β-Cell–Specific E2f1 Deficiency Impairs Glucose Homeostasis, β-Cell Identity, and Insulin Secretion

Type 2 diabetes (T2D) is a progressive metabolic disorder characterized by chronic hyperglycemia resulting from inadequate pancreatic β-cell response to peripheral insulin resistance. Usually, β-cells respond to obesity and aging-associated loss of insulin sensitivity by increasing insulin secretion to compensate for rising glycemia. However, when insulin secretion and action are imbalanced, the emerging chronic hyperglycemic state progressively leads to massive β-cell dysfunction and decreased mass (1). Importantly, single-cell transcriptomic analysis of human pancreatic islet cells from patients with T2D and healthy individuals has revealed the complex identity of endocrine cells associated with the pathophysiology (2–4). In addition, recent studies have demonstrated a loss of transcriptional maturity (5) and cell-type–specific regulatory profiles underlying T2D pathogenesis (6). Moreover, histology studies of human diabetic pancreata have found a significant increase in bihormonal insulin-positive/glucagon-positive cells, suggesting an altered β-cell identity (7,8). Accordingly, recent studies using β-cell lineage–tracing murine models have demonstrated that islet cells can transdifferentiate directly toward another islet cell fate and/or dedifferentiate toward a progenitor-like cell (9). In this context, both genetic and epigenetic mechanisms are important to maintain adult β-cell fate and function in mice and humans (10–13). However, the molecular actors controlling β-cell mass, identity maintenance, and cellular plasticity still need to be identified.

Gene transcription and chromatin states are tightly regulated to ensure the appropriate transcriptome for a specific cell type. Many transcription factors have been identified as key regulators of β-cell identity and function, including Pdx1 (14), Pax6 (15), Nkx6.1 (16), and Nkx2.2 (17). Interestingly, pleiotropic transcription factors are also involved in the control of β-cell functions and glucose homeostasis, suggesting their important roles in activating or repressing gene transcription. Among those, members of the E2F transcription factor (E2F1–E2F8) family play critical roles in cell survival and proliferation by regulating the expression of genes involved in cell-cycle progression (18). The transcriptional activity of E2F1, the founder member of the family, is regulated by several protein complexes, including the retinoblastoma tumor suppressor family (pRB, p107, p130), cyclin-dependent kinases (e.g., CDK4) and their regulatory partner cyclins (Ccn), and the family of CDK inhibitors (e.g., p16^Ink4A encoded by the Cdkn2a locus). Interestingly, the role of the cell-cycle machinery goes beyond the unique regulation of cell proliferation. Indeed, modulation of the expression levels of these cell-cycle regulators revealed an important role for these proteins in glucose homeostasis (19) and diabetes development through the control of β-cell mass and function (20,21). We and others recently identified molecular cross talk between E2f1/pRb and glucagon-like peptide-1 (Glp1) pathways (22,23). Although E2F1 gene expression is decreased in human T2D islets (24), the causal effect of E2F1 deficiency on impaired β-cell mass and function and T2D development is not elucidated. In particular, the cellular and molecular mechanisms underlying the contribution of E2F1 as a transcription factor to β-cell identity and/or plasticity in mice and humans remain unknown. We recently demonstrated that the germ line deletion of E2f1 in obese and diabetic db/db mouse models, despite lowering liver steatosis, does not protect against diabetes or obesity (25). Interestingly, we observed decreased plasma insulin levels and increased plasma glucose and glucose intolerance in db/db::E2f1^−/− mice compared with db/db::E2f1^+/+ controls (25). These metabolic alterations in a diabetic background raise the possibility that E2f1 may contribute to islet morphology and cell identity and function in a cell-autonomous manner. To test this hypothesis, we generated mice lacking E2f1 in β-cells and found that E2f1 is mandatory to maintain β-cell identity gene expression in both mouse models and human islets. By combining cellular and mouse models with pharmacological approaches, we identified E2f1 as a critical transcription factor necessary to maintain proper β-cell gene expression and function while repressing non–β-cell transcriptional programs.

A detailed description of the following procedures can be found in the Supplementary Material section.

In vivo experiments were performed in compliance with the French ethical guidelines for studies on experimental animals (animal house agreement no. A 59-35015; authorization for animal experimentation no. 59-350294; project approval by our local ethics committee [Lille, France; no. CEEA-482012 and no. APAFIS-2915-201511300923025v4]). All experiments were performed with male mice. E2f1^−/− (stock no. 002785; Jackson Laboratory), CMV-CDK4^R24C (26), Cdkn2a^−/− (27), db/db::E2f1^−/− (25), E2f1 floxed (E2f1^fl/fl) (Taconic Biosciences, Germantown, NY), and RIP-Cre^/+ mice were previously described (25,28). MIP-CreERT mice were obtained from the Jackson Laboratory [stock no. 024709 (29)]. For lineage-tracing experiments, Rosa26R-td-tomato (stock no. 007914; Jackson Laboratory) and Rosa26R-ECFP [Gt(ROSA)26Sortm2(ECFP)Cos (30)] mice were used. Intraperitoneal glucose (IPGTT) and insulin tolerance tests were performed in 6-, 8-, and 12-week-old male mice (E2f1^fl/fl and E2f1^β−/−) or 5-month-old (E2f1^fl/fl and E2f1^{MIP-CreERT β−/−}) littermates as previously described (31) after 16- (IPGTT) or 5-h fasting (insulin tolerance test). Glycemia was measured using Accu-Chek Performa (Roche Diagnostics). Circulating insulin levels were measured using a mouse insulin ELISA kit (Mercodia).

