Excitotoxicity and Overnutrition Additively Impair Metabolic Function and Identity of Pancreatic β-Cells

The loss of β-mass and function in response to metabolic stress is a major determinant of type 2 diabetes (T2D) (1). While multiple mechanisms, including glucolipotoxicity, excitotoxicity, inflammation, endoplasmic reticulum (ER) stress, and oxidative stress (2–5), have been implicated in metabolic stress–induced β-cell failure, the molecular and cellular mechanisms that actually cause the loss of β-cell function and development of T2D are not understood.

Excitotoxicity refers to the pathological process in excitable cells in which overstimulation leads to a sustained increase in intracellular Ca²⁺ concentration ([Ca²⁺]_i), resulting in disrupted homeostasis, loss of cell function, or cell death (6). In pancreatic β-cells, metabolic stress–induced increases in [Ca²⁺]_i activate Ca²⁺/calmodulin-dependent kinases (CaMKs), calcineurin (a Ca²⁺-dependent phosphatase), and other Ca²⁺-dependent proteins, causing alterations in β-cell gene expression that negatively impact both β-cell mass and function (7). Increases in [Ca²⁺]_i have been described in rat islets cultured in high glucose (8), in mouse islets from obese (db/db) mice (9), in islets of mice fed a high-fat diet (HFD) (10), and in β-cells that exhibit chronic membrane depolarization (11). Both the verapamil-induced blockage of Ca²⁺ influx (12) and genetic knockdown of Cavβ3, a Ca²⁺ channel subunit (13), by reducing [Ca²⁺]_i, attenuate β-cell loss and diabetes in animal models, suggesting that an increase in [Ca²⁺]_i is a fundamental determinant of stress-induced β-cell failure.

Overnutrition, by elevating circulating free fatty acids (FFAs), contributes to insulin resistance, increasing insulin biosynthesis and secretion (14). The prolonged exposure of β-cells to FFAs elicits multiple responses, including the activation of ER stress, oxidative stress, and inflammatory signaling pathways (15,16). Most notably, FFA-induced oxidative stress triggers the release of Ca²⁺ from ER stores, increasing [Ca²⁺]_i, accentuating ER stress, and inducing apoptosis (17).

Sex also influences the response of β-cells to stress (18). Women are less likely than men to develop T2D and require a higher BMI to do so (19). Increased estrogen receptor signaling (20), sex-specific differences in islet DNA methylation status (21), and differences in the expression of islet-enriched transcription factors (TFs) and genes involved in cell cycle regulation (22) have all been suggested as causes for these differences.

To obtain a systems-wide understanding of the effects of both excitotoxicity and overnutrition on β-cell function and gene expression, we used mice lacking Abcc8, a critical subunit of the ATP-dependent K⁺ channel (K_ATP). Previously, we have shown that β-cells from these mice exhibit chronic membrane depolarization and increases in [Ca²⁺]_i that cause impairments in islet morphology, glucose tolerance, and β-cell identity (11,23). Because the loss of β-cell function in mice lacking Abcc8 develops slowly over several months (23), the individual and combined effects of excitotoxicity and overnutrition on β-cell function and gene expression were determined prior to the onset of hyperglycemia and glucotoxicity (11). Additionally, by using a recently described Ins2^Apple allele, we avoided confounding effects of the MIP-GFP transgene (22).

The Abcc8^tm1.1Mgn (23) and Ins2^Apple (22) alleles were bred into and maintained as C57BL/6J congenic lines (stock 000664; The Jackson Laboratory). At weaning (3–4 weeks of age), mice were fed either regular chow (RC) (4.5% fat content) (5L0D; PicoLab) or HFD (60% fat content) (D12492; Research Diets, Inc.) for 5 weeks. Verapamil (1 mg/mL) (V4629; Sigma-Aldrich) was administered through the drinking water during the period of HFD feeding. All animal experimentation was performed under the oversight of the Vanderbilt University Institutional Animal Care and Use Committee.

Intraperitoneal glucose tolerance tests (GTTs) were performed following a 16-h overnight fast. Blood glucose concentrations were measured at 0, 15, 30, 60, and 120 min after administering d-glucose (2 mg/g body mass). Insulin tolerance testing was performed following a 4-h morning fast by administering 0.1 units/mL insulin (in Dulbecco’s PBS) (Humulin R; Eli Lilly and Company) and measuring blood glucose concentrations at 0, 15, 30, 60, and 120 min.

Islets were isolated following injection of 0.6 mg/mL Collagenase P (Roche) into the pancreatic bile duct followed by Histopaque-1077 (Sigma-Aldrich) fractionation and handpicking. For FACS and RNA sequencing, islets from two to four mice were pooled for each sample. Islets were cultured in low-glucose DMEM (11966–025; Gibco) containing 1 g/L glucose and supplemented with 10% FBS and penicillin/streptomycin (100 mg/mL) (Gibco) at 37°C with 5% CO₂ infusion and 95% humidity. Palmitic acid (PA) (P0500; Sigma-Aldrich) was diluted in 50% ethanol to 100 mmol/L and conjugated to an FA-free BSA (A6003; Sigma-Aldrich) to generate a 5 mmol/L PA/5% FA-free BSA stock solution. Experimental media concentrations of the compounds used were 100 μmol/L for tolbutamide (T0891; Sigma-Aldrich), 0.5 mmol/L for PA, and 50 μmol/L for verapamil (V4629; Sigma-Aldrich).

