Chronic β-Cell Depolarization Impairs β-Cell Identity by Disrupting a Network of Ca²⁺-Regulated Genes

In type 2 diabetes (T2D), pancreatic β-cells fail to respond appropriately to metabolic stresses brought on by age, obesity, and genetic risk factors. The mechanisms by which chronic metabolic stress, including insulin resistance, glucotoxicity, and lipotoxicity (1–3), impair β-cell function are not understood. Although metabolic stress is usually considered to be exogenous to the β-cell, chronic stimulation leads to changes within the β-cell, impairing function. One such factor is chronic elevation in the concentration of intracellular Ca²⁺ ([Ca²⁺]_i), sometimes called excitotoxicity (4), which may be triggered by sustained β-cell depolarization resulting from chronic stimulation.

Ca²⁺ is a ubiquitous second messenger that is central to regulating cellular dynamics of many cell types, including β-cells. Genetic and pharmacological perturbations that stimulate or impair Ca²⁺signaling have dramatic effects on β-cell function. For instance, the disruption of calcineurin, a Ca²⁺-dependent phosphatase, or Ca²⁺/calmodulin-dependent protein kinase II or IV, two Ca²⁺-dependent kinases, profoundly impairs β-cell function, likely by modulating the activity of Ca²⁺-responsive transcription factors such as NFAT, CREB, and TORC2 (5–9). Conversely, the constitutive activation of calcineurin or calmodulin, a Ca²⁺ binding protein, also causes marked β-cell dysfunction (3,10,11).

Acutely, glucose metabolism induces ATP-sensitive potassium (K_ATP) channel closure, membrane depolarization, opening of voltage-gated Ca²⁺channels, a rise in [Ca²⁺]_i, and insulin secretion. However, sustained elevation in [Ca²⁺]_i has multiple effects on β-cell function that can be adaptive or maladaptive. β-Cell proliferation induced by glucose metabolism (12) is an example of an adaptive response to sustained elevations in [Ca²⁺]_i. However, chronically elevated [Ca²⁺]_i can also induce maladaptive responses because prevention of Ca²⁺ influx in the setting of insulin resistance prevents β-cell death (13). In either case, mice lacking K_ATP channels exhibit disrupted islet morphology, characterized by α-cells being located in the islet core (14,15), suggesting loss of β-cell identity or impairments in cell adhesion.

Here, we show that β-cells in Abcc8^−/− mice exhibit chronic membrane depolarization and a sustained elevation in [Ca²⁺]_i and dysregulation of more than 4,200 genes, many of which are involved in cell adhesion, Ca²⁺binding and Ca²⁺signaling, and maintenance of β-cell identity. We also report that Abcc8^−/− mice exhibit β-cell to pancreatic polypeptide (PP)–cell trans-differentiation and have increased expression of Aldh1a3, a gene recently suggested as a marker of dedifferentiating β-cells. In addition, we show that S100a6 and S100a4, two EF-hand Ca²⁺ binding proteins, are acutely regulated in β-cells by membrane depolarization agents, suggesting that they may be markers for β-cell excitotoxicity. Finally, we performed a computational analysis to predict components of the gene regulatory network that may govern the observed gene expression changes and found that one of the predicted regulators, Ascl1, is directly regulated by Ca²⁺ influx in β-cells.

All animal experimentation was approved by the Vanderbilt Institutional Animal Care and Use Committee. Abcc8^−/− (Abcc8^tm1.1Mgn, MGI: 2388392), RIP-Cre (Tg.Ins2-cre^25Mgn, MGI: 2176227), and R26^LSL.YFP (Gt [ROSA]26Sor^tm[EYFP]Cos, MGI: 2449038) mice were congenic C57BL/6 and genotyped as previously described (16–18). MIP-GFP mice (Tg[Ins1-EGFP/GH1]^14Hara, MGI: 3583654) were congenic CD-1 and genotyped as previously described (19).

After a 16-h fast, male Abcc8^+/+ and Abcc8^−/− C57BL/6 mice were given intraperitoneal injection of d-glucose (2 mg/g body weight). Blood glucose was measured using a BD Logic glucometer.

Adult Abcc8^+/+;MIP-GFP and Abcc8^−/−;MIP-GFP mice were given Splenda (2%) or a combination of verapamil (1 mg/mL; Sigma-Aldrich, V4629) and Splenda in their drinking water for 3 weeks. Splenda was used to mask the taste of verapamil.