Immunofluorescence and immunohistochemistry were performed exactly as described previously (31). Briefly, pancreatic tissues were fixed in 10% formalin, embedded in paraffin, and sectioned at 5 μm. For immunofluorescence microscopy analyses, after antigen retrieval using citrate buffer, 5-μm formalin-fixed paraffin-embedded pancreatic sections were incubated with primary antibodies as indicated. Immunofluorescence staining was revealed by using a fluorescein-isothiocyanate–conjugated anti-rabbit antibody (for cleaved caspase-3, Alexa-conjugated anti-mouse [for glucagon], anti-rabbit [for Glut-2], or anti–guinea pig [for insulin] secondary antibodies were used to identify specifically primary antibodies). Nuclei were stained with DAPI (cat. no. D9542; Sigma-Aldrich). For morphometric analysis, three to ten animals from each genotype were analyzed, and images were processed and quantified using ImageJ software by an observer blinded to experimental groups. Morphometric analysis, including islet density and β-cell fraction and mass, was performed as previously described (32). β-Cell size was measured using immunofluorescence staining of pancreatic sections using an anti–Glut-2 antibody (cat. no. 07-1402-1; Merck) from five mice from each genotype, and acquisition was performed on a spinning disk confocal microscope (Zeiss). Twenty islets per mouse were analyzed, and five mice from each genotype were used to analyze β-cell size. Quantification was then performed using a custom-made script with ImageJ software (available upon request). β-Cell proliferation was measured using BrdU staining on formalin-fixed paraffin-embedded pancreatic sections. Apoptosis was quantified with immunofluorescence staining using anticleaved caspase 3.

For mouse islet studies, pancreata were digested by type V collagenase (cat. no. C9263; Sigma-Aldrich) (1.5 mg/mL) for 10 min at 37°C as described previously (31). Briefly, after digestion and separation in a density gradient medium, islets were purified by handpicking under a macroscope and cultured for 16 h before subsequent analysis. For expression studies, mouse isolated islets were snap frozen in liquid nitrogen before RNA extraction. Human pancreatic tissues were harvested from brain-dead adult human donors without diabetes (Supplementary Table 1 lists donor information). Isolation and islet culture were performed as described (33). Human islets were treated for 48 h with DMSO 0.1% or HLM006474 (34,35) at 10 μmol/L. Data are expressed as a ratio of total insulin content. For mRNA and protein quantification, human islets were isolated as described above and snap frozen for further processing. Min6 cells (AddexBio) were cultured in DMEM (Gibco) with 15% fetal bovine serum, 100 mg/mL penicillin-streptomycin, and 55 mmol/L β-mercaptoethanol. Min6 cells were treated with HLM006474 (10 μmol/L) or DMSO 0.1% (35) for 48 h before glucose-stimulated insulin secretion (GSIS) assay or RNA extraction.

Total RNA was extracted from cells and tissues using TRIzol reagent (Life Technologies) as described previously (22). mRNA expression was measured after reverse transcription by quantitative RT-PCR (qRT-PCR) with FastStart SYBR Green master mix (Roche) using a LightCycler Nano or LC480 instrument (Roche). qRT-PCR results were normalized to endogenous cyclophilin reference mRNA levels. Results are expressed as the relative mRNA level of a specific gene expression using the formula 2^−ΔCt. The complete list of primers is presented in Supplementary Table 2. For RNA sequencing (RNA-seq), total RNA was extracted from Min6 cells or pancreatic islets using the RNeasy Plus Microkit (Qiagen) following manufacturer instructions. Using P < 0.05 adjusted for multiple comparisons as the threshold for differential gene expression analyses, we then performed pathway analysis using Ingenuity Pathway Analysis (IPA) (Ingenuity Systems; Qiagen), Metascape (36), and Gene Set Enrichment Analysis (https://software.broadinstitute.org/gsea/).

Chromatin immunoprecipitation (ChIP) experiments were performed on formaldehyde-fixed Min6 cells. Briefly, 20 × 10⁶ Min6 cells or 100 × 10⁶ Min6 cells (for E2f1 ChIP sequencing [ChIP-seq]) were treated with formaldehyde at a final concentration of 1% to crosslink DNA and protein complexes for 10 min. The reaction was stopped by the addition of glycine (0.125 mol/L) over 5 min. Cells were lysed and DNA-protein complexes were sheared using the Bioruptor Pico (reference no. B01060010; Diagenode) for 8 min. The sheared chromatin was immunoprecipitated with either the nonspecific antibody immunoglobulin G (cat. no. sc2025; Santa Cruz Biotechnology), trimethylation of lysine 4 in histone H3 (H3K4me3) (cat. no. 61379; Active Motif), H3K27me3 (cat. no. 61017; Active Motif), acetylation of lysine 27 in histone H3 (H3K27ac) (cat. no. 39685; Active Motif), or anti-flag (Clone M2; Sigma-Aldrich). Additional bioinformatic analyses of the public data sets and those of this study were performed using the open web-based platform Galaxy Europe (https://usegalaxy.eu). A list of ChIP-seq and RNA-seq data sets used in this study can be found in Supplementary Table 3. A detailed description of RNA-seq and ChIP-seq procedures can be found in the Supplementary Material.

Data are presented as mean ± SEM. Data are derived from multiple experiments unless stated otherwise. Statistical analysis was performed using a two-tailed unpaired t test or one- or two-way ANOVA with the Tukey post hoc test comparing all groups with one another using GraphPad Prism 9.0 software. Differences were considered statistically significant at P < 0.05 (*P < 0.05, **P < 0.01, and ***P < 0.001).

