FDG (ThermoFisher) for 1 h in a 37°C water bath, washed once in cell staining buffer (BioLegend) and then incubated with 1:250 ENTPD3 antibody (ectonucleotidases-ab, Université Laval, Quebec, QC, Canada) (ESM Table 2) for 30 min on ice, followed by 1:200 anti-mouse IgG-APC conjugated secondary antibody and 1:1000 Zombie near-infrared live-dead stain (BioLegend) for 30 min on ice protected from light. Approximately 5000–10,000 live, ENTPD3⁺/C₁₂FDG⁺ (SA-βgal⁺) or ENTPD3⁺/C₁₂FDG⁻ (SA-βgal⁻) beta cells were sorted on a FACSMelody sorter directly into RIPA lysis buffer containing protease inhibitor and phosphatase inhibitor cocktail and used in ELISA for proinsulin and insulin. To perform image cytometry, dissociated islets were stained to detect SA-βgal activity with C₁₂FDG as above, followed by washing in cell staining buffer and incubation for 5 min in 1 µg/ml propidium iodide (ThermoFisher) as a live/dead marker. Approximately 20 µl of cell suspension (≥1.5 × 10⁴ cells/ml) was analysed by image cytometry using the Cellometer Spectrum platform and software, which imaged and analysed fluorescence from 360–640 cells per sample. Quadrant plots were generated using FCS Express software version 7.0 (DeNovo Software) to visualise the propidium iodide⁻/C₁₂FDG⁺ (Live, SA-βgal⁺) islet cell population from each sample.

Immunohistochemistry

Formalin-fixed paraffin-embedded human donor pancreas sections were obtained from the Network for Pancreas Organ Donors with Diabetes (nPOD) and subjected to fluorescent immunostaining as previously described [5]. nPOD donors used in this study were: donor 6229 (31-year-old female, control); and donor 6196 (26-year-old female, type 1 diabetes of 15 years duration) (ESM Table 1). Sections were stained with monoclonal rabbit p21 antibody and monoclonal proinsulin antibody (ESM Table 2), Alexa488-conjugated or Alexa594-conjugated secondary antibodies (Jackson Immunoresearch) diluted 1:250 in PBS and counterstained with DAPI. Mounted slides were imaged on a Leica TCS SP5 confocal microscope (Leica Microsystems, Hesse, Germany). All images were acquired using the same exposure setting across samples and pixel intensity was kept within the dynamic range. Quantification of proinsulin in p21⁺ and p21⁻ islet cells was manually performed using ImageJ version 1.54p (National Institutes of Health) and by extracting integrated pixel intensities (average pixel intensity normalised for cell area) for the proinsulin channel on these cells. Proinsulin⁺ cells were scored in ten islets from the control donor and seven islets from the donor with long-standing type 1 diabetes.

Public dataset analyses

Publicly available single-cell RNA-seq datasets from the HPAP (https://hpap.pmacs.upenn.edu/ accessed 28 May 2025) [17, 18, 23] were analysed using CellxGene viewer. Datasets were subset on pancreatic beta cells from control donors, autoantibody-positive donors and donors with type 1 diabetes, and gene expression per cell was visualised for CDKN1A, PCSK1, PCSK2, CPE and PCSK1N using bivariate plots. Bulk RNA-seq data for FACS-sorted beta cells (n=4 donors) [24] previously analysed and compared with identified differentially expressed genes with EndoC-βH5 cells (n=5 independent batches of cells) using DESeq2 [15] were plotted to compare gene expression (normalised read counts) of selected cyclin-dependent kinase inhibitor genes and DNA repair genes between the cell types. Genes identified as significantly differentially expressed between cell types [15] were noted with an asterisk, while other genes were not significant. To identify islet protein associations with ageing, we used the HumanIslets.com phenotyping data platform [16] involving human islet proteomics datasets [25] (accessed 28 May 2025). Proteins that had a significant association with donor age were identified after controlling for sex, BMI, HBA_1c and diabetes status, using 120 donor islet preparation datasets meeting these criteria. Plots were extracted from Humanislets.com for cyclin-dependent kinase inhibitor 2A (CDKN2A; p16^INK4A and p14^ARF proteins) and cyclin-dependent kinase inhibitor 1A (CDKN1A; p21 protein). We also extracted plots showing the associations between CDKN2A and CDKN1A with insulin content per islet equivalent (IEQ).