Pancreata were fixed in 4% paraformaldehyde, frozen, and sectioned at a depth of 8 μm. Immunofluorescence staining was performed as previously described (20). Antibodies are listed in the Supplementary Data. Images were acquired using an Olympus FV-1000 confocal microscope, pseudocolored using ImageJ, and are representative of the phenotype observed in at least three animals. Cell death was determined using the In Situ Cell Death Detection Kit (Roche, 11684795910).

Pancreata were injected with 0.6 mg/mL collagenase P (Roche, 11213865001) into the pancreatic bile duct. Dissociated tissue was fractionated using Histopaque-1077 (Sigma-Aldrich, 10771), followed by hand-picking of islets. For FACS and RNA sequencing (RNA-Seq), islets from four to seven mice were pooled per sample. For quantitative RT-PCR (qRT-PCR), islets from a single mouse were used per sample.

Islets were isolated from pancreata of 7- to 10-week-old Abcc8^+/+;MIP-GFP and Abcc8^−/−;MIP-GFP mice, and electrophysiological recordings were performed as previously described (21).

Islets were isolated from pancreata of 9- to 11-week-old Abcc8^+/+and Abcc8^−/− mice, and imaging of cytoplasmic Ca was performed as previously described (22).

Wild-type islets were incubated for 24 h in DMEM (Gibco, 11966-025) containing 5.6 mmol/L glucose, 10% FBS (Gibco, 16140–071), and 1% penicillin-streptomycin (Gibco, 15140-122). Experimental media contained 100 μmol/L tolbutamide (Sigma-Aldrich, T0891) or 20 mmol/L KCl (Sigma-Aldrich, P5405), with or without 50 μmol/L verapamil.

Islets were dissociated in Accumax (Sigma-Aldrich, A7089) containing 10 units/mL DNase (Invitrogen, AM2222). After filtration (35-μm strainer) and centrifugation, cells were resuspended in Flow Cytometry Buffer (R&D Systems, FC001) containing 2 units/mL DNase, 0.5 mol/L EDTA, and 7-aminoactinomycin D (1:1,000; ThermoFisher, A1310). GFP⁺/7-aminoactinomycin D^– cells were analyzed and isolated using an Aria II (BD Biosciences) and collected in Homogenization Solution (Promega, TM351; containing 1-thioglycerol).

RNA was isolated from FACS-purified β-cells and whole islets using Maxwell 16 LEV simplyRNA Tissue Kit (Promega, TM351). RNA samples were analyzed using the Agilent 2100 Bioanalyzer. Only samples with an RNA integrity number >7 were used.

RNA samples from FACS-purified β-cells were amplified using SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Clontech, 634888; eight PCR cycles). cDNA libraries were constructed using the Low Input Library Prep Kit (Clontech, 634947). Paired-end sequencing of four replicates each of Abcc8^+/+;MIP-GFP and Abcc8^−/−;MIP-GFP samples was performed on an Illumina NextSeq 500 (75-nucleotide reads). Approximately 900 million raw reads were processed using TrimGalore 0.4.0. STAR (23) was used to align sequences to mm10 (Genome Reference Consortium Mouse Reference 38 [GRCm38]) and GENCODE comprehensive gene annotations (Release M8). HTSeq was used for counting reads mapped to genomic features (24), and DESeq2 was used for differential gene expression analysis (25).

iRegulon (26) was used to predict gene regulatory networks. Default search parameters were used (20 kb centered on the transcription start site, 7 species conservation). The enrichment score threshold was 3.0.

The High Capacity cDNA Reverse Transcription Kit (ThermoFisher, 4368814) was used to convert whole-islet RNA to cDNA. qRT-PCR was performed on an Applied Biosystems 7900HT using 2× SYBR Green PCR Master Mix (ThermoFisher, 4309155). Samples were analyzed in triplicate and normalized to Hprt expression. Primer sequences are listed in the Supplementary Data.

Statistical significance was determined using the two-tailed Student t test. Data are represented as mean ± SEM. A threshold of P < 0.05 was used to declare significance.

