ChREBPβ is the predominant nuclear isoform after prolonged exposure to high glucose
To understand the mechanisms and consequences of the dynamic relationship between activation of ChREBPα and the positive feedback induction of ChREBPβ (Supplementary Fig. 1), it was critically important to devise and use tools that distinguish the two isoforms (Supplementary Fig. 2). Since ChREBPβ is a truncated version of ChREBPα, and lacks the N-terminal low glucose inhibitory (LID) domain11, an antibody directed against an N-terminal epitope identifies only ChREBPα, and an antibody directed against the C-terminal region recognizes both ChREBPα and ChREBPβ (Supplementary Figs. 1–3). We found ChREBP in the nucleus of INS-1-derived 832/13 cells (hereafter INS-1 cells), as expected, using an antibody recognizing the C-terminus and thus both isoforms after 24 h in 20 mM glucose (Fig. 1a), consistent with prior reports17,18. Interestingly, when we used an N-terminal antibody, ChREBPα appeared mostly cytoplasmic, with no discernable difference between low and high glucose (Fig. 1b). Thus, the predominant nuclear ChREBP isoform was ChREBPβ, which seemed at odds with the canonical view that ChREBPα translocates to the nucleus in response to increased glucose metabolism19. To confirm that ChREBPβ was the nuclear isoform, we employed islets from floxed ChREBPβ mice10. Dispersed islet cells transduced with either an adenovirus expressing LacZ, as a control, or Cre recombinase to remove ChREBPβ, were incubated in 6 or 20 mM glucose for 48 h. As seen in Fig. 1c, d, the N-terminal antibody (ChREBPα) was mostly cytoplasmic, even in 20 mM glucose, while the C-terminal antibody showed bright nuclear staining after culture in 20 mM glucose, indicating the nuclear presence of ChREBPβ. Importantly, after Cre-mediated deletion of ChREBPβ, there was no nuclear ChREBP immunolabeling. To examine this further, we performed a time course experiment with adenovirally-expressed Flag-tagged ChREBPα, and an antibody against the Flag epitope in INS-1 cells after exposure to 2 or 20 mM glucose (Fig. 1e, f). Flag-tagged ChREBPα was mostly cytoplasmic when cultured in 2 mM glucose. After replacing the media with 20 mM glucose, Flag-tagged ChREBPα migrated rapidly into the nucleus within 5 min, in concert with previous observations17,18. However, after 3 h in 20 mM glucose ChREBPα appeared cytoplasmic, and remained cytoplasmic at 6 and 24 h (Fig. 1f, g). These studies illustrate that ChREBPα moves transiently into the nucleus in response to glucose stimulation, but rapidly returns (minutes) to the cytoplasm. In contrast, ChREBPβ, induced by high glucose via ChREBPα, becomes both elevated and nuclear.
Fig. 1: ChREBPβ is the main nuclear isoform after prolonged exposure to high concentrations of glucose. a, b Ins-1 cells were cultured in low (2 mM) or high (20 mM) glucose for 24 h. Cells were fixed and immunostained with the C-terminus antibody for ChREBP, which recognizes both ChREBPα and ChREBPβ (a), or the N-terminus antibody for ChREBP, specific for ChREBPα (b). The results are representative of three independent experiments. c, d Islets from floxed ChREBPβ mice were isolated and transduced with adenoviruses expressing LacZ as a control, or Cre recombinase to excise ChREBPβ and cultured in 6 mM or 20 mM glucose. After 48 h, cells were fixed and immunostained with the N-terminal (c) or C-terminal (d) antibody; nuclei were stained with DAPI. Shown is a representative result of 3 independent experiments. e, f Ins-1 cells were transduced with an adenovirus expressing flag-tagged ChREBPα and treated with 2 mM (e) or 20 mM glucose (f) for the indicated times. Cells were fixed and immunostained with an anti-Flag antibody and nuclei were stained with DAPI. The results from (f) were quantified in (g). The results are representative of three independent experiments, *, mean, P < 0.05. h, i Ins-1 cells were cultured overnight in 2 mM glucose and then glucose was added to a total of 20 mM for the indicated times; chromatin was isolated and processed for chromatin immunoprecipitation using antibodies recognizing the C-terminus, the N-terminus, or IgG. DNA was amplified by qPCR using primers specific for regions near the ChoREs on the Pklr (h), or Mlxipl (ChREBPβ) (i) gene promoters. The data are the means ± SEM of the percent input after subtraction of the IgG control. n = 4; *p < 0.05; ***P < 0.0005; ****P < 0.0001 using two-way ANOVA. Full size image