Purified β-cells were obtained as previously described (15) in which live cells expressing red fluorescence were sorted with a 100-μm nozzle using the FACSAria II instrument (BD Biosciences). Cells were collected in chilled Homogenization Solution from the Maxwell 16 LEV simplyRNA Tissue Kit (TM351; Promega), and RNA was isolated as directed. For quantitative PCR (qPCR), reverse transcription was done using a High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). A total of 2 ng cDNA was used in real-time qPCR with Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) using a CFX96 Real-Time PCR system (Bio-Rad Laboratories). Primers are listed in Supplementary Table 1.

RNA samples were analyzed using an Agilent 2100 Bioanalyzer, and only those samples with an RNA integrity number of seven or above were used. cDNA synthesis and amplification were performed using the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Takara Bio, Inc.) using 10 cycles of PCR. cDNA libraries were constructed using the Low Input Library Prep Kit (Takara Bio, Inc.). An Illumina NovaSeq 6000 instrument was used to produce paired-end, 150-nucleotide reads for each RNA sample. Paired-end sequencing of 31 samples produced ∼1.55 billion raw sequencing reads. The Spliced Transcripts Alignment to a Reference (STAR) application (16) was used to perform sequence alignments to the mm10 (GRCm38) mouse genome reference and GENCODE comprehensive gene annotations (release M17). Overall, 80–88% of the raw sequencing reads were uniquely mapped to genomic sites, resulting in 1.3 billion usable reads. HTSeq was used for counting reads mapped to genomic features (17), and DESeq2 was used for differential gene expression analysis (18). P_adj <0.05 cutoff was used to define differentially expressed genes. Gene ontology (GO) analysis of differentially expressed genes was performed using Metascape (19).

Oxygen consumption rates (OCRs) of isolated islets were determined using a Seahorse XF96 respirometer (Agilent Technologies), as described (24). A total of 10–20 islets/well were loaded onto a Cell-Tak (Corning) precoated XF96 spheroid plate (Agilent Technologies) and preincubated for 2 h at 37°C without CO₂ in a Seahorse assay DMEM (Agilent Technologies) supplemented with 3 mmol/L glucose, 1 mmol/L pyruvate, and 2 mmol/L glutamine. After measuring basal OCR, glucose (20 mmol/L), oligomycin (5 mmol/L), carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone (FCCP) (1 mmol/L), or antimycin A/rotenone (2.5 mmol/L) was added at indicated time points to modulate mitochondrial OCR response. For mtDNA copy number measurements, islet DNA was isolated with a DNeasy kit (Qiagen) and analyzed by qPCR with primers for mtDNA-encoded gene mt-CO1 and nuclear gene Ndufv1. Relative copy number was calculated as 2^([Ct(mtDNA) * E] − [Ct(nuclear DNA) * E]), where E is efficiency of corresponding qPCR determined from standard curves and Ct is threshold cycle.

RNA-sequencing data are available in ArrayExpress (https://www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-8643.

To compare the effects of excitotoxicity and overnutrition on pancreatic β-cells, we fed C57BL/6J (wild-type [WT]) and Abcc8 knockout (KO) mice either RC or HFD for 5 weeks. Both groups contained male and female animals (n = 7–8 of each sex), and all animals gained weight on HFD (Fig. 1A). Blood glucose measurements after 5 weeks showed that the KO animals had lower fasting blood glucose in comparison with WT mice (Fig. 1B), consistent with previous observations that Abcc8 KO mice have impaired glucagon secretion (25). In contrast, fed blood glucose levels were higher in the HFD-KO mice compared with the RC-KO and HFD-WT animals (Fig. 1C). Treatment with verapamil, a Ca²⁺ channel blocker, during HFD lowered the fed blood glucose concentration in both the HFD-KO and HFD-WT mice (Fig. 1C).

After 5 weeks, HFD-KO mice exhibited greater glucose intolerance compared with the RC-KO, RC-WT, and HFD-WT mice (Fig. 1D and F). Verapamil treatment improved glucose tolerance in the HFD-WT mice, while in HFD-KO mice, the effect of the drug was apparent only at 120 min after glucose administration (Fig. 1E and F). Both the RC-KO and HFD-KO animals had increased insulin sensitivity compared with the RC-WT and HFD-WT mice, respectively (Supplementary Fig. 1A and C). However, while verapamil increased insulin sensitivity of HFD-WT compared with the RC-WT mice, it had no effect on the HFD-KO mice (Supplementary Fig. 1B and C). These findings indicate that Abcc8 KO mice are more insulin sensitive, have a higher fed blood glucose concentration, and are more intolerant of glucose on an HFD than are WT animals. Interestingly, coadministration of verapamil during HFD feeding normalized blood glucose concentration in the KO animals independent of improvements in insulin sensitivity. Together, these findings indicate that excitotoxicity increases the susceptibility of β-cells to the negative effects of overnutrition.

