CBX4 stabilizes NEUROD1 through SUMO-ubiquitin crosstalk to delay β-cell senescence

Linming Su; Yan Liu; Qianxing Hu; Hongjin Chen; Yimeng Qin; Chenying Xu; Kun Gao; Yuncai Zhou; Yue Yang; Jianxing Liu; Yi Pan; Yanfeng Zhang; Fengjiao Huo; Xiaohang Zhou; Hailiang Liu; Liang Jin; Fangfang Zhang

doi:10.1038/s41467-026-77536-7

. 2026 Sep 7;17:10589. doi: 10.1038/s41467-026-77536-7

CBX4 stabilizes NEUROD1 through SUMO-ubiquitin crosstalk to delay β-cell senescence

Linming Su ^1,^#, Yan Liu ^1,^#, Qianxing Hu ^1,^#, Hongjin Chen ¹, Yimeng Qin ¹, Chenying Xu ¹, Kun Gao ¹, Yuncai Zhou ², Yue Yang ¹, Jianxing Liu ¹, Yi Pan ¹, Yanfeng Zhang ¹, Fengjiao Huo ³, Xiaohang Zhou ⁴, Hailiang Liu ^3,^✉, Liang Jin ^1,^✉, Fangfang Zhang ^1,^✉

PMCID: PMC13645805 PMID: 42844274

Abstract

Aging-associated β-cell senescence is a key driver of type 2 diabetes mellitus (T2DM). However, its associated molecular mechanisms remain poorly explored. This study identifies Chromobox 4 (CBX4), a Polycomb group protein with SUMO E3 ligase activity, as an essential regulator of β-cell aging and function. The results show that CBX4 expression progressively downregulates in both aged and diabetic pancreatic islets. Targeted deletion of Cbx4 in β-cell accelerates cellular senescence, impairs glucose tolerance, and compromises mitochondrial function. CBX4 interacts directly with the transcription factor NEUROD1 via its SUMO-interacting motif 1 (SIM1). CBX4 promotes SUMOylation of NEUROD1 at lysine 38 and drives its assembly into nuclear condensates via phase separation. These processes collectively antagonize NEDD4-mediated ubiquitination at the same site, stabilizing NEUROD1. The preserved NEUROD1 enhances Ins2 transcription and suppresses Camk2n1 expression, thus activating CaMKII signaling, promoting calcium influx, and maintaining mitochondrial bioenergetics. Furthermore, the study identifies STAT5A as a transcriptional regulator upstream of Cbx4. These findings reveal a STAT5A-CBX4-NEUROD1 signaling axis associated with β-cell aging in T2DM and suggest CBX4 as a promising therapeutic target to maintain β-cell function in metabolic disorders.

Subject terms: Calcium and phosphate metabolic disorders, Sumoylation, Chronic inflammation

As pancreatic β-cells age, CBX4 levels fall. Restoring CBX4 stabilizes the transcription factor NEUROD1 via SUMOylation and phase separation, counteracting its degradation to rejuvenate β-cell function and offering a new diabetes treatment avenue.

Introduction

Aging is associated with progressive deterioration of protein function, organelle integrity, cellular function, and tissue homeostasis, which contributes to the development of metabolic disorders and neurodegenerative diseases¹. T2DM is a metabolic condition characterized by pancreatic β-cell dysfunction, in which inadequate insulin secretion fails to compensate for peripheral insulin resistance, leading to disrupted glucose regulation². The accumulation of senescent β-cell correlates with the higher incidence of T2DM in aging populations, but the mechanisms associated with β-cell senescence to T2DM development remain poorly defined.

Senescent β-cell show several hallmark features, including disturbed intracellular calcium ([Ca²⁺]_cyt) homeostasis, reduced expression of essential β-cell identity and functional genes such as Pdx1, NeuroD1, and MafA, impaired mitochondrial function, elevated inflammatory signaling, and decreased insulin secretion^3–5. Increasing evidence indicates that maintaining mitochondrial quality is vital for delaying β-cell senescence and sustaining β-cell function, and it plays a key role in T2DM pathophysiology^6–8. Furthermore, dysregulation of [Ca²⁺]_cyt has been shown to impair mitochondrial function in β-cell⁹. Despite this, the exact mechanisms linking calcium signaling, β-cell aging, and T2DM in both mice and human β-cell remain unexplored.

Polycomb repressive complex 1 (PRC1) is a key epigenetic regulator of gene silencing^10,11, and Chromobox (CBX) family members serve as molecular readers that repress transcription associated with cellular senescence¹². Among these, CBX4 is distinguished by its dual function: it not only participates in canonical PRC1-dependent transcriptional repression but also possesses intrinsic SUMO E3 ligase activity¹³. Through SUMOylation, CBX4 modulates the stability, subcellular localization, and protein–protein interaction (PPI) networks of its substrates^14,15, processes that are essential for cellular aging^16–18, metabolic reprogramming¹⁴, and tumor progression^19–21. Moreover, genome-wide association studies (GWAS) have identified CBX4 as a susceptibility locus for diabetes²². Despite these findings, the functional role of CBX4 in pancreatic β-cell and its contribution to β-cell senescence–associated T2DM is mainly unexplored²³.

This study shows that CBX4 expression is reduced in pancreatic islets from both aging and diabetes-induced mouse models. Using an inducible, β-cell-specific Cbx4 knockout mouse model, the study indicates that Cbx4 deficiency accelerates β-cell senescence, impairs mitochondrial function, and reduces insulin secretion. CBX4 maintains NEUROD1 protein stability by promoting its sequestration into nuclear condensates and enhancing SUMOylation at lysine 38, preventing NEDD4-mediated ubiquitination. Stabilized NEUROD1 directly binds to the Camk2n1 promoter and represses its transcription, leading to high phosphorylated CAMKII levels, increased calcium influx, and enhanced mitochondrial oxidative capacity. These molecular events delay β-cell aging and support insulin secretory function. Furthermore, STAT5A is identified as an upstream regulator of Cbx4 expression, establishing a mechanistic link between CBX4 dysregulation and aging-associated T2DM pathogenesis. This study identifies CBX4 as a pivotal molecular mediator of β-cell senescence and metabolic dysfunction, providing a strong mechanistic foundation for therapeutic strategies to maintain β-cell integrity and function.

Results

Downregulation of CBX4 during β-cell senescence

To examine the potential relationship between CBX4 and β-cell aging, this study conducted a meta-analysis of publicly available single-cell RNA sequencing (scRNA-seq) datasets from pancreatic islets of healthy donors and T2DM patients, obtained from the Human Pancreas Analysis Program (HPAP) database²⁴ (https://hpap.pmacs.upenn.edu/explore/download?matrix/, Supplementary Fig. 1A). Analysis was restricted to β cell, which segregated into ten transcriptionally distinct clusters (Fig. 1A; Supplementary Fig. 1B). The study applied the SenMayo senescence gene signature to these scRNA-seq datasets, identifying clusters 6, 7, and 9 as senescent β-cell populations (Fig. 1B). These senescent clusters showed substantially lower CBX4 expression than non-senescent clusters (Fig. 1C), indicating that CBX4 may function as an aging-associated gene in β-cell.

Fig. 1 — A Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction of pancreatic β cells from healthy volunteers and patients with T2DM in the HPAP database. B SenMayo was used to predict senescent cells within the β-cell clusters, red line circled the senescent β-cell cluster. C The feature plot of CBX4 expression in β cells. D Immunoblot and quantification of CBX4 and β-Actin proteins in the islets from C57BL/6J mice (2-month-old, 12-month-old, and 18-month-old; n = 10 mice). Immunoblot and quantification of CBX4, P16^INK4a and β-Actin proteins in senescent MIN6 cells (E), senescent mouse islets (F), and senescent human islets (G) induced by 450 μM H₂O₂ (n = 3 biological samples). Immunoblot and quantification of CBX4, P16^INK4a and β-Actin proteins in senescent MIN6 cells (H), senescent mouse islets (I), and senescent human islets (J) induced by 200 nM doxorubicin (n = 3 biological samples). Immunoblot and quantification of CBX4, P16^INK4a and β-Actin proteins in islets of the 15 weeks *db/db* mice (K, n = 5 mice) and in the islets of the 20 weeks HFD-fed mice (L, n = 5 mice). (M-N) Isolated islets and exocrine glands from the pancreas of female and male mice (4-month-old; n = 5 mice) showed that CBX4 protein (M) and mRNA (N) levels were enriched in the islets. The protein content of CBX4 was calculated using ImageJ. Isolated β cells and α cells from islets of female and male mice (4-month-old; n = 10 mice) showed that CBX4 protein (O) and mRNA (P) levels were enriched in the β-cell. The protein content of CBX4 was calculated using ImageJ. Data were analyzed using a two-sided t-test with Welch’s correction to compare two groups. Values are presented as mean ± SEM. M islet, mouse islet; H islet, human islet.

CBX4 expression was examined in pancreatic islets isolated from young (2 months), middle-aged (12 months), and aged (18 months) mice. Higher blood glucose levels and increased insulin resistance were observed in the 12- and 18-month-old groups (Supplementary Fig. 1C, D). Consistent with the scRNA-seq findings, CBX4 expression was significantly reduced in islets from middle-aged and aged mice and inversely correlated with the senescence marker p16^INK4a or P21^Cis1, encoded by the Cdkn2a and Cdkn1a locus respectively (Fig. 1D and Supplementary Fig. 1E).

To further validate the association between CBX4 and cellular senescence, senescence was induced in MIN6 cells, isolated islets, and human islets using 450 μM H₂O₂ or 200 nM doxorubicin, as per the established protocols³. Under these conditions, both CBX4 mRNA and protein levels significantly decreased, while p16^INK4a expression increased (Fig. 1E–J and Supplementary Fig. 1F–K). Since β-cell senescence is known to contribute to T2DM development, CBX4 levels were also examined in metabolic disease models. Similarly, CBX4 levels were also reduced in the islets of high-fat diet (HFD)-fed mice and db/db mice, with a corresponding increase in p16^INK4a expression (Fig. 1K, L and Supplementary Fig. 1L, M). These results suggest that lower CBX4 expression in pancreatic islets is closely associated with the development of aging-associated T2DM.

Further comparative analysis revealed that CBX4 expression in mouse pancreatic islets was approximately 24-fold higher than in exocrine tissue (Fig. 1M, N). This enrichment was further confirmed by fluorescence-activated cell sorting (FACS), based on β and α cells, followed by quantitative assessment of CBX4 expression, which showed preferential expression of CBX4 in β-cell (Fig. 1O, P and Supplementary Fig. 1N). These results establish that CBX4 is highly enriched in β-cell and suggest that its downregulation contributes to β-cell senescence and functional decline during aging and T2DM progression.

Compromised Cbx4 induces β-cell senescence in vitro and in vivo

To observe the relationship between CBX4 expression and β-cell senescence, this study modulated Cbx4 levels in MIN6 cells and mouse primary islets by transfecting with either a Cbx4 coding sequence plasmid (oe-Cbx4) or a Cbx4-specific siRNA (si-Cbx4) for 48 h, followed by exposure to 450 μM H₂O₂ for another 24 h. Cbx4 expression increased by approximately 200-fold in the oe-Cbx4 group and decreased by nearly 70% in the si-Cbx4 group (Supplementary Fig. 2A, B). As shown in Fig. 2A, B, silencing of Cbx4 in MIN6 cells significantly upregulated senescence-associated genes P16^INK4a and P21^Cis1. Consistent with this, senescence-associated β-galactosidase (SA-β-gal) staining revealed a substantial increase in SA-β-gal-positive cells after Cbx4 knockdown (Fig. 2C). Mitochondrial function was examined using cellular respiration assays, which showed that Cbx4 depletion significantly impaired mitochondrial activity (Fig. 2D), whereas Cbx4 overexpression increased the ATP/ADP ratio (Supplementary Fig. 2C). Similarly, Cbx4 knockdown caused functional deficits in β-cell. Glucose-stimulated insulin secretion (GSIS) assays and quantitative RT-PCR analyses showed that reduced Cbx4 expression significantly repressed insulin secretion (Fig. 2E, F), insulin biosynthesis (Supplementary Fig. 2D, E), and total insulin content (Supplementary Fig. 2F, G). However, MIN6 cells overexpressing Cbx4 showed a significant increase in insulin granule abundance (Supplementary Fig. 2H). Consistent results were observed in mouse primary islets, where Cbx4 overexpression enhanced glucose response during both the first and second phases of insulin secretion, whereas Cbx4 knockdown attenuated glucose-stimulated responses (Fig. 2G, H).

Fig. 2 — *Oe-Cbx4* or *si-Cbx4* was transfected into MIN6 cells and mouse primary islets and induced an aging model. A, B Immunoblot of senescence-related markers (n = 3 biological samples). (C) SA-β-gal staining and quantification were performed to test MIN6 cell senescence (Scale bar: 100 μm, n = 3 biological samples). D OCR was measured (n = 3 biological samples). GSIS assays were performed to test insulin secretion in MIN6 cells (E) and mouse islets (F, 100 islets per group). G Real-time perfusion was performed in the islets (30 islets per group, n = 3 replications). H The phase I and phase II AUC of islet perfusion are shown (n = 3 replications). Oe-*CBX4* or *si-CBX4* was transfected into Endoc-βH5 cells and human primary islets to induce an aging model. I, J Immunoblot senescence-related genes protein levels (n = 3 biological samples). K, L GSIS was performed to measure insulin secretion (n = 3 biological samples). M Immunoblots of CBX4 protein levels were tested in the islets (n = 7 mice). N Representative SA-β-gal staining images of pancreatic tissues (n = 6 mice). O Flow cytometry was performed to test the percentage of SA-β-gal-positive cells in the islets (n = 7 mice). P Immunoblotting of senescence-related genes in the islets (n = 6 mice). OCR (Q, n = 6 mice)and ATP/ADP ratio assays (R, n = 7 mice) were measured in the islets isolated from 18-M-old mice. S *Cbx4* KO primary islets were isolated and pooled from each individual mouse (n = 4 mice). Real-time perfusion was performed in response to 2.8 mM glucose and 16.7 mM glucose (n = 3 biological samples). T TEM of pancreatic β-cell. Insulin granules are indicated with red squares. Scale bars: 2 μm. Statistics of insulin granules; n = 20–27 fields per group. U GTT was performed on 6-M-old mice (n = 8 mice). Data were analyzed using two-way ANOVA, and a two-sided t-test correction was applied to compare two groups. Values are presented as mean ± SEM. AUC area under the curve, FCCP carbonyl cyanide p-trifluoromethoxyphenylhydrazone.

To observe translational relevance in human β-cell systems, CBX4 expression was modified in EndoC-βH5 cells and human primary islets, followed by treatment with 450 μM H₂O₂. CBX4 expression increased approximately 120-fold after overexpression and decreased by roughly 70% after knockdown (Supplementary Fig. 2I, J). Consistent with observations in mouse models, CBX4 overexpression significantly repressed the expression of the senescence markers P16^INK4a and P21^Cis1 (Fig. 2I, J), while enhancing insulin secretion (Fig. 2K, L), insulin content (Supplementary Fig. 2K, L), and insulin biosynthesis (Supplementary Fig. 2M, N). These findings indicate that CBX4 attenuates β-cell senescence in both mouse and human islets in vitro.

To further validate the in vivo role of Cbx4 in β cells, this study developed β-cell-specific Cbx4 knockout mice (Cbx4 KO) by crossing Cbx4^fl/fl mice with RIP-Cre transgenic mice (Supplementary Fig. 2O, P). In these animals, Cbx4 expression was nearly abolished in pancreatic islets, with only slight reductions in the brain and no significant changes in adipose tissue, kidney, or liver (Fig. 2M and Supplementary Fig. 2Q). Both male and female Cbx4 KO mice maintained on a normal diet for 20 months exhibited mildly higher blood glucose levels without significant differences in body weight compared with control mice (Supplementary Fig. 2R, S). Cellular senescence in vivo analysis revealed significant SA-β-gal staining in the pancreas of Cbx4 KO mice as early as 4 months of age, with further intensification at 18 months (Fig. 2N). A significantly higher number of SA-β-gal–positive cells was detected within the islets of Cbx4 KO mice at both ages compared with controls (Fig. 2O and Supplementary Fig. 2T). Consistent with these results, protein levels of the senescence markers P16^INK4a and P21^Cis1 were substantially elevated in islets from Cbx4 KO mice after both 4- and 18-month feeding periods (Fig. 2P). Genes associated with the senescence-associated secretory phenotype (SASP) were significantly upregulated in Cbx4-deficient islets (Supplementary Fig. 2U).