The data sets and resources generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

To determine whether E2f1 regulates insulin-producing β-cell fate and function in a cell-autonomous manner, we generated β-cell–specific E2f1-deficient mice by crossing E2f1^fl/fl with RIP-Cre (Rip-Cre^/+) mice (28). qRT-PCR showed a 91% tissue-specific reduction in E2f1 expression in pancreatic islets isolated from E2f1^β−/− mice (Supplementary Fig. 1A). Twelve-week-old E2f1^β−/− mice displayed normal body weight (Supplementary Fig. 1B) and fasting glycemia when fed a chow diet (Supplementary Fig. 1C) compared with age-matched control E2f1^fl/fl and Rip-Cre (Rip-Cre^/+) male littermates. No significant difference was observed between E2f1^β−/− mice and E2f1^fl/fl control littermates regarding body weight, fasting glycemia, glucose tolerance, or insulinemia in 6- (Supplementary Fig. 2A–E) and 8-week-old animals (Supplementary Fig. 2F–J). However, when challenged with a bolus of glucose, 12-week-old E2f1^β−/− mice exhibited glucose intolerance (Fig. 1A and B), primarily because of decreased insulin secretion in response to in vivo glucose injection (Fig. 1C) rather than defective insulin sensitivity (Supplementary Fig. 2K).

We next sought to evaluate the metabolic consequences of modulating E2f1 expression in adult β-cells. To this end, we generated β-cell–specific E2f1-deficient mice (E2f1^{MIP-CreERT β−/−}) by crossing E2f1^fl/fl with the tamoxifen-inducible model MIP-CreERT (MIP-CreERT^/+) mice (29). We first confirmed efficient recombination upon tamoxifen treatment in pancreatic islets isolated from E2f1^fl/fl and E2f1^{MIP-CreERT β−/−} mice (Supplementary Fig. 2L). As in E2f1^β−/− mice, the genetic deletion of E2f1 in adult β-cells revealed glucose intolerance 1 month after tamoxifen administration (Supplementary Fig. 2M and N), a phenotype that was aggravated 3 months after tamoxifen gavage (Fig. 1D and E). Interestingly, insulinemia during IPGTT was also impaired in E2f1^{MIP-CreERT β−/−} mice 20 min after glucose injection (Fig. 1F). Altogether, these data suggest that the β-cell–specific E2f1 deficiency, triggered either early during development or in fully mature adult β-cells, impairs glucose tolerance and insulin secretion in response to glucose.

Because E2f1^β−/− mice displayed impaired insulin secretion in response to glucose, we investigated the contribution of E2f1 to islet morphology. We observed that islet density (Fig. 2A), islet size (Fig. 2B), β-cell fraction (Fig. 2C), β-cell mass (Fig. 2D), and β-cell size (Fig. 2E and Supplementary Fig. 3A) were comparable between pancreata from 3-month-old E2f1^β−/− and E2f1^fl/fl mice. We then analyzed proliferation and apoptosis in E2f1^β−/− and E2f1^fl/fl pancreata. Apoptotic cells were not detected in pancreata from control and β-cell–specific E2f1-deficient mice, as shown by cleaved caspase 3 immunofluorescence staining (Supplementary Fig. 3B and C). BrdU staining of E2f1^β−/− and E2f1^fl/fl pancreatic sections revealed that the number of proliferative cells was decreased in E2f1^β−/− islets (Fig. 2F and Supplementary Fig. 3D). Remarkably, an immunofluorescence analysis of sections from control and E2f1^β−/− pancreata demonstrated a significant increase in glucagon-positive cells (Fig. 2G and H), whereas the number of somatostatin-positive cells was not modified in E2f1^β−/− islets (Fig. 2I and J). These results suggest that loss of E2f1 expression could induce α-cell hyperplasia associated with lower endocrine cell proliferation.

To confirm the role of E2f1 in maintaining the physiological ratio of α- to β-cells, we performed immunofluorescence staining of insulin and glucagon in the pancreas of wild-type and global E2f1 knockout mice (E2f1^−/−). The detailed analysis of E2f1^−/− pancreata revealed a decreased proportion of insulin-positive β-cells and a concomitant expansion of the glucagon-positive α-cell percentage per islet in chow-fed 16-week-old animals, as demonstrated by the quantification of insulin- and glucagon-positive cells in E2f1^+/+ and E2f1^−/− pancreata (Supplementary Fig. 4A–C). To test whether bona fide regulators of E2F1 activity could rescue the altered ratio of α- to β-cells observed in E2f1^−/− mice, we used two different genetically engineered mouse models (Supplementary Fig. 4D). We first used the R24C mouse model of CDK4 hyperactivation (CMV-CDK4^R24C) that constitutively expresses a CDK4-mutated protein that restores β-cell mass and function during diabetes development (37,38). As previously observed (38), the ratio of α- to β-cells was decreased in CMV-Cdk4^R24C pancreata (Supplementary Fig. 4C and E). We then generated compound-mutated mouse models with both E2f1 deficiency and overactive CDK4 (E2f1^−/−::CMV-Cdk4^R24C). Interestingly, the decreased ratio of α- to β-cells observed in CMV-Cdk4^R24C pancreata was blunted in E2f1^−/−::CMV-Cdk4^R24C mice, with a concomitant increase in glucagon immunofluorescence staining (Supplementary Fig. 4C and F). To sustain such findings, we then replicated these observations in an alternative model using mice deficient in Cdkn2a [Cdkn2a^−/− (27)], an upstream regulator of the E2F1-CDK4-pRb signaling pathway involved in β-cell function (39–41). Although glucagon- and insulin-positive immunofluorescence cell staining was similar between Cdkn2a^+/+ and Cdkn2a^−/− pancreata (Supplementary Fig. 4C and G), E2f1^−/−::Cdkn2a^−/− compound-mutated mice displayed an altered ratio of α- to β-cells, with a decrease in insulin-positive and increase glucagon-positive cell numbers (Supplementary Fig. 4C and H). In addition, our recent observation of decreased plasma insulin levels and increased plasma glucose and glucose intolerance in db/db::E2f1^−/− mice compared with in db/db::E2f1^+/+ controls (25) prompted us to evaluate islet morphology in these diabetic mouse models. Accordingly, in the db/db background, E2f1-deficient mice displayed a further increase in α-cell number per islet, while exhibiting a lower β-cell count compared with db/db::E2f1^+/+ controls (Supplementary Fig. 5A and B). Altogether, these results suggest a specific role of E2f1 in maintaining pancreatic β-cell numbers under normal conditions but also in a diabetic environment associated with glucose intolerance, insulin resistance, and obesity. We may then hypothesize that E2f1 expression in β-cells may be necessary to maintain insulin secretion and glucose homeostasis associated with a normal ratio of α- to β-cells.