To determine how excitotoxicity, overnutrition, and both stresses affect gene expression, we performed RNA sequencing on FACS-purified β-cells. All mice were identical in strain (C57BL/6J) and age (8–9 weeks old), but differed by sex, the presence or absence of Abcc8, and diet (RC vs. HFD). Sample clustering by principal component analyses of data from 31 samples (Supplementary Table 2) showed clear separation of WT and KO samples and indicated that the effect of the Abcc8 KO was much greater than that of the HFD (Supplementary Fig. 2). After pooling the data from both sexes, we performed three differential expression (DE) analyses, as summarized in Supplementary Table 3.

To determine effects of excitotoxicity, we first compared the RC-KO and RC-WT data sets. Similar to our previous report, which used an MIP-GFP transgene (11), we observed a profound alteration in the β-cell transcriptome with a total of 7,393 genes being affected (3,957 upregulated genes [URGs] and 3,436 downregulated genes [DRGs]) (Fig. 2A and B and Supplementary Table 4). The magnitude of gene dysregulation was similar to our prior study (R = 0.72) (Supplementary Fig. 3) with differences in dysregulated gene sets attributed to leaky growth hormone from the MIP-GFP transgene (22).

GO term and pathway enrichment analysis of protein coding URGs and DRGs revealed that the URGs were enriched in mitochondrial genes involved in oxidative phosphorylation, mitochondrial organization, multiple metabolic pathways, and lysosomal genes (Fig. 2C and Supplementary Table 5). In contrast, DRGs were associated with microtubule cytoskeleton, insulin secretion, chromatin organization, transcription, FoxO and Mapk signaling, and cell junction organization. Several of the top URGs are critical for neural and β-cell development, including TFs (Ascl1, Fev, and Neurog3) and growth factors (Nog, Wif1, and Igf2) (Fig. 2D). Other top URGs include gastrin (Gast), a putative marker of dedifferentiating β-cells (26), the EF-hand domain Ca²⁺-binding protein S100a6, voltage-gated K⁺ channels (Kcnc2 and Kcns3), and Ca²⁺ channels (Slc24a33 and Cacng3), which are likely involved in compensatory regulation of ion flow in the absence of functional K_ATP channels (Fig. 2D). Top DRGs are involved in insulin secretion (Nnat and Ins1), cell junction formation (Cldn8 and Pcdh15), potassium ion transport (Trpm5 and Hcn1), response to vascular endothelial growth factor (Kdr and Flt1), and gene transcription (Npas4 and Egr4). These results indicate that β-cell excitotoxicity increases expression of many developmentally important TFs and genes required for mitochondrial energy production while also broadly downregulating genes involved in insulin secretion, chromatin maintenance, and cytoskeletal function.

Next, we determined the effects of overnutrition on WT β-cells by comparing the HFD-WT and RC-WT data sets. This analysis revealed 2,372 affected genes (1,320 URGs and 1,052 DRGs) (Fig. 3A and B and Supplementary Table 4). Functional enrichment analysis indicated that URGs were involved in ER protein processing, ER stress, unfolded protein responses, glycan biosynthesis, and cell cycle regulation. In contrast, DRGs were involved in chromatin organization and response to hormone stimulus, as well as mammalian target of rapamycin signaling pathways (Fig. 3C and Supplementary Table 5). The top URGs included hormone receptors (Ptger3 and Oxtr), immune cell surface proteins (Cd74 and H2-Eb2), and cell cycle regulators (Cdc20 and Ccnb1) (Fig. 3D). Top DRGs included receptors (Erbb2 and Hspg2), extracellular matrix proteins (Olfm2 and Hspg2), secreted growth factors (Igfbp5 and Angptl7), and TFs (Trnp1 and Epas1). These results indicate that overnutrition causes increased expression of genes involved in ER protein processing and β-cell proliferation and downregulation of genes involved in mammalian target of rapamycin signaling and chromatin maintenance.

To determine the combined effects of excitotoxicity and overnutrition, we compared the HFD-KO and RC-WT data sets and identified 8,836 dysregulated genes (4,322 URGs and 4,514 DRGs) (Fig. 4A and B and Supplementary Table 4). URGs were involved in oxidative phosphorylation, the citric acid cycle, and nucleotide metabolism, whereas DRGs were involved in cytoskeleton, insulin secretion, cell projection and cell junction organization, and chromatin and transcriptional regulation (Fig. 4C and Supplementary Table 5). Many of the top URGs were also increased in the RC-KO (Ascl1 and Stc2) and HFD-WT (Cd74 and Cdc20) mice (Fig. 4D). Interestingly, genes involved in lipid uptake (Fabp3 and Apoe), stimulation of ketogenesis, and impairment of glycolysis (Hmgcs2 and Pdk4) were only upregulated in response to the combined stresses. The top DRGs included genes involved in cell adhesion (Gjd4 and Dlagap2), acid transporters (Slc28a2 and Slc7a11), and many genes that were also downregulated in either RC-KO (Npy and Nnat) or HFD-WT (Igfpb5 and Atf5) mice. These findings indicate that the combination of excitotoxicity and overnutrition causes a further increase in the expression of β-cell genes involved in energy metabolism and ATP production, as well as genes linked to a decrease in glucose and an increase in FA-derived ketone utilization as an energy source. At the same time, genes associated with cytoskeleton and cell junction organization are downregulated.