Statistical analysis

Data are expressed as mean ± SD and all statistical analysis was done with GraphPad Prism version 10.4.2 (DotMatics). No sample randomisation or blinding was done. Aggregated data were tested for normality/lognormality and data that did not pass the normality test were analysed by non-parametric Mann–Whitney test. This was only done for results with CPE in islets and PC2 in EndoC-βH5 cells. For each human islet preparation, or EndoC-βH5 cells, typically 3–5 independently treated wells of islets or cells per condition were analysed and we considered each human donor islet preparation or vial of EndoC-βH5 cells as an independent experiment. Specific ‘n’ sample size per group was plotted for each experiment and the number of donor islet experiments or vials of EndoC-βH5 cell experiments are described in figure legends. To report intra-donor and within-experiment heterogeneity between treated wells of islets and cells, we also analysed and compared each islet sample replicate across all donors or independently treated wells of cells for each vial for EndoC-βH5 experiments. The term ‘islet sample replicates’ or ‘independently treated well of cells per vial’ are here considered as ‘technical’ replicates with respect to each islet donor or vial of EndoC-βH5 cells as they are repeated measures of samples taken from the same islet donor or vial of EndoC-βH5 cells, respectively. Unpaired two-tailed t tests, Mann–Whitney tests or two-way ANOVAs were used to determine statistical significance and results were considered significant at p<0.05.

Results

DNA damage-mediated senescence leads to increased PC1/3 in human islets

We first investigated how DNA damage-mediated senescence impacts the expression of enzymes and proteins involved in proinsulin processing in human islets. Specificity of antibodies for recombinant PC1/3, CPE, and proSAAS proteins were confirmed by western blotting and showed sensitivity down to 3–15 ng (ESM Fig 1a–d, ESM Table 2). Next, we used our established DNA damage-mediated senescence model in human islets [12, 19], entailing a 48 h drug treatment period followed by a 4–6 day washout period to examine stable, established senescence phenotypes. This mode of senescence induction leads to upregulation of p21 but not p16^INK4A, reflecting the signature of cyclin-dependent kinase inhibitor expression found in beta cells in type 1 diabetes donor pancreas [11], and involves persistent DNA damage response, SASP and pro-survival gene expression [12, 20, 26]. DNA damage-mediated senescence was induced by this approach in islet preparations from ten adults without diabetes (five male and five female donors; ESM Table 1).

We previously validated senescence phenotypes in this model using orthogonal approaches such as western blot analysis, qRT-PCR, flow cytometry, RNA-seq and mutli-plex protein secretion assays [12, 19] and confirmed senescence for each donor islet experiment in this study. First, we confirmed stable upregulation of p21 at 4–6 days post-drug removal (ESM Fig. 2a), consistent with our previous data showing a persistent DNA damage response via ataxia telangiectasia mutated (ATM) activation, which persists 4 days after drug washout [12]. Second, islets also developed a SASP as indicated by increased secretion of a flagship SASP factor, growth and differentiation factor 15 (GDF15), in the conditioned media (ESM Fig. 2b) [12], in accordance with our previous report showing increased GDF15 secretion from islets isolated from an adult donor with type 1 diabetes as compared with a control donor without diabetes [20]. Third, flow cytometry confirmed that p21 is upregulated in live beta cells (insulin⁺ cells) [19, 20], with ~12–30% of beta cells developing a p21^High phenotype and elevated p21 levels in the insulin-expressing cells of bleomycin-treated islets shown by MFI (ESM Fig. 3a–c). The percentage of live islet cells is modestly reduced from approximately 99% in controls to 76% in bleomycin-induced senescence (ESM Fig. 3a). This is similar to the p21^High phenotype we previously identified in residual beta cells in an adult donor with type 1 diabetes [20]. Fourth, results from our previous RNA-seq experiment on DNA damage senescent and control islets [12], including an islet preparation from donor R434 used in this study, demonstrated that the islets developed a p53-driven transcriptional response. This response was characterised by decreased expression of cell cycle and proliferation genes (MKI67, PCNA, CDK1, CCND1, TOP2A), increased expression of p53 cell cycle arrest and target genes (PHLDA3, CDKN1A, ZMAT3, DDB2, DINOL) and SASP genes (CXCL1, TNFRSF10C, CXCL5, IGFBP4, GDF15), and elevated expression of pro-survival BCL-2 family gene BCL2L1. Notably, while CDKN1A is upregulated, CDKN2A (encoding p16^INK4A) is not upregulated in this model [11, 12, 26]. Finally, we monitored changes in SA-βgal activity in dissociated islet cells following DNA damage and senescence induction using the fluorogenic SA-βgal substrate C₁₂FDG, which has been used to quantify SA-βgal⁺ senescent islet cells in mice [27] and humans [28]. The percentage of C₁₂FDG^High islet cells was higher in the islets from a 57-year-old male donor than in islets from a 21-year-old male donor (ESM Table 1, ESM Fig. 4a), as expected, consistent with the age-related increase in p16^INK4A-expressing SA-βgal⁺ senescent beta cells [28, 29]. However, we found that the C₁₂FDG^High population of live islet cells was reduced during DNA damage-mediated senescence compared with controls (ESM Fig. 4b) and this was also the case when we examined C₁₂FDG^High live beta cells following DNA damage and senescence induction (ESM Fig. 4c, d). This finding is consistent with the lack of increase in CDKN2A in this model and suggested that the proportion of SA-βgal⁺ beta cells does not increase in response to DNA damage and p21 upregulation in islets. Taken together, our data indicate that in response to sublethal DNA damage with bleomycin, human islets activate a p53 transcriptional program involving p21 as the main cyclin-dependent kinase inhibitor and also develop a SASP and pro-survival gene expression without an increase in the per cent of SA-βgal⁺ cells.