Mitochondrial function in isolated islets was evaluated using Seahorse XFe96 analysis, which showed reduced basal oxygen consumption rates (OCR) in response to glucose in Cbx4-deficient islets compared with controls (Fig. 2Q). Consistent with this, primary islets from Cbx4 KO mice showed a reduced ATP/ADP ratio, further indicating impaired mitochondrial function after Cbx4 knockdown (Fig. 2R). Functional assessment of β cells revealed that Cbx4-deficient islets showed impaired GSIS during perfusion assays (Fig. 2S). Moreover, ultrastructural analysis revealed a 4-fold reduction in insulin granule abundance in islets from Cbx4 KO mice compared with control animals (Fig. 2T). These defects were associated with impaired glucose tolerance, as demonstrated by glucose tolerance testing (GTT) (Fig. 2U).

These data highlight that Cbx4 deletion in β cells accelerates cellular senescence, disrupts mitochondrial and secretory function, and impairs glucose homeostasis, underscoring a crucial role for Cbx4 in maintaining β-cell integrity and function both in vitro and in vivo.

Obesity-induced β-cell senescence is rejuvenated by Cbx4

To evaluate whether Cbx4 silencing exacerbates β-cell senescence under metabolic stress, Cbx4 KO mice and control mice were fed an HFD for 20 weeks starting at 6 weeks of age (Supplementary Fig. 3A). After 2 weeks of HFD, Cbx4 KO mice had higher blood glucose levels (Supplementary Fig. 3B) and increased body weight (Supplementary Fig. 3C). Cbx4 KO mice also had higher HOMA-IR indices (Supplementary Fig. 3D) and modestly higher circulating insulin levels (Supplementary Fig. 3E). After 8 and 16 weeks of HFD, Cbx4 KO mice showed significantly impaired glucose tolerance (Fig. 3A and Supplementary Fig. 3F, G) and reduced insulin sensitivity relative to control mice (Fig. 3B and Supplementary Fig. 3H, I). Based on these metabolic abnormalities, SA-β-gal staining was substantially increased in the pancreas of Cbx4 KO mice after 8 and 16 weeks of HFD feeding (Supplementary Fig. 3J). Quantitative analysis of isolated islets showed that the proportion of SA-β-gal-positive cells increased from 6.48 to 9.37% after 4 weeks of HFD and from 23.5 to 34.5% after 20 weeks of HFD in Cbx4-deficient mice (Fig. 3C and Supplementary Fig. 3K). This increase in senescent β cells was associated with significant upregulation of SASP genes (Fig. 3D). Western blot analyses further confirmed increased expression of the senescence markers P16^INK4a and P21^Cis1 in islets from Cbx4 KO mice after both 4 and 20 weeks of HFD (Fig. 3E). Mitochondrial function was evaluated using Seahorse XFe96 analysis, which revealed significant impairment in islets isolated from Cbx4 KO mice fed an HFD (Fig. 3F). Consistent with this, a reduced ATP/ADP ratio was observed in primary islets after 20 weeks of HFD feeding (Fig. 3G). Functional assessment of β cells showed that Cbx4 KO mice on an HFD showed reduced insulin secretion (Fig. 3H, I), insulin biosynthesis (Supplementary Fig. 3L), and a decreased number of insulin granules within β cells (Fig. 3J). These in vivo findings indicate that β-cell-specific deletion of Cbx4 accelerates obesity-induced β-cell senescence and compromises islet function under metabolic stress.

Fig. 3 — AUC of the blood glucose level was calculated via IPGTT (2 g/kg, A, n = 6–10 mice) or IPITT (0.75 U/kg, B, n = 7 mice) in male and female *Cbx4* KO mice treated with 8-week or 16-week HFD feed. C Flow cytometry was used to detect SA-β-gal positive cells in the islets isolated from 4-w or 20-w HFD-treated mice (n = 3 biological samples). D SASP gene expression was analyzed by qRT-PCR (n = 3 mice). E Immunoblotting of senescence-related genes in the islets isolated from male and female *Cbx4* KO mice maintained on HFD for 4 weeks or 20 weeks (n = 7 mice). F OCR was measured in the islets isolated from 12-w HFD-treated mice (n = 6 mice). G ATP/ADP ratio assays were performed in the islets isolated from 12-w HFD-treated mice (n = 5 mice). Islets isolated from male and female *Cbx4* KO mice and their WT littermates maintained on HFD for 8 weeks (H, n = 5 mice) or 16 weeks (I, n = 5 mice) were used for GSIS to test insulin secretion. J TEM of pancreatic β-cell at 12-w HFD treated. Scale bars, 2 μm, Statistics of insulin granules. n = 20–27 fields per group. (K) IPGTT (2 g/kg) were performed in *oe-Cbx4*/HFD mice (n = 10 mice). L Immunoblotting of senescence-related genes in the islets isolated from *oe-Cbx4*/HFD mice (n = 5 mice). (M-O) SASP genes (M, n = 5 mice), OCR (N, n = 7 mice) and ATP/ADP ratio (O, n = 5 mice) were measured in the islets isolated from *oe-Cbx4*/HFD mice. P *oe-Cbx4*/HFD and HFD-control mice primary islets were isolated and pooled from each individual mouse (n = 4 mice). Real-time perfusion was performed in response to 2.8 mM glucose, 16.7 mM glucose. And AUC (n = 4 mice). Data were analyzed using two-way ANOVA, and a two-sided t-test correction was used to compare two groups. Values are presented as mean ± SEM. AUC area under the curve, FCCP carbonyl cyanide p-trifluoromethoxyphenylhydra zone.

To evaluate whether restoring Cbx4 expression could reverse obesity-associated β-cell senescence, C57BL/6J mice were fed an HFD for 16 weeks to induce metabolic stress. Approximately 1 × 10¹² AAV-RIP-Cbx4 particles, designed to induce β-cell–specific overexpression of Cbx4 under the control of the Insulin 2 promoter, were delivered via pancreatic intraductal infusion in male C57BL/6J mice. These animals, referred to as oe-Cbx4/HFD mice, were maintained on the HFD for an additional 8 weeks (Supplementary Fig. 3M). Seven days after viral delivery, Cbx4 expression in pancreatic islets increased by approximately 12-fold compared with control mice (Supplementary Fig. 3N, O). Metabolic analysis showed that oe-Cbx4/HFD mice showed significantly improved glucose tolerance (Fig. 3K) and enhanced insulin sensitivity (Supplementary Fig. 3P). Based on these results, circulating insulin levels were reduced (Supplementary Fig. 3Q), and HOMA-IR indices were significantly decreased (Supplementary Fig. 3R). Histological analysis revealed a significant reduction in SA-β-gal staining in pancreatic tissue from oe-Cbx4/HFD mice compared with controls (Supplementary Fig. 3S). Similarly, expression of the senescence markers P16^INK4a and P21^Cis1, as well as SASP-associated cytokines, including IL-6, IL-8, and IL-1α, was significantly reduced in islets from oe-Cbx4/HFD mice (Fig. 3L, M and Supplementary Fig. 3T). Restoration of Cbx4 expression also improved mitochondrial function, as evidenced by increased oxygen consumption and a higher ATP/ADP ratio in isolated islets (Fig. 3N, O). Moreover, islet perifusion assays showed that oe-Cbx4/HFD mice revealed significantly enhanced glucose-stimulated insulin secretion during both the first and second phases compared with control animals (Fig. 3P). These results demonstrate that β-cell-specific overexpression of Cbx4 alleviates obesity-induced β-cell senescence, restores mitochondrial and secretory function, and improves systemic insulin sensitivity, showing the therapeutic potential of Cbx4 in maintaining β-cell integrity under metabolic stress.

CBX4 interacts with NEUROD1 to attenuate β-cell senescence

Next, to elucidate the molecular mechanism by which Cbx4 modulates β-cell senescence, this study performed co-immunoprecipitation (Co-IP) assays using an anti-CBX4 antibody in MIN6 cells, followed by mass spectrometry to identify interacting proteins. A total of 348 proteins were enriched in the CBX4-associated complex, among which NEUROD1 ranked the top ten candidate interactors (Supplementary Fig. 4A). Consistent with this result, Co-IP analysis revealed a reduced interaction between CBX4 and NEUROD1 in pancreatic islets isolated from Cbx4 KO mice compared with controls (Fig. 4A) and in MIN6 cells after siRNA-mediated knockdown of Cbx4 (Fig. 4B). To determine whether this interaction was direct, an in vitro pull-down assay was performed using recombinant GFP-tagged CBX4 and mCherry-tagged NEUROD1 proteins purified from E. coli. These experiments confirmed a direct physical association between CBX4 and NEUROD1, with binding intensity increasing in a concentration-dependent manner for both proteins (Fig. 4C; Supplementary Fig. 4B). Bimolecular fluorescence complementation (BiFC) assays were also performed in MIN6 cells to observe the intracellular localization of the CBX4-NEUROD1 interaction. In this system, CBX4 and NEUROD1 were fused to the N-terminal (NYFP) and C-terminal (CYFP) fragments of yellow fluorescent protein (YFP), respectively. Interaction between the two proteins resulted in reconstitution of functional YFP (Fig. 4D, top), revealing that the CBX4-NEUROD1 complex was predominantly localized within the nucleus (Fig. 4D, bottom). These results establish a direct nuclear interaction between CBX4 and NEUROD1.

Fig. 4 — A Endogenous co-IP experiments were performed on islets isolated from *Cbx4* KO or control mice (n = 8 mice). B Exogenous co-IP experiments were conducted on MIN6 cells transfected with *Cbx4* plasmid or *si-Cbx4* (n = 3 biological samples). (C) In vitro co-IP was carried out using purified GFP-tagged CBX4 (1 μg) and mCherry-tagged NEUROD1 (0 μg, 0.25 μg, 0.5 μg, 1 μg; n = 3 biological samples). D BiFC assays were performed to test the binding ability between NEUROD1 and CBX4 (n = 3 biological samples). Schematic illustration was drawn by figdraw (ID: TPSTI99497). Immunoblotting and quantification of senescence-related genes in the islets isolated from *Cbx4* KO, and *Cbx4* KO & AAV-*RIP-NeuroD1* mice (E, n = 7 mice); GSIS was performed to assess insulin secretion (F, n = 5 mice); OCR (G, n = 3 mice) and ATP/ADP ratio were measured to evaluate mitochondrial function (H, n = 3–5 mice). I SA-β-gal staining and quantification (Scale bar: 100 μm, n = 3 biological samples) showed that *oe-NeuroD1* MIN6 cells restored SA-β-gal levels induced by *si-Cbx4*. J OCR was measured in MIN6 cells after *Cbx4* overexpression or *NeuroD1* knockdown (n = 5 biological samples). FCCP carbonyl cyanide p-trifluoromethoxyphenylhydrazone. K GSIS was performed to assess insulin secretion (n = 3 biological samples). Data were analyzed using two-way ANOVA. Values are presented as mean ± SEM.

NEUROD1 is a key transcription factor required for the maintenance of mature β-cell identity, and its expression declines during β-cell senescence^25,26. Based on this, this study observed a significant reduction in NEUROD1 levels in pancreatic islets isolated from 12- and 18-month-old mice (Supplementary Fig. 4C). To determine whether CBX4 regulates β-cell senescence through NEUROD1, adeno-associated virus (flo) particles were used to mediate β-cell-specific overexpression of NeuroD1 under the control of the Insulin 2 promoter. Specifically, 1 × 10¹² AAV-RIP-NeuroD1 particles were administered via pancreatic intraductal infusion into 18-month-old Cbx4 KO male mice. Restoration of NeuroD1 expression in β cells significantly alleviated senescence phenotypes (Fig. 4E) and enhanced insulin biosynthesis and secretion compared with untreated Cbx4 KO mice (Fig. 4F and Supplementary Fig. 4D). Mitochondrial function was further evaluated using Seahorse XFe96 analysis, which showed that NeuroD1 overexpression restored the OCR profile compromised by Cbx4 deficiency (Fig. 4G). Simultaneously, the ATP/ADP ratio in primary islets from Cbx4 KO mice increased significantly after NeuroD1 re-expression, approaching levels observed in control animals (Fig. 4H).

Besides this, rescue experiments in MIN6 cells yielded consistent results. Overexpression of NeuroD1 significantly reduced the proportion of SA-β-gal positive cells (Fig. 4I) and suppressed the induction of senescence markers P21^Cis1 and P16^INK4a caused by siRNA-mediated Cbx4 knockdown (Supplementary Fig. 4E). Cell respiration assays further demonstrated that NeuroD1 overexpression restored mitochondrial function (Fig. 4J) and normalized the ATP/ADP ratio (Supplementary Fig. 4F), both of which were impaired after Cbx4 silencing. Further, NeuroD1 re-expression rescued defects in insulin biosynthesis and secretion induced by Cbx4 knockdown (Fig. 4K and Supplementary Fig. 4G).

These results indicate that CBX4 delays β-cell senescence by directly interacting with NEUROD1, thus stabilizing NEUROD1-dependent transcriptional programs and maintaining mitochondrial integrity and insulin secretory capacity. These findings prompted further investigation into the molecular mechanisms underlying CBX4’s role in NEUROD1-mediated regulation of β-cell homeostasis.

CBX4 stabilizes NEUROD1 via SIM1-dependent SUMOylation and phase separation-mediated protection from NEDD4

Considering the crucial role of NEUROD1 in maintaining β-cell identity and function, its expression was evaluated after modulation of CBX4. Although NeuroD1 mRNA levels remained unchanged, NEUROD1 protein abundance was markedly increased in MIN6 cells transfected with a Cbx4 expression plasmid (Fig. 5A and Supplementary Fig. 5A). However, NEUROD1 protein levels were selectively reduced in pancreatic islets isolated from Cbx4 knockout mice (Fig. 5B; Supplementary Fig. 5B), indicating post-transcriptional regulation by CBX4. CBX4 contains a chromodomain that recognizes H3K27me3 and two SUMO-interacting motifs (SIMs) required for its SUMO E3 ligase activity. To dissect the structural requirements for NEUROD1 stabilization, a series of CBX4 mutants was generated, including a chromodomain mutant (CDM; F11A/W35L), SIM1 deletion (ΔSIM1), SIM2 deletion (ΔSIM2), and a combined SIM1/SIM2 deletion (ΔSIM1&2) (Supplementary Fig. 5C). Transient expression of these constructs in MIN6 cells revealed a significant reduction in NEUROD1 protein levels in cells expressing the ΔSIM1 mutant, whereas deletion of SIM2 or disruption of the chromodomain had no significant effect (Fig. 5C). Co-immunoprecipitation assays further demonstrated that deletion of SIM1 substantially weakened the interaction between CBX4 and NEUROD1, compared with full-length CBX4, CDM, or ΔSIM2 constructs (Fig. 5D). These observations indicate that the SIM1 domain is essential for CBX4-mediated stabilization of NEUROD1.