To better categorize the many different transcriptional responses, we performed a meta-analysis of genes dysregulated in three comparisons (Fig. 5A). Major functional categories shared among URGs included oxidative phosphorylation, mitochondrial organization, metabolic pathways, and oxidative stress response, and shared downregulated pathways were chromatin organization, cytoskeleton and cell organization, and DNA damage response (Fig. 5B and C). The large overlap in genes dysregulated in response to both excitotoxicity and overnutrition (Fig. 6A) suggests that Ca²⁺-mediated nuclear responses are involved in many of the responses of β-cells to overnutrition. Of the 620 URGs that were similarly affected by excitotoxicity (RC-KO vs. RC-WT), overnutrition (HFD-WT vs. RC-WT), or excitotoxicity and overnutrition (HFD-KO vs. RC-WT), the most highly affected genes were Mc5r, a melanocortin receptor, Aldh1a3, an oxidoreductase for which expression correlates with β-cell failure (27), and Gabra4, a GABA receptor subunit that potentiates insulin secretion (28). In contrast, 522 DRGs were shared among all three comparisons with Trnp1, a regulator of cell cycle progression (29), tribbles pseudokinase 3 (Trib3), a multifunctional signaling protein involved in coordinating stress-adaptive metabolic responses (30), and glucagon receptor (Gcgr) being the most highly downregulated.

Functional gene enrichment analysis showed that overlapping stress URGs are involved in ER protein processing, glycan biosynthesis, metabolic pathways, oxidative phosphorylation, and lysosomes (Fig. 6B and Supplementary Table 5). Included were genes involved in carbohydrate metabolism (Me3 and Mdh1), amino acid metabolism (Gatm and Oat), FA β-oxidation (Acad11 and Acadvl), components of complexes I–V of mitochondrial electron transport chain (Uqcrfs1, Atp5d, Cox6b1, and Ndufc2), and mitochondrial rRNA proteins (Mrps12 andMrpl51) (Supplementary Fig. 4). Common URGs also include oxidoreductases (Ald1a3 and Aass), secreted proteins (Gc and Vgf), redox homeostasis maintenance genes (Gsto2 and Gpx3), lysosome (Ctsh and Dapl1), ER protein folding (Ppib and Selenos), and vesicle traffic (Rgs8) genes. Notably, genes involved in Ca²⁺ signaling (Camk1d and Mapkapk3) and TFs that are activated by Ca²⁺ signaling (Mef1c and Nfatc1) are among common URGs (Supplementary Fig. 4). Other upregulated TFs include Fev, Bach2, Etv1, Ppargc1a, and Bhlha15. Overlapping stress-induced URGs also contain genes involved in DNA damage cell-cycle checkpoint (Check1 and Ccng1), apoptosis regulation (Bcl2 and Endog), potassium ion transport (Kcnk13 and Slc12a2), receptors (Gabra4 and Gfra4), extracellular matrix (Col8a2 and P3h2), and immune response (H2-Eb1 and Tnfrsf11b).

DRGs common to all three comparisons are involved in transcription, chromatin modification, protein phosphorylation, as well as adherens junctions and FoxO1 signaling pathways (Fig. 6B and Supplementary Table 5). Downregulated TFs included known regulators of β-cell identity and function (Myt1, Thra, Myt1l, and Stat5a) and many for which the role in β-cells has not been studied (Mesp2, Phf21b, Otub2, and Chd7) (Supplementary Fig. 5). Over 40 proteins involved in epigenetic regulation were decreased, including chromatin-modifying enzymes (Kdm6b, Jmjd1c, and Kat2b), DNA methylation enzymes (Dnmt3a and Tet3), and miRNA-processing proteins (Ago1 and Tnrc6c). Common DRGs also included those involved in regulation of circadian rhythms (Prkab, Prkag2, and Nr1d1), the DNA damage response (Plk3, Taok1, and Primpol), Bmp and Wnt signaling (Acvr1c, Bmpr2, and Amer1/2), cell junction and polarity (Nectin1, Cldn4, Dlgap3, and Pard3), and cilia morphogenesis (Alms1, Cep162, Rfx3, and Ulk4) (Supplementary Fig. 5). In addition, common DRGs were for receptors (Ffar1 and Trpc1), kinases (Prkab2 and Jak), the phosphatidylinositol 3-kinase/AKT/FoxO1 signaling pathway (Akt3, Insr, and Foxo1), Ca²⁺ transport (Trpc1 and Grin2c), and proteins involved in amino acid transport (Slc7a11 and Slc36a1). A total of 114 long noncoding RNAs (lncRNAs) were also reduced in all three comparisons (Supplementary Fig. 6).