To evaluate the fate of E2f1-deficient β-cells and the potential origin of the increased number of α-cells in E2f1^β−/− mice, we performed lineage-tracing experiments. To track the destiny of β-cells, the expression of the tomato tracer was monitored in Rip-Cre^/+::tdTomato and E2f1^β−/−::tdTomato mice. As expected, tomato-positive cells were colocalized with insulin-positive cells in controls and E2f1^β−/−::tdTomato pancreata (Fig. 3A). Interestingly, we could not detect any colocalization of tomato- or glucagon-positive cells in controls or in β-cell–specific E2f1-deficient pancreata (Fig. 3B). β-Cell lineage-tracing results were further confirmed in E2f1^β−/− mice by using the ECFP reporter mouse strain model (30) (Fig. 3C and D). These data suggest that, although the number of α-cells is increased upon E2f1 deficiency, loss of E2f1 expression in β-cells does not trigger their conversion to α-cells.

E2F1 is a transcription factor that controls gene expression in several cellular systems. In order to gain a global view of the transcriptional mechanisms associated with the loss of E2f1 expression in β-cells, we performed RNA-seq in control (E2f1^fl/fl) and E2f1^β−/− isolated islets. As expected, the floxed region of the E2f1 gene spanning exons 2 and 3 was not covered in E2f1^β−/− isolated islets compared with E2f1^fl/fl isolated islets, indicating specific and efficient gene deletion through the Cre recombinase activity in E2f1^β−/− isolated islets (Supplementary Fig. 6A). Transcriptome analysis revealed that 692 annotated genes were differentially expressed across the two groups (adjusted P < 0.05). Interestingly, a majority of genes were upregulated in E2f1^β−/− isolated islets (493 genes) (Fig. 4A and Supplementary Table 4), with only 199 downregulated genes associated with loss of E2f1 expression in β-cells (Fig. 4A and Supplementary Table 4). This first observation suggests that E2f1 expression is necessary not only to activate but also to repress gene transcription within β-cells. Analyses of the RNA-seq data of downregulated genes with IPA software revealed an enrichment of gene networks involved in insulin secretion pathway or maturity-onset diabetes of the young signaling (Fig. 4B). Conversely, upregulated genes in E2f1^β−/− isolated islets were mostly associated with inflammatory response, including the hepatic fibrosis signaling pathway, iNOS signaling, and toll-like receptor signaling (Fig. 4C). RNA-seq data were then analyzed using the upstream regulator analysis function of IPA to identify potential contributors that could be associated with the transcriptional reprogramming observed in E2f1^β−/− isolated islets. Among the most significant upstream regulators of the upregulated genes, tumor necrosis factor, interferon-γ, and Myd88 were identified (Supplementary Table 5). Conversely, analysis of the downregulated genes suggested that Xbp1, Hnf1a, Neurod1, and Pdx1, transcription factors previously shown to control β-cell identity or function, were found to be potential upstream regulators that were predicted to be inhibited (Supplementary Table 5). To better understand the relationship between the observed metabolic phenotype (i.e., glucose intolerance and defective insulin secretion), increased glucagon-positive cells, and E2f1 deficiency, we filtered gene sets to focus on β- and α-cell–specific genes being conserved between zebrafish, mice, and humans (42). Notably, a total of 15 genes from 109 conserved genes (82 and 27 for α- and β-cell genes, respectively) were differentially expressed. Interestingly, most of the conserved β-cell markers were found to be decreased in E2f1^β−/− isolated islets (Fig. 4D and Supplementary Fig. 6B), whereas most of the conserved α-cell genes were upregulated upon E2f1 deficiency (Fig. 4D and Supplementary Fig. 6C). Gene Set Enrichment Analysis further confirmed an enriched signature of increased expression of α-cell markers (Fig. 4E) and, conversely, a decreased expression of β-cell markers in E2f1^β−/− isolated islets (Fig. 4F). We further confirmed these changes in the expression of selected α- and β-cell markers at the transcriptional level by qPCR in an independent experiment; accordingly, we observed a strong decrease in transcript levels of Pdx1, Ins2, Pcsk9, Foxo1, and Glp1r in E2f1^β−/− islets (Supplementary Fig. 6D). Conversely, α-cell–specific Aristaless-related homeobox (Arx) mRNA levels were increased in E2f1^β−/− islets (Supplementary Fig. 6E). Altogether, these results suggest that the β-cell–specific deletion of E2f1 induces a transcriptional reprogramming characterized by a loss of β-cell identity genes associated with increased expression of non–β-cell markers.