To determine whether ChREBPβ becomes the major isoform bound to DNA, we performed a time course chromatin immunoprecipitation (ChIP) experiment using antibodies directed against either the N-terminus (recognizing ChREBPα only) or the C-terminus (recognizing both isoforms; Supplementary Fig. 3). We selected the mouse Pklr promoter as a target because it contains a well-studied ChoRE5,12,20,21. The recruitment over time of ChREBPα to both the Pklr ChoRE region and the ChREBPβ (Mlxipl) ChoRE region increased for the first 30 min, and sharply decreased and plateaued for the remainder of the experiment (Fig. 1h, i). The recruitment of both isoforms, as determined by the C-terminal antibody, increased and decreased, similar to the signal from the N-terminal antibody, but then increased for the duration of the 18 h experiment. These observations strongly suggested that recruitment of ChREBPβ continued throughout the time course of the experiment, whereas recruitment of the ChREBPα isoform increased only during the first 30 min of glucose treatment, broadly consistent with the experiments in Fig. 1a–f. These results are consistent with a model in which ChREBPβ becomes the major nuclear ChREBP isoform bound to DNA after prolonged exposure to high concentrations of glucose. Together, these observations strongly suggest that ChREBPα is transiently nuclear, and that ChREBPβ is the primary nuclear isoform after prolonged treatment with high concentrations of glucose.
An alternative explanation of the results in Fig. 1 could be that the epitope of the N-terminal antibody (recognizing only ChREBPα) becomes “masked” by the assembly of the transcriptional apparatus. In this scenario, the glucose-dependent recruitment of large co-activator protein complexes might sterically hinder the binding of antibodies, rendering ChREBPα “invisible”. To address this possibility, we used CRISPR/Cas9 editing to add fluorescent tags to the 5′ and 3′ ends of full-length ChREBPα in INS-1 cells. mCherry was attached to the N-terminal LID domain, which identifies ChREBPα exclusively (Red cells, see schematic Fig. 2a). We also generated cells labeled with eGFP on the C-terminus (Green cells), representing both ChREBPα and ChREBPβ, and doubly labeled cells (Red/Green cells), in which ChREBPα appeared red/green, and ChREBPβ appeared green. The cells were FACS sorted and validated via PCR, RT-PCR and Western blots (Supplementary Figs. 4–6). Red and Green cells were cultured in increasing concentrations of glucose (2–20 mM) for 72 h. Live-cell imaging demonstrated cytoplasmic localization for the ChREBPα regardless of glucose concentration, and nuclear localization for ChREBPβ at the highest concentration (Fig. 2b–d; Supplementary Fig. 7), independently confirming the results in the preceding paragraphs.
Fig. 2: Cellular localization of genetically edited and tagged ChREBP isoforms in response to glucose. a Design of labeled ChREBP isoforms. mCherry and eGFP were integrated into the genome using the CRISPR/Cas9 method for gene editing in frame with exon 1a and exon 17. b, c Confocal live imaging was performed on Red or Green INS-1 cells after incubation with the indicated concentrations of glucose for 72 h. d The percent nuclear green fluorescence from c was determined. Data are the means ± SEM, n = 3, *P < 0.05; ****p < 0.0001 using one-way ANOVA. e, f Red/Green INS-1 cells were incubated in 2 or 20 mM glucose or the media was changed from 2 to 20 mM glucose as indicated. Confocal live-image microscopy was performed and the percent nuclear green or red fluorescence was determined. The data presented in (e) represent the aggregate of 3 independent experiments. f Representative images from the indicated times and treatments. DAPI nuclear staining is included in the 2 mM image and excluded in the others for clarity. See also related Supplementary Figs. 4–8 and Supplementary Movie 1. Full size image