We next performed western blotting for proinsulin-processing enzymes PC1/3, PC2, CPE and proSAAS on control and senescent islets lysates from each donor (n=10 donors in total). We performed relative quantifications of each protein at the level of each islet donor experiment to show inter-donor variation and at the level of individually treated samples of islets (islet sample replicates) across all donors, to reveal intra-donor variation. On average per islet donor, senescent islets showed increased levels of PC1/3 (p<0.01) without levels of PC2, CPE or proSAAS being affected (Fig. 1a and ESM Fig. 5). There were inter-donor differences in the changes during senescence, with some donor islet preparations showing no differences compared with controls and others showing increases or decreases in PC2, CPE or proSAAS (Fig. 1a). Analysis of mean fold-change per islet sample replicate, which we treated as a technical replicate with respect to each donor (n=44 samples per treatment group) revealed higher PC1/3 levels (p<0.001), PC2 levels (p<0.01) and CPE levels (p<0.01) in senescent islets samples compared with controls (Fig. 1b). A sex-stratified analysis showed an increase of PC1/3 in senescent samples (p<0.05) in islets from male but not female donors (Fig. 1c). The same analysis at the level of each well of islets across all donors (n=22 wells of islets from five male donors; n=22 wells of islets from five female donors) demonstrated increases in PC1/3 (p<0.01) and PC2 (p<0.01) in senescent islets from male donors and CPE (p<0.05) in senescent islets from female donors, compared with their control counterparts (Fig. 1d).

Fig. 1.

Proinsulin-processing factor levels are sustained during DNA damage-induced senescence in human islets. (a) Western blot analysis of PC1/3, PC2, CPE and ProSAAS levels normalised to vinculin or β-actin in human islets following induction of DNA damage-mediated senescence with 50 µmol/l bleomycin vs vehicle control (0.1% vol./vol. DMSO). Data are means ± SD from n=10 islet donor preparations. Preparations of islets from male and female donor are colour-coded with donor ages shown in years. (b) Western blot analysis of PC1/3, PC2, CPE and ProSAAS levels normalised to vinculin or β-actin in human islets following induction of DNA damage-mediated senescence with 50 µmol/l bleomycin compared with vehicle control (0.1% vol./vol. DMSO). Data are means from n=44 islet sample replicates per treatment group, representing islets from the same ten different male and female donors as in (a). (c) Western blot relative quantifications as for (a) except analysed by islet donor sex. Data are mean ± SD of n=5 male and n=5 female donors. (d) Western blot relative quantifications as for (b) except analysed by islet donor sex. Data are mean ± SD of n=22 islet sample replicates per treatment group, representing five male and five female donors. ***p<0.001, **p<0.01, *p<0.05, by two-tailed t test (a, b) or two-way ANOVA with multiple-testing (c, d)

Proinsulin and insulin content is sustained during DNA damage senescence in human islets

We next assessed whether senescence affects proinsulin or insulin content in human islets (Fig. 2). We performed ELISAs on the islet lysate for proinsulin or insulin from the same validated control-treated islet and bleomycin-induced senescent islet protein samples used for western blots of the proinsulin-processing factors shown in Fig. 1 (n=10 donors in total). Analysis of the relative levels of proinsulin and insulin, or the proinsulin:insulin ratio (PI/I ratio), at the donor level showed no significant differences between senescent and control islets (Fig. 2a) and this was also the case when data were analysed at the per-well level of islet sample replicates when comparing treatment groups (Fig. 2b) (n=39 control islet samples, n=38 senescent islet samples). Since insulin content can vary based on the number of beta cells in each sample, we also measured insulin levels by flow cytometry on the live islet cell population following induction of DNA damage senescence (Fig. 2c). p21 MFI was higher in the insulin⁺ cell population of the DNA damage senescent islets compared with control islets as expected (Fig. 2d) but insulin MFI in the insulin⁺ gated population of islet cells did not differ between DNA damage senescent islets and control islets (Fig. 2e) (n=3 donor islet preparations).

Fig. 2.