Analysis of genes that were dysregulated only in the presence of both stresses (1,098 URGs and 1,407 DRGs) (Fig. 6A) showed a further increase in oxidative phosphorylation, translation, and nucleotide metabolism genes and a decrease in microtubule-based processes, RNA transport, cilia, and nuclear pore organization genes (Supplementary Fig. 7 and Supplementary Table 5). Importantly, an increase in genes that impair glucose utilization for energy production (Hmgcs2 and Pdk4) was observed only when excitotoxicity and overnutrition were combined, suggesting that the two stresses additively cause metabolic inflexibility.

To determine how sex affects the response of β-cells to excitotoxicity and overnutrition, we reanalyzed our glucose homeostasis measurements and transcriptome data to extract these differences. As shown in Supplementary Figs. 8 and 9, male WT and KO mice gained more weight and had higher fed glucose concentrations and greater glucose intolerance than females of the same genotypes after 5 weeks on HFD. Verapamil treatment during HFD improved glucose tolerance in both the WT and KO males, but not in females of the same genotypes (Supplementary Fig. 9). However, while verapamil improved insulin tolerance in both the WT males and females, it had no effect on the KO mice (Supplementary Fig. 10). These findings indicate that male mice are more susceptible to negative effects of increased [Ca²⁺]_i than are female animals.

To determine how sex affects β-cell stress responses, we performed four different female-versus-male pairwise comparisons on the RNA-sequencing data sets (Supplementary Tables 3 and 4). Comparison of the RC-WT, RC-KO, and HFD-KO data sets in this manner yielded 140, 36, and 126 sex-specific genes, respectively (P_adj < 0.05) (Supplementary Table 3). The lower number of sex-specific changes in the RC-KO and HFD-KO mice, compared with the RC-WT mice, suggests that the marked perturbation of the β-cell transcriptome that occurs in Abcc8 KO mice hinders the detection of sex-related differences. However, 2,618 were differentially expressed between female and male HFD-WT data sets. Overlaying all of the differentially expressed genes from the four pairwise comparisons revealed seven core sex-enriched genes, with Xist, Fmo1, and Kdm6a being female enriched and Kdm5d, Uty, Eif2s3y, and Ddx3y being male enriched (Supplementary Fig. 11A).

GO analysis of sex-enriched genes from HFD-WT comparison revealed that female-enriched pathways included oxidative phosphorylation, proteasome degradation, spliceosome, glutathione metabolism, and adrenergic signaling. Male-enriched pathways included ER to Golgi vesicle traffic, autophagy, and cell cycle (Supplementary Fig. 11B and Supplementary Table 5). Among the top genes expressed higher in females on HFD were neuropeptides (Npy and Pyy), hormones (Gcg and Sst), as well as developmental endocrine TFs (Neurog3, Mafb, Fev, Arx, and Hhex) (Supplementary Fig. 11C). In males, the more abundantly expressed genes included Mc5r and Aldh1a3, cell proliferation genes (Mki76 and Ccna), DNA damage-response genes (Pole and Fanca), and transcriptional regulators (Chd5, Bach2, and Txnip). Overall, transcriptional response of male β-cells to HFD indicates a greater increase in β-cell proliferation, secretory function, autophagy, and associated DNA repair and ER-associated protein degradation pathways. The transcriptional response of female β-cells to HFD shows an increase in mitochondrial function, glutathione antioxidant defense, adrenergic signaling, and changes in β-cell identity.

Because our in vivo analyses revealed the modulation of genes involved in mitochondrial energy production and the maintenance of β-cell identity, we further analyzed Ppargc1a, Bach2, Thra, and Myt1 in cultured islets by RT-PCR (Fig. 7A). Ppargc1a is a transcriptional coregulator that is central to activation of mitochondrial energy metabolism, FA β-oxidation, and mitochondrial biogenesis (31). Bach2 belongs to a family of TFs induced by oxidative stress that may play a role in immune-mediated β-cell apoptosis (32). Myt1 and Thra, a thyroid hormone nuclear receptor, both contribute to the function of mature β-cells (33,34). Cultured WT mouse islets were treated with tolbutamide, a K_ATP channel inhibitor, PA, or a combination of both agents alone and with verapamil. After 24 h, the expression of Ppargc1a and Bach2 was increased, and Thra and Myt1 decreased, in response to tolbutamide alone. These changes were greatly accentuated when tolbutamide and PA were combined (Fig. 7B) and largely negated with the addition of verapamil. While the effect of PA by itself was generally small, its combination with tolbutamide was strongly additive, particularly for Ppargc1a. These findings provide additional evidence for rapid changes in key β-cell TFs and FA signaling in response to a rise in [Ca²⁺]_i.