Previous studies have shown that single β-cells from mice lacking K_ATP channels, in contrast to normal mice, exhibit chronically elevated [Ca²⁺]_i and continuous action potential firing (16,27,28), whereas intact islets exhibit electrical bursting and Ca²⁺ oscillations (29,30). To definitively establish Abcc8^−/− mice as a model for chronically elevated β-cell [Ca²⁺]_i, we measured membrane potential at high (11 mmol/L) and low (2 mmol/L) glucose concentrations (Fig. 1A and B). Abcc8^−/− β-cells exhibit action potentials in high glucose but fail to undergo membrane potential polarization in low glucose (Fig. 1B). Quantification of the changes in Abcc8^−/− β-cell membrane potentials confirms that Abcc8^−/− β-cells show no difference in membrane potential between high and low glucose, contrary to Abcc8^+/+ β-cells (Fig. 1C and D). However, Abcc8^−/− β-cells exhibit a chronic elevation in [Ca²⁺]_i, as shown by Ca²⁺-imaging at both low and high glucose (Fig. 1E). Quantification of the area under the curve shows that Abcc8^−/− β-cells have significantly higher [Ca²⁺]_i than Abcc8^+/+ β-cells at each time interval measured (Fig. 1F). In addition, by examining individual traces rather than averaged data and imaging over a longer period, we observe Ca²⁺ oscillations in Abcc8^−/− islets under low glucose conditions, as previously reported (29,30), whereas Abcc8^+/+ islets exhibit consistently low Ca²⁺ levels (Supplementary Fig. 1). This indicates that although Abcc8^−/− β-cells exhibit variable and oscillating [Ca²⁺]_i under euglycemic conditions, their mean [Ca²⁺]_i is elevated compared with Abcc8^+/+ β-cells.

Interestingly, despite having chronically elevated β-cell [Ca²⁺]_i, Abcc8^−/− mice have long been known to have surprisingly small disturbances in their plasma glucose concentrations (16). First, and as we previously reported (31), they exhibit outright hypoglycemia in the fasted state (Supplementary Fig. 3A). Second, they exhibit a trend toward mild hyperglycemia in the fed state (Supplementary Fig. 3B).

During prolonged metabolic stress, β-cells can lose expression of functional markers and convert to other endocrine cell types (32). However, β-cell dedifferentiation has not been studied in the context of chronically elevated [Ca²⁺]_i. Thus, we performed β-cell lineage tracing using Abcc8^−/−; RIP-Cre; Rosa26^LSL.YFP/+ mice. Although there was no evidence for β- to α-cell or β- to δ-cell trans-differentiation (Supplementary Fig. 2A and B), we observed an increase in yellow fluorescent protein (YFP)/PP coexpression (Fig. 2C). Most YFP/PP coexpressing cells are polyhormonal, expressing both PP and insulin (Fig. 2A), but a few cells (0.24% of all YFP-expressing cells and 10% of YFP/PP-coexpressing cells) (Supplementary Fig. 2C) no longer express insulin (Fig. 2B). Administration of a Ca²⁺channel blocker to Abcc8^−/− mice resulted in a lower number of insulin/PP-coexpressing cells, but the difference was not significant (Supplementary Fig. 2D). The loss of β-cell identity correlates with impaired glucose tolerance not attributable to β-cell death (Supplementary Fig. 3A and C). Complete dedifferentiation (e.g., YFP expression without hormone immunostaining), was only observed at a very low rate and was similar in both wild-type and knockout animals (0.19% and 0.21%, respectively) (Supplementary Fig. 2E).

To determine how chronically elevated β-cell [Ca²⁺]_i affects gene expression, we performed RNA-Seq using FACS-purified β-cells from Abcc8^−/−;MIP-GFP mice at 8–9 weeks of age. Because the human growth hormone mini-gene within the MIP-GFP allele causes both human growth hormone expression and activation of STAT5 signaling (33), we used MIP-GFP–expressing mice as controls. Principal component and gene clustering analyses (Supplementary Fig. 4B and C), performed on RNA-Seq data from FACS-purified cells (Supplementary Fig. 4A), indicates that most of the top 500 differentially expressed genes cluster as expected. Differential expression analysis (Fig. 3B and Supplementary Table 1) revealed 4,208 differentially expressed genes (2,152 downregulated and 2,056 upregulated) in Abcc8^−/− β-cells (adjusted P < 0.05). Approximately 90% are protein coding, 3% are noncoding RNAs, 2% are pseudogenes, and 3% are other types of processed transcripts (Fig. 3A).