To assess the effect of E2F1 activity on β-cell identity markers in Min6 cells and human islets, we made use of the E2F paninhibitor HLM006474 previously shown to inhibit the binding of E2Fs to their DNA target genes and E2F1 transcriptional activity (35). Consistent with our previous findings in HEK293 cells (35), 48-h treatment of Min6 cells with this inhibitor triggered a decrease in E2f1 transcriptional activity, as measured by transient transfection experiments using an E2F reporter gene (Fig. 5A). In addition, treatment of Min6 cells with this inhibitor induced a marked decrease in GSIS (Fig. 5B) and in the expression of several β-cell markers, including Ins1, Pdx1, and Pax4 (Fig. 5C). Treatment of human islets with the E2F inhibitor for 48 h also decreased GSIS (Fig. 5D) and β-cell marker expression levels, with a concomitant increase in the expression of α-cell genes (Fig. 5E). Therefore, the pharmacological inhibition of E2F activity impairs β-cell function and gene expression in both mouse cell lines and human islets, suggesting that E2F1 activity is also required in human islets to maintain proper insulin secretion and β-cell identity genes, as observed in mice.

Our transcriptome analysis revealed 2.5-fold greater up- than downregulation of global gene expression in E2f1^β−/− isolated islets. Comparative analysis of log10 transcripts per million levels of up- and downregulated gene values demonstrated that the expression of upregulated genes in E2f1^β−/− islets was significantly increased upon E2f1 deficiency compared with wild-type controls, whereas the expression of downregulated genes was not significantly modulated (Fig. 6A). Considering that E2f1 could play a dual role in the regulation of gene expression in pancreatic β-cells, we postulated that this mechanism could be related to a distinct epigenomic profile within promoter of genes that are up- and downregulated in E2f1^β−/− islets. Using a recently published chromatin-state segmentation model (12), we probed the active/repressive levels of these promoters by monitoring several epigenome marks, such as active and poised promoters (H3K4me3) and enhancers (H3K27ac). Intersecting publicly available data of ChIP followed by next-generation sequencing (ChIP-seq) from healthy C57Bl6/J mouse pancreatic islets [Genomic Spatial Event no. 110648 (12)] and our RNA-seq data, we grouped up- and downregulated genes according to their chromatin state (Fig. 6B and Supplementary Table 4). We observed that 61% of the upregulated genes in E2f1^β−/− pancreatic islets were associated with a silent chromatin state in healthy C57Bl6/J mouse pancreatic islets characterized by bivalent H3K4me3/H3K27me3 and polycomb-repressed (H3K27me3) marks. Conversely, 82% of the downregulated genes showed enrichment in an active chromatin state characterized by RNA-Pol2 recruitment and H3K4me3 and H3K27ac histone marks.

Although H3K4me3, H3K27ac, and H3K27me3 ChIP-seq data were available for mouse islets, we then performed ChIP-seq experiments in Min6 cells as a surrogate for β-cells. H3K4me3, H3K27ac, and H3K27me3 ChIP-seq signals were thus interrogated within promoters (centered to transcription start site [TSS] ± 1 kb) of up- and downregulated genes both in Min6 cells and mouse pancreatic islets (Fig. 6C and D and Supplementary Fig. 7A and B). H3K4me3 as well as H3K27ac signals were stronger within promoters of genes that are downregulated in E2f1^β−/− isolated islets compared with upregulated genes, both in Min6 cells and mouse pancreatic islets (Fig. 6C and Supplementary Fig. 7A and B). Conversely, H3K27me3 ChIP-seq signals were lower in the promoter regions of downregulated genes compared with upregulated genes (Fig. 6D and Supplementary Fig. 7A and B). Altogether, these results suggest that the transcriptome of E2f1^β−/− islets is associated with a specific epigenomic pattern that may be involved in the control of E2f1-mediated transcriptional activation or repression.

Having demonstrated that the transcriptome of E2f1^β−/− isolated islets was altered compared with controls, we next asked whether these up- and downregulated genes identified in isolated islets from E2f1^β−/− mice were direct or indirect E2f1 target genes. To reach this aim, we performed ChIP-seq experiments to determine chromatin-bound E2f1 target genes.

In a first series of experiments, ChIP-seq experiments targeting endogenous E2f1 were conducted in pancreatic islets but failed to profile the endogenous E2f1 cistrome because of a lack of antibody specificity (data not shown). Alternatively, E2f1 was overexpressed in Min6 cells as a surrogate for β-cells through transfection of a plasmid encoding E2f1 fused to flag tag (E2f1-flag). As expected, the E2f1-flag protein was overexpressed in Min6 cells transfected with a plasmid encoding E2f1-flag compared with those transfected with an empty plasmid (Supplementary Fig. 8A). These results prompted us to perform a ChIP-seq analysis using an anti-flag antibody to profile E2f1-flag cistrome at a genome-wide level in Min6 cells.