In β-cells, mitochondrial metabolism is essential for coupling glucose metabolism to insulin secretion. To determine whether the observed increases in genes involved in mitochondrial respiration, organization, and FA β‐oxidation reflect actual changes in mitochondrial function, we compared mitochondrial OCR and mitochondrial biogenesis of RC-WT, RC-KO, HFD-WT, and HFD-KO islets. OCR measurements (Fig. 8A) showed that basal respiration rate reflecting the cell baseline metabolic energy production was progressively increased in HFD-WT, RC-KO, and HFD-KO islets (Fig. 8B). Similarly, metabolically stressed islets also had increased spare respiratory capacity, indicating increased ability to respond to increased energy demand (Fig. 8C) and increased mtDNA copy number (Fig. 8F). These results are consistent with corresponding increases in both mitochondrial respiration and biogenesis gene expression. However, the change in OCR in response to glucose is significantly decreased in RC-KO islets and almost completely ablated in HFD-KO islets (Fig. 8D), indicating that excitotoxicity and overnutrition additively impair mitochondrial glucose metabolism, most likely due to an increased FA β-oxidation, a rise in metabolic inflexibility, and mitochondrial damage. Consistent with an additive increase in mitochondrial damage, mitochondrial coupling efficiency is decreased and proton leak is increased in HFD-KO islets (Fig. 8E and F). These findings directly indicate that excitotoxicity and overnutrition additively impair mitochondrial glucose metabolic coupling function.

To gain a systems-wide understanding of pancreatic β-cell failure, we explored the additive effects of excitotoxicity and overnutrition on β-cell function and gene expression. Among the many changes that occur in β-cells in response to metabolic stress, we identified several critical alterations that may be tipping points for the development of T2D.

It has long been known that C57BL/6J (WT) mice develop insulin resistance and impaired β-cell function on an HFD (35). Our studies strongly suggest that an increase in [Ca²⁺]_i, as occurs in the Abcc8 KO β-cells, increases the propensity of β-cell to fail in response to HFD and, conversely, verapamil, which blocks calcium entry and protects against the loss of β-cell function. RC-KO mice, which are euglycemic at 8–9 weeks of age (11,23), exhibit greater insulin sensitivity than RC-WT mice, as has been reported for mice lacking Kcnj11, another essential K_ATP channel component (36). While HFD predictably caused insulin resistance in WT mice, Abcc8 KO mice, like Kcnj11 KO animals (37), remained insulin sensitive on HFD, indicating that observed impairments in glucose homeostasis are mostly due to loss of β-cell function. Furthermore, while verapamil had no effect on insulin sensitivity in the KO mice, it increased the insulin sensitivity of WT mice on an HFD. Thus, our results not only confirm that Ca²⁺ channel blockers attenuate the development of obesity-induced insulin resistance (38), but they also indicate that the protective effects of these agents extend to β-cells. Moreover, they also indicate that an increase in [Ca²⁺]_i, besides contributing to obesity-induced insulin resistance (39), has a negative effect on β-cell function. Indeed, given the pleiotropic metabolic effects of dysregulated Ca²⁺ homeostasis, it is noteworthy that the effects of Ca²⁺ channel blockers in both type 1 diabetes and T2D are now being investigated (40,41).

Excitotoxicity and overnutrition each have a major impact on β-cell gene expression, and together, the two stresses affect the expression of 11,952 unique genes, or nearly three-quarters (72%) of all genes expressed in β-cells. We also observed an overlap between genes and pathways that were dysregulated in response to both stresses, suggesting that increased [Ca²⁺]_i may be involved in mediating the nuclear responses of other metabolic stresses.

The largest category of genes upregulated by excitotoxicity and/or overnutrition are those involved in mitochondrial function, metabolism, and oxidative phosphorylation. Mitochondrial metabolism is a major determinant of insulin secretion from pancreatic β-cells, and mitochondrial dysfunction plays a key role in development of T2D (42). Increased [Ca²⁺]_i, in both normal and pathological states, invariably leads to increased mitochondrial Ca²⁺ uptake that stimulates mitochondrial Ca²⁺-sensitive metabolic enzymes and oxidative phosphorylation in the electron transport chain (43). However, a persistent increase in mitochondrial [Ca²⁺] and respiration may lead to an increase in reactive oxygen species production, collapse of the mitochondrial membrane potential, and mitochondrial dysfunction (6). Consistently, we observed that even though mitochondrial biogenesis and basal respiration are increased in KO islets, they also exhibit decreased coupling efficiency and increased proton leak upon addition of HFD, indicating mounting mitochondrial damage. Normally, damaged mitochondria are replaced by the combination of mitophagy, a lysosomal-based degradation process, and the production of new mitochondria that occurs through mitochondrial biogenesis. A balance between these processes is essential for normal β-cell function (44). Excitotoxicity increases the expression of multiple mitochondrial and lysosomal genes, as well as master regulators of both mitochondrial (Ppargc1a) and lysosomal (Tfeb) biogeneses, potentially maintaining biogenesis/mitophagy balance. However, the combination of both excitotoxicity and overnutrition shifts this balance, as there is a further increase in mitochondrial energy metabolism, oxidative stress, and DNA damage-response genes, indicating an increase in reactive oxygen species and mitochondrial damage, while expression of lysosomal genes is not further changed, and several mitophagy-associated genes, such as Clec16a and Prkn (45), become downregulated. These results suggest that the combination of excitotoxicity and overnutrition overwhelms the ability of β-cells to replace metabolically damaged mitochondria.