We next performed time-lapse confocal microscopy in INS-1 cells bearing fluorescent tags on both ends of ChREBP (Red/Green cells), while culturing the cells in 2 mM or 20 mM glucose, or after transitioning the cells from low to high glucose. We quantified the nuclear localization of red fluorescence (representing ChREBPα) and exclusively green fluorescence (representing ChREBPβ). After culture for 24 h in 2 mM glucose, nearly all of the fluorescence was cytoplasmic (Fig. 2e, f, left panels). Culture in 20 mM glucose for 24 h resulted in predominantly green nuclear fluorescence, indicating ChREBPβ as the principal nuclear form (Fig. 2e, f right panels). The green signal (ChREBPβ) was mostly nuclear throughout the time of acquisition. In contrast, the red signal (ChREBPα) was present in approximately 50% of the nuclei with a more random distribution, illustrating that ChREBP shuttles between the nucleus and the cytoplasm, and that increased glucose accelerates the rate of nuclear entry17. When the glucose in the medium was changed from 2 mM to 20 mM (Fig. 2e, f, middle panels), there was a rapid nuclear localization of ChREBPα (red and green fluorescence) followed by a separation of florescent signals after approximately 30 min, agreeing with our observations in Fig. 1 (and see Supplementary Fig. 8a and Supplementary Movie 1). Interestingly, we found that nuclear localization of either isoform required ongoing protein translation as it was inhibited by cycloheximide (Supplementary Fig. 8b). Together, these results suggested a model wherein ChREBPα is mostly cytoplasmic at low glucose, but with increased glucose metabolism, ChREBPα transiently becomes more nuclear to induce the production of ChREBPβ. ChREBPβ remains predominantly nuclear due to the absence of the LID domain (including the NES) and is more potently and constitutively active when compared to ChREBPα11. With sustained high levels of glucose, this positive feedback mechanism continues to produce more ChREBPβ, ensuring that ChREBPβ becomes the major isoform recruited to ChoREs (Supplementary Fig. 1b).
ChREBPβ expression and nuclear localization correlate with hyperglycemia and diabetes
ChREBPβ was visualized in the nucleus of mouse β-cells in vivo, using our validated C-terminal antibody approach under quiescent, metabolically stressful, and diabetic conditions (Fig. 3a). The N-terminal antibody, representing ChREBPα only, was mostly cytoplasmic in all the conditions tested (Fig. 3a). In C57BL/6 mice, we observed a gradation of ChREBPβ expression: no detectable expression on a standard chow diet; modest abundance after one week on a high-fat diet, representing a normal physiological adaptive response10, and very high expression levels in diabetic db/db mice (Fig. 3a). In addition, human islets labeled with a RIP-ZsGreen-expressing adenovirus were sorted to obtain pure β-cells22. The ratio of ChREBPβ to ChREBPα mRNA expression was significantly higher in subjects with T2D compared to non-diabetic control donors (Fig. 3b). Furthermore, we found that treatment of db/db mice with adipsin, which preserves β-cell mass in diabetic db/db mice23, decreased ChREBPβ abundance in a manner proportionate to the improved glycemia and plasma insulin levels, concordant with the idea that ChREBPβ expression contributes to the glucose toxicity seen in the db/db mouse model of T2D (Fig. 3c–f).