Fig. 5 — NEUROD1 protein levels were assessed in MIN6 cells (A, n = 3 biological samples), and in islets (B, n = 5 mice) and in MIN6 cells transfected with Flag-*Cbx4* or *Cbx4* mutant (C, n = 3 biological samples). D Co-IP was performed to test the binding ability between CBX4 and NEUROD1 (n = 3 biological samples). Endogenous co-IP experiments were performed in islets (E, n = 7 mice), in MIN6 cells transfected with *Cbx4* or *si-Cbx4* (F, n = 3 biological samples) and in MIN6 cells transfected with *Cbx4* or a *Cbx4* mutant (G, n = 3 biological samples) to test NEUROD1 SUMOylation. H Representative IF staining of NEUROD1 (turquoise) and SUMO1 (red) in MIN6 cells (n = 3 biological samples). I Flag-tagged *Cbx4* was co-overexpressed with GFP-tagged *SUMO1*, *SUMO2*, or *SUMO3*, and HA-tagged *Ubc9*. Endogenous co-IP experiments were performed using anti-Flag antibody (n = 3 biological samples). J Co-IP was performed using an anti-HA antibody to test NEUROD1 SUMOylation (n = 3 biological samples). NEUROD1 protein levels in MIN6 cells induced by CHX (K, n = 3 biological samples), in MIN6 cells induced by MG132 (L, n = 3 biological samples). M MIN6 cells were co-transfected with *si-Cbx4*, the cells underwent Co-IP using an antibody against NEUROD1 (n = 3 biological samples). N MIN6 cells were transfected with oe-*Cbx4* or *si-Cbx4*. The cells were then subjected to Co-IP using an antibody against NEUROD1 (n = 3 biological samples). O MIN6 cells were transfected with *si-Nedd4* or co-transfected with *si-Nedd4* and *si-Cbx4*. Co-IP was then performed using an antibody against NEUROD1 (n = 3 biological samples). P MIN6 cells were transfected with NEUROD1 mutants and also with *oe-Nedd4*, the cells were subjected to Co-IP using an antibody against Flag (n = 3 biological samples). GFP-tagged CBX4 co-transfected with mcherry- tagged NEUROD1 (Q, n = 3 biological samples), or GFP-tagged CBX4 co-transfected with mcherry-tagged NEDD4 (R, n = 3 biological samples) into MIN6 cells and imaged with LSM 800 with airyscan, Scale bar 10 μM. Data were analyzed using two-way ANOVA. Values are presented as mean ± SEM.

To determine whether NEUROD1 stabilization depends on the SUMO E3 ligase activity of CBX4, SUMOylation assays were performed using pancreatic islets from Cbx4 knockout and control mice. Co-immunoprecipitation analysis revealed a significant reduction in NEUROD1 SUMOylation in Cbx4-deficient islets compared with controls (Fig. 5E). Based on these results, SUMOylation assays in MIN6 cells transfected with oe-Cbx4 or si-Cbx4 overexpression or knockdown constructs showed that NEUROD1 was conjugated with SUMO1, as detected by immunoblotting under denaturing conditions (Fig. 5F). SUMO1 conjugation to NEUROD1 was abolished in cells expressing the ΔSIM1 mutant (Fig. 5G), further supporting the requirement of SIM1 for NEUROD1 SUMOylation.

Immunofluorescence analysis showed that Cbx4 overexpression not only increased NEUROD1 protein abundance but also promoted colocalization of NEUROD1 with SUMO1 within nuclear condensates, indicative of SUMO-dependent phase separation (Fig. 5H). Considering that UBC9 is the sole E2-conjugating enzyme required for SUMOylation, its involvement was further examined. In the presence of HA-tagged UBC9, all three SUMO isoforms (GFP-SUMO1, SUMO2, and SUMO3) conjugated to NEUROD1. Co-expression of Cbx4 and Ubc9 considerably enhanced NEUROD1 SUMO1 conjugation relative to Cbx4 expression alone (Fig. 5I). In comparison, co-expression of ΔSIM1 with Ubc9 failed to improve NEUROD1 SUMOylation (Supplementary Fig. 5D), confirming the indispensability of SIM1 in this process.

To examine the specific SUMOylation sites on NEUROD1, bioinformatic predictions were performed using the GPS-SUMO and JASSA computational tools^27,28. Overlapping predictions from both platforms identified four candidate lysine residues: K38, K39, K139, and K286 (Supplementary Fig. 5E). Site-directed mutagenesis was used to substitute each lysine residue with arginine (R), generating single-point mutants (wild-type (WT), K38R (Mut1), K39R (Mut2), K139R (Mut3), K286R (Mut4)) and combinatorial mutants. These constructs were co-transfected with a Cbx4 expression plasmid into MIN6 cells. SUMOylation assays showed that substitution of K38, K39, K139, or K286 with arginine reduced SUMO1 conjugation compared with WT NEUROD1, with the K38R mutation producing the most pronounced reduction (Fig. 5J). These data demonstrate that CBX4 promotes NEUROD1 SUMOylation in a SIM1-dependent manner, with lysine 38 identified as the primary SUMOylation site responsible for NEUROD1 stabilization.

Previous studies have shown that SUMO-modified proteins can be targeted for ubiquitin-dependent degradation or that SUMOylation can antagonize ubiquitination²⁹. To observe how Cbx4 preserves NEUROD1 protein stability, cycloheximide (CHX) chase assays were performed in control and si-Cbx4-treated MIN6 cells. Under CHX treatment, depletion of Cbx4 considerably accelerated NEUROD1 degradation, indicating reduced protein stability (Fig. 5K). Furthermore, the stabilizing effect of Cbx4 on NEUROD1 was abolished by co-treatment with the 26S proteasome inhibitor MG132, whereas silencing of Cbx4 significantly increased the accumulation of endogenous NEUROD1 in the presence of MG132 (Fig. 5L). Based on these observations, extensive ubiquitination of NEUROD1 was detected in MIN6 cells transfected with si-Cbx4 (Fig. 5M). In comparison, high expression of Cbx4 significantly prolonged the half-life of endogenous NEUROD1 (Supplementary Fig. 5F). These observations indicate that Cbx4 promotes SUMOylation of NEUROD1 while suppressing its ubiquitin-mediated degradation.

To identify the E3 ubiquitin ligase responsible for NEUROD1 turnover, the UbiBrowser database was queried, and Co-IP/MS analysis of NEUROD1-associated proteins was performed. This approach identified neural precursor cell expressed developmentally down-regulated 4 (NEDD4) as a candidate E3 ligase for NEUROD1 (Supplementary Fig. 5G, H). Co-immunoprecipitation assays further showed that Cbx4 overexpression significantly reduced the interaction between NEUROD1 and NEDD4 in MIN6 cells (Fig. 5N). Simultaneous knockdown of Nedd4 and Cbx4 attenuated NEUROD1 ubiquitination compared with Cbx4 knockdown alone (Fig. 5O). Moreover, NEDD4-mediated ubiquitination of NEUROD1 depended on lysine 38 (Fig. 5P), suggesting direct competition between SUMOylation and ubiquitination at this residue.

To elucidate how CBX4-mediated SUMOylation spatially limits NEDD4 accessibility to NEUROD1, we investigated the role of CBX4 in regulating NEUROD1 phase separation. We considered our earlier observation of NEUROD1 puncta formation and the emerging concept of biomolecular condensation, which led us to investigate the role of CBX4 in regulating NEUROD1 phase separation. Analysis using the PONDR algorithm revealed that NEUROD1 harbors extensive intrinsically disordered regions (IDRs) (Supplementary Fig. 5I), which are known to promote biomolecular condensate condensation³⁰. In vitro droplet formation assays showed that increasing concentrations of CBX4 enhanced NEUROD1 droplet formation (Supplementary Fig. 5J). In cellular assays, GFP-tagged CBX4 formed distinct nuclear condensates into which mCherry-NEUROD1 was efficiently recruited (Fig. 5Q). In comparison, NEDD4 failed to form droplets and was excluded from CBX4-positive nuclear bodies (Supplementary Fig. 5K and Fig. 5R). Moreover, Co-IP assays results showed that condensates disrupted by 1.5% 1,6-hexanediol (1,6-HD) can reduce NEUROD1 SUMOylation induced by CBX4 (Supplementary Fig. 5L).

In summary, CBX4 interacts with NEUROD1 via its SIM1 domain and catalyzes SUMOylation at lysine 38, a modification that directly competes with NEDD4-mediated ubiquitination at the same site. Simultaneously, CBX4 mediates the formation of nuclear condensates that selectively recruit the intrinsically disordered NEUROD1 while excluding NEDD4, providing spatial protection against proteasomal degradation and enhancing SUMOylation efficiency. These mechanisms establish a dual regulatory strategy by which CBX4 maintains NEUROD1 stability through both chemical modification and physical sequestration.

CBX4-NEUROD1 axis programs calcium signaling to maintain mitochondrial function and suppress β-cell senescence

Based on our finding that CBX4 stabilizing NEUROD1 at the post-translational level, this study next examined how NEUROD1 stabilization is translated into transcriptional programs that regulate β-cell function and senescence. To comprehensively define NEUROD1 downstream targets regulated by Cbx4 during β-cell senescence, Cleavage Under Targets and Tagmentation (CUT&Tag) assays were performed to map genome-wide NEUROD1 binding profiles. Nuclei were isolated from MIN6 cells stably transduced with lentiviral vectors expressing either Cbx4 (len-Cbx4) or a control construct. This strategy enabled high-resolution identification of NEUROD1 chromatin occupancy under conditions of Cbx4 upregulation. In len-Cbx4 MIN6 cells, CUT&Tag profiling revealed significant enrichment of NEUROD1 binding near transcription start sites (TSSs) (Fig. 6A and Supplementary Fig. 6A), with approximately 45.4% of NEUROD1 peaks localized within promoter regions (Fig. 6B). Consistent with previous observations, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of promoter-associated peaks selectively upregulated after Cbx4 overexpression showed significant enrichment in pathways related to insulin secretion (Fig. 6C). Similarly, Gene Ontology (GO) analysis of the CUT&Tag dataset revealed strong enrichment in biological processes associated with calcium ion-regulated exocytosis and regulation of calcium-dependent exocytotic signaling (Supplementary Fig. 6C), implicating calcium signaling as a key downstream effector of the CBX4-NEUROD1 axis.

Fig. 6 — A–E MIN6 cells infected with lentivirus *Cbx4* (len-Cbx4), cells were collected for CUT&Tag assay to screen downstream targets of NEUROD1 (n = 3 biological samples). A The violin plot shows the average reads density of NEUROD1 in control and len-*Cbx4*- MIN6-cell, minima (2.089, 2.539), maxima (9.514, 10.515), centre (6.218, 6.904), bounds of box (upper edge: 7.886, 8.826; lower edge: 4.521, 5.263), whiskers (upper: −0.5265, −0.0815; lower: 12.09225, 13.27975), 25% percentile (4.521, 5.263) and 75% percentile (7.886, 8.826). B The distribution of NEUROD1 on the genome. C KEGG analysis of the promoter region of the NEUROD1 distribution. D Integrative Genomics Viewer tracks of CUT&Tag showing enriched NEUROD1 in the promoters of *Ins2* and *Camk2n1*, and (E) relative fold enrichment of peak area. Chip-qPCR was performed to test the binding ability between *NeuroD1* and *Ins2* promoter (F, n = 3 biological samples) or between *NeuroD1* and *Camk2n1* promoter (G, n = 3 biological samples) in the MIN6 cells. H *Cbx4* bound to insulin promoter depend on *NeuroD1* as determined by dual luciferase assay (n = 3 biological samples). I Relative luciferase activity in MIN6 cells (n = 3 biological samples). Western blot of CAMK2N1, CAMKⅡ and P-CAMKⅡ protein levels in the islets (J, n = 7 mice), in the MIN6 cells transfected with *oe-Cbx4* or *si-Cbx4* (K, n = 3 biological samples), and in the MIN6 cells transfected with *Cbx4*, *Cbx4* mutant (L, n = 3 biological samples). M Ca²⁺ concentration was examined by FLIPR induced by 16.7 mM glucose (n = 3 biological samples). N Ca²⁺ concentration were examined in the MIN6 cells via living cell workstation, scale bar, 5 μm (n = 3 biological samples). O Representative staining of mitochondria with JC-1 to reflect the MMP, scale bar, 20 μm (n = 3 biological samples). P A schematic illustration of the mechanism by which CBX4/NEUROD1 axis delays β-cell aging. Schematic illustration was drawn by figdraw (ID: ISI0A11191). Data were analyzed using two-way ANOVA. len lentivirus.

Among the most prominent findings, differential NEUROD1 binding peaks were identified within the promoter regions of Insulin 2 (Ins2) and calcium/calmodulin-dependent protein kinase II inhibitor 1 (Camk2n1) (Fig. 6D). Two promoter-associated peaks reached statistical significance, indicating divergent regulation of these genes (Fig. 6E). CUT&Tag assays using an anti-RNA polymerase II antibody further revealed increased Pol II occupancy at the Ins2 promoter and reduced Pol II occupancy at the Camk2n1 promoter in len-Cbx4 MIN6 cells (Fig. 6D, E). Quantitative RT-PCR analysis of genes identified by CUT&Tag confirmed Ins2 as the most strongly upregulated transcript and Camk2n1 as the most downregulated transcript in response to Cbx4 overexpression (Supplementary Fig. 6D). Overexpression of NeuroD1 restored the expression changes of Ins2 and Camk2n1 induced by Cbx4 knockdown, further supporting NEUROD1-dependent transcriptional regulation (Supplementary Fig. 6E). High-resolution NEUROD1 CUT&Tag mapping enabled precise identification of NEUROD1 binding motifs within the promoter regions of Ins2 and Camk2n1 (Supplementary Fig. 6F). Chromatin immunoprecipitation followed by qPCR (ChIP-qPCR) further demonstrated that Cbx4 overexpression significantly enhanced NEUROD1 occupancy at both promoters (Supplementary Fig. 6G, H). Mutation of the SUMO-interacting motif 1 (SIM1) domain of CBX4 abolished this enhancement, indicating that SIM1-dependent interactions are required for NEUROD1 chromatin recruitment (Fig. 6F, G). Luciferase reporter assays revealed that Cbx4 overexpression promoted NEUROD1-mediated transcriptional regulation of Ins2 and Camk2n1 promoters, whereas disruption of the SIM1 domain attenuated this effect (Fig. 6H, I).

Camk2n1 encodes a specific endogenous inhibitor of Ca²⁺/calmodulin-dependent protein kinase II (CaMKII), a key regulator of calcium influx and intracellular calcium signaling^31–33. Examination of pancreatic islets from Cbx4 KO mice revealed elevated CAMK2N1 protein levels along with a significant reduction in phosphorylated CaMKII (Fig. 6J). In comparison, Cbx4 overexpression in MIN6 cells suppressed CAMK2N1 expression and increased CaMKII phosphorylation (Fig. 6K). This effect was abrogated by SIM1 mutation, indicating that CBX4-NEUROD1 interaction is required for CaMKII activation (Fig. 6L). Furthermore, knockdown of NeuroD1 reversed the Cbx4-induced increase in CaMKII phosphorylation (Supplementary Fig. 6I), confirming NEUROD1 as a crucial mediator of this pathway. Consistent with CaMKII activation, Cbx4 overexpression significantly increased cytosolic calcium concentrations ([Ca²⁺]_cyt) in response to high glucose stimulation (16.7 mM, Fig. 6M). However, NeuroD1 knockdown significantly reduced glucose-induced increases in [Ca²⁺]_cyt (Supplementary Fig. 6J). Based on the previous functional data demonstrating enhanced mitochondrial activity after Cbx4 overexpression, this study examined mitochondrial calcium dynamics. Previous reports indicate that CaMKII-mediated increases in cytosolic calcium enhance mitochondrial calcium uptake ([Ca²⁺]_mt)³⁴. In line with this model, mitochondrial calcium uptake was significantly elevated in len-Cbx4 MIN6 cells after glucose stimulation, whereas NeuroD1 knockdown attenuated Cbx4-induced [Ca²⁺]_mt accumulation (Fig. 6N). Moreover, mitochondrial membrane potential (MMP) analysis with the JC-1 dye showed enhanced mitochondrial polarization in Cbx4-overexpressing cells, an effect reversed upon NeuroD1 depletion (Fig. 6O).

These results indicate that CBX4 enhances NEUROD1 binding to the Ins2 promoter to stimulate insulin production while enhancing NEUROD1 binding to the Camk2n1 promoter to suppress Camk2n1 transcription, thus relieving inhibition of CaMKII. This synergistic transcriptional regulation elevates CaMKII activity, mitochondrial calcium uptake, preserves mitochondrial function, delays β-cell senescence, and promotes insulin secretion (Fig. 6P).