Ppargc1a, a central transcriptional regulator of mitochondrial biogenesis, energy metabolism, and FA β-oxidation, is also increased in β-cells in response to both excitotoxicity and overnutrition. PPARGC1A interacts with multiple TFs and chromatin modifiers to regulate metabolic reprogramming in response to diet and oxidative stress and is implicated in pathogenesis of T2D (46). Ppargc1a is necessary for normal β-cell function (47), and its overexpression causes β-cell dysfunction (48). In this study, we show that expression of Ppargc1a is induced in cultured WT islets in response to increases in [Ca²⁺]_i and is further increased with the addition of PA. This finding further implicates [Ca²⁺ ]_i as an important regulator of mitochondrial metabolism in β-cells and suggests that an increase in [Ca²⁺]_i causes the rapid upregulation of Ppargc1a, most likely through the activation of Ca²⁺-dependent CaN/CaMK/MAPK/AMPK signaling pathways. Similar signaling pathways are activated in skeletal muscle cells in response to exercise (49). In support of this hypothesis, excitotoxicity and overnutrition both upregulate Mef2c, a known target of Ca²⁺ signaling and a known activator of Paprgc1a in muscle, several CaMK and MAPKs, and multiple Ppargc1a target genes (Supplementary Fig. 12A). Similar to the activation of Ppargc1a, we also confirm an additive effect of increased [Ca²⁺]_i and PA on the upregulation of Bach2, an oxidative stress–responsive TF, and the downregulation of both Myt1 and Thra, two other TFs important for the function of mature β-cells. However, less is known about how these other important genes are regulated and how they may contribute to normal β-cell function and the maintenance of β-cell identity.

Coupling of glucose metabolism to insulin secretion is an important aspect of mitochondrial function in β-cells. We find that the combination of excitotoxicity and overnutrition increases expression of genes that contribute to metabolic inflexibility or the decreased ability to use glucose as an energy source, another key feature of β-cell failure (50). Similar to overnutrition alone, excitotoxicity alone causes an increase in genes involved in FA β-oxidation, indicating that β-cells partially switch to utilization of fat as a fuel in response to an increased [Ca²⁺]_i, a process known as glucose sparing (Supplementary Fig. 12B). This switch is reflected in reduced ability of KO islets to increase mitochondrial respiration in response to glucose. However, in addition to FA β-oxidation, combination of excitotoxicity and overnutrition causes increases in Pdk4, a kinase that inhibits pyruvate flux into the tricarboxylic acid cycle, promoting FA and ketone body utilization (51), and Hmgcs2, a rate-limiting enzyme in ketone body production that is activated by FAs (52). Because these changes would be expected to impair glycolytic flux and cause metabolic inflexibility (Supplementary Fig. 12C), they may explain the observed collapse of mitochondrial glucose response in HFD-KO islets. Therefore, our data suggest a tipping point at which the combination of an increase in [Ca²⁺]_i and an elevation in FAs causes the β-cell to cease relying on glycolysis and to switch to FAs and ketones as their fuel source, impairing their ability to sense and respond to changes in the blood glucose concentration.

Glycosylation is a process during which glycans (mono- or oligosaccharides) are attached to proteins in the ER and Golgi that serve as a quality control signal in ER protein folding (53). In β-cells, increases in protein glycosylation cause ER stress, eventually leading to apoptosis (54). We found that overnutrition in particular increased expression of genes associated with ER protein folding and N- and O-linked protein glycosylation, suggesting that the stability, localization, trafficking, and function of many receptors, ion channels, nutrient transporters, and TFs are also adversely affected, likely contributing to development of T2D (55).

Excitotoxicity, overnutrition, and the combination of both stresses downregulate genes important for cell organization and secretory function of β-cells, including cell adhesion, cell junctions, cilia, cytoskeleton, and vesicular trafficking genes. We have previously identified impairments in islet architecture of Abcc8 KO mice (11), suggesting that critical cell-to-cell contacts, which are necessary for insulin secretion, may become impaired (56). The downregulation of synaptic vesicle-targeting proteins, GTPases, and cytoskeletal proteins has been shown to affect insulin exocytosis (57). Similarly, alterations in β-cell polarity may occur as genes associated with the apical domain (Pard3) and associated primary cilia (Alms1 and Cep162) and lateral domain (Dlgap3 and Nectin1) are downregulated, and genes associated with the vasculature-facing basal domain (Col8a2, Ntn4, and Ppfia3) are increased (58). These changes indicate that Ca²⁺ signaling in β-cells is crucial for maintaining cell polarity and cytoskeleton dynamics and that a chronic increase in [Ca²⁺]_i impairs the ability of these cells to secrete insulin.