Fig. 3: ChREBPβ expression and nuclear localization correlates with adaptive expansion of β-cells and with glucose toxicity in diabetes. a The N-terminal or the C-terminal antibodies recognizing ChREBP were used to stain pancreatic tissue slices from C57Bl/6 mice fed on a standard chow diet, or fed a high-fat diet for 1 week, or from db/db diabetic mice. All micrographs represent at least 3 independent experiments. b Ratio of expression of ChREBPβ to ChREBPα FPKM from RNAseq performed from FACS-sorted human β-cells isolated from non-diabetic (n = 4) and Type 2 diabetic subjects (n = 7). Each data point represents a different donor. Data are means ± SEM; *p < 0.05 using one-way ANOVA. c, d The N-terminal or C-terminal antibody recognizing ChREBP were used to immunolabel pancreatic tissue slices from diabetic db/db mice treated with control AAV (GFP) or with AAV expressing adipsin for 24 weeks. Correlation between percentages of nuclear ChREBPβ in insulin-positive β-cells and blood glucose levels (e) or blood insulin levels (f), where each data point represents an individual mouse. All micrographs represent at least 3 independent experiments. Full size image
We next explored nuclear localization of ChREBPβ in human β-cells in vivo using a minimal human islet transplant model in immunosuppressed mice24. In this experiment, human islets were transduced with adenoviruses and then transplanted under the kidney capsules of streptozotocin-induced diabetic immunocompromised mice (Fig. 4a). 1500 islet equivalents (IEQs) were sufficient to normalize glucose levels (Fig. 4b). 500 IEQs transduced with a control Cre-expressing adenovirus was a minimal mass of β-cells sufficient to keep the animals alive, but hyperglycemic (around 400 mg/dL). When 500 IEQs were transduced with an adenovirus expressing ChREBPα, which activates the Nrf2 antioxidant pathway8, blood glucose, plasma insulin, and glucose tolerance all approached normal levels (Fig. 4b–e). At the end of the experiment, a uninephrectomy was performed and glucose levels rose to diabetic levels, confirming that the transplanted human β-cells provided the only insulin in the recipient mice (Fig. 4b). Kidneys containing islet grafts were fixed and immunolabeled for insulin and ChREBP using the C-terminal antibody (Fig. 4f). This revealed abundant nuclear ChREBPβ in β-cells transduced with control, Cre-expressing adenovirus, but nuclear labeling was nearly absent in the ChREBP-α-treated β-cells. In addition, there was a strong correlation between glucose levels prior to the harvesting of the graft and nuclear ChREBPβ abundance (Fig. 4g). Thus, nuclear ChREBPβ expression is proportionate to glucose levels in human β-cells in vivo.
Fig. 4: ChREBPβ nuclear localization correlates with blood glucose levels in β-cells from transplanted human islets. a Schematic of experimental design using the STZ-diabetic immunocompromised marginal islet mass model (Created with BioRender.com). Three groups of mice each received human islets from the same four to five human islet donors: Cre-transduced 500 IEQs (n = 4), ChREBPα-transduced 500 IEQs (n = 5), or 1500 untreated control IEQs (n = 5). The 500 IEQ groups were treated with an adenovirus expressing either Cre as a negative control or ChREBPα 24 h prior to transplantation. Glucose tolerance test was performed at day 42, a unilateral nephrectomy was performed at day 43. b Glucose levels were measured. c Circulating insulin measured using a human insulin-specific assay. d Intraperitoneal glucose tolerance test. e Area under the curve (AUC) for the three groups in (d). Values are means ± SEM. *p < 0.05; **p < 0.01 compared to 1500 IEQ using two-way ANOVA. f Representative images of insulin and C-terminal ChREBP immunohistochemistry from islet grafts of at least 3 different mice. g Correlation between percentages of nuclear ChREBPβ from at least 1000 insulin-positive β-cells from the grafted human islets and blood glucose levels, where each data point represents an individual transplanted mouse. Full size image
Whereas ChREBPβ is required for adaptive β-cell proliferation and expansion of β-cell mass, depletion of ChREBPβ protects from glucolipotoxicity
To more deeply explore the physiological role of ChREBPβ in adult pancreatic β-cells, floxed ChREBPβlox/lox mice were crossed with MIP-Cre-ERT mice to generate a β-cell-specific tamoxifen-inducible knock out of ChREBPβ (iβKOβ)10,25, (Fig. 5a and Supplementary Fig. 9, 10). iβKOβ mice were injected with tamoxifen or oil for 5 consecutive days (75 μg/g body weight). Two days later, they were placed on either a control chow diet or a high-fat diet (HFD) for either one or four weeks. As expected, the oil-injected mice showed nuclear ChREBP labeling with the C-terminal antibody after a high-fat diet, whereas this labeling was not observed in the tamoxifen-injected mice, confirming the high knock-out efficiency in iβKOβ mice, and supporting the notion that ChREBPβ becomes nuclear in vivo in response to high-fat feeding (Supplementary Fig. 9a). Three weeks after the first tamoxifeninjection, mice on a HFD displayed a significant increase in non-fasting blood glucose levels compared to vehicle-treated littermate controls at 3 weeks (Fig. 5b). After one week, glucose tolerance was impaired on a HFD, but there was no significant difference between the oil and tamoxifen-injected groups (Fig. 5c, d). However, Ki67 immunolabeling in β-cells after one week on a HFD was significantly lower in tamoxifen-injected iβKOβ mice compared to littermate controls, demonstrating the necessity for ChREBPβ for adaptive β-cell proliferation (Fig. 5e, f). After 1 month on a HFD, the tamoxifen-injected mice became more glucose intolerant compared with littermate oil-injected mice (Fig. 5g, h). Concordantly, β-cell mass was significantly lower in tamoxifen-injected iβKOβ mice fed a HFD for a month compared to iβKOβ mice injected with oil and fed a HFD, but was not different from β-cell mass in chow-fed control mice (Fig. 5i). Plasma insulin levels remained the same in mice on a HFD diet despite the increased glucose levels (Fig. 5j). Female mice were largely protected from the effects of ChREBPβ depletion (Supplementary Fig. 10c–g). Together, these findings illustrate that ChREBPβ is necessary for adaptive β-cell mass expansion.