To correlate gene expression with the functional abnormalities of Abcc8^−/− islets, we examined genes known to be highly enriched in mature β-cells and observed that Ins1, Slc2a2, Neurod1, Gck, Glp1r, Npy, several synaptotagmins, Tph1, Tph2, and Egfr are all downregulated (Fig. 3C). Conversely, Ppy is upregulated in Abcc8^−/− β-cells, consistent with our observation of polyhormonal cells.

Because Abcc8^−/− β-cells have a persistent increase in [Ca²⁺]_i, we examined genes with well-defined roles in Ca²⁺ binding or signaling, such as Myo7a, a Ca²⁺/calmodulin-binding myosin motor protein, Pcp4, a calmodulin binding protein known to protect neurons from Ca²⁺-induced toxicity (34), and Cacng3, a subunit of a voltage-dependent Ca²⁺ channel, and found their expression was elevated. Similarly, we observed that Nfatc1, Mef2c, and Mef2d, three calcium-regulated transcription factors; Camk1d, Camkk1, and Camkk2, three kinases involved in calcium-signaling; and S100a1, S100a3, S100a4, S100a6, and S100a13, five EF-hand Ca²⁺ binding proteins, were also upregulated (Fig. 3C).

Quantitation of islet cell types in Abcc8^−/− mice revealed fewer β-cells, more α-, δ-, and PP cells, and a progressive redistribution of non–β-endocrine cells to the islet core (Supplementary Fig. 5). To correlate gene expression with this disrupted islet architecture, we examined expression of cell adhesion molecules, hypothesizing that a reduction in such genes could result in loss of islet structure. Consistently, we observed a reduction in multiple cell adhesion molecules, including Cldn1, Cldn3, Cldn8, Cldn23, and Ocln, genes involved in tight junctions; Cdh1, Cdh7, Cdh8, Cdh13, and Cdh18, encoding cadherins; Cdhr1, a cadherin-related protein; Pcdhb15 and Pcdhb22, encoding protocadherins; and Vcam1, a vascular cell adhesion molecule (Fig. 3C).

Aldh1a3, a retinaldehyde dehydrogenase recently suggested to be a marker for β-cell dedifferentiation (35), is 27-fold upregulated in Abcc8^−/− β-cells by RNA-Seq (Fig. 3C). In addition, six other genes (Serpina7, Aass, Asb11, Penk, Fabp3, and Tcea1) enriched in dedifferentiated β-cells (35) are upregulated in Abcc8^−/− β-cells (Fig. 3C). Coimmunostaining with insulin indicates that although ALDH1A3 is expressed in only 0.4% of Abcc8^+/+ β-cells, it is present in 29.1% of Abcc8^−/− β-cells (Fig. 4A and B). However, this fraction was reduced to 11.4% in Abcc8^−/− mice administered the Ca²⁺ channel blocker verapamil (Fig. 4A and B), further suggesting that a chronic increase in [Ca²⁺]_i impairs the maintenance of cell identity.

To identify genes that could serve as markers of β-cell excitotoxicity, we analyzed S100a6 and S100a4. Both genes are appealing targets because of their functions as EF-hand Ca²⁺ binding proteins, the association of S100a6 with insulin secretion (36), and the increased expression of members of the S100 gene family in islets from humans with hyperglycemia (37). Moreover, they are among the most highly upregulated genes in Abcc8^−/− β-cells, with S100a6 and S100a4 increasing 37- and 5-fold, respectively. S100A6 is only expressed in 2.8% of Abcc8^+/+ β-cells but is expressed in 38.4% of Abcc8^−/− β-cells (Fig. 4C and D). In addition, qRT-PCR using whole-islet RNA confirms the upregulation of S100a6 and S1004 in Abcc8^−/− islets (Fig. 5A).

To determine whether the observed changes in S100a6 and S100a4 expression in Abcc8^−/− β-cells are the result of chronically elevated [Ca²⁺]_i and not some other mechanism, we treated wild-type islets with KCl or tolbutamide and found that membrane depolarization is associated with an increase in both S100a6 and S100a4 expression (Fig. 5B–E), mirroring the expression pattern in Abcc8^−/− β-cells. These changes were reversed when islets were treated with both a depolarizing agent and the Ca²⁺channel inhibitor verapamil (Fig. 5B–E). Moreover, the expression of S100a6 and S100a4 was decreased when Ca²⁺ influx is pharmacologically inhibited in Abcc8^−/− mice (Fig. 5F and G). These results suggest that S100a6 and S100a4 expression in Abcc8^−/− β-cells is tightly correlated with [Ca²⁺]_i. Finally, coimmunostaining with S100A6 and ALDH1A3 revealed that 15.4 ± 2.1% of insulin-expressing cells in Abcc8^−/− mice express both S100A6 and ALDH1A3 (Fig. 4E).