Among the E2f1-bound genomic regions, 94.5% were localized within the promoter (≤1 kb), whereas other genomic locations were marginally represented (Supplementary Fig. 8B), confirming that E2f1 is a transcription factor mainly bound within a GC-rich proximal promoter (43). Indeed, a DNA motif enrichment analysis was performed within E2f1-enriched promoter regions, showing that the E2f motif was the most significant enriched motif within these regions and suggesting that E2f1 was directly bound on the promoter (Supplementary Fig. 8C). Next, we consolidated our results in an independent experiment by performing a ChIP-qPCR analysis targeting two E2f1-flag–bound promoters (from Ezh2 and Ccne1 genes) detected by ChIP-seq (Supplementary Fig. 8D and E, respectively). As expected, ChIP-qPCR performed with an anti-flag antibody led to a significant enrichment of E2f1-flag within both Ezh2 (Supplementary Fig. 8F) and Ccne1 (Supplementary Fig. 8G) promoters compared with ChIP-qPCR conducted with an unspecific antibody (immunoglobulin G). In addition, no E2f1-flag enrichment was detected within an unspecific intergenic chromatin region (negative control) (Supplementary Fig. 8H), validating our E2f1 cistrome analysis.

After ChIP-qPCR–based validation of our anti-flag antibody approach in Min6 cells (Supplementary Fig. 8), the E2f1-flag ChIP-seq signal was interrogated within the promoter (centered to TSS ± 1 kb) of dysregulated genes identified in pancreatic islets isolated from E2f1^β−/− mice to identify E2f1-bound (i.e., direct E2f1 target genes) and E2f1-unbound (i.e., indirect E2f1 target genes) genes. Among the upregulated genes in E2f1^β−/− islets, we identified 197 genes harboring E2f1-bound proximal promoters and 297 genes with E2f1-unbound proximal promoters (Supplementary Table 6). Among the downregulated genes, 95 genes harbored an E2f1-bound proximal promoter, whereas 106 genes were identified as E2f1-unbound proximal promoters (Supplementary Table 6). This ChIP-seq and RNA-seq intersecting analysis was exemplified by a series of genes, including Nkx6–1 (Supplementary Fig. 9A) and Tcf7 (Supplementary Fig. 9B), displaying E2f1 enrichment in the close vicinity of TSS. Regarding E2f1-unbound promoters, we observed that Ins1 (Supplementary Fig. 9C) and Gzma (Supplementary Fig. 9D), although differentially regulated at the mRNA level, were not bound by E2f1 at the chromatin level. Altogether, these results suggest that E2f1 could have a dual role in the regulation of gene expression by acting either as a transcriptional activator or a repressor in pancreatic islets, through both indirect and direct mechanisms at the chromatin level.

To better appreciate the potential role of E2f1 in the control of gene expression, a canonical pathway analysis was performed using the different clusters of genes identified above. Our analysis revealed that E2f1-undound downregulated genes were mostly associated with the insulin secretion signaling pathway (Supplementary Fig. 9E–H). IPA could not reveal any enrichment of pathways involved in insulin secretion in E2f1-bound promoters. However, a detailed analysis of ChIP-seq data from down- and upregulated genes revealed that E2f1 was associated with the chromatin region of several genes involved in β-cell function, such as Glis3, Glp1r, Manf, Nkx6–1, Ppp1r1a, and Wfs1, or β-cell dysfunction, such as Bcl11a and Cxcl14 (Supplementary Fig. 9A and Supplementary Table 6).

Because E2f1-bound genes in Min6 cells are predicted to be potential E2f1 direct target genes, we next investigated whether there were differences in the epigenomic activation level of these promoters. Therefore, we conducted a chromatin hidden Markov model (ChromHMM) analysis, allowing characterization of chromatin states based on differential enrichment levels of histone marks as well as transcription factors at a genome-wide level (44). ChromHMM analysis was thus performed by combining ChIP-seq data obtained in Min6 cells from both E2f1-flag and a selection of histone marks related to the active/repressive states of promoters (i.e., H3K4me3, H3K9ac, H3K27ac, and H3K27me3) and enhancers (i.e., H3K4me1, H3K27ac and H3K27me3). By applying a 16-state ChromHMM analysis, we demonstrated that only three states were enriched with E2f1-flag (states 6, 14, and 15) (Fig. 7A). Whereas states 14 and 15 displayed the typical features of an active promoter (i.e., H3K4me3, H3K9ac, and/or H3K27ac), the chromatin state related to state 6 was more associated with a repressed promoter (H3K4me3 and H3K27me3). The TSS-centered genomic distribution analysis of these 16 chromatin states confirmed that states 6, 14, and 15 corresponded to a proximal promoter of genes (Supplementary Fig. 10A).

To go further in the characterization of the role of E2f1 in β-cells, we next decided to analyze the differential transcriptomic data from E2f1^β−/− isolated islets in light of this ChromHMM analysis by determining whether the promoter of deregulated genes belonged to one of the chromatin states described above. This analysis showed that 23.4% of the upregulated genes (109 of 465 genes) harbored a state 6 promoter, indicating that the expression of these genes could be repressed by E2f1 in β-cells in vivo (Supplementary Table 7). Conversely, promoters of downregulated genes were more associated with states 14 (46.5%; 87 of 187 genes) (Supplementary Table 7) and 15 (41.1%; 77 of 187 genes) (Supplementary Table 7), indicating that E2f1 could play a positive role in the regulation of the expression of these genes in β-cells. Although pathway analysis of the different chromatin states of the direct down- and upregulated genes could not precisely demonstrate a specific enrichment of pathways involved in insulin secretion or β-cell function (Supplementary Fig. 10B–G), we could observe that most of the non–β-cell markers, such as Cxcl12 (Fig. 7B) and the β-cell–disallowed genes Gab1 (Fig. 7C), Mgll, Oat (Supplementary Table 7), Slc25a33 (Fig. 7D), and Yap1 (Fig. 7E), were mostly associated with segment 6 and were upregulated in E2f1^β−/− isolated islets (Supplementary Table 7). Interestingly, genes related to β-cell function, such as Glis3 (Fig. 7F), Glp1r (Fig. 7G), Manf (Fig. 7H), Nkx6–1 (Supplementary Table 7), P2ry1 (Fig. 7I), and Wfs1 (Supplementary Table 7), were downregulated and associated with state 14 or 15 (Supplementary Table 7). Altogether, these results support a potential dual active and repressive role for E2f1 in transcriptional regulation within β-cells. Importantly, our data suggest that E2f1 modulates β-cell and non–β-cell transcriptional programs in pancreatic islets through direct epigenomic based–mechanisms as well as indirect processes, probably through the upstream regulation of β-cell–enriched transcriptional regulators, such as Pdx-1 or Hnf1a.