Fig. 5: ChREBPβ is required for adaptive β-cell proliferation and expansion of β-cell mass, and loss of ChREBPβ prevents glucolipotoxicity. a Schematic showing generation of β-cell specific, inducible ChREBPβ knockout mice (iβKOβ, created with BioRender.com). b Blood glucose levels after the indicated treatments and times. c, d Glucose tolerance test and area under the curve measurements after 1 week on a chow or high-fat diet (HFD). e, f Percent Ki67-postive/insulin-positive cells in pancreata from iβKOβ mice after 1 week of HFD. g, h Glucose tolerance test and area under the curve measurements after 1 month on a chow or high-fat diet (HFD). i, j β-cell mass and Plasma insulin were measured after 1 month on a chow or high-fat diet (HFD). Data are means ± SEM; N = 4 or 5 as indicated; *p < 0.05; **p < 0.01 using two-way ANOVA. k Islets were isolated from floxed ChREBPβ mice, dispersed and transduced with adenoviruses expressing either GFP or Cre recombinase. Cells were cultured as indicated and subjected to a TUNEL assay after 48 h and immunostained for insulin and DAPI. l Percentage of insulin-positive/TUNEL-positive cells. Data are means ± SEM; N = 7; *p < 0.05; **p < 0.01 using two-way ANOVA. Full size image
We repeated these experiments, crossing ChREBPβlox/lox mice with INS-1-CreHerr mice26, which express Cre under control of the insulin promoter at or near embryonic day 8.5, to generate an embryonic β-cell-specific knock out of ChREBPβ (eβKOβ), and performed the same set of experiments on two-month-old mice (Supplementary Figs. 11–13). These experiments recapitulated the results in iβKOβ mice, showing that ChREBPβ was necessary for β-cell expansion after a HFD in male (but not female) mice. Thus, ChREBPβ is dispensable for normal β-cell development under non-metabolically-stressed conditions, but is required for adaptive proliferation and expansion of β-cell mass in response to a HFD in adult male mice.
To explore underlying mechanisms, mRNA was isolated from islets of male eβKOβ mice fed one week on a chow or HFD (Supplementary Fig. 12e–g). Whereas the knockdown efficiency of ChREBPβ in the Cre-positive mice both on chow or HFD is evident and significant, ChREBPα remained unchanged between Cre-negative and Cre-positive groups, as did β-cell markers, consistent with the idea that ChREBPβ is not necessary for β-cell differentiation or maintenance of β-cell phenotype. Myc, a cell cycle regulator required for adaptive expansion of β-cell mass, and an essential factor for ChREBP activity21,27,28, was downregulated in both chow and HFD after depletion of ChREBPβ (Supplementary Fig. 12g). This suggests that the lack of proliferation seen in Cre-positive iβKOβ and eβKOβ β-cells may be caused by a failure to induce Myc in response to the HFD. Altogether, the data from the iβKOβ and eβKOβ mice demonstrate that ChREBPβ plays a key role in adaptive β-cell proliferation but is unlikely to play a role in normal pre- and postnatal β-cell expansion during development.
Depletion of ChREBPβ protects against glucolipotoxicity
To further explore whether ChREBPβ might play a role in glucolipotoxicity, we cultured dispersed islet cells from ChREBPβlox/lox mice after transduction with adenoviruses expressing either control GFP or Cre recombinase in low glucose or high glucose plus palmitate (Fig. 5k, l). Culturing islet cells in glucolipotoxic conditions led to marked increases in β-cell death as assessed using TUNEL assay (average ~50%). Strikingly, deletion of ChREBPβ completely prevented cell death. Thus, although transient increases in ChREBPβ are necessary for adaptive β-cell expansion, sustained increases in ChREBPβ are a key driver of glucolipotoxic β-cell death.