To determine whether differentially expressed genes in Abcc8^−/− β-cells might share common upstream regulators, we used iRegulon, a program that identifies common DNA binding motifs in coexpressed genes (26). Analysis of the top 500 upregulated genes revealed binding sites for ASCL1, CEBPG, and RARG in many genes, with some genes, including Aldh1a3, containing binding sites for all three of these transcriptional regulators (Fig. 6A, Supplementary Table 2, and Supplementary Fig. 6). Interestingly, all three of these predicted regulators are upregulated in Abcc8^−/− β-cells (Fig. 6C). Conversely, analysis of the top 500 downregulated genes predicts binding sites for TEAD1, HNF1A, and ZFP647 (Fig. 6B, Supplementary Table 3, and Supplementary Fig. 7). However, the expression of these regulators is unchanged in Abcc8^−/− β-cells (Fig. 6C).

Because Ascl1 is 22-fold upregulated in Abcc8^−/−;MIP-GFP β-cells (Fig. 3B) and 24-fold upregulated in Abcc8^−/− islets (Fig. 6D) and has been shown by chromatin immunoprecipitation (ChIP) sequencing or ChIP PCR to bind near 51% of the predicted targets (38–40), we studied its responsiveness to changes in [Ca²⁺]_i in isolated islets. Consistent with Ascl1 being regulated by [Ca²⁺]_i, its expression increases in response to membrane depolarization, an effect that is reversed when Ca²⁺ influx is inhibited by verapamil, both in isolated islets (Fig. 6E) and in mice (Fig. 6F). These results support the idea that ASCL1 could play a central role in regulating gene expression in β-cells with chronically elevated [Ca²⁺]_i.

Abcc8^−/− β-cells have been previously shown to have altered [Ca²⁺]_i homeostasis, as would be expected in the absence of K_ATP channels (16,27–30). We similarly observed that Abcc8^−/− β-cells exhibit both persistent depolarization and an elevated mean [Ca²⁺]_i compared with Abcc8^+/+ islets (Fig. 1). Interestingly, although Abcc8^−/− β-cells exhibit Ca²⁺ oscillations in substimulatory glucose (Supplementary Fig. 1), as has been previously reported (29,30), we did not consistently observe oscillations in membrane potential, as has been observed in β-cells lacking K_ATP channels. This may be the result of the short duration of the recordings or to our use of the patch-clamp technique rather than intracellular microelectrodes, which limits the detection of oscillations in membrane potential (29). Regardless, Abcc8^−/− β-cells experience more persistent membrane depolarization and elevated mean [Ca²⁺]_i compared with wild-type β-cells, and the downstream effects on gene expression are likely to be the same. We also observed a transient drop in [Ca²⁺]_i in Abcc8^−/− islets after glucose stimulation (Fig. 1E) that was not seen in Abcc8^+/+ islets. This drop is the result of brief membrane hyperpolarization caused by transient activation of the sodium potassium ATPase (29) and has been observed previously in islets treated with K_ATP channel inhibitors (41) as well as in other K_ATP channel–deficient mice (29).

By performing RNA-Seq on purified pancreatic β-cell populations from both Abcc8^−/− and Abcc8^+/+ mice, we found that the expression of 4,208 genes was perturbed. Some of these genes, including S100a6, Myo7a, Pcp4, Cacng3, Mef2c, and Camk1d, are known to be involved in calcium binding and signaling (Fig. 3C). The S100 gene family, for instance, modulates the activity of other proteins on calcium binding (42). Moreover, S100A6 specifically promotes Ca²⁺-stimulated insulin release (36). Camk1d, Camkk1, and Camkk2, three protein kinases, Myo7a, a myosin motor protein, and Mef2d and Mef2c, are involved in Ca²⁺/calmodulin-dependent signaling or are regulated by [Ca²⁺]_i (43,44). However, the alteration of some Ca²⁺-regulated genes does not imply that all of the gene expression changes we observed are caused by changes in [Ca²⁺]_i. Although we have demonstrated that changes in S100a6, S100a4, and Ascl1 expression are tightly linked to membrane potential (Figs. 5B and E and 6E)_, a similar analysis has not been performed for the remaining genes. Thus, although it is likely that a substantial fraction of the dysregulated genes are the result of altered [Ca²⁺]_i homeostasis, other mechanisms, such as paracrine and/or neuronal signaling, may also be involved. Additional studies are necessary to determine which changes are directly attributable to chronically elevated [Ca²⁺]_i and which are the result of other signaling mechanisms and/or compensatory adaptations.