Insulin content is sustained in senescent beta cells and islets. (a) Relative proinsulin and insulin content and PI/I ratio, measured by ELISA in islets following DNA damage-mediated senescence induction with 50 µmol/l bleomycin compared with vehicle control, as for Fig. 1a. Data are mean ± SD of n=10 islet donor preparations. (b) Relative proinsulin and insulin content measured by ELISA in samples as for (a) except analysed by islet sample replicate treatments across all ten donors. Data are mean ± SD of n=39 islet sample replicates in vehicle controls (0.1% vol./vol. DMSO) from ten islet donors and n=38 biological replicates from ten islet donors in bleomycin-induced senescent islets. (c) Schematic of flow cytometry analysis of p21 and insulin in human islets following DNA damage-induced senescence. Human islets were treated with bleomycin to induce DNA damage senescence, or with vehicle (0.1% vol./vol. DMSO); after senescence establishment at day 5 after bleomycin washout, islets were dissociated and cells were fixed and stained for p21 and insulin for flow cytometry. (d) p21 expression (MFI) in the live insulin⁺ cell population in bleomycin-treated islets compared with vehicle-treated control islets. Data are mean ± SD of n=3 human donor islet donor preparations with >5000 live insulin⁺ cells scored per sample. (e) Insulin expression (MFI) in live islet cell populations in bleomycin-treated compared with vehicle control islets. Data are mean ± SD of n=3 human donor islet preparations with >5000 live insulin⁺ cells scored per sample. (f) Schematic of FACS sorting of endogenous senescent beta cells by SA-βgal activity. Human islets were dissociated, incubated with fluorogenic substrate C₁₂FDG, stained to detect beta cells with ENTPD3 antibody, then sorted into ENTPD3⁺/C₁₂FDG^High (SA-βgal⁺) and ENTPD3⁺/C₁₂FDG^Low (SA-βgal⁻) beta cell populations from each donor for insulin ELISAs. For each sample, 5000–10,000 live ENTPD3⁺ cells were sorted. (g) Percentages of live ENTPD3⁺ islet cells that were C₁₂FDG^Low and C₁₂FDG^High from each donor. Data are mean ± SD from n=3 biological replicates of islets for each donor. (h) ELISA determination of insulin content (normalised per 100 sorted cells) of ENTPD3⁺/C₁₂FDG^Low and ENTPD3⁺C₁₂FDG^High cells. Data are mean ± SD from n=3 human islet donor preparations. *p<0.05 by two-tailed t test. bleo, bleomycin; cyto, cytometry; Veh, vehicle

To further explore the relationships between cyclin-dependent kinase inhibitors and insulin content, we used the HumanIslets.com data portal [16] to examine islet proteomics datasets [25] and identify associations between protein abundance and islet phenotypes. Computational analysis to identify proteins significantly associated with donor age (as an indicator for accumulation of senescent cells) demonstrated 22 proteins with a positive association across 120 donors after controlling for donor sex, BMI, HbA_1c and diabetes status (ESM Fig. 6a, b). These proteins included CDKN2A (p16^INK4A and p14^ARF) but not CDKN1A (p21) (ESM Fig. 6b), suggesting that CDKN2A-expressing islet cells accumulate with age, consistent with previous reports [28, 29]. However, neither CDKN2A nor CDKN1A protein abundance showed a significant association with normalised islet insulin content per IEQ (ESM Fig. 6c).

Endogenous SA-βgal⁺ senescent beta cells have similar insulin content to SA-βgal⁻ beta cells in human islets

Given that our chemical DNA damage model involves a p21⁺ senescence phenotype without upregulation of CDKN2A [11, 12, 26] or increase in the percentage of live SA-βgal⁺ cells (ESM Fig. 4b, d), we next monitored endogenous SA-βgal⁺ beta cells as a different subpopulation of senescent cells and examined differences in insulin content compared with SA-βgal⁻ beta cells. To this end we used C₁₂FDG to sort live human beta cells with the surface marker ENTPD3 [21] based on a C₁₂FDG^High (SA-βgal⁺) vs C₁₂FDG^Low (SA-βgal⁻) phenotype and performed ELISAs for insulin on the lysates (Fig. 2f). We noted a wide range of endogenous C₁₂FDG^High beta cells across islet preparations from three middle-aged adult donors aged 45–54 years, with approximately ~30% to ~80% C₁₂FDG^High beta cells (Fig. 2g and ESM Table 1). However, the insulin content of C₁₂FDG^High beta cells was not significantly different from that of C₁₂FDG^Low beta cells in these donors (Fig. 2h) (n=3 donor islet preparations). Taken together, these data support the notion that both DNA damaged p21-expressing senescent beta cells and endogenous SA-βgal⁺ senescent beta cells show similar insulin content when compared with their non-senescent counterparts.

p21⁺ beta cells have similar proinsulin content to p21⁻ beta cells in type 1 diabetes

To explore the relationship between insulin production and senescence in type 1 diabetes, we examined publicly available single-cell RNA-seq (scRNA-seq) datasets from HPAP [17, 18]. We assessed whether there was a correlation between expression of CDKN1A and proinsulin-processing genes PCSK1, PCSK2, CPE and PCSK1N in human beta cells from control donors (n=27,160 cells), autoantibody-positive donors (n=8,523 cells) and donors with type 1 diabetes (n=715 cells) by the bivariate plots for each combination. No correlations were identified between any of these comparisons (ESM Fig. 7). We next asked whether proinsulin content was different between endogenous p21⁺ vs p21⁻ beta cells in pancreas tissue from control donors and donors with long-standing type 1 diabetes where residual beta cells are still present. Endogenous p21⁺ and p21⁻ islet cells expressing proinsulin were found in both a 31-year-old female normoglycaemic control donor and a 26-year-old female donor with type 1 diabetes who had been diagnosed at age 11 years (15 years diabetes duration) and had residual beta cells indicated by 0.16 nmol/l C-peptide (Fig. 3a and ESM Table 1), consistent with the prior study of p21⁺/insulin⁺ cells in type 1 diabetes [11]. The proinsulin signal intensity per cell was significantly lower in the donor with type 1 diabetes as compared with the control donor without diabetes, in both p21⁺ (p<0.001) and p21⁻ cells (p<0.001) (Fig. 3b), as expected [5, 6]. However, the levels of proinsulin per cell were no different between p21⁺ and p21⁻ cells in each donor (Fig. 3b). Although these comparisons were made on n=1 control donor and type 1 diabetes donor, the result can be considered indicative. In summary, our results suggest that DNA damage-induced senescence in islets or p21⁺ senescent beta cells in type 1 diabetes have sustained insulin production.