ChREBPβ overexpression in vivo results in β-cell death and diabetes
To test if overexpressing ChREBPβ in β-cells caused β-cell death in vivo, MIP-Cre-ERT mice were crossed to mice containing a Lox-Stop-Lox Flag-tagged ChREBPβ cassette residing in the Rosa26 locus, termed iβOEβ mice (Fig. 6a, Supplementary Fig. 14). Cre-mediated recombination resulted in the expression of flag-tagged ChREBPβ, confirmed by immunoblots and RT-PCR. Tamoxifen-mediated recombination was restricted to β-cells (Supplementary Fig.. 14b–d). Furthermore, immunostaining of pancreata displayed a marked induction of Cre in insulin-positive β-cells from the iβOEβ mice after tamoxifen treatment (Supplementary Fig. 14e).
Fig. 6: Overexpression of ChREBPβ leads to β-cell death, glucose intolerance and diabetes. a LSL-ChREBPβ mice were bred with MIP-Cre-ERT mice to generate inducible β-cell-specific overexpressing ChREBPβ mice, termed iβOEβ (created with BioRender.com; see also Supplementary Fig. 9). Presented are measurements from male mice (see also Supplementary Fig. 10 for female mice). b, c Glucose tolerance test and area under the curve (AUC) one week after vehicle oil or tamoxifen treatment. d Weekly measures of whole body weight. e Plasma Insulin levels after one week. f Non-fasting blood glucose levels for the indicated times and treatment groups. g β-cell mass (BCM) one week after vehicle oil or tamoxifen treatment. h, i Glucose tolerance test and area under the curve (AUC) one month after vehicle oil or tamoxifen treatment. j β-cell mass (BCM) one month after vehicle oil or tamoxifen treatment. Data are means ± SEM, N = 4–6; *p < 0.05; **P < 0.001; ****P < 0.0001 using two-way ANOVA. k, l Percent Ki67-postive and insulin-positive cells in pancreata from male homozygous iβOEβ mice 1 week after oil or tamoxifen injection. m, n Percent TUNEL-positive and insulin-positive cells in pancreata from male homozygous iβOEβ mice 1 week after oil or tamoxifen treatment. Data are means ± SEM from five mice/group. *p < 0.05 using two-way ANOVA. Full size image
Seven days after tamoxifen-injection, iβOEβ male mice displayed no significant change in glucose tolerance, body weight, plasma insulin or non-fasting glucose (Fig. 6b–f). However, heterozygotes and homozygotes displayed significant decreases in β-cell mass after one week (Fig. 6g). Strikingly, within 30 days after the last injection of tamoxifen, male iβOEβ mice became diabetic, evidenced by increased non-fasting blood glucose levels, impaired glucose tolerance and a marked decrease in β-cell mass, all in a gene-dose-dependent manner (Fig. 6f, h–j). Induction of ChREBPβ was associated with a significant increase in Ki67 staining, and a concomitant increase in TUNEL staining, perhaps reflecting simultaneous attempts at β-cell proliferation and cell death, or DNA damage repair activity (Fig. 6k–n). By contrast, females were partially protected, with only homozygous mice displaying significant glucose intolerance one month after tamoxifen treatment, (Supplementary Fig. 15). Thus, inducible overexpression of ChREBPβ in mice mimics glucose toxicity with β-cell destruction resulting in diabetes.