Recent studies have suggested that β-cell dedifferentiation contributes to the development of T2D (32,35). Among the genes affected in Abcc8^−/− β-cells are many known to be involved in maintaining β-cell identity and/or function, such as Ins1, Slc2a2, Neurod1, Gck, and Syt10 (Fig. 3C). However, some of the transcription factors previously associated with β-cell dedifferentiation (32), including Mafa, Pdx1, Nkx6.1, FoxO1, Ngn3, Oct4, and Nanog, are unchanged. This difference may explain the modest loss of β-cell identity observed in Abcc8^−/− mice, with only 2.21% of β-cells undergoing trans-differentiation to insulin/PP polyhormonal or PP monohormonal cells (Fig. 2C). Although treatment with a Ca²⁺ channel blocker did not statistically reverse the insulin/PP polyhormonal cells, its use resulted in a decrease in the percentage of Abcc8^−/− β-cells expressing the dedifferentiation marker ALDH1A3 (Fig. 4B), further suggesting that a chronic increase in [Ca²⁺]_i may contribute to loss of β-cell identity. In this regard, it is important to distinguish our results, which reflect the loss of K_ATP channels, from studies using K_ATP channel gain-of-function mutants (45,46). The current study examines the effects of elevated [Ca²⁺]_i in a nearly euglycemic setting, whereas the latter reflects decreased [Ca²⁺]_i in a hyperglycemic setting.

Our studies identified two genes, S100a6 and S100a4, that were highly upregulated in Abcc8^−/− β-cells and were also acutely regulated by treatment with depolarizing agents and a Ca²⁺channel blocker in isolated islets (Fig. 5B and E). The administration of a Ca²⁺ channel blocker to Abcc8^−/− mice only partially rescued the expression of S100a6 or S100a4 (Fig. 5F and G), but this may be the result of inadequate dosing or a treatment duration that was too short to overcome the effects of the Abcc8 gene knockout. However, it is also possible that these genes are only partially regulated by Ca²⁺ influx in vivo. Regardless, our data strongly suggest that S100a6 and S100a4 are regulated by Ca²⁺ in the β-cell and that they may serve as markers for β-cell excitotoxicity. In support of this, members of the S100 gene family, including S100A3, S100A6, S100A10, S100A11, and S100A16, have been associated with hyperglycemia in human islets (37), and S100A6 was specifically shown to be positively correlated with BMI in β-cells from donors with T2D (47). Neither gene was altered in two recent single-cell sequencing studies of people with T2D (47,48), but the sequencing depth was very shallow.

Perturbations in Ca²⁺ signaling also provide a compelling explanation for the disrupted islet morphology observed in K_ATP-channel knockout mice (Supplementary Fig. 5). [Ca²⁺]_i is required for maintenance of tight junctions and adherens junctions, and focal adhesion disassembly occurs in response to elevated [Ca²⁺]_i (49). Accordingly, among the downregulated genes are many that encode cell adhesion molecules, such as Ocln, Tln1, and Cdh1 (Fig. 3C), suggesting that a Ca²⁺-dependent gradual breakdown of cell-to-cell contacts may be responsible for the altered islet morphology.