STAT5A downregulates Cbx4 expression during β-cell senescence

Considering the significant role of CBX4 in modulating β-cell senescence, this study examined potential upstream regulators of CBX4 expression. Previous reports have shown that H3K27me3 can regulate CBX4 transcription in brown adipose tissue¹⁴. To check whether a similar mechanism occurs in β cells, MIN6 cells were treated with the H3K27me3 inhibitor UNC1999.Immunoblot analysis did not reveal significant changes in CBX4 expression, indicating that H3K27me3 does not contribute to CBX4 regulation in β cells (Supplementary Fig. 7A, B). Next, transcription factors with potential binding sites within the Cbx4 promoter were predicted using the PROMO and JASPAR databases (score > 10). A total of 12 candidate factors were identified, including Pou2f2 (POU domain, class 2), Hoxd9 (homeobox D9), Sp1 (trans-acting transcription factor 1), Ar (androgen receptor), Zic1 (zinc finger protein of the cerebellum 1), Maz (MYC-associated zinc finger protein), Stat4 (signal transducer and activator of transcription 4), Zic2 (zinc finger protein of the cerebellum 2), Nkx2-5 (NK2 homeobox 5), Pou5f1 (POU domain, class 5, transcription factor 1), Max (Max protein), and Stat5a (signal transducer and activator of transcription 5A, (Supplementary Fig. 7C). Among these candidates, Stat5a showed the most significant reduction in MIN6 cells rendered senescent by treatment with either 450 μM H₂O₂ or 200 nM doxorubicin (Fig. 7A, B).

Based on these observations, STAT5A protein levels were decreased in senescent pancreatic islets and MIN6 cells (Fig. 7C, D and Supplementary Fig. 7D, E). STAT5A is recognized as an important regulator of β-cell function during T2DM progression^35,36, and its expression is known to be suppressed by oxidative stress^37,38. In line with these reports, STAT5A protein levels were significantly reduced in islets from HFD-fed and db/db mice (Fig. 7E, F). To further examine the regulatory relationship between STAT5A and CBX4, MIN6 cells were transfected with a Stat5a overexpression plasmid (oe-Stat5a), a control pcDNA3.1 vector, si-NC, or si-Stat5a for 48 h. Stat5a overexpression increased transcript levels by approximately 210-fold, whereas si-Stat5a achieved an estimated 70% knockdown efficiency. Significantly, CBX4 expression increased by approximately 5.5-fold after Stat5a overexpression and decreased by nearly 60% upon Stat5a silencing (Supplementary Fig. 7F). CBX4 protein abundance was elevated in oe-Stat5a-treated MIN6 cells (Fig. 7G). Similar results were observed in EndoC-βH5 cells (Fig. 7H and Supplementary Fig. 7G). Stat5a overexpression effectively restored CBX4 expression that had been suppressed by H₂O₂ or doxorubicin treatment (Fig. 7I, J).

To evaluate whether STAT5A regulates Cbx4 expression in vivo, adeno-associated virus vectors enabling β-cell-specific Stat5a overexpression under the Insulin 2 promoter (AAV-RIP-Stat5a) were developed. A total of 1 × 10¹² viral particles were delivered to 8-week-old C57BL/6J mice via pancreatic intraductal infusion. This approach induced Stat5a expression increased in islets compared with control virus-treated mice (Supplementary Fig. 7H). After AAV-RIP-Stat5a administration, CBX4 protein levels were significantly higher in pancreatic islets (Fig. 7K). Moreover, mice receiving either AAV-RIP-Stat5a or a control virus were fed for 16 weeks with HFD, reaching body weights of 45–50 g. HFD feeding exacerbated β-cell senescence, as evidenced by overexpression of P16^INK4a and P21^Cis1. In comparison, Stat5a overexpression alleviated HFD-induced β-cell senescence and restored CBX4 expression (Fig. 7L).

To elucidate the molecular mechanism by which STAT5A regulates Cbx4 transcription during β-cell senescence, reporter constructs containing the WT Cbx4 promoter or predicted STAT5A-binding-site mutants (Cbx4-Mut1 and Cbx4-Mut2) were generated based on JASPAR predictions (Supplementary Fig. 7I, J). Dual-luciferase assays showed that Stat5a overexpression significantly enhanced transcriptional activity driven by the WT Cbx4 promoter, whereas mutation of the R1 binding region (Cbx4-Mut1) significantly attenuated this effect (Supplementary Fig. 7K). These results were further validated by ChIP-qPCR analyses in MIN6 cells and in islets from AAV-RIP-Stat5a-treated mice (Fig. 7M, N). EMSA further confirmed direct STAT5A binding to the Cbx4 promoter in both MIN6 cells (Fig. 7O) and mouse islets after AAV-RIP-Stat5a administration (Fig. 7P). These results revealed that age-associated downregulation of STAT5A contributes to reduced CBX4 expression during β-cell senescence, identifying STAT5A as an upstream transcriptional regulator of CBX4 in aging pancreatic β cells.

Discussion

Aging-related β-cell senescence is a major, complex factor in T2DM; however, its associated molecular mechanisms are poorly understood³⁹. This study identifies CBX4, a Polycomb group protein with intrinsic SUMO E3 ligase activity, as a key protector against β-cell aging. CBX4 expression declines progressively in pancreatic islets during physiological aging and in diabetic states, and β-cell-specific deletion of Cbx4 accelerates senescence and compromises β-cell function. In comparison, reconstitution of CBX4 expression reduces senescence-associated defects. Mechanistic analyses reveal that CBX4 directly associates with NEUROD1 and stabilizes its protein by coordinating site-specific SUMOylation at lysine 38 with phase-separation-dependent spatial sequestration. This dual mechanism counteracts NEDD4-mediated ubiquitination and proteasomal degradation of NEUROD1. Stabilized NEUROD1 promotes insulin gene transcription and represses Camk2n1 expression, thus maintaining mitochondrial bioenergetic capacity and insulin secretory function. Moreover, STAT5A is identified as an upstream transcriptional regulator of Cbx4, and its age-associated reduction triggers this pathogenic signaling cascade. These findings define the STAT5A-CBX4-NEUROD1 signaling axis as a key molecular association between β-cell aging and metabolic dysfunction, positioning CBX4 as a potential therapeutic target in T2DM (Fig. 8).

Fig. 8 — In young islets, STAT5A promotes the transcription of *Cbx4*. *Cbx4* promotes the sumoylation of NEUROD1 at lysine 38, which is dependent on its SIM1 region. The nuclear condensates of CBX4 recruit NEUROD1 but exclude the E3 ubiquitin ligase NEDD4, thereby stabilizing the protein expression of NEUROD1. The preserved NEUROD1 enhances *Ins2* transcription and suppresses *Camk2n1* expression, thus activating CaMKII signaling, promoting calcium influx, and maintaining mitochondrial bioenergetics, which finally facilitate CBX4 to delay β-cell senescence, and improves insulin secretion. During aging, reduced STAT5A expression leads to downregulation of *Cbx4*, resulting in exacerbated β-cell senescence and repressed insulin secretion. Schematic illustration was drawn by figdraw (ID: STYIYaeeed).

As an integral component of the canonical PRC1, CBX4 is unique among Polycomb proteins in possessing SUMO E3 ligase activity and has been involved in the regulation of cell proliferation, differentiation, and senescence¹². Genetic studies have further identified CBX4 as a diabetes susceptibility locus²². Meta-analyses of publicly available scRNA-seq datasets from human pancreatic islets revealed reduced CBX4 expression within senescent β-cell clusters. Despite these observations, the functional contribution of CBX4 to β-cell biology and its role in senescence-driven T2DM pathogenesis had remained poorly defined²³. The current findings provide direct evidence that CBX4 delays β-cell aging and preserves β-cell functional integrity by stabilizing NEUROD1, a transcription factor indispensable for insulin synthesis and endocrine lineage maintenance^26,40. Indeed, CBX4 is a multifunctional PRC1 component with broad SUMO E3 ligase activity and has multiple substrates. In our study, co-immunoprecipitation combined with mass spectrometry identified interactions between CBX4 and multiple senescence-associated proteins, including lamin A⁴¹, HMGB1/2⁴², and PARP1⁴³. Moreover, based on previous reports, the heterochromatin protein KRAB-associated protein 1 (KAP1) and YAP1 are substrate of CBX4^44,45. It has also been reported KAP1 as a sensor for regulating the DNA damage of β-cell and Yes-associated Protein 1(YAP1) affects inflammation in β-cell⁴⁶, which further confirms that CBX4 is an important regulatory factor in controlling the aging of β-cell⁴⁷. Further exploration of the interaction between CBX4 and other potential substrates may provide additional insight into its multifaceted roles in β-cell aging and homeostasis.

Previously, NEUROD1 was regulated by noncoding RNAs^25,48 and by post-translational modifications, including O-GlcNAcylation⁴⁹. However, the regulation of NEUROD1 through ubiquitin-like modifications has not been described. This study demonstrates that CBX4 increases NEUROD1 protein abundance without altering its transcript levels, supporting a post-translational regulatory mechanism. CBX4 interacts with NEUROD1 through its SIM1 domain and promotes SUMOylation at lysine 38. NEDD4 was identified as the E3 ubiquitin ligase responsible for targeting NEUROD1 at the same residue. Competitive binding between CBX4 and NEDD4 determines the modification state of NEUROD1, with CBX4-driven SUMOylation preventing ubiquitination and enhancing protein stability. This mechanism is consistent with previous observations that intrinsically disordered regions of CBX4 promote the recruitment of interacting partners into biomolecular condensates³⁰. In line with this model, CBX4 overexpression enhanced NEUROD1-SUMO1 co-localization within nuclear condensates, forming a favorable microenvironment for SUMOylation and NEUROD1 stabilization.

NEUROD1 has recently been identified as an age-associated gene in β cells^3,50; however, the mechanisms by which NEUROD1 controls β-cell senescence and insulin secretory capacity remain poorly defined. This study explores direct downstream targets of NEUROD1 that mediate its effects on β-cell aging and function. CUT&Tag profiling was used to characterize transcriptional targets of the CBX4–NEUROD1 regulatory axis. These analyses showed that Cbx4 overexpression increased NEUROD1 occupancy at the Camk2n1 promoter, leading to transcriptional repression of this gene. Camk2n1 encodes a selective inhibitor of Ca²⁺/CaMKII, a kinase essential for maintaining mitochondrial Ca²⁺ homeostasis^51,52. Camk2n1 has been reported to act as a negative regulator of insulin-stimulated glucose uptake⁵³. Disruption of mitochondrial Ca²⁺ homeostasis is known to precipitate mitochondrial dysfunction, a key driver of β-cell senescence and impaired insulin secretion^7,54. In this regard, the current findings indicate that NEUROD1, acting in concert with CBX4, promotes mitochondrial function, ATP-dependent [Ca²⁺]_mt influx, and attenuates β-cell senescence by transcriptionally suppressing Camk2n1.

Although CBX4 is a key anti-aging factor, the mechanisms underlying its age-dependent downregulation remain unclear. Previous work by Chen Q¹⁴ demonstrated that CBX4 expression is regulated by H3K27me3 modifications in brown preadipocytes. In comparison, these findings indicate that H3K27me3 does not influence CBX4 expression in pancreatic islets. Instead, STAT5A was identified as a direct transcriptional regulator of Cbx4 through promoter binding. STAT5A is recognized as an age-associated gene⁵⁵ and has been shown to play an important role in preserving β-cell function during diabetes progression^36,56,57. The current data demonstrate that both aging and diabetic conditions are associated with reduced STAT5A expression. Moreover, β-cell–specific overexpression of Stat5a in mice, as well as Stat5a overexpression in MIN6 cells, effectively restored age-associated reductions in Cbx4 expression. These results suggest the molecular basis of CBX4 downregulation during aging. Recent research by Zhou J further identified STAT5 as a key regulator of tumor immunity⁵⁸, indicating that future studies may explore the role of CBX4 in pancreatic islet inflammation during diabetes progression.

In conclusion, these results establish CBX4 as a pivotal molecular association between β-cell senescence and T2DM and provide the first evidence that NEUROD1 protein abundance is regulated by an ubiquitin-like post-translational modification. Two interrelated regulatory axes are proposed: the CBX4–NEUROD1–INS2 axis regulating insulin biosynthesis and the CBX4-NEUROD1-CAMK2N1 axis controlling cellular senescence and insulin secretion. These pathways play essential roles in preserving β-cell functional integrity under conditions of aging and metabolic stress. This study represents an important step toward elucidating the contribution of CBX4 to diabetes and age-associated metabolic disorders.

Limitations of the study

This study highlights a significant protective role for CBX4 in regulating β-cell aging and function across multiple experimental models. However, several key limitations remain for future research. Firstly, while NEUROD1 contains IDRs, and CBX4 has been shown to promote the formation of NEUROD1-enriched nuclear condensates, the exact structural features and functional significance of this phase separation remain unclear. Future work should explore whether NEUROD1 IDRs are both necessary and sufficient for condensate formation and investigate how CBX4 affects the biophysical properties, stability, and dynamic behavior of these condensates. Furthermore, although key findings were validated in human islets and EndoC-βH5 cells, expanding these analyses to a larger, more diverse group of human donor islets stratified by age and diabetic status would improve clinical relevance and enable evaluation of the potential for therapeutic targeting of the CBX4-NEUROD1 pathway.

Methods

Single-cell RNA sequencing (scRNA-seq) data retrieval

The expression of CBX4 in human islet cells and its relationship to SenMayo score⁵⁹ were analyzed using a scRNA-seq expression matrix from islets isolated from healthy volunteers and patients with T2DM (Data obtained from the Human Pancreas Analysis Program (HPAP) Database, (https://hpap.pmacs.upenn.edu/explore/download?matrix)²⁴. ScRNA-seq data retrieval was processed using the R package Seurat⁶⁰. The gene expression data were normalized, and the Uniform Manifold Approximation and Projection (UMAP) analysis and the R package ggplot2 were used for visualization.

Animal care

Care of all animals was within institutional animal-care committee guidelines, and all procedures were approved by the animal ethics committee of China Pharmaceutical University (Permit Number: 10072) and were in accordance with the international laws and policies (EEC Council Directive 86/609, 1987). All animals were on the C57BL/6 background except for db/- mice and db/db mice, which were on the BKS background. Mice were housed in a pathogen-free animal facility and maintained in a temperature-controlled room (22 °C), with humidity at 55% and on a 12 h light-dark cycle (lights on from 6 a.m. to 6 p.m.). Unless otherwise stated, animals were fed normal chow diet (D12450J, 10% calories from fat), and water ad libitum. High-fat diet (HFD, D12494, 60% calories from fat) fed male mice for at least 16 weeks to obtain senescent mice induced by obesity. The number of mice is either indicated in the Figure legends.

Mouse models

We generated Cbx4 conditional knockout mice via the CRISPR/Cas9 system from Cyagen (suzhou, China). The Cbx4 gene (NCBI Reference Sequence: NM_007625.2) is on mouse chromosome 11, with 5 exons, starting at exon 1 and ending at exon 5. Exons 1 and 2 will be targeted for conditional knockout to inactivate the gene. A loxP site were inserted 0.9 kb upstream of exon 1. The targeting vector were engineered using PCR with a BAC clone template. Cas9 and gRNA were co-injected with the vector into fertilized eggs to produce mice. The offspring was genotyped by PCR and sequencing to obtain homozygous Cbx4 knockout mice (Cbx4^fl/fl). Ins2-cre (RIP-cre, Strain #:003573) mice wereobtained from Jackson laboratory. Then, Cbx4^fl/fl mice were crossed with RIP-cre mice to select RIP-Cre:Cbx4^fl/wt. Finally, these mice were crossed with Cbx4^fl/fl mice to delete Cbx4 in β cells. Off-target effects were identified by PCR and sequencing of tail DNA. RIP-Cre:Cbx4^fl/fl (Cbx4 KO) mice were used for further research, and Cbx4^fl/fl (control) mice were used as littermate controls. Both male and female mice have been used. The primers are listed in Supplementary Table 1.

MIN6 cells culture

The mouse pancreatic β-cell line MIN6 was cultured in DMEM (Gibco) supplemented with 15% FBS (Gibco, Burlington, ON, USA), 100 IU/mL penicillin, 100 μg/ml streptomycin, and 50 μmol/l β-mercaptoethanol (Sigma Aldrich, St. Louis, MO, USA) at 37 °C in a humidified atmosphere containing 5% CO₂.

Human EndoC-βH5 cells culture

Human EndoC-βH5 cells were obtained from Univercell Biosolutions (Toulouse, France). Human EndoC-βH5 cells were cultured using the described methods^61,62. Briefly EndoC-βH5 were cultured on βCoat® coated plastic plates in Ulti-β1 culture media. Cells were propagated every 7 days using trypsin (0.05%, Thermofisher).