To determine whether ChREBPβ impacts β-cell function if overexpressed early in development, Lox-Stop-Lox Flag-tagged ChREBPβ mice were crossed with INS-1-CreHerr mice26, to generate embryonically expressed β-cell-specific knock-in flag-tagged ChREBPβ mice, termed eβOEβ mice (Supplementary Figs. 16a–c). At 3 weeks of age, body weight was not significantly different between the Cre-negative and Cre-positive littermates, and only homozygous eβOEβ mice showed a significant increase in non-fasting or fasting blood glucose levels (Supplementary Fig. 16d, e). By 8 weeks both heterozygous and homozygous eβOBβ Cre-positive mice displayed increased non-fasting glucose levels (Supplementary Figs. 16e). At 8 weeks of age, the male mice overexpressing either one or two copies of ChREBPβ had severely impaired glucose tolerance, increased fasting and non-fasting glucose levels, and homozygous animals had severe diabetes and weight loss (Supplementary Fig. 16d–h). Concordantly, insulin levels were significantly decreased in homozygous mice, and β-cell mass was nearly absent for both heterozygous and homozygous mice (Supplementary Fig. 16i, j). Furthermore, the islet architectures of the Cre-positive mice are clearly disturbed with necrotic centers evident within islets (Supplementary Fig. 17a), with no obvious difference in somatostatin or glucagon immunolabeling, and normal alpha cell mass (Supplementary Fig. 17b). The overall graded response between heterozygotes and homozygotes demonstrated a gene dosage effect. Insulin tolerance was very similar between the Cre-negative and Cre-positive groups, indicating that the diabetic phenotype was due to catastrophic loss of β-cell mass rather than insulin intolerance in peripheral tissues (Supplementary Fig. 17c–g; see also Supplementary Movies 2 and 3). Results in female eβOEβ mice were similar to those in male mice (Supplementary Fig. 18). Thus prolonged and very high overexpression of ChREBPβ leads to β-cell death, mimicking glucotoxicity.
Overexpression of ChREBPβ promotes a signature of increased proliferation, apoptosis, and dedifferentiation
To more deeply explore the relationship between ChREBPβ overexpression and β-cell proliferation versus death, we performed RNAseq using INS-1 cells transduced with a control adenovirus (GFP) or an adenovirus overexpressing ChREBPβ and cultured for 48 h in 2 mM or 11 mM glucose (Fig. 7). Six differential gene expression (DGE) analyses were conducted to unbiasedly compare all possible combinations between the 4 groups. Gene Ontology (GO) terms enriched by differentially expressed genes (Supplementary Table 4) from each DGE were processed using the ViSEAGO R package that helps capture the biological background from the complexity of the experimental design with multiple comparisons29. ViSEAGO computed the semantic similarity between 471 enriched GO terms, and identified 45 clusters, and 5 superclusters of GO terms (Fig. 7a). Supercluster (1) includes clusters populated by apoptosis, cell death and proliferation clusters. Other superclusters include transmembrane transport (2) regulation of metabolic processes (3 and 4) and cell differentiation (5). Figure 7b depicts a volcano plot comparing GFP and ChREBPβ-treated INS-1 cells cultured in 11 mM glucose. GO pathway analysis revealed that the top two pathways associated with ChREBPβ overexpression in 11 mM glucose were proliferation and apoptosis (Fig. 7c). Nearly every apoptosis marker was upregulated by ChREBPβ, both in 2 mM and 11 mM glucose (Fig. 7d). In addition, Txnip, which drives β-cell glucose toxicity and is a major target gene of ChREBP30, was highly upregulated by ChREBPβ in both 2 mM and 11 mM glucose (Fig. 7e). Key cell cycle regulator genes were generally induced by increased glucose, and overexpression of ChREBPβ led to an even greater increase in their expression (Fig. 7f). By contrast, β-cell identity markers were markedly decreased by ChREBPβ overexpression (Fig. 7g). Thus, overexpression of ChREBPβ behaves much like overexpression of Myc, with a signature supporting both proliferation and apoptosis in β-cells, and with a general effect of decreasing β-cell identity while increasing the transcription of most of the genes examined31,32.