Another important category of genes downregulated in metabolically stressed β-cells is those involved in transcriptional control. Besides decreases in many TFs that are necessary for function of mature β-cells (Myt1, Thra, Pbx1, Stat5a, and Nrf1), we also identified several other stress-inhibited TFs (Mesp2, Klf7, Nfia, Ikzf3, and Chd7) that may be critical for maintaining β-cell identity and function. Similarly, many chromatin modifiers, including histone methyltransferases and acetyltransferases, were downregulated in response to both stresses. Among these is Dnmt3a, a DNA methyltransferase important for silencing of developmental or “disallowed” metabolic genes in mature β-cells (59). Consistent with this, several disallowed genes (60) (Slc16a, Oat, and Aldob) are upregulated in response to excitotoxicity and/or overnutrition. Finally, maintenance of the epigenetic and transcriptional landscape of β-cells is also regulated by lncRNAs (61), and we identified multiple lncRNAs that are downregulated in response to these two metabolic stresses.

Male rodents have a greater propensity for β-cell failure than do females (18). In this study, we analyzed the effects of excitotoxicity and/or HFD on β-cell function and gene expression in both sexes. We found that female animals (both WT and Abcc8 KOs) withstand overnutrition better than males. These results are consistent with previous data on HFD-WT mice (62) and may reflect the protective influence of female sex hormones on β-cell function and metabolism (63). Interestingly, verapamil improved insulin sensitivity and glucose clearance in both the HFD-WT and HFD-KO males, but had little effect in females, suggesting that males are more negatively affected by a stress-induced increase in [Ca²⁺]_i than females. Testosterone is known to increase [Ca²⁺]_i in multiple tissues (64) and may also predispose β-cells for dysfunction associated with further [Ca²⁺]_i increase due to increased metabolic load.

Analysis of sex differences on a transcriptome level in WT mice confirmed our previous findings that female β-cells express higher amounts of several TFs important for β-cell function including Mlxipl, Nkx2-2, and Hnf1b (22). Interestingly, we were only able to detect a few sex-related changes in β-cells from both RC- and HFD-Abcc8 KO mice, suggesting that the massive gene expression changes that occur in the Abcc8 KO mice may mask the detection of sex-related differences. However, overlaying the differentially expressed genes from all four female to male transcriptome comparisons allowed us to identify three core female-enriched β-cell genes (Xist, Fmo1, and Kdm6a) and four male-enriched genes (Kdm5d, Uty, Eif2s3y, and Ddx3y). All of the male-enriched genes are located on the Y chromosome, whereas only two of the female-enriched genes (Xist and Kdm6a) are located on the X chromosome. Kdm6a, Uty, and Kdm5d all code for histone demethylases that may be involved in sex-specific epigenetic regulation of gene expression (65,66). Our discovery that Fmo1, an autosome-located gene, is enriched in female β-cells is interesting, as this gene encodes a flavin-containing monooyxgenase 1, a drug-metabolizing enzyme that regulates energy balance (67). Another member of the same family, Fmo4, was also upregulated in females on HFD. Overall, the most profound sex differences revealed by our analysis were in the β-cell transcriptome on HFD (2,618 genes). Female β-cells express higher levels of genes involved in oxidative phosphorylation, suggesting they may have higher energy metabolism than do male β-cells. Genes involved in prevention of oxidative damage, such as glutathione peroxidases, are also higher in female cells, suggesting a mechanism enabling female β-cells to tolerate overnutrition better than males. Female β-cells also express more of neuropeptides Npy and Pyy, both of which could be protective against β-cell damage (68,69), and the adrenergic receptor Adrb2, signaling through which can increase insulin secretion. Consistent with this, it was recently reported that pancreas-specific loss of Adrb2 causes glucose intolerance and impaired glucose-stimulated insulin secretion only in female mice (70). Intriguingly, genes associated with endocrine progenitor (Neurog3 and Mafb) and α-cells (Gcg and Arx) and δ-cells (Sst and Hhex) are upregulated in HFD-fed female β-cells, suggesting changes in cell identity. Male mice fed HFD exhibit higher expression of β-cell genes involved in protein secretion and cell proliferation, consistent with the well-established fact that male β-cells exhibit greater proliferation when fed HFD (71).

We have identified many different genes and pathways in β-cells that are affected by metabolic stress. The broad-based nature of the changes we observe suggests that β-cell failure in T2D is not due to the failure of a single cellular process, but instead is a complex, multifaceted, and additive process in which Ca²⁺ signaling plays a crucial role.

Acknowledgments. The authors thank the Vanderbilt Technologies for Advanced Genomics (VANTAGE) Core and the Vanderbilt University Medical Center Flow Cytometry Shared Resource for assistance in performing cell sorting and RNA sequencing and the Vanderbilt Islet Procurement and Analysis Core for help with isolating islets.

Funding. This research was supported by institutional and philanthropic funds provided by Vanderbilt University. The Vanderbilt Islet Procurement and Analysis Core is supported by DK-020593. VANTAGE is supported by grants P30-CA-68485, P30-EY-08126, and G20-RR-030956. Vanderbilt University Medical Center Flow Cytometry Shared Resource is supported by grants P30-CA-68485 and DK-058404.

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

Author Contributions. A.B.O. and M.A.M. designed the study. A.B.O., J.S.S., and K.D.D. performed experiments. A.B.O. analyzed the data. J.-P.C. performed RNA-sequencing data processing, alignment, and DE analyses. A.B.O. and M.A.M. wrote and edited the manuscript. J.S.S. edited the manuscript. M.A.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.