Fig. 3.

Proinsulin is maintained in p21⁺ senescent beta cells in pancreas tissue from a control donor and a donor with long-standing type 1 diabetes. (a) Representative immunofluorescence staining of p21 (red), proinsulin (green) and DAPI (white) in pancreas sections from a donor with no diabetes (control, 31 years old, female, nPOD 6229) and a donor with long-standing type 1 diabetes (26 years old, female, diabetes of 15 years duration, nPOD 6196). Scale bar, 50 µm. Zoomed insets show representative p21⁺ (red outline) and p21⁻ (white outline) proinsulin⁺ islet cells. (b) Quantification of proinsulin staining intensity in proinsulin⁺/p21⁺ and proinsulin⁺/p21⁻ islet cells. Data are mean ± SD of n=83 p21⁺ cells and n=148 p21⁻ cells from ten islets randomly selected from the non-diabetic control donor (black circles and bars), n=50 p21⁺ cells and n=61 p21⁻ cells from seven islets randomly selected from the donor with long-standing type 1 diabetes (red circles and bars). ***p<0.001 by two-way ANOVA. a.u., arbitrary units; ND, no diabetes; T1D type 1 diabetes

DNA damage reduces CPE, proSAAS and insulin content in EndoC-βH5 cells

Middle-aged adult human donor islets contain a heterogeneous mixture of SA-βgal⁺ senescent and SA-βgal⁻ non-senescent beta cells at baseline (ESM Fig. 4) due to ageing [28, 29]. Therefore, it is not possible to achieve equivalent induction of DNA damage in both populations of cells in the same experiment to directly compare them in our DNA damage model. DNA damage, p21 activation and SASP could have different effects on insulin production in beta cells that are already growth-arrested with high p16^INK4A expression and SA-βgal activity as compared with those that are not. To determine how the DNA damage response impacts insulin production in a clonal human beta cell model system that has undergone p16^INK4A-mediated growth arrest and has SA-βgal activity we turned to the fetus-derived EndoC-βH5 beta cell line [15]. The removal of transforming simian virus 40 (SV40) large T antigen and human telomerase reverse transcriptase (hTERT) transgenes in the earlier generations of EndoC-βH2 and EndoC-βH3 cells using Cre-LoxP deletion leads to a senescent growth arrest involving elevated p16^Ink4a, SA-βgal activity and senescence transcriptional signatures [29, 30]. Surprisingly, growth-arrested senescent EndoC-βH2 and EndoC-βH3 cells have increased insulin content and secretory function [28, 29, 31]. However, transgene-excised growth-arrested EndoC-βH2 and EndoC-βH3 cells lack DNA damage, p21 upregulation and SASP [28, 29], making these cells a model of age-related proliferation arrest involving p16^INK4A and SA-βgal activity, and not of senescence in response to DNA damage. EndoC-βH5 cells are like the EndoC-βH3 after transgene excision but further incorporate a thymidine kinase suicide gene to enable elimination of residual proliferating cells [15]. We previously showed that EndoC-βH5 cells are capable of mounting a stress response to DNA damage similar to that in adult human donor islets, with elevated CDKN1A but not CDKN2A expression, SASP factor secretion, and pro-survival BCL-2 family gene upregulation [12]. Notably, 2 days after removal of the DNA damaging agent bleomycin, these stressed EndoC-βH5 cells had impaired insulin secretion [12]. This intriguing phenotype prompted us to explore whether DNA damage in the setting of a growth-arrested p16^INK4A-expressing state in human beta cells results in impaired insulin production.