Isolation and culture of primary islets

After euthanasia, mouse islets were isolated using the intraductal collagenase technique, and the digestion process was terminated with Hanks’ Balanced Salt Solution (composition: 137.93 mM NaCl, 5.33 mM KCl, 4.17 mM NaHCO₃, 1.26 mM CaCl₂, 0.493 mM MgCl₂, 0.407 mM MgSO₄, 0.441 mM KH₂PO₄, 0.338 mM Na₂HPO₄, 5.56 mM D-glucose; pH 7.4) supplemented with 0.1% bovine serum albumin, 2.5 mM glucose, and 1% penicillin/streptomycin. The islets were then handpicked and incubated overnight in 1640 medium containing 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. After overnight incubation, the islets were seeded in plates for subsequent experiments.

Human islets were obtained from the Human Islet Resource Center (HIRC, China), Tianjin First Central Hospital, People of the Republic of China, with informed research consents. All human studies were conducted according to the principles of the Declaration of Helsinki and approved by Ethics Committee of the Tianjin First Central Hospital⁶³. Written informed consent was obtained from all subjects. High purity islets (>80%) were collected and cultured in CMRL-1066 medium (Corning, Manassas, VA, USA), supplemented with 10% Human Serum Albumin (Baxter, Vienna, Austria), 100 U/ml penicillin and 100 μg/ml streptomycin at 37 °C in 5% CO2. After overnight incubation, the islets were seeded in plates for subsequent experiments. Donor information is listed in Supplementary Table 2, with every subject utilized in each analysis. Sex and/or gender determinations were gathered from the patient’s medical records in accordance with local hospital policies. Due to limited group size, the impact of sex and/or gender on the variables assessed in our studies were not performed.

Induction of senescent cells model

MIN6 cells, EndoC-βH5 cells, or islets were treated with 200 nM doxorubicin or 450 μM H₂O₂ in medium for 24 h. Following this initial treatment, the cells were cultured in compound-free medium for an additional 48 h to obtained senescent cells model. Then, senescent cells were used for the subsequent research.

Transfection of genes

Isolated mouse islets (100 islets), MIN6 cells (approximately 5 × 10⁵), or EndoC-βH5 cells (approximately 5 × 10⁵) were seeded in six-well plates and cultured in antibiotic-free medium for 4 h. Subsequently, the cells were transfected with LipoBooster 3000 reagent (YEASEM, C0533) following the manufacturer’s instructions. For plasmid transfection, 2 μg of Cbx4 plasmid, CDM, ΔSIM1, ΔSIM2, or ΔSIM1&2 was used, while for siRNA-mediated knockdown, 50 nM si-Cbx4 or si-Nedd4 was applied. After transfection 48 h, the cells were treated with either 200 nM doxorubicin or 450 μM H₂O₂ in medium for 24 h, then cells were cultured in compound-free medium for an additional 48 h. After which they were subjected to assessments of aging-related functions.

Islet perifusion assay

For each perifusion assay, perifusion chambers were loaded with 150 IEQ of islets. Islets were first perifused with Krebs’ buffer (135 mM NaCl, 3.6 mM KCl, 0.5 mM NaH₂PO₄, 1.5 mM CaCl₂, 2 mM NaHCO₃, 10 mM HEPES, and 0.1% BSA, pH 7.4) containing 2.5 mM glucose for 30 min. This was followed by perifusion with Krebs’ buffer containing 2.8 mM or 16.7 mM glucose for 12 and 30 min, respectively, under a flow rate of 30 μl/min. Fractions were collected every 3 min. After the assay, islets were incubated with acid-ethanol (1.5% HCl in 70% ethanol) at −20 °C overnight, followed by homogenization. The homogenized lysate was centrifuged at 10,000 × g at 4 °C, and the supernatant was used for insulin content measurement using a mouse insulin ELISA kit (ExCell Bio, Shanghai, China), according to the manufacturer’s instructions. Insulin secretion at each time point was quantified using the same ELISA kit and normalized to the total insulin content. To quantify biphasic insulin secretion, phase I and phase II were manually separated based on the plateau formation of phase II and quantified by calculating the area under the curve (AUC).

Insulin secretion assay

MIN6 cells (2 × 10⁵ cells well⁻¹) or isolated islets (15 islets well⁻¹) were seeded in 48-well plates and ectopically expressed Cbx4 as above for 48 h for glucose-stimulated insulin secretion (GSIS) assay. MIN6 cells or the islets were pre-incubated overnight in KRBH balanced buffer containing 0.2% BSA supplemented with 2.8 mM glucose, and were incubated for 2 h in 2.8 mM or 16.7 mM glucose. Immediately after incubation an aliquot of the medium was removed for analysis of insulin, and the cells were incubated in acid-ethanol for insulin content determination by mice insulin ELISA kit (ExCell Bio, Shanghai, China), according to the manufacturer’s instructions.

Mitochondrial respiration analysis

The mitochondrial respiration of islets was measured by high-resolution respirometry using the Seahorse Bioscience XFe96 Extracellular Flux Analyzer at 37 °C. Islet single cells were seeded in an XFe96 cell culture microplate (Agilent) at a density of approximately 1 × 10⁴ cells per well. The cells were equilibrated in DMEM medium transfected with Cbx4 or si-Cbx4. The baseline oxygen consumption rate (OCR) was recorded, followed by sequential injections of pharmacological inhibitors through the ports in the XF Assay cartridges: oligomycin (5 μM), an ATP synthase inhibitor used to measure oxygen consumption coupled to oxidative phosphorylation (OXPHOS); carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP) (5 μM), an uncoupling agent that induces maximum electron transport and thus enables the measurement of maximum OXPHOS respiratory capacity; and a mixture of antimycin A (5 μM) and rotenone (5 μM), which are inhibitors of mitochondrial complexes I and III, respectively. These injections were performed during continuous oxygen measurements using the Seahorse XFe96 Extracellular Flux Analyzer (Agilent) according to the manufacturer’s instructions.

Measurement of ATP/ADP ratio in primary islets

Primary mice islets cultured overnight were washed twice with Kreb’s buffer. For each sample, 20 IEQ islets were dissociated into single cells by TrypLE (12604013, Thermo Fisher Scientific), followed by incubation in Kreb’s buffer containing 2.8 or 16.7 mM glucose for 30 min. ATP/ADP ratios were measured using the ADP/ATP Ratio Assay kit (MAK135, Sigma-Aldrich) according to the manufacturer’s protocol.

Mitochondrial membrane potential assay

The mitochondrial membrane potential (MMP) is determined using JC-1 staining (C2003S; Beyotime, China). MIN6 cells were incubated with JC-1 for 30 min at 37 °C. The cells were directly examined by confocal microscope (Zeiss LSM 800).

Mitochondrial calcium imaging of MIN6 cells

MIN6 cells were infected with lentivirus for 72 h. Then, 2 μl of Mito-Tracker reagent was added per 10,000 cells, and the cells were cultured at 37 °C for 16 h. The culture medium containing Mito-Tracker was then discarded. The cells were washed three times with Hanks solution under light-protected conditions. Next, MIN6 cells were incubated for 60 min at 37 °C with Rhod-2/AM working solution (R1244; Thermo Fisher Scientific) at a final concentration of 4 μM. Afterward, the Rhod-2/AM was removed, and the cells were washed three times with Hanks solution without probe. To thoroughly remove residual Rhod-2/AM working solution, Hanks buffer was added to cover the cells, followed by incubation at 37 °C for approximately 30 min to ensure complete esterification of Rhod-2/AM within the cells. MIN6 cells were imaged in a humidified 37 °C cell culture chamber with 5% CO₂ mounted on a Zeiss LSM 800 inverted confocal microscope. Continuous image acquirement with 10 s interval started 2 min of glucose supplementation to reach 16.7 mM final concentration until fluorescent plateaus was reached.

Intracellular calcium imaging of MIN6 cells

MIN6 cells were seeded on a 35 mm glass-bottom dish (MatTek) and then infected with specific lentivirus for 72 h. Cells were then washed twice with Kreb’s buffer containing 2.8 mM glucose, and subsequently loaded with Kreb’s buffer containing 5.6 µM Fluo-4 AM (F14201; Thermo Fisher Scientific) and 2.8 mM glucose for 45 min. Following this, MIN6 cells were washed twice again with Kreb’s buffer containing 2.8 mM glucose, reaching a final volume of 150 μl in each well. The plate was then transferred to a BioTek Cytation 5. Fluorescence readings were taken for 10 s to establish the baseline, after which 25 μl of test compound solutions (8×) were added to the corresponding wells. Fluorescence readings were then recorded for 32 s. Background fluorescence was automatically subtracted from all Fluo-4 AM fluorescence measurements.

Senescence-associated β-galactosidase staining of MIN6 cells

Senescent MIN6 cells was assessed using a Senescence-associated β-Galactosidase Staining Kit (C0602; Beyotime, China) following the manufacturer’s instructions. MIN6 cells were washed three times with PBS, fixed with 4% paraformaldehyde for 15 min at room temperature, and then incubated for 16 h at 37 °C in darkness with the working solution containing 0.05 mg/ml 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal).

Senescence-associated β-galactosidase staining of pancreas

Pancreatic tissues were washed three times with PBS and fixed at room temperature for 15 min. Subsequently, they were incubated with SA-β-gal solution at 37 °C for 16 h. Ice-cold PBS was used to stop the enzymatic reaction. The stained intact pancreatic tissues were photographed using a digital camera.

Flow cytometric analysis of islets senescence

Isolated mice islets were digested with 0.25% Trypsin-EDTA at 37 °C for 15 min and resuspended in FACS buffer (2% FBS in PBS). A fluorescent substrate (Enzo Life Sciences EnzKit 130-0010) was used according to the manufacturer’s instructions, and the incubation time with the substrate was optimized to 1 h at 37 °C for single islet cells. A flow cytometer (FACScan®; BD Biosciences) equipped with FlowJo v10 software (BD Biosciences) was used to gate cells based on forward scatter. Propidium iodide was used to exclude dead cells, and cells were analyzed based on acidic β-galactosidase activity.

Bimolecular fluorescence complementation (BiFC)

For BiFC assay, 100 multiplicity of infection (MOI) of lentivirus-CBX4-NYFP and lentivirus-NEUROD1-CYFP were infected into MIN6 cells with polybrene (4 μg/ml), 72 h after transfection, the cells were stained with Hoechst (Beyotime, C1027). Fluorescent images were captured by a Zeiss LSM800 microscope. The fluorescence from YFP was excited with a 488 nm laser and collected in a range of 500–600 nm. The Hoechst dye was visualized using the same settings as DAPI.

Lentiviral production

Cbx4 overexpression and sh-NeuroD1 and sh-Cbx4 plasmids were constructed in the pHAGE-CMV vector using the One Step Cloning Kit (C113-02, Vazyme Biotech Ltd Nanjin, China). Lentiviral particles were produced by transient transfection of HEK293T cells with the packaging plasmids pMD2.G (Addgene #12259) and psPAX2 (Addgene #12260), together with the Cbx4 overexpression, sh-NeuroD1, or sh-Cbx4 plasmid in the lentiCRISPRv2 vector (len-Cbx4, len-shNeuroD1, len-shCbx4), using LipoBooster 3000 reagent (YEASEM, C0533). Forty-eight hours after transfection, the culture medium was collected, and lentiviral particles were concentrated by ultracentrifugation. The lentiviral titer for each construct was determined by infecting HEK293T cells and performing FACS analysis.

Lentivirus infection

Min6 cells were seeded in six-well plates at a density of about 5 × 10⁵ cells with 2 × 10⁸ TU/ml of lentivirus encoding either Cbx4 (len-Cbx4, 5 µl ml⁻¹), sh-Cbx4 (len-shCbx4, 5 µl ml⁻¹), or sh-NeuroD1 (len-shNeuroD1, 5 µl ml⁻¹) plus polybrene for 72 h. For comparison, a scramble control (lentiviral particles without targeting any specific region) was used as the control. The Oligo sequences of shRNA were listed in the Supplementary Table 6.

Pancreatic intra-ductal AAV infusion in mice

The procedure for retrograde infusion of adeno-associated virus (AAV) into the pancreatic duct has been described previously⁴⁸. Briefly, the male C57BL/6J mice were anesthetized with isoflurane. A volume of 1 × 10¹² particles of AVV-RIP-Cbx4, AVV-RIP-Stat5a, or AAV-RIP-NeuroD1 dissolved in 0.15 ml of normal saline (NS) or NS alone was infused at a rate of 6 µl per minute for 25 min using a R462 perfusion pump. After infusion, the hole in the duodenum created by the catheter was closed with a 7-0 suture. Upon completion of the infusion, the clamps were released. The exterior abdominal wound was closed using 7 mm wound clips. Mice were allowed to recover on a heating pad set at 37 °C until fully recovered. They were given free access to food and water after surgery. At 72 h post-injection, islets were lysed to extract total RNA or protein to assess overexpression efficacy. AVV-RIP-Cbx4, AVV-RIP-Stat5a, and AAV-RIP-NeuroD1 were obtained from WZ Biosciences Inc.

Mouse metabolic assays

After 12 h fasting treatment, mice fasting blood glucose (FBG) levels and fasting serum insulin (FINS) levels were examined via using a glucometer (OMRON, Japan) and by ELISA (ExCell Bio, Shanghai, China), respectively. The homeostatic model assessment indices of insulin resistance (HOMA-IR) was calculated with the equation (FBG (mmol/l)xFINS (mIU/l))/22.5. To perform the glucose tolerance tests, 2 g/kg glucose (Sigma-Aldrich, St Louis, MO, USA) was intraperitoneal (i.p.) injected into mice, whereas 0.75 U/kg insulin (Novolin R, Novo Nordisk, Bagsvaerd, Denmark) was i.p. injected into mice for insulin tolerance tests. Blood glucose levels were examined at 0, 15, 30, 60, 90 and 120 min after glucose or insulin injection. The AUC, calculated by the conventional trapezoid rule, are given as the incremental area under the curve.

RNA isolation and qRT-PCR analysis

Primary islets, MIN6 cells and Endoc-βH5 cells were cultured and treated as described above. Total RNA was extracted using Trizol (Invitrogen) and an RNeasy kit (QIAGEN, Duesseldorf, Germany) as previously described⁶⁴. Reverse transcription reaction was performed using EasyScript® All-in-One First-Strand cDNA Synthesis SuperMix for qPCR (One-Step gDNA Removal, Accurate Biology, China) and diluted cDNA was used for qRT-PCR analysis using SYBR Green ProTaq HS Premix qPCR Kit (Accurate Biology, China) with the appropriate primers listed in Supplementary Table 3. Relative expression of genes was determined using a comparative method (2^−△CT). Gapdh were used as an internal standard for mRNAs.

Western blot analysis

Primary islets (800 islets per group), MIN6 cells, and Endoc-βH5 cells were lysed with RIPA buffer (Beyotime) containing 1% PMSF (Sigma). The protein concentration was quantified using a BCA kit (Vazyme Biotech Ltd. Nanjing, China). The protein samples were separated by SDS-PAGE and transferred onto PVDF membranes, followed by blocking with 5% skim milk for 2 h and 16 h incubation at 4 °C with primary antibodies. After three washes with TBST, the membranes were incubated for 1 h with the appropriate secondary antibodies and developed using chemiluminescent substrates. All antibodies used in this study are listed in Supplementary Table 4. Blot images were captured using the Tanon 3500 system, and the intensity was quantified using ImageJ (v1.8.0).

Plasmid and transient transfections

The coding sequences for mouse chromobox 4 (Cbx4, NM_007625.3), neurogenic differentiation 1 (NeuroD1, NM_010894.3), small ubiquitin-like modifier 1 (Sumo1, NM_009460.2), small ubiquitin-like modifier 2 (Sumo2, NM_133354.2), small ubiquitin-like modifier 3 (Sumo3, NM_001301673.1), ubiquitin-conjugating enzyme E2I (Ube2i, Ubc9, NM_001177609.1), and signal transducer and activator of transcription 5A (Stat5a, NM_001164062.2), as well as human CBX4 (NM_003655.3), were amplified by PCR from full-length mouse cDNA and then cloned into the pcDNA 3.1 vector (Addgene, Watertown, MA, USA).