Bioinformatics analysis of genes whose expression was increased in the Abcc8^−/− mice identified ASCL1, CEBPG, and RARG as possible upstream regulators (Fig. 6A). ASCL1, also known as MASH1, is important for neuronal commitment and differentiation (50). Retinoic acid receptors, including RARG, play a role in both endocrine cell development and maintenance of proper insulin secretion and β-cell mass (51). Finally, CEBPB, a closely related protein to CEBPG, represses the insulin promoter under conditions of chronically elevated glucose, and its accumulation in β-cells increases vulnerability to endoplasmic reticulum stress (52,53). Although our computational analysis is purely predictive, it is partially validated by the fact that 51% of the ASCL1 target genes and 60% of the RARG binding sites were established by ChIP-Seq or ChIP-PCR (38–40,54). Importantly, S100a4 is among the ASCL1 targets that have been experimentally validated (38–40), and our in-depth analysis of Ascl1 strongly supports its regulation by [Ca²⁺]_i, consistent with our model. Conversely, analysis of genes that are downregulated by [Ca²⁺]_i predicts binding sites for TEAD1, HNF1A, and ZFP647 (Fig. 6B). Although Tead1 gene expression is unaffected in Abcc8^−/− β-cells, the activity of TEAD1, a member of the Hippo pathway that interacts with YAP/TAZ to promote proliferation, is directly inhibited by Ca²⁺ (55). Furthermore, because a gene knockout of Hnf1a results in β-cell failure and diabetes (56), a predicted decrease in its activity could explain the decrease in genes involved in β-cell function in Abcc8^−/− β-cells.

Our studies, together with findings of others (15,16,27–30), indicate that the absence of functional K_ATP channels in Abcc8^−/− mice causes a chronic elevation of β-cell [Ca²⁺]_i, an impairment of β-cell identity, alterations in islet cell numbers and morphology, and very significant perturbations in β-cell gene expression. Although the elevation of [Ca²⁺]_i in Abcc8^−/− β-cells, caused by the total absence of K_ATP-channels, is persistent and severe, a less severe or persistent elevation in mean [Ca²⁺]_i may also occur in patients with prediabetes or T2D, when hyperglycemia chronically stimulates insulin secretion, or when a sulfonylurea is given as therapy for T2D. Interestingly, although Abcc8^−/− mice exhibit impaired glucose tolerance at 12 weeks of age, they do not go on to develop T2D, at least in the time frame studied here. This suggests that the impairments in β-cell identity and islet morphology are insufficient themselves to cause β-cell failure. However, β-cell excitotoxicity, acting in combination with other metabolic stresses, may accelerate the loss of functional β-cells, perhaps by dedifferentiation or glucolipotoxicity. Regardless, our findings provide new insights into transcriptional responses that occur within β-cells when they are subjected to chronically elevated [Ca²⁺]_i and suggest a model (Fig. 7) that involves the coordinated interactions of a network of Ca²⁺-responsive genes and upstream transcriptional modulators. This network may exist to balance both the positive and negative effects that [Ca²⁺]_i is known to have on β-cell function and mass.

RNA-Seq data are available in ArrayExpress (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-4726.

Acknowledgments. The authors thank the Vanderbilt Islet Procurement and Analysis Core for performing multiple islet isolations, Vanderbilt Technologies for Advanced Genomics (VANTAGE) for performing cell sorting and RNA-Seq, the Vanderbilt Cell Imaging Shared Resource for helping to acquire images, the Vanderbilt Tissue Pathology Shared Resource for processing and sectioning paraffin-embedded tissues, and the Vanderbilt Transgenic Mouse/ESC Shared Resource for reconstituting mice with the Abcc8 null allele. The authors also thank Guoqiang Gu and Roland Stein of Vanderbilt University for reading the manuscript and Jody Peters, Katherine Boyer, and Lauren Stiehle, also of Vanderbilt University, for technical assistance.

Funding. These studies were partly supported by National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases grants DK-72473 and DK-89523 to M.A.M. and DK-020593 to Vanderbilt Islet Procurement and Analysis Core, Vanderbilt Cell Imaging Shared Resource, and Vanderbilt Transgenic Mouse/ESC Shared Resource and by National Cancer Institute grant CA-68485 to Vanderbilt Cell Imaging Shared Resource and Vanderbilt Transgenic Mouse/ESC Shared Resource.

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

Author Contributions. J.S.S. and M.A.M. designed the study and wrote the manuscript. J.S.S., H.W.C., and J.T.O. performed experiments and analyzed data. J.-P.C. performed RNA-Seq data processing, alignment, and differential gene expression analysis and edited the manuscript. M.T.D., P.K.D., and D.A.J. performed membrane potential measurements and calcium imaging and analyzed the data. A.B.O. contributed to study design and edited the manuscript. M.A.M. supervised and funded the experiments. 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.