To understand differences in the response to DNA damage between islet beta cells and EndoC-βH5 cells, we first used public RNA-seq datasets [15] to compare cyclin-dependent kinase inhibitor genes and non-homologous end joining (NHEJ) pathway DNA damage and repair genes between these cell types. NHEJ is a major pathway for DNA double-strand break repair in non-dividing cells [32] and disruption or polymorphisms in components of NHEJ repair leads to beta cell apoptosis or senescence in mouse models [33, 34]. EndoC-βH5 cells had lower basal CDKN1A and higher basal CDKN2A (encoding p16^INK4A and p14^ARF), CDKN2B (encoding p15^INK4B) and CDKN2C (encoding p18^INK4C) as compared with FACS-sorted adult beta cells (Fig. 4a), consistent with the high p16^INK4A expression observed in the transgene-excised EndoC-βH2 and EndoC-βH3 cells [28, 29]. EndoC-βH5 cells also expressed much higher levels of NHEJ repair genes LIG4 and XRCC5 compared with adult islet beta cells (Fig. 4a). We next used bleomycin to induce a DNA damage response involving p21 according to our previous EndoC-βH5 cell model [12]. Induction of the DNA damage response with bleomycin treatment for 48 h followed by drug washout and culture for 4–6 days in three independent experiments was validated by western blot for stable upregulation of p21 and ELISA for secreted SASP factor GDF15 in the media at day 6 post-drug washout (ESM Fig. 8a–c), consistent with our previous validation [12]. Remarkably, DNA damaged EndoC-βH5 cells showed significantly lower levels of CPE (p<0.05) and proSAAS (p<0.001) when examined at the level of independent experiments (Fig. 4b and ESM Fig. 8a–c) (n=3 independent experiments). Analysis at the level of independently treated samples across experiments revealed heterogeneity in responses, with reduced PC1/3 (p<0.05), CPE (p<0.001) and proSAAS (p<0.001) levels (Fig. 4c) (n=14 cell samples per group). Consistent with this, DNA damaged EndoC-βH5 cells had reduced insulin content compared with control vehicle-treated cells (p<0.01) without an effect on proinsulin content or PI/I ratio (Fig. 4d). Analysis at the level of independent samples revealed reduced insulin (p<0.001) and proinsulin content (p<0.05) relative to control samples (Fig. 4e). Thus, the DNA damage response triggered by bleomycin led to reduced proinsulin-processing factor levels and reduced insulin content in EndoC-βH5 cells.

Fig. 4.

DNA damage response leads to reduced expression of proinsulin-processing enzymes and lower insulin levels in EndoC-βH5 human beta cells. (a) RNA-seq data of normalised read counts, comparing FACS-purified adult islet beta cells (n=4 donors) with EndoC-βH5 cells (n=5 independent batches), for cyclin-dependent kinase inhibitor genes and NHEJ DNA repair pathway genes. *p<0.05 for difference between groups, as previously reported [15]. (b) Western blot analysis of PC1/3, PC2, CPE and proSAAS in EndoC-βH5 cells treated with 35 µmol/l bleomycin to induce DNA damage response compared with vehicle-treated control cells (0.07% vol./vol. DMSO). Data are mean ± SD from n=3 independent experiments with different vials of EndoC-βH5 cells. (c) Western blot analysis of PC1/3, PC2, CPE and proSAAS as in (b) except analysed by independently treated wells of cells per treatment across three experiments. Data are mean ± SD, n=14 wells of cells per group from three independent batches of EndoC-βH5 cells. *p<0.05, **p<0.01, ***p<0.001 by two-tailed t test. (d) ELISAs for relative proinsulin and insulin content and PI/I ratio in bleomycin-induced DNA damage and vehicle control EndoC-βH5 cells. Data are mean ± SD from n=3 independent experiments with different vials of EndoC-βH5 cells. (e) ELISAs for relative proinsulin and insulin content in bleomycin-induced DNA damage and vehicle control EndoC-βH5 cells as in (d) except analysed by independently treated wells of cells across all experiments. Data are mean ± SD of n=14 wells of cells per group from three independent batches of EndoC-βH5 cells. For (b–d), *p<0.05, **p<0.01, ***p<0.001 by two-tailed t tests

Discussion

Residual beta cells in type 1 diabetes show evidence of defects in proinsulin processing, with altered proinsulin levels [2, 3, 6], reduced PC1/3 levels [5] and persistent proinsulin secretion [4], yet the mechanistic basis for these observations remains to be established. Beta cells with a senescence stress signature involving DNA damage and p21 activation accumulate during the development of type 1 diabetes [10, 11]. However, it is not known how this stress response affects insulin production. Human beta cell function declines during ageing [35], yet naturally occurring p16^INK4a-expressing human beta cells and p16^INK4A-overexpressing beta cells have improved identity, functional markers, insulin content and insulin secretion [28, 29], suggesting that p16^INK4A-mediated senescence during ageing per se is not a major driver of islet functional decline. Although the senescence stress response found in beta cells during type 1 diabetes shares some features in common with that in naturally senescent p16⁺ beta cells, there are also distinct phenotypes in the former not found in the latter, including markers of DNA damage response, high p21 levels, SASP and pro-survival signalling [14, 29]. Therefore, the effect of the DNA damage-mediated senescence response on human beta cell insulin production remains to be addressed.