In the current study, we show that the transcription factor E2F1 plays an essential role in maintaining β-cell function and identity through the control of transcriptomic and epigenetic programs within the pancreatic islets. Our study is the first to demonstrate a direct cell-autonomous contribution of E2f1 in controlling not only in vivo insulin secretion but also β-cell identity maintenance, without affecting pancreatic islet cell number or β-cell mass. The role of E2F1 in the control of cell cycle and proliferation has been extensively studied, particularly in the context of cancer (18). However, its function in nonproliferating fully differentiated cells, including β-cells, remains to be precisely deciphered. The observation that cell-cycle regulators, including E2F1, are expressed in cells that are not proliferating suggests that they may be involved in adaptive pathways that are independent of cellular proliferation. We and others have demonstrated that this pathway plays a key role in postnatal β-cell proliferation (45–47), glucose homeostasis, and insulin secretion (20,48,49). Studies have also revealed that the Cdk4-E2F1-pRb pathway controls the fate of pancreatic progenitors through the transcriptional control of the expression of Ngn3 and Pdx1 as well as Pdx1 protein stability (50,51). Most of these studies were performed using germ line E2f1-deficient mice, which precludes ascertainment of a cell-autonomous role of E2f1 in the endocrine pancreas and β-cell functions. To better appreciate the specific role of E2F1 in β-cells, we specifically knocked down E2f1 in β-cells using Cre/loxP technology. Although we cannot rule out an early role for E2f1 in pancreatic progenitors because of our use of RIP-Cre mice (28), our data demonstrate that E2f1 expression within β-cells is necessary to maintain a proper gene expression program to regulate insulin secretion and glucose homeostasis. Moreover, we observe that the specific E2f1 deficiency in adult mature β-cells results in glucose homeostasis impairments and failure in GSIS. Because E2F1 overexpression stimulates β-cell proliferation and function (48), our results further suggest that targeting E2f1 might be of interest in maintaining β-cell functions in the context of diabetes.

The complexity of E2f1 biology resides in the fact that it can positively or negatively regulate the expression of its target genes. Our data suggest that E2f1 may be part of a repressor complex that regulates β-cell identity. In line with this, we observed that genes mostly upregulated in E2f1^β−/− isolated islets, including Gab1, Slc25a33, and Yap1, were identified as β-cell–disallowed genes [i.e., genes that are repressed in β-cells (52) or genes involved in the inflammatory response, such as Cxcl12, Cxcl14, and Irf5]. This suggests that the repressive effects of E2f1 are key in maintaining β-cell identity and subsequent β-cell function. Accordingly, repression of non–β-cell programs is crucial for maintaining β-cell identity. Pdx1 (14), Pax6 (15), Nkx6.1 (16), and Nkx2.2 (17) are key transcription factors required to maintain gene repression of non–β-cell programs. Interestingly, Nkx6.1 mRNA levels are decreased in E2f1^β−/− isolated islets, and the Nkx6.1 promoter is directly bound by E2f1, suggesting that E2f1 could modulate β-cell function through Nkx6.1 regulation. Chromatin regulators and epigenomic features play important roles in the control of β-cell identity and plasticity (53–55). Indeed, the modulation of islet-enriched transcription factor activity, including Pdx1 and Nkx2.2, involves their interaction with coregulators, such as Dnmt1, Dmnt3a, or Hdac1 (56–58), or their accessibility to chromatin, as demonstrated in the β-cell–specific deletion of protein arginine methyltransferase 1 (Prmt1), which results in the loss of β-cell identity and diabetes development (59). In addition, recent studies also indicate that loss of polycomb silencing in human and mouse β-cells contributes to the loss of β-cell identity in diabetes (12). Indeed, the β-cell–specific deletion of Eed, a component of the polycomb repressive complex-2 (PRC2), triggers β-cell dysfunction and dedifferentiation and diabetes development associated with chromatin state–associated transcriptional dysregulation (12). In addition, PRC2 loss induces β-cell plasticity through epigenomic reprogramming at both active and silent genes, suggesting that maintaining proper and specific histone marks and chromatin states at precise loci is crucial to maintaining normal β-cell function and avoiding T2D development. Interestingly, most of the upregulated genes in E2f1^β−/− pancreatic islets were characterized by bivalent H3K4me3/H3K27me3 and polycomb-repressed (H3K27me3) marks. In addition, epigenomic interventions triggering ectopic acetylation and gene derepression contribute to β-cell dysfunction. Indeed, blocking histone deacetylase (HDAC) activity through the use of the HDAC inhibitor SAHA impairs glucose intolerance in mice fed a high-fat diet (12). Here, we show that the loss of E2f1 function triggers transcriptional dysregulation of specific genes that are in a bivalent (i.e., H3K4me3/H3K27me3 and polycomb-repressed [H3K27me3] marks) and active (i.e., RNA-Pol2 recruitment and H3K4me3 and H3K27ac histone marks) state in healthy pancreatic islets. This suggests a potential ectopic acetylation in the promoter region of upregulated genes in E2f1^β−/− islets and decreased activation of the promoter region of downregulated genes. Interestingly, we demonstrate in this study that Ezh2, a member of the PRC2 complex, is also a direct E2f1 target gene. Although E2f1 is a ubiquitous transcription factor that is weakly expressed within β-cell compared with bona fide β-cell genes, we speculate that this transcriptional regulator may cooperate with β-cell transcription factors and/or the chromatin machinery, such as HDAC or the PRC2 complex, to integrate some signals necessary for β-cell maintenance in physiological conditions. Because E2f1 regulates gene expression through its interaction with repressor complexes, including pRb, SWI/SNF, and HDACs (18), our finding that E2f1 mediates repression of non–β-cell programs deserves deeper investigation to identify the E2f1 complexes that can trigger these transcriptomic and epigenomic effects in β-cells and their patho(physio)logical consequences.