For the CBX4 mutants, the following mutants were inserted into the pcDNA 3.1 vector: CDM (F11 and W35L double mutant); SIM1 mutant (ΔSIM1, amino acids Δ263–266); SIM2 mutant (ΔSIM2, amino acids Δ460–463), SIM1&SIM2 mutant (ΔSIM1&2, amino acids Δ263–266 and amino acids Δ460–463). For the NEUROD1 mutant variants, the following mutations were inserted into the pcDNA 3.1 vector: Mut1 (lysine 38 to arginine), Mut2 (lysine 39 to arginine), Mut3 (lysine 139 to arginine), Mut4 (lysine 286 to arginine). All plasmids were confirmed to be correct by sequencing. All the primer sequences for PCR are listed in Supplementary Table 5.

The siRNA of Cbx4, NeuroD1 and Nedd4 were obtained from Gene Pharma (Shanghai, China). The Oligo sequences of shRNA or siRNA were listed in the Supplementary Table 7.

Luciferase assay

MIN6 cells were plated at a concentration of 2 × 10⁵ cells per well in 24-well plates and cultured for 24 h. To generate the reporter constructs, the complete promoters of mouse Camk2n1, Ins2, Cbx4 promoter, containing the wild-type NEUROD1 or STAT5A binding sites, were cloned upstream of the firefly luciferase coding sequence. 1 μg of Camk2n1 promoter, Ins2 promoter, Cbx4 promoter, or promoter mutant, along with Cbx4 overexpression plasmid, ΔSIM1 plasmid, or Stat5a overexpression plasmid was transfected into MIN6 cells using LipoBooster 3000 reagent (YEASEM, C0533). At 48 h post-transfection, dual luciferase reporter assays were performed using a Luciferase Assay System (Vazyme, Nanjing, China). Transfection data represent at least three independent experiments, each performed in triplicate. The primers for the wild-type and mutation sites are listed in Supplementary Table 5.

Electrophoretic mobility shift assay (EMSA)

Incubate the Stat5a protein with a probe composed of a Cbx4 promoter fragment for analysis via native polyacrylamide gel electrophoresis (PAGE). Amplify a 200-bp probe via PCR and purify it using a kit (TaKaRa). Generate negative control probes by amplifying 16S rRNA. Subsequently, combine the Stat5a protein and DNA probes with binding buffer (10 mM Tris base, 50 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol, 0.05% Nonidet P-40, 2.5% glycerol, pH 7.5) and incubate at 37 °C for 30 min. Afterward, run the samples on a 6% native polyacrylamide gel in 0.5×TBE buffer (44.5 mM Tris base, 44.5 mM boric acid, 1 mM EDTA, pH 7.5) at 200 V for 45 min. Then, stain the gel with ethidium bromide in 0.5×TBE for 20 min, and subsequently capture the image. The EMSA primers were listed in the Supplementary Table 5.

Chromatin immunoprecipitation qPCR (ChIP-qPCR) assay

For the ChIP-qPCR, MIN6 cells were fixed with 37% formaldehyde for 10 min, followed by sonication for 3 s per round, repeated for 30 rounds, to fragment the chromatin. The chromatin was incubated with anti-NEUROD1 or anti-STAT5A antibody at 4 °C overnight, and then immunoprecipitated using Protein A/G agarose beads. The purified DNA was amplified by real-time PCR using primer pairs that spanned the predicted NEUROD1 or STAT5A binding sites. The primer sequences were listed in Supplementary Table 3.

Co-IP assays

According to the manufacturer’s protocol, immunoprecipitation assays were performed using protein A/G agarose beads (Beyotime, China). In brief, total cell extracts were incubated with the specific primary antibodies and protein A/G agarose beads at 4 °C for 16 h. The next day, the immunoprecipitates were washed at least three times with SDS lysis buffer (Beyotime, China) and then further resolved by SDS-PAGE for subsequent Western blot analysis. Antibodies are listed in Supplementary Table 4.

Co-IP/MS

To identify proteins recruited by CBX4, HA-tagged CBX4 was overexpressed in MIN6 cells. A control group was transfected with an empty vector. After 48 h of transfection, the cells were lysed using NP-40 lysis buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% NP-40, 1% Triton X-100, 10% glycerol, 2 mM EDTA, 1 mM NaF, 1 mM Na₃VO₄), and a mixture of protease inhibitors (PIC) was added. HA-tagged CBX4 was immunoprecipitated using anti-HA beads, followed by SDS-PAGE to separate the CBX4-enriched proteins. The recovered proteins associated with HA or the control were resolved by gel electrophoresis and Coomassie brilliant blue staining. The eluted solutions were subjected to MS analysis (Shanghai Applied Protein Technology Co., Ltd) on a Q Exactive mass spectrometer (Proxeon Biosystems, Thermo Fisher Scientific).

CUT&Tag profiling

To dissect the DNA binding profile of NEUROD1, NEUROD1 CUT&Tag was performed after MIN6 cells transduced with len-Cbx4 for 72 h. A CUT&Tag assay kit (Vazyme, TD903) was used to perform this assay⁶⁵, according to the manufacturer’s instruction. In brief, approximately 200,000 cells were harvested by cell scraper and immobilized with ConA-coated magnetic beads. Cells were incubated in diluted anti-NEUROD1 antibody or anti-polyⅡ (1:50 dilution in antibody buffer) overnight. Permeabilized cells were then incubated with the anti-mouse antibody (1:100 dilution) in 50 μl of Dig-Wash buffer with rotation at room temperature for 1 h. After washing three times with Dig-Wash buffer, cells were incubated with pAG-Tnp (1:50 dilution in Dig-300 buffer) for 1 h at room temperature with rotation. Cells were washed three times with Dig-300 buffer and resuspended in 50 μl of fragmentation buffer and incubated at 37 °C for another 1 h. To release the DNA fragments, 5 μl Proteinase K, 100 μl Buffer L/B and20 μl DNA Extract Beads were added to the reaction system, and then incubated at 55 °C for 10 min. After washing beads with Buffer WA and Buffer WB subsequently, the attached DNA was collected using sterilized water. TruePrep® Index Kit V2 for lllumina (VazymeTD202) was used for library amplification, PCR reaction products were purified using VAHTS DNA Clean Beads (Vazyme, N411) according to the manufacturer’s protocol. After quality control, libraries were quantified and pooled for sequencing by Novogene company.

For sequencing data analysis, we applied a commonly used pipeline. In brief, high qualityand adaptor-removed clean data were submitted to Bowtie2 (v2.25) for alignment using mm10 assembly. Mapped reads that contain abundant redundant duplicates were taggedby Picard tools (v2.27.1) and then indexed by Samtools (v1.6). Deeptools (v3.5.1) was used to plot heatmap, profile or correlation according to the documentation’s instruction and output data matrix for drawing violin plot with R package ggpubr (v0.4.0). R-basedtools ChIPpeakAnno (v3.15) was used for peaks distribution analysis. The raw data of CUT & Tag were uploaded in GEO database, and the GEO accession numbers was GSE289234.

The CUT & Tag data of H3k27Ac (SRX1098144) and H3K4me3 (SRX5817359) in the β cells were obtained from GEO database.

GO and KEGG pathway enrichment analysis

To identify Gene Ontology (GO) terms that are significantly enriched compared to the entire genomic background, the hypergeometric test in ClusterProfiler was used to perform enrichment analysis of the biological process, molecular function, and cellular component for each set of DEGs (Differentially Expressed Genes). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment results were visualized using graphical representations generated by the ClusterProfiler software (version 4.4.4).

Immunofluorescence cell staining

Cells were passaged into 35 mm glass-bottom culture dishes (MatTek) cultured overnight. The cells were washed once with PBS, then fixed with a 4% paraformaldehyde solution in PBST (PBS containing 0.05% Tween-20) at room temperature for 15 min. After two washes with PBST, cells were permeabilized with 0.5% Triton X-100 in PBST for 20 min at room temperature. After one wash with PBST, cells were blocked with a solution of 1% BSA in PBST for 30 min at room temperature. Then, the blocking solution was replaced with 1 ml of the blocking solution containing the desired primary antibody (diluted at 1:200), and incubated for 1 h at room temperature or overnight at 4 °C. After four washes with PBST, the appropriate Alexa Fluor-conjugated secondary antibodies (diluted at 1:1000 in the blocking solution) were applied and incubated for 1 h at room temperature. After three washes with PBST, cells were incubated with 0.5 μg/ml DAPI for 15 min. After four washes with PBST, the cells were mounted using an anti-fade reagent (Invitrogen). The slides were analyzed using a confocal laser scanning microscope (Zeiss LSM800) at ×20 or ×40 magnification. The antibodies are listed in Supplementary Table 4.

Protein purification

The pET-28a-Flag-TEV-CBX4-EGFP and pET-28a-His-TEV-NEUROD1-mCherry plasmids were transformed into the E. coli BL21-RIL strain for the expression of recombinant proteins. Bacteria were cultured in 2×YT medium at 37 °C until the OD₆₀₀ nm reached 0.8, and then induced with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at 16 °C for 16 h. The pellet from a 2-liter bacterial culture was resuspended in 80 ml of binding buffer (50 mM Tris-HCl, pH 7.5, 500 mM NaCl) and lysed by sonication for 20 min on ice. The sonication was performed with 5-s pulses and 5-s pauses at 70% of the maximum power. After centrifugation at 14,500 x g for 30 min to remove cellular debris, the supernatant was loaded onto a Ni-NTA column. The column was then washed with 10 column volumes of Buffer A (binding buffer supplemented with 50 mM imidazole) to remove non-specifically bound proteins. Additionally, 10 column volumes of high-salt buffer (16.7 mM Tris-HCl, pH 7.5, 2 M NaCl) were used to reduce the amount of RNA bound to CBX4 or NEUROD1. The target proteins were eluted with 4 column volumes of binding buffer supplemented with 250 mM imidazole. Finally, all bound proteins were eluted with 4 column volumes of binding buffer supplemented with 500 mM imidazole. The purity of the proteins was verified by Coomassie blue staining and Western blotting. The proteins were then concentrated using Amicon Ultra-15 (Millipore) centrifugal filter units and further purified at room temperature with buffer (25 mM Tris-HCl, pH 7.5, 500 mM NaCl, 10% glycerol). The purity of the proteins was analyzed by SDS-PAGE. The purified proteins were aliquoted, quickly frozen with liquid nitrogen, and stored at −80 °C. Once thawed, the aliquots were not refrozen.

In vitro LLPS assay

The purified mCherry-NEUROD1 and EGFP-CBX4 proteins were used for in vitro droplet formation experiments. In brief, the proteins were added to various droplet formation buffers containing different gradients of NaCl. Buffer I (50 mM Tris-HCl, pH 7.5, 10% glycerol, 1 mM DTT) was used to dilute NaCl. Protein solutions with varying concentrations were prepared and incubated for 10 min in a 96-well plate to promote the formation of protein droplets. The samples were imaged using a Zeiss LSM 800 microscope. Droplets were captured using 594 nm and 488 nm lasers. A 63x oil immersion objective was used.

Statistics and reproducibility

Statistical analyses were carried out using GraphPad Prism 8. Statistical differences were evaluated using two-tailed unpaired Student’s t test for comparisons between two groups or ANOVA and appropriate post hoc analyses for comparisons of more than two groups. For ITT, GTT, random blood glucose, and calcium imaging studies, two-way ANOVA with multiple comparisons was used for statistical analysis. Furthermore, area under curve was calculated for each mouse and evaluated by two-tailed unpaired Student’s t-test for statistical differences between two groups. p value of <0.05 was considered statistically significant. Statistical parameters and corresponding p values for data shown in each panel were included in the figures. Key differentially expressed genes obtained by CUT&Tag were validated via qRT-PCR analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(6.3MB, pdf)}

Reporting Summary^{(4.2MB, pdf)}

Transparent Peer Review file^{(1.5MB, pdf)}

Source data

Source Data^{(122.3KB, xlsx)}

Acknowledgements

We thank Shusen Wang and Human Islet Resource Center (HIRC, China) for providing human islets. We thank Xiaonan Ma for providing technical assistance with the Carl Zeiss LSM 800 microscope on the Public Experimental Platform of China Pharmaceutical University. We thank Yumeng Shen (Public Platform of State Key Laboratory of Natural Medicines, China Pharmaceutical University) for her assistance with flow cytometry analysis.

Author contributions

L.S, Y.L., Q.H., and H.C. performed the experiments; C.X, Y.Q., and K.G. performed some of the animal experiments; X.Z analyzed the human islets of single-cell sequencing. F.H. and J.L. collected the human samples; Y.Z., Y.Y., Y.P., and Y.Z. analyzed the data; H.L., F.Z., and L.J. designed the project; and F.Z. and L.J. interpreted the data and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Natural Science Foundation of China (grant nos. 82370804 and 82100858 to F.F.Z., 82373925 and 82070801 to L.J., 82504329 to J.X.L.); Key Projects of the Xinjiang Production and Construction Corps Natural Science Fund (2025DA013), Jiangsu Province Youth Science and Technology Talent Support Program [JSTJ-2024-104], the Natural Science Foundation of Jiangsu Province (BK20221520 to L.J. and BK20200569 to F.F.Z.); the Shihezi University Carthamus Tinctorius Industry Research Institute Open Project (CZ2024KFG04), the 111 project (grant B16046 to L.J.); the Priority Academic Development Program of Jiangsu Higher Education Institutions (PADP to L.J. and 2632023TD03 to F.F.Z.); the China Postdoctoral Science Foundation (2022T150726 to F.F.Z.); and Postgraduate Research & Practice Innovation Program of Jiangsu (KYCX25_1067, KYCX25_1065).

Data availability

Requests for assistance in obtaining data utilized in this study should be directed to the corresponding author, Dr. Liang Jin (ljstemcell@cpu.edu.cn). The raw data of CUT & Tag were uploaded in the GEO database, and the GEO accession number was GSE289234. We used a scRNA-seq expression matrix with islets isolated from healthy volunteers and patients with T2DM (Data from the HPAP Database, https://hpap.pmacs.upenn.edu/explore/download?matrix). No original code was utilized for the analyses reported herein. The protein mass spectrometry raw data has been updated in the Integrated Proteome Resources (PXD079253, https://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD079253). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Linming Su, Yan Liu, Qianxing Hu.

Contributor Information

Hailiang Liu, Email: hailiang_1111@tongji.edu.cn.

Liang Jin, Email: ljstemcell@cpu.edu.cn.

Fangfang Zhang, Email: zhangff@cpu.edu.cn.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77536-7.