Here, we showed that insulin production in human islet beta cells is sustained during chemically induced DNA damage-mediated senescence involving p21 as well as in endogenous senescent beta cells detected by either SA-βgal or p21. Sustained levels of proinsulin-processing enzymes including PC1/3 during the DNA damage response differs from the response to ER stress during IFN signalling [9] and thus our study identified differences between the impact of ER stress and DNA damage on insulin production in human islets. Using ten different human donor islet preparations with five from each sex (ESM Table 1), we found some differences in the effect of DNA damage-mediated senescence on PC1/3 levels in islets from male donors. We employed several complementary approaches to confirm the findings of our western blot analyses on islets. First, flow cytometry on DNA damage-mediated senescent human islets showed that insulin expression levels were comparable with those of control islets in the insulin⁺ (beta) cell population. Second, we used FACS sorting of endogenous SA-βgal⁺ senescent beta cells to show that this subpopulation of naturally occurring senescent cells also have comparable insulin content when compared with SA-βgal⁻ non-senescent beta cells. Third, examination of proteomics datasets from Humanislets.com showed no associations between CDKN1A or CDKN2A and islet insulin content per IEQ, and HPAP datasets revealed no correlative relationships between CDKN1A and proinsulin-processing genes in human beta cells from control donors, autoantibody-positive donors and donors with type 1 diabetes. Fourth, we showed that although proinsulin staining was reduced in a type 1 diabetes donor vs a no-diabetes control donor, as expected, endogenous p21⁺ senescent beta cells had similar proinsulin levels to p21⁻ cells in both contexts. Taken together, these findings support the conclusion that insulin production is sustained in human islet beta cells in a chemically induced DNA damage-mediated senescence model and in endogenous senescent human beta cells.

We used the EndoC-βH5 beta cell model system to address the question of how DNA damage affects insulin production in a clonal human beta cell population already in a growth-arrested state with high p16^INK4A and SA-βgal levels. DNA damage to EndoC-βH5 cells and concomitant activation of p21 and SASP coincided with reduced CPE and proSAAS as well as reduced insulin content. Unlike DNA damaged human islets containing p21⁺ beta cells, which did not show insulin secretory defects in static assays [12], insulin secretory defects occurred 2 days following DNA damage and p21 induction in EndoC-βH5 cells, but decreased insulin content was not observed at that timepoint [12]. Our results here build on this prior work by showing that the response to DNA damage involving p21 upregulation and SASP leads to reduced insulin production in these cells. The mechanism(s) explaining why EndoC-βH5 cells show a different effect on insulin production during senescence as compared with adult islets remains to be determined but may stem from differences in cyclin-dependent kinase inhibitor gene expression profiles in these cells. RNA-seq datasets revealed that EndoC-βH5 cells had higher basal CDKN2A, CDKN2B and CDKN2C expression whereas FACS-purified beta cells showed higher basal CDKN1A expression. In contrast, NHEJ pathway genes were more highly expressed in EndoC-βH5 cells than in islet beta cells, suggesting that baseline differences in DNA repair are not likely to explain the impaired insulin production phenotype in DNA damaged EndoC-βH5 cells. As CDKN2A has been linked with beta cell ageing and maturation state [28, 29], it was unexpected to find that CDKN2A expression was higher in EndoC-βH5 cells than in adult islet beta cells. EndoC-βH5 cells have lower beta cell identity gene expression and are fetus-derived [15], suggesting they are more immature than adult islet beta cells and would therefore be expected to express less CDKN2A compared with adult islet beta cells. However, heterogeneity in CDKN2A and p16^INK4A protein expression at the single beta cell-level in human islets [28, 29] is likely to be lost in the clonal EndoC-βH5 cells, and could also explain this difference. Additional studies will be required to delineate the contributions of CDKN2A vs CDKN1A to the effects of DNA damage on insulin production in these cells.

Our study had important limitations to consider. First, we observed extensive heterogeneity in the responses to DNA damage-mediated senescence in islets, both at the donor level and islet sample level, with some donor preparations and identically treated islet samples from the same donor showing different responses in the levels of proteins measured. Larger sample size will be required to better elucidate the inter-donor heterogeneity as well as intra-donor islet sample heterogeneity in response to DNA damage, and stratify the donor islet preparations that exhibited reductions in processing factors, proinsulin and/or insulin content during DNA damage-mediated senescence from those that showed increased levels. Second, we were limited in the age of the donors used for islet preparations available for study. As with most human islet studies, all our islet preparations were from middle-aged donors (40–50 years, and one donor aged 28 years) whereas islet samples from paediatric (infant to 17 years) or younger adult donors (ages 18–30 years) would be more representative of the age and maturation stage of islets in the context of type 1 diabetes in children and young adults. It is possible that DNA damage-mediated senescence may have a different impact on insulin production in islet beta cells in younger-aged donor islets, as compared with the middle-aged/older donor islets studied here. Third, although we quantified relative differences in proinsulin, insulin and processing factors, these assays could not determine whether there were differences in the major forms of proinsulin or enzymatic activities of PC1/3, PC2 and CPE during senescence. Similarly, we could not determine whether the naturally occurring heterogeneity in PC1/3 levels among beta cells [5, 36] is altered by senescence. Finally, we recognise that although our chemically induced DNA damage model of senescence mirrors some of the senescence phenotypes in donors with type 1 diabetes [11, 20], it is nevertheless an incomplete and artificial reflection of senescent human beta cell populations in vivo, which exhibit heterogeneity and are still being elucidated [13]. Future studies will be required to explore how senescence affects other aspects of insulin synthesis and regulation, including nutrient-stimulated secretion or presentation of hybrid insulin peptides or neoantigens [37, 38], which occur during other beta cell stress responses [39, 40].