Fig. 7: ChREBPβ initiates programs of cell cycle regulation and apoptosis when overexpressed in Ins-1 cells. Ins-1 cells were transduced with an adenovirus expressing GFP or ChREBPβ and cultured for 48 h in 2 or 11 mM glucose. RNA was collected and RNA-seq was performed. a Heatmap of the cluster of GO terms. The 471 GO terms (pathways) enriched by differentially expressed genes across the 6 DGEs from the 4 groups (see text) are grouped into 45 clusters of GO terms. The hierarchical tree scoring the distances between each of the 45 clusters originates superclusters that are boxed in red. The superclusters highlight the main cellular functions common to the different DGEs. b, c Volcano plot and gene ontology of biological processes affected by ChREBPβ compared to GFP-treated INS-1 cells cultured in 11 mM glucose. d Heat maps of the average expression levels of apoptosis regulators e Txnip mRNA expression; means of 3 independent measures, error bars are SEM; *p < 0.05. f, g Heat maps of the averages of differentially expressed cell cycle activators or β-cell marker genes. Values represent the average of three independent experiments and genes are sorted by the Pierson correlation of nearest neighbors. Full size image
Rescue from ChREBPβ-mediated β-cell death by ChREBPα and Nrf2
Since overexpression of ChREBPα does not result in β-cell death, but rather augments glucose-stimulated β-cell proliferation via activation of the antioxidant Nrf2 pathway5,8, we queried whether co-expression of ChREBPα with ChREBPβ could rescue β-cells from ChREBPβ-mediated cell death. We first tested the hypothesis that the ratio of ChREBPα:ChREBPβ is an important determinant of β-cell apoptosis. INS-1 cells were transduced with a constant MOI (150) of adenoviruses expressing either ChREBPα or ChREBPβ. Increasing MOIs of each virus were compared, using LacZ-expressing adenovirus to maintain a constant viral load (Fig. 8a). Cell death, as measured by Annexin V staining and flow cytometry, was evident when the ratio of ChREBPβ to ChREBPα was greater than one. In addition, titration of ChREBPα into ChREBPβ-expressing cells reduced β-cell death. We next isolated islets from Lox-stop-Lox ChREBPβ mice, and induced ChREBPβ using a Cre adenovirus in the absence or presence of an adenovirus expressing ChREBPα, and then measured β-cell death after 48 h by counting TUNEL- and insulin-positive cells. There was almost universal apoptosis in cells overexpressing ChREBPβ, but co-expression of ChREBPα with ChREBPβ rescued β-cell death (Fig. 8b, c). Overexpression of ChREBPα activates the antioxidant Nrf2 pathway8. Thus, we explored whether CDDO-Me, an Nrf2 activator, might also rescue β-cells from isolated ChREBPβ-mediated cell death using islets from Lox-stop-Lox ChREBPβ mice. This proved to be true (Fig. 8b, c). Thus, overexpression of ChREBPα or activation of Nrf2 rescues murine β-cells from the cytotoxicity of ChREBPβ overexpression.
Fig. 8: ChREBPα and activation of Nrf2 mitigates ChREBPβ-mediated β-cell death. a The indicated amounts of ChREBPα and ChREBPβ were transduced into INS-1-derived 832/13 cells and 48 h later apoptosis was measured with Annexin V staining. Error bars are SEM, N = 3, *p < 0.05. b Islets were isolated from Lox-Stop-Lox ChREBPβ mice, dispersed and transduced with the indicated adenovirus or 10 µM CDDO-Me for 48 h. Cells were immunostained or processed for TUNEL assay as indicated. c Percent of TUNEL+/insulin+ cells from (b). d Human islets were dispersed and transduced with the indicated adenovirus or 10 µM CDDO-Me for 48 h and immunostained as indicated. e Percent of TUNEL+/insulin+ cells from (d). f Percent of Ki67+/insulin+ cells from (d). Data are the means ± SEM, n = 3–4, *p < 0.05, **p < 0.01 using one-way ANOVA; ns not significant. Full size image
We asked if ChREBPα could rescue ChREBPβ-mediated β-cell death in human β-cells. ChREBPβ was overexpressed using an adenovirus in dispersed human islets. Overexpression of ChREBPβ resulted in pronounced β-cell apoptosis, as assessed by insulin immunolabeling and TUNEL assay (Fig. 8d, e). We also immunolabeled the same cells for Ki67 and observed a marked Ki67 labeling in β-cells, likely reflecting DNA damage in dying β-cells [Fig. 8f; refs. 2,33]. Overexpression of ChREBPα had no effect on either proliferation or cell death. However, co-expression of ChREBPβ and ChREBPα resulted in a complete absence of cell death, but retention of robust β-cell proliferation. In separate experiments, ChREBPβ-transduced human islet cells were treated with CDDO-Me, an Nrf2 activator (Fig. 8e, f). CDDO-Me reduced ChREBPβ-mediated TUNEL staining and induced or permitted robust Ki67 immunolabeling, strongly suggesting that the rescue effect of ChREBPα was at least in part due to activation of the Nrf2 antioxidant pathway.