References

1.Sun, H. et al. IDF Diabetes Atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diab. Res Clin. Pract.183, 109119 10.1016/j.diabres.2021.109119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Campbell, J. E. & Newgard, C. B. Mechanisms controlling pancreatic islet cell function in insulin secretion. Nat. Rev. Mol. Cell Biol.22, 142–158 10.1038/s41580-020-00317-7 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Aguayo-Mazzucato, C. et al. Acceleration of β cell aging determines diabetes and senolysis improves disease outcomes. Cell Metab.30, 129–142.e124 10.1016/j.cmet.2019.05.006 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Carapeto, P. et al. Exercise activates AMPK in mouse and human pancreatic islets to decrease senescence. Nat. Metab.6, 1976–1990 10.1038/s42255-024-01130-8 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Li, J. et al. A single-cell transcriptomic atlas of primate pancreatic islet aging. Natl. Sci. Rev.8, nwaa127 10.1093/nsr/nwaa127 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Walker, E. M. et al. Retrograde mitochondrial signaling governs the identity and maturity of metabolic tissues. Science388, eadf2034 10.1126/science.adf2034 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Yang, Y. et al. The mitochondrial enzyme pyruvate carboxylase restricts pancreatic β-cell senescence by blocking p53 activation. Proc. Natl. Acad. Sci. USA121, e2401218121 10.1073/pnas.2401218121 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Zhou, Y. et al. β-Cell miRNA-503-5p induced by hypomethylation and inflammation promotes insulin resistance and β-cell decompensation. Diabetes73, 57–74 10.2337/db22-1044 (2024). [DOI] [PubMed] [Google Scholar]
9.Gregg, T. et al. Pancreatic β-cells from mice offset age-associated mitochondrial deficiency with reduced KATP channel activity. Diabetes65, 2700–2710 10.2337/db16-0432 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Piunti, A. & Shilatifard, A. The roles of Polycomb repressive complexes in mammalian development and cancer. Nat. Rev. Mol. Cell Biol.22, 326–345 10.1038/s41580-021-00341-1 (2021). [DOI] [PubMed] [Google Scholar]
11.Pu, L. & Sung, Z. R. PcG and trxG in plants - friends or foes. Trends Genet.31, 252–262 10.1016/j.tig.2015.03.004 (2015). [DOI] [PubMed] [Google Scholar]
12.Blackledge, N. P. & Klose, R. J. The molecular principles of gene regulation by Polycomb repressive complexes. Nat. Rev. Mol. Cell Biol.22, 815–833 10.1038/s41580-021-00398-y (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Kagey, M. H., Melhuish, T. A. & Wotton, D. The polycomb protein Pc2 is a SUMO E3. Cell113, 127–137 10.1016/s0092-8674(03)00159-4 (2003). [DOI] [PubMed] [Google Scholar]
14.Chen, Q., Huang, L., Pan, D., Zhu, L. J. & Wang, Y. X. Cbx4 sumoylates Prdm16 to regulate adipose tissue thermogenesis. Cell Rep.22, 2860–2872 10.1016/j.celrep.2018.02.057 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Roscic, A. et al. Phosphorylation-dependent control of Pc2 SUMO E3 ligase activity by its substrate protein HIPK2. Mol. Cell24, 77–89 10.1016/j.molcel.2006.08.004 (2006). [DOI] [PubMed] [Google Scholar]
16.Sanyal, S., Mondal, P., Sen, S., Sengupta Bandyopadhyay, S. & Das, C. SUMO E3 ligase CBX4 regulates hTERT-mediated transcription of CDH1 and promotes breast cancer cell migration and invasion. Biochem. J.477, 3803–3818 10.1042/BCJ20200359 (2020). [DOI] [PubMed] [Google Scholar]
17.Chen, Y. H., Zhang, X., Ko, K. Y., Hsueh, M. F. & Kraus, V. B. CBX4 Regulates Replicative Senescence of WI-38 Fibroblasts. Oxid. Med. Cell Longev.2022, 5503575 10.1155/2022/5503575 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Hoffmann, M. J. et al. Expression changes in EZH2, but not in BMI-1, SIRT1, DNMT1 or DNMT3B are associated with DNA methylation changes in prostate cancer. Cancer Biol. Ther.6, 1403–1412 10.4161/cbt.6.9.4542 (2007). [DOI] [PubMed] [Google Scholar]
19.Pan, Y., Li, Q., Cao, Z. & Zhao, S. The SUMO E3 ligase CBX4 is identified as a poor prognostic marker of gastric cancer through multipronged OMIC analyses. Genes Dis.8, 827–837 10.1016/j.gendis.2020.08.010 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Hu, K., Yao, L., Xu, Z., Yan, Y. & Li, J. Prognostic value and therapeutic potential of CBX family members in ovarian cancer. Front. Cell Dev. Biol.10, 832354 10.3389/fcell.2022.832354 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Zhou, J. et al. Prognosis and immune infiltration of chromobox family genes in sarcoma. Front. Oncol.11, 657595 10.3389/fonc.2021.657595 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Song, Y. et al. Identification of susceptibility loci for cardiovascular disease in adults with hypertension, diabetes, and dyslipidemia. J. Transl. Med.19, 85 10.1186/s12967-021-02751-3 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Hu, Q. et al. CBX4 regulation of senescence and associated diseases: molecular pathways and mechanisms. Pharm. Res.215, 107705 10.1016/j.phrs.2025.107705 (2025). [DOI] [PubMed] [Google Scholar]
24.Shapira, S. N., Naji, A., Atkinson, M. A., Powers, A. C. & Kaestner, K. H. Understanding islet dysfunction in type 2 diabetes through multidimensional pancreatic phenotyping: the Human Pancreas Analysis Program. Cell Metab.34, 1906–1913 10.1016/j.cmet.2022.09.013 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Liu, Y. et al. Circular RNA circGlis3 protects against islet β-cell dysfunction and apoptosis in obesity. Nat. Commun.14, 351 10.1038/s41467-023-35998-z (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Bohuslavova, R. et al. NEUROD1 reinforces endocrine cell fate acquisition in pancreatic development. Nat. Commun.14, 5554 10.1038/s41467-023-41306-6 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Zhou, J. et al. iUUCD 2.0: an update with rich annotations for ubiquitin and ubiquitin-like conjugations. Nucleic Acids Res.46, D447–D453 10.1093/nar/gkx1041 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Beauclair, G., Bridier-Nahmias, A., Zagury, J. F., Saïb, A. & Zamborlini, A. JASSA: a comprehensive tool for prediction of SUMOylation sites and SIMs. Bioinformatics31, 3483–3491 10.1093/bioinformatics/btv403 (2015). [DOI] [PubMed] [Google Scholar]
29.Huang, H. T. et al. Ubiquitin-specific proximity labeling for the identification of E3 ligase substrates. Nat. Chem. Biol.20, 1227–1236 10.1038/s41589-024-01590-9 (2024). [DOI] [PubMed] [Google Scholar]
30.Wu, L. et al. CBX4 contributes to HIV-1 latency by forming phase-separated nuclear bodies and SUMOylating EZH2. EMBO Rep.23, e53855 10.15252/embr.202153855 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Vest, R. S., Davies, K. D., O’Leary, H., Port, J. D. & Bayer, K. U. Dual mechanism of a natural CaMKII inhibitor. Mol. Biol. Cell18, 5024–5033 10.1091/mbc.e07-02-0185 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Lin, J. et al. Senolytics prevent caveolar Ca(V) 3.2-RyR axis malfunction in old vascular smooth muscle. Aging Cell22, e14002 10.1111/acel.14002 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Gheibi, S. et al. Reduced expression level of protein phosphatase PPM1E serves to maintain insulin secretion in type 2 diabetes. Diabetes72, 455–466 10.2337/db22-0472 (2023). [DOI] [PubMed] [Google Scholar]
34.LaMoia, T. E. et al. Cytosolic calcium regulates hepatic mitochondrial oxidation, intrahepatic lipolysis, and gluconeogenesis via CAMKII activation. Cell Metab.36, 2329–2340.e2324 10.1016/j.cmet.2024.07.016 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Millette, K., Rodriguez, K., Sheng, X., Finley, S. D. & Georgia, S. Exogenous lactogenic signaling stimulates beta cell replication in vivo and in vitro. Biomolecules12, 215 (2022). [DOI] [PMC free article] [PubMed]
36.Chen, H. et al. Augmented Stat5 signaling bypasses multiple impediments to lactogen-mediated proliferation in human β-cells. Diabetes64, 3784–3797 10.2337/db15-0083 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Zhang, Y. et al. MEHP promotes liver fibrosis by down-regulating STAT5A in BRL-3A hepatocytes. Chemosphere295, 133925 10.1016/j.chemosphere.2022.133925 (2022). [DOI] [PubMed] [Google Scholar]
38.Zhang, Y. et al. Mono-2-ethylhexyl phthalate (MEHP) promoted lipid accumulation via JAK2/STAT5 and aggravated oxidative stress in BRL-3A cells. Ecotoxicol. Environ. Saf.184, 109611 10.1016/j.ecoenv.2019.109611 (2019). [DOI] [PubMed] [Google Scholar]
39.Murakami, T., Inagaki, N. & Kondoh, H. Cellular senescence in diabetes mellitus: distinct senotherapeutic strategies for adipose tissue and pancreatic β cells. Front Endocrinol.13, 869414 10.3389/fendo.2022.869414 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Pavlinkova, G. & Smolik, O. NEUROD1: transcriptional and epigenetic regulator of human and mouse neuronal and endocrine cell lineage programs. Front. Cell Dev. Biol.12, 1435546 10.3389/fcell.2024.1435546 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Osorio, F. G. et al. Splicing-directed therapy in a new mouse model of human accelerated aging. Sci. Transl. Med.3, 106ra107 10.1126/scitranslmed.3002847 (2011). [DOI] [PubMed] [Google Scholar]
42.Hu, P. et al. N-Homocysteinylation of HMGB1/2 promotes corpus cavernosum endothelial senescence in erectile dysfunction. Int. J. Biol. Sci.21, 6723–6744 10.7150/ijbs.119514 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Nehme, J. et al. Converting cell death into senescence by PARP1 inhibition improves recovery from acute oxidative injury. Nat. Aging4, 771–782 10.1038/s43587-024-00627-x (2024). [DOI] [PubMed] [Google Scholar]
44.Ren, X. et al. Maintenance of nucleolar homeostasis by CBX4 alleviates senescence and osteoarthritis. Cell Rep.26, 3643–3656.e3647 10.1016/j.celrep.2019.02.088 (2019). [DOI] [PubMed] [Google Scholar]
45.Gu, Y. et al. CBX4 counteracts cellular senescence to desensitize gastric cancer cells to chemotherapy by inducing YAP1 SUMOylation. Drug Resist. Update77, 101136 10.1016/j.drup.2024.101136 (2024). [DOI] [PubMed] [Google Scholar]
46.Geravandi, S. et al. The Hippo terminal effector YAP boosts enterovirus replication in type 1 diabetes. Nat. Commun.16, 8882 10.1038/s41467-025-64508-6 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Oleson, B. J., Naatz, A., Proudfoot, S. C., Yeo, C. T. & Corbett, J. A. Role of protein phosphatase 1 and inhibitor of protein phosphatase 1 in nitric oxide-dependent inhibition of the DNA damage response in pancreatic β-cells. Diabetes67, 898–910 10.2337/db17-1062 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Zhang, F. et al. The long non-coding RNA βFaar regulates islet β-cell function and survival during obesity in mice. Nat. Commun.12, 3997 10.1038/s41467-021-24302-6 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Andrali, S. S., Qian, Q. & Ozcan, S. Glucose mediates the translocation of NeuroD1 by O-linked glycosylation. J. Biol. Chem.282, 15589–15596 10.1074/jbc.M701762200 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Rao, Y. et al. NeuroD1 induces microglial apoptosis and cannot induce microglia-to-neuron cross-lineage reprogramming. Neuron109, 4094–4108.e4095 10.1016/j.neuron.2021.11.008 (2021). [DOI] [PubMed] [Google Scholar]
51.Zhang, J. et al. Novel CaMKII-δ Inhibitor Hesperadin Exerts Dual Functions to Ameliorate Cardiac Ischemia/Reperfusion Injury and Inhibit Tumor Growth. Circulation145, 1154–1168, 10.1161/circulationaha.121.055920 (2022). [DOI] [PubMed] [Google Scholar]
52.Madreiter-Sokolowski, C. T. et al. Enhanced inter-compartmental Ca(2+) flux modulates mitochondrial metabolism and apoptotic threshold during aging. Redox Biol.20, 458–466, 10.1016/j.redox.2018.11.003 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Yip, M. F. et al. CaMKII-mediated phosphorylation of the myosin motor Myo1c is required for insulin-stimulated GLUT4 translocation in adipocytes. Cell Metab.8, 384–398 10.1016/j.cmet.2008.09.011 (2008). [DOI] [PubMed] [Google Scholar]
54.Li, X. et al. The MDM2-p53-pyruvate carboxylase signalling axis couples mitochondrial metabolism to glucose-stimulated insulin secretion in pancreatic β-cells. Nat. Commun.7, 11740 10.1038/ncomms11740 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Yang, J. et al. Identification of human age-associated gene co-expressions in functional modules using liquid association. Oncotarget9, 1063–1074 10.18632/oncotarget.23148 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Dalgaard, L. T., Billestrup, N. & Nielsen, J. H. STAT5 activity in pancreatic β-cells. Expert Rev. Endocrinol. Metab.3, 423–439 10.1586/17446651.3.4.423 (2008). [DOI] [PubMed] [Google Scholar]
57.Jackerott, M. et al. STAT5 activity in pancreatic beta-cells influences the severity of diabetes in animal models of type 1 and 2 diabetes. Diabetes55, 2705–2712 10.2337/db06-0244 (2006). [DOI] [PubMed] [Google Scholar]
58.Zhou, J. et al. STAT5 and STAT3 balance shapes dendritic cell function and tumour immunity. Nature643, 519–528 10.1038/s41586-025-09000-3 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Saul, D. et al. A new gene set identifies senescent cells and predicts senescence-associated pathways across tissues. Nat. Commun.13, 4827 10.1038/s41467-022-32552-1 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Satija, R., Farrell, J. A., Gennert, D., Schier, A. F. & Regev, A. Spatial reconstruction of single-cell gene expression data. Nat. Biotechnol.33, 495–502 10.1038/nbt.3192 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Scharfmann, R. et al. Development of a conditionally immortalized human pancreatic β cell line. J. Clin. Investig.124, 2087–2098 10.1172/jci72674 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Blanchi, B. et al. EndoC-βH5 cells are storable and ready-to-use human pancreatic beta cells with physiological insulin secretion. Mol. Metab.76, 101772 10.1016/j.molmet.2023.101772 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
63.((2019–2020)) Centers for disease control and prevention. About prediabets & type 2 diabetes. https://www.cdc.gov/diabetes-prevention/about-prediabetes-type-2/index.html.
64.Zhang, F. F. et al. Obesity-induced reduced expression of the lncRNA ROIT impairs insulin transcription by downregulation of Nkx6.1 methylation. Diabetologia63, 811–824 10.1007/s00125-020-05090-y (2020). [DOI] [PubMed] [Google Scholar]
65.Kaya-Okur, H. S. et al. CUT&Tag for efficient epigenomic profiling of small samples and single cells. Nat. Commun.10, 1930 10.1038/s41467-019-09982-5 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(6.3MB, pdf)}

Reporting Summary^{(4.2MB, pdf)}

Transparent Peer Review file^{(1.5MB, pdf)}

Source Data^{(122.3KB, xlsx)}

Data Availability Statement

[CR1] 1.Sun, H. et al. IDF Diabetes Atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diab. Res Clin. Pract.183, 109119 10.1016/j.diabres.2021.109119 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Campbell, J. E. & Newgard, C. B. Mechanisms controlling pancreatic islet cell function in insulin secretion. Nat. Rev. Mol. Cell Biol.22, 142–158 10.1038/s41580-020-00317-7 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Aguayo-Mazzucato, C. et al. Acceleration of β cell aging determines diabetes and senolysis improves disease outcomes. Cell Metab.30, 129–142.e124 10.1016/j.cmet.2019.05.006 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR4] 4.Carapeto, P. et al. Exercise activates AMPK in mouse and human pancreatic islets to decrease senescence. Nat. Metab.6, 1976–1990 10.1038/s42255-024-01130-8 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR5] 5.Li, J. et al. A single-cell transcriptomic atlas of primate pancreatic islet aging. Natl. Sci. Rev.8, nwaa127 10.1093/nsr/nwaa127 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR6] 6.Walker, E. M. et al. Retrograde mitochondrial signaling governs the identity and maturity of metabolic tissues. Science388, eadf2034 10.1126/science.adf2034 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Yang, Y. et al. The mitochondrial enzyme pyruvate carboxylase restricts pancreatic β-cell senescence by blocking p53 activation. Proc. Natl. Acad. Sci. USA121, e2401218121 10.1073/pnas.2401218121 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Zhou, Y. et al. β-Cell miRNA-503-5p induced by hypomethylation and inflammation promotes insulin resistance and β-cell decompensation. Diabetes73, 57–74 10.2337/db22-1044 (2024). [DOI] [PubMed] [Google Scholar]

[CR9] 9.Gregg, T. et al. Pancreatic β-cells from mice offset age-associated mitochondrial deficiency with reduced KATP channel activity. Diabetes65, 2700–2710 10.2337/db16-0432 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Piunti, A. & Shilatifard, A. The roles of Polycomb repressive complexes in mammalian development and cancer. Nat. Rev. Mol. Cell Biol.22, 326–345 10.1038/s41580-021-00341-1 (2021). [DOI] [PubMed] [Google Scholar]

[CR11] 11.Pu, L. & Sung, Z. R. PcG and trxG in plants - friends or foes. Trends Genet.31, 252–262 10.1016/j.tig.2015.03.004 (2015). [DOI] [PubMed] [Google Scholar]

[CR12] 12.Blackledge, N. P. & Klose, R. J. The molecular principles of gene regulation by Polycomb repressive complexes. Nat. Rev. Mol. Cell Biol.22, 815–833 10.1038/s41580-021-00398-y (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Kagey, M. H., Melhuish, T. A. & Wotton, D. The polycomb protein Pc2 is a SUMO E3. Cell113, 127–137 10.1016/s0092-8674(03)00159-4 (2003). [DOI] [PubMed] [Google Scholar]