In conclusion, our findings suggest that insulin production is sustained during the DNA damage response and in p21-expressing senescent beta cells in adult islets, which differs from the outcomes of chronic ER stress in islets. Senescent beta cells may be a previously unappreciated source of residual insulin production in type 1 diabetes.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CDKN1A

Cyclin-dependent kinase inhibitor 1A

CDKN2A

Cyclin-dependent kinase inhibitor 2A

CPE

Carboxypeptidase E

ENTPD3

Ectonucleoside triphosphate diphosphohydrolase 3

ER

Endoplasmic reticulum

GDF15

Growth and differentiation factor 15

HPAP

Human Pancreas Analysis Program

IEQ

Islet equivalent

MFI

Median fluorescence intensity

NHEJ

Non-homologous end joining

nPOD

Network for Pancreas Organ Donors with Diabetes

PC1/3

Prohormone convertase 1/3

PC2

Prohormone convertase 2

PI/I

Proinsulin:insulin ratio

proSAAS

Proprotein convertase subtilisin/kexin type 1 inhibitor

RIPA

Radioimmunoprecipitation assay

SA-βgal

Senescence-associated β-galactosidase

SASP

Senescence-associated secretory phenotype

UPR

Unfolded protein response

Acknowledgements

We thank the organ donors and their families. We thank the Network for Pancreatic Organ Donors (nPOD) (University of Florida, USA) for fixed pancreatic tissue sample procurement and distribution. We also thank the Alberta Diabetes Institute IsletCore for providing human islets for research with the assistance of the Human Organ Procurement and Exchange (HOPE) program, Trillium Gift of Life Network (TGLN) and other Canadian organ procurement organisations. The authors thank the NIDDK-funded Integrated Islet Distribution Program (IIDP) (RRID:SCR _014387) at City of Hope, NIH grant no. U24DK098085 and the Breakthrough T1D (formerly JDRF)-funded IIDP Islet Award Initiative for providing human islets for research for this study. This manuscript used data acquired from the database (https://hpap.pmacs.upenn.edu/) of the Human Pancreas Analysis Program (HPAP; RRID:SCR_016202; PMID: 31127054; PMID: 36206763). HPAP is part of a Human Islet Research Network (RRID:SCR_014393) consortium (UC4-DK112217, U01-DK123594, UC4-DK112232 and U01-DK123716). This work also includes data and/or analyses from HumanIslets.com funded by the Canadian Institutes of Health Research, Breakthrough T1D Canada, and Diabetes Canada (5-SRA-2021-1149-S-B/TG 179092) with data from islets isolated by the Alberta Diabetes Institute IsletCore with the support of the Human Organ Procurement and Exchange program, Trillium Gift of Life Network, BC Transplant, Quebec Transplant and other Canadian organ procurement organisations with written informed donor consent as approved by the Human Research Ethics Board at the University of Alberta (Pro00013094). The graphical abstract was created in BioRender.com.

Data availability

All data related to the manuscript are included in the text or in the ESM.

Funding

CP was supported by a Canadian Islet Research Network NSERC-CREATE MSc studentship award. JP was supported by a PhD studentship in Health Research from Research Manitoba. NRM was supported by a Canadian Islet Research Network-Breakthrough T1D postdoctoral fellowship award. CBV was supported by grants from the Canadian Institutes of Health Research and Breakthrough T1D (TDP-186359; 3-COE-2022-1103-M-B). Y-CC was supported by an Advanced Post-doctoral Fellowship from Breakthrough T1D (3-APF-2022–1141-A-N) and a Research Excellence, Diversity, and Independence (REDI) Early Career Transition Award from the Canadian Institutes of Health Research and Breakthrough T1D (197485). The laboratory of PJT was supported by a Diabetes Canada End Diabetes 2022 grant (OG-3-22-5694-PT), grants from the Canadian Institutes of Health Research (PJT-479641 and PJT-485915) and Breakthrough T1D Canada (4-SRA-2023-1184-S-N), the 2023 Allen Rouse Basic Science Career Award from the Manitoba Medical Services Foundation and the Canada Foundation for Innovation John Evan Leaders Fund award (43168).

Authors’ relationships and activities

CBV receives a salary award from the BC Children’s Hospital Research Institute. The authors declare that there are no other relationships or activities that might bias, or be perceived to bias, their work.

Contribution statement

CP, JP, NRM and PJT designed the experiments. CP, JP, NRM and YCC performed the experiments with input and guidance from PJT. CBV, PJT and CP wrote the manuscript with editing by JP, NRM, YCC and CBV. All authors approved the final manuscript for submission. PJT is the guarantor of this work.