The current study has a number of limitations. First, the use of RipCre transgenic mice to delete E2f1 in the β-cells may have also contributed to adverse effects on glucose homeostasis. Although we observed that these mice do not have a metabolic phenotype (Fig. 1) (22,31), it cannot be excluded that the transgene per se could have exerted specific deleterious functions in β-cells.

Second, overexpression studies in Min6 and pharmacological treatment with the E2F inhibitor may display adverse effects and induce off-target actions, such as nonspecific inhibition of other cell-cycle/E2f-related proteins or modulation of cofactor activity as a result of overexpression of transcription factors (60).

Third, the use of bulk RNA-seq on total pancreatic islets does not allow us to perfectly address the effect of E2f1 deletion on the β-cell transcriptome. Transcriptomic analyses using sorted β-cells or single-cell RNA-seq will provide direct evidence regarding the alteration of β-cell identity within specific pancreatic cell populations.

Finally, we observed increased α-cell–positive staining associated with decreased endocrine cell proliferation in E2f1^β−/− islets, and lineage-tracing experiments demonstrated that α-cells do not originate from β-cells. The origin of the supernumerary α-cells observed in E2f1^β−/− islets remains unknown and requires further investigation.

In summary, the present data highlight that E2f1 transcriptional activity within pancreatic islets is key for maintaining glucose homeostasis and insulin secretion through the regulation of key β-cell identity genes and the repression of non–β-cell programs, both in mouse and human islets. The observation that E2F1 levels are decreased in human T2D islets (24) suggests that reduced E2F1 expression or activity may contribute to β-cell failure in diabetes.

Acknowledgments. The authors thank Raphael Scharfmann and members of INSERM UMR1167 and U1283 for helpful discussions and Céline Gheeraert for excellent help with ChIP experiments; the Experimental Resources platform at the Université de Lille, especially Cyrille Degraeve, Yann Lepage, Mélanie Besegher, and Julien Devassine for animal care; the Department of Histology of the Lille Medicine Faculty, particularly M.H. Gevaert and R.M. Siminski, for histological preparations; and the France Genomique Consortium.

Funding. Human islets were provided through JDRF award 31-2008-416 (European Consortium for Islet Transplantation Islet for Basic Research program). This work was supported by grants from the France Genomique Consortium (grant ANR-10-INSB-009); EGID (grant ANR-10-LABX-46); Equipex 2010 (grant ANR-10-EQPX-07-01); the LIGAN-PM Genomics platform, a French state fund managed by the Agence Nationale de la Recherche under the frame program Investissements d’Avenir I-SITE Université de Lille Nord-Europe (ULNE) (grant ANR-16-IDEX-0004) (J.K.-C., F.P., P.F., A.B., and J.-S.A.); Agence Nationale pour la Recherche (Betaplasticity grant ANR-17-CE14-0034) (P.F., P.C., and J.-S.A.); European Foundation for the Study of Diabetes (J.-S.A.); European Commission, INSERM, CNRS, and Institut Pasteur de Lille (grant CPER CTRL Melodie) (E.C., B.P., and J.-S.A.); Fondation pour la Recherche Médicale (grant EQU202103012732) (J.-S.A.); Association pour la Recherche sur le Diabète (J.-S.A.); Université de Lille (F.O., C.B., X.G., N.R., and J.-S.A.); I-SITE ULNE (EpiRNAdiab gustain grant) (J.-S.A.); Conseil Régional Hauts de France and Métropole Européenne de Lille (X.G., N.R., and J.-S.A.); Fonds Européen de Développement Régional (N.R., P.F., and J.-S.A.); and Société Francophone du Diabète (S.A.H. and J.-S.A.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.O., C.B., M.E.F., E.C., L.R., X.G., M.M., C.C., N.R., E.B., and S.A.H. contributed to the in vivo and cellular experiments. F.O., C.B., M.E.F., E.C., N.R., P.F., A.B., and J.-S.A. wrote and/or edited the manuscript. F.O., C.B., M.M., E.D., S.A., L.B., M.D., and A.B. performed the RNA-seq and ChIP-seq experiments and analyses. P.-D.D., Z.B., P.M., L.F., J.K.-C., F.P., B.P., and P.C. provided reagents and data and discussed the results of the study. J.-S.A. designed the study, supervised the project, and contributed to experiments and/or their analysis and to the funding of this project. J.-S.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.