[CR14] 14.Chen, Q., Huang, L., Pan, D., Zhu, L. J. & Wang, Y. X. Cbx4 sumoylates Prdm16 to regulate adipose tissue thermogenesis. Cell Rep.22, 2860–2872 10.1016/j.celrep.2018.02.057 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.Roscic, A. et al. Phosphorylation-dependent control of Pc2 SUMO E3 ligase activity by its substrate protein HIPK2. Mol. Cell24, 77–89 10.1016/j.molcel.2006.08.004 (2006). [DOI] [PubMed] [Google Scholar]

[CR16] 16.Sanyal, S., Mondal, P., Sen, S., Sengupta Bandyopadhyay, S. & Das, C. SUMO E3 ligase CBX4 regulates hTERT-mediated transcription of CDH1 and promotes breast cancer cell migration and invasion. Biochem. J.477, 3803–3818 10.1042/BCJ20200359 (2020). [DOI] [PubMed] [Google Scholar]

[CR17] 17.Chen, Y. H., Zhang, X., Ko, K. Y., Hsueh, M. F. & Kraus, V. B. CBX4 Regulates Replicative Senescence of WI-38 Fibroblasts. Oxid. Med. Cell Longev.2022, 5503575 10.1155/2022/5503575 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Hoffmann, M. J. et al. Expression changes in EZH2, but not in BMI-1, SIRT1, DNMT1 or DNMT3B are associated with DNA methylation changes in prostate cancer. Cancer Biol. Ther.6, 1403–1412 10.4161/cbt.6.9.4542 (2007). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Pan, Y., Li, Q., Cao, Z. & Zhao, S. The SUMO E3 ligase CBX4 is identified as a poor prognostic marker of gastric cancer through multipronged OMIC analyses. Genes Dis.8, 827–837 10.1016/j.gendis.2020.08.010 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Hu, K., Yao, L., Xu, Z., Yan, Y. & Li, J. Prognostic value and therapeutic potential of CBX family members in ovarian cancer. Front. Cell Dev. Biol.10, 832354 10.3389/fcell.2022.832354 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Zhou, J. et al. Prognosis and immune infiltration of chromobox family genes in sarcoma. Front. Oncol.11, 657595 10.3389/fonc.2021.657595 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Song, Y. et al. Identification of susceptibility loci for cardiovascular disease in adults with hypertension, diabetes, and dyslipidemia. J. Transl. Med.19, 85 10.1186/s12967-021-02751-3 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Hu, Q. et al. CBX4 regulation of senescence and associated diseases: molecular pathways and mechanisms. Pharm. Res.215, 107705 10.1016/j.phrs.2025.107705 (2025). [DOI] [PubMed] [Google Scholar]

[CR24] 24.Shapira, S. N., Naji, A., Atkinson, M. A., Powers, A. C. & Kaestner, K. H. Understanding islet dysfunction in type 2 diabetes through multidimensional pancreatic phenotyping: the Human Pancreas Analysis Program. Cell Metab.34, 1906–1913 10.1016/j.cmet.2022.09.013 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Liu, Y. et al. Circular RNA circGlis3 protects against islet β-cell dysfunction and apoptosis in obesity. Nat. Commun.14, 351 10.1038/s41467-023-35998-z (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Bohuslavova, R. et al. NEUROD1 reinforces endocrine cell fate acquisition in pancreatic development. Nat. Commun.14, 5554 10.1038/s41467-023-41306-6 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Zhou, J. et al. iUUCD 2.0: an update with rich annotations for ubiquitin and ubiquitin-like conjugations. Nucleic Acids Res.46, D447–D453 10.1093/nar/gkx1041 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Beauclair, G., Bridier-Nahmias, A., Zagury, J. F., Saïb, A. & Zamborlini, A. JASSA: a comprehensive tool for prediction of SUMOylation sites and SIMs. Bioinformatics31, 3483–3491 10.1093/bioinformatics/btv403 (2015). [DOI] [PubMed] [Google Scholar]

[CR29] 29.Huang, H. T. et al. Ubiquitin-specific proximity labeling for the identification of E3 ligase substrates. Nat. Chem. Biol.20, 1227–1236 10.1038/s41589-024-01590-9 (2024). [DOI] [PubMed] [Google Scholar]

[CR30] 30.Wu, L. et al. CBX4 contributes to HIV-1 latency by forming phase-separated nuclear bodies and SUMOylating EZH2. EMBO Rep.23, e53855 10.15252/embr.202153855 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Vest, R. S., Davies, K. D., O’Leary, H., Port, J. D. & Bayer, K. U. Dual mechanism of a natural CaMKII inhibitor. Mol. Biol. Cell18, 5024–5033 10.1091/mbc.e07-02-0185 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR32] 32.Lin, J. et al. Senolytics prevent caveolar Ca(V) 3.2-RyR axis malfunction in old vascular smooth muscle. Aging Cell22, e14002 10.1111/acel.14002 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Gheibi, S. et al. Reduced expression level of protein phosphatase PPM1E serves to maintain insulin secretion in type 2 diabetes. Diabetes72, 455–466 10.2337/db22-0472 (2023). [DOI] [PubMed] [Google Scholar]

[CR34] 34.LaMoia, T. E. et al. Cytosolic calcium regulates hepatic mitochondrial oxidation, intrahepatic lipolysis, and gluconeogenesis via CAMKII activation. Cell Metab.36, 2329–2340.e2324 10.1016/j.cmet.2024.07.016 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Millette, K., Rodriguez, K., Sheng, X., Finley, S. D. & Georgia, S. Exogenous lactogenic signaling stimulates beta cell replication in vivo and in vitro. Biomolecules12, 215 (2022). [DOI] [PMC free article] [PubMed]

[CR36] 36.Chen, H. et al. Augmented Stat5 signaling bypasses multiple impediments to lactogen-mediated proliferation in human β-cells. Diabetes64, 3784–3797 10.2337/db15-0083 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Zhang, Y. et al. MEHP promotes liver fibrosis by down-regulating STAT5A in BRL-3A hepatocytes. Chemosphere295, 133925 10.1016/j.chemosphere.2022.133925 (2022). [DOI] [PubMed] [Google Scholar]

[CR38] 38.Zhang, Y. et al. Mono-2-ethylhexyl phthalate (MEHP) promoted lipid accumulation via JAK2/STAT5 and aggravated oxidative stress in BRL-3A cells. Ecotoxicol. Environ. Saf.184, 109611 10.1016/j.ecoenv.2019.109611 (2019). [DOI] [PubMed] [Google Scholar]

[CR39] 39.Murakami, T., Inagaki, N. & Kondoh, H. Cellular senescence in diabetes mellitus: distinct senotherapeutic strategies for adipose tissue and pancreatic β cells. Front Endocrinol.13, 869414 10.3389/fendo.2022.869414 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Pavlinkova, G. & Smolik, O. NEUROD1: transcriptional and epigenetic regulator of human and mouse neuronal and endocrine cell lineage programs. Front. Cell Dev. Biol.12, 1435546 10.3389/fcell.2024.1435546 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Osorio, F. G. et al. Splicing-directed therapy in a new mouse model of human accelerated aging. Sci. Transl. Med.3, 106ra107 10.1126/scitranslmed.3002847 (2011). [DOI] [PubMed] [Google Scholar]

[CR42] 42.Hu, P. et al. N-Homocysteinylation of HMGB1/2 promotes corpus cavernosum endothelial senescence in erectile dysfunction. Int. J. Biol. Sci.21, 6723–6744 10.7150/ijbs.119514 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR43] 43.Nehme, J. et al. Converting cell death into senescence by PARP1 inhibition improves recovery from acute oxidative injury. Nat. Aging4, 771–782 10.1038/s43587-024-00627-x (2024). [DOI] [PubMed] [Google Scholar]

[CR44] 44.Ren, X. et al. Maintenance of nucleolar homeostasis by CBX4 alleviates senescence and osteoarthritis. Cell Rep.26, 3643–3656.e3647 10.1016/j.celrep.2019.02.088 (2019). [DOI] [PubMed] [Google Scholar]

[CR45] 45.Gu, Y. et al. CBX4 counteracts cellular senescence to desensitize gastric cancer cells to chemotherapy by inducing YAP1 SUMOylation. Drug Resist. Update77, 101136 10.1016/j.drup.2024.101136 (2024). [DOI] [PubMed] [Google Scholar]

[CR46] 46.Geravandi, S. et al. The Hippo terminal effector YAP boosts enterovirus replication in type 1 diabetes. Nat. Commun.16, 8882 10.1038/s41467-025-64508-6 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR47] 47.Oleson, B. J., Naatz, A., Proudfoot, S. C., Yeo, C. T. & Corbett, J. A. Role of protein phosphatase 1 and inhibitor of protein phosphatase 1 in nitric oxide-dependent inhibition of the DNA damage response in pancreatic β-cells. Diabetes67, 898–910 10.2337/db17-1062 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR48] 48.Zhang, F. et al. The long non-coding RNA βFaar regulates islet β-cell function and survival during obesity in mice. Nat. Commun.12, 3997 10.1038/s41467-021-24302-6 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Andrali, S. S., Qian, Q. & Ozcan, S. Glucose mediates the translocation of NeuroD1 by O-linked glycosylation. J. Biol. Chem.282, 15589–15596 10.1074/jbc.M701762200 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR50] 50.Rao, Y. et al. NeuroD1 induces microglial apoptosis and cannot induce microglia-to-neuron cross-lineage reprogramming. Neuron109, 4094–4108.e4095 10.1016/j.neuron.2021.11.008 (2021). [DOI] [PubMed] [Google Scholar]

[CR51] 51.Zhang, J. et al. Novel CaMKII-δ Inhibitor Hesperadin Exerts Dual Functions to Ameliorate Cardiac Ischemia/Reperfusion Injury and Inhibit Tumor Growth. Circulation145, 1154–1168, 10.1161/circulationaha.121.055920 (2022). [DOI] [PubMed] [Google Scholar]

[CR52] 52.Madreiter-Sokolowski, C. T. et al. Enhanced inter-compartmental Ca(2+) flux modulates mitochondrial metabolism and apoptotic threshold during aging. Redox Biol.20, 458–466, 10.1016/j.redox.2018.11.003 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR53] 53.Yip, M. F. et al. CaMKII-mediated phosphorylation of the myosin motor Myo1c is required for insulin-stimulated GLUT4 translocation in adipocytes. Cell Metab.8, 384–398 10.1016/j.cmet.2008.09.011 (2008). [DOI] [PubMed] [Google Scholar]

[CR54] 54.Li, X. et al. The MDM2-p53-pyruvate carboxylase signalling axis couples mitochondrial metabolism to glucose-stimulated insulin secretion in pancreatic β-cells. Nat. Commun.7, 11740 10.1038/ncomms11740 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR55] 55.Yang, J. et al. Identification of human age-associated gene co-expressions in functional modules using liquid association. Oncotarget9, 1063–1074 10.18632/oncotarget.23148 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Dalgaard, L. T., Billestrup, N. & Nielsen, J. H. STAT5 activity in pancreatic β-cells. Expert Rev. Endocrinol. Metab.3, 423–439 10.1586/17446651.3.4.423 (2008). [DOI] [PubMed] [Google Scholar]

[CR57] 57.Jackerott, M. et al. STAT5 activity in pancreatic beta-cells influences the severity of diabetes in animal models of type 1 and 2 diabetes. Diabetes55, 2705–2712 10.2337/db06-0244 (2006). [DOI] [PubMed] [Google Scholar]

[CR58] 58.Zhou, J. et al. STAT5 and STAT3 balance shapes dendritic cell function and tumour immunity. Nature643, 519–528 10.1038/s41586-025-09000-3 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR59] 59.Saul, D. et al. A new gene set identifies senescent cells and predicts senescence-associated pathways across tissues. Nat. Commun.13, 4827 10.1038/s41467-022-32552-1 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR60] 60.Satija, R., Farrell, J. A., Gennert, D., Schier, A. F. & Regev, A. Spatial reconstruction of single-cell gene expression data. Nat. Biotechnol.33, 495–502 10.1038/nbt.3192 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR61] 61.Scharfmann, R. et al. Development of a conditionally immortalized human pancreatic β cell line. J. Clin. Investig.124, 2087–2098 10.1172/jci72674 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR62] 62.Blanchi, B. et al. EndoC-βH5 cells are storable and ready-to-use human pancreatic beta cells with physiological insulin secretion. Mol. Metab.76, 101772 10.1016/j.molmet.2023.101772 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR63] 63.((2019–2020)) Centers for disease control and prevention. About prediabets & type 2 diabetes. https://www.cdc.gov/diabetes-prevention/about-prediabetes-type-2/index.html.

[CR64] 64.Zhang, F. F. et al. Obesity-induced reduced expression of the lncRNA ROIT impairs insulin transcription by downregulation of Nkx6.1 methylation. Diabetologia63, 811–824 10.1007/s00125-020-05090-y (2020). [DOI] [PubMed] [Google Scholar]

[CR65] 65.Kaya-Okur, H. S. et al. CUT&Tag for efficient epigenomic profiling of small samples and single cells. Nat. Commun.10, 1930 10.1038/s41467-019-09982-5 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

CBX4 stabilizes NEUROD1 through SUMO-ubiquitin crosstalk to delay β-cell senescence

Linming Su

Yan Liu

Qianxing Hu

Hongjin Chen

Yimeng Qin

Chenying Xu

Kun Gao

Yuncai Zhou

Yue Yang

Jianxing Liu

Yi Pan

Yanfeng Zhang

Fengjiao Huo

Xiaohang Zhou

Hailiang Liu

Liang Jin

Fangfang Zhang

Abstract

Introduction

Results

Downregulation of CBX4 during β-cell senescence

Fig. 1. CBX4 is downregulated in senescent β-cell.

Compromised Cbx4 induces β-cell senescence in vitro and in vivo

Fig. 2. β-cell-specific deletion of Cbx4 aggravates cell senescence and impairs insulin secretion.

Obesity-induced β-cell senescence is rejuvenated by Cbx4

Fig. 3. Senescent β-cell induced by obesity are rejuvenated by Cbx4.

CBX4 interacts with NEUROD1 to attenuate β-cell senescence

Fig. 4. CBX4 interacts with NEUROD1 to delay β-cell senescence.

CBX4 stabilizes NEUROD1 via SIM1-dependent SUMOylation and phase separation-mediated protection from NEDD4

Fig. 5. CBX4 promotes NEUROD1 sumoylation at lysine 38 blocking its ubiquitination via phase separation.

CBX4-NEUROD1 axis programs calcium signaling to maintain mitochondrial function and suppress β-cell senescence

Fig. 6. Cbx4 deficiency prevents mitochondrial function accelerated β-cell dysfunction.

STAT5A downregulates Cbx4 expression during β-cell senescence

Fig. 7. STAT5A reduced Cbx4 expression during β-cell senescence.

Discussion

Fig. 8. Schematic illustration of the mechanism through which the aging-induced reduction of Cbx4 exacerbates β-cell senescence and impairs insulin secretion.

Limitations of the study

Methods

Single-cell RNA sequencing (scRNA-seq) data retrieval

Animal care

Mouse models

MIN6 cells culture

Human EndoC-βH5 cells culture

Isolation and culture of primary islets

Induction of senescent cells model

Transfection of genes

Islet perifusion assay

Insulin secretion assay

Mitochondrial respiration analysis

Measurement of ATP/ADP ratio in primary islets

Mitochondrial membrane potential assay

Mitochondrial calcium imaging of MIN6 cells

Intracellular calcium imaging of MIN6 cells

Senescence-associated β-galactosidase staining of MIN6 cells

Senescence-associated β-galactosidase staining of pancreas

Flow cytometric analysis of islets senescence

Bimolecular fluorescence complementation (BiFC)

Lentiviral production

Lentivirus infection

Pancreatic intra-ductal AAV infusion in mice

Mouse metabolic assays

RNA isolation and qRT-PCR analysis

Western blot analysis

Plasmid and transient transfections

Luciferase assay

Electrophoretic mobility shift assay (EMSA)

Chromatin immunoprecipitation qPCR (ChIP-qPCR) assay

Co-IP assays

Co-IP/MS

CUT&Tag profiling

GO and KEGG pathway enrichment analysis

Immunofluorescence cell staining

Protein purification

In vitro LLPS assay

Statistics and reproducibility

Reporting summary

Supplementary information

Source data