Research Article

Pancreatic β cell enhancers regulate rhythmic transcription of genes controlling insulin secretion

See allHide authors and affiliations

Science  06 Nov 2015:
Vol. 350, Issue 6261, aac4250
DOI: 10.1126/science.aac4250

The clockwork of insulin release

In healthy people, blood glucose levels are maintained within a narrow range by several physiological mechanisms. Key among them is the release of the hormone insulin by pancreatic β cells, which occurs when glucose levels rise after a meal. In response to insulin, blood glucose is taken up by tissues that need fuel, such as muscle. β cells can anticipate the body's varying demand for insulin throughout the 24-hour day because they have their own circadian clock. How this clock controls insulin release has been unclear. Perelis et al. now show that the activity of transcriptional enhancers specific to β cells regulates the rhythmic expression of genes involved in the assembly and trafficking of insulin secretory vesicles (see the Perspective by Dibner and Schibler).

Science, this issue p. 10.1126/science.aac4250; see also p. 628

Structured Abstract

INTRODUCTION

The circadian clock is a molecular oscillator that coordinates behavior and physiology in anticipation of the daily light cycle. Desynchrony of circadian cycles, through genetic or environmental perturbation, contributes to metabolic disorders such as cardiovascular disease, obesity, and type 2 diabetes. We previously showed that disruption of the clock transcription factors CLOCK and BMAL1 in the pancreas causes hypoinsulinemic diabetes in mice. The mechanism(s) linking clock dysfunction to pancreatic β cell failure and the means by which CLOCK and BMAL1 affect glucose metabolism in the whole organism are not well understood.

RATIONALE

The circadian system helps to maintain glucose homeostasis across the sleep-wake cycle. This system requires cross-talk between the master clock in the central nervous system, which coordinates feeding and sleep, and peripheral tissue clocks, which synchronize behavior with the storage, mobilization, and synthesis of glucose. Although it is clear that clocks within distinct organs participate in glucose turnover, the molecular basis for time-of-day variation in organismal glucose responsiveness is still not understood. Here, we combined genome-wide analyses with gene targeting in mice to study the impact of the cell-autonomous clock on β cell function.

RESULTS

We found that cell-autonomous expression of CLOCK and BMAL1 in pancreatic islets isolated from wild-type mice generates robust 24-hour rhythms of glucose- and potassium chloride–stimulated insulin secretion ex vivo. About 27% of the β cell transcriptome exhibited circadian oscillation. Many of these transcripts correspond to genes coding for proteins that are involved in the assembly, trafficking, and membrane fusion of vesicles that participate in insulin secretion. Chromatin immunoprecipitation sequencing revealed that CLOCK and BMAL1 regulate cycling genes in β cells by binding at distal regulatory elements distinct from those controlling the circadian transcription of metabolic gene networks within the liver. The regulatory sites of cycling genes in the β cell resided primarily within transcriptionally active enhancers that were also bound by the pancreatic transcription factor PDX1. Finally, we found that in islets from adult mice, Bmal1 ablation either in vivo or ex vivo abrogates nutrient-responsive insulin secretion, demonstrating clock control of pancreatic β cell function throughout adult life.

CONCLUSION

Our results show that local clock-driven genomic rhythms program cell function across the light-dark cycle, including the priming of insulin secretion within limited time windows each day. Cell type–specific transcriptional regulation by the clock localizes to rhythmic enhancers that are unique to the β cell. Thus, our findings uncover a transcriptional process through which the core clock aligns physiology with the light cycle, revealing pathways that are important in both health and disease states such as type 2 diabetes.

β cell–specific enhancers control the rhythmic transcription of genes linked to insulin secretion.

Peripheral clocks maintain glucose homeostasis across the sleep-wake cycle by gating β cell insulin secretion through genome-wide transcriptional control of the assembly and trafficking of insulin secretory vesicles. Clock transcription factors bind within cell type–specific enhancer neighborhoods of cycling genes, revealing the mechanisms that synchronize rhythmic metabolism at transcriptional and physiologic levels across the light-dark cycle.

Abstract

The mammalian transcription factors CLOCK and BMAL1 are essential components of the molecular clock that coordinate behavior and metabolism with the solar cycle. Genetic or environmental perturbation of circadian cycles contributes to metabolic disorders including type 2 diabetes. To study the impact of the cell-autonomous clock on pancreatic β cell function, we examined pancreatic islets from mice with either intact or disrupted BMAL1 expression both throughout life and limited to adulthood. We found pronounced oscillation of insulin secretion that was synchronized with the expression of genes encoding secretory machinery and signaling factors that regulate insulin release. CLOCK/BMAL1 colocalized with the pancreatic transcription factor PDX1 within active enhancers distinct from those controlling rhythmic metabolic gene networks in liver. We also found that β cell clock ablation in adult mice caused severe glucose intolerance. Thus, cell type–specific enhancers underlie the circadian control of peripheral metabolism throughout life and may help to explain its dysregulation in diabetes.

The mammalian circadian system is organized hierarchically and is driven by cellular transcriptional oscillators that coordinate behavior and metabolism with the light-dark cycle. Specifically, CLOCK/BMAL1 within the forward limb of the clock induces the expression of repressors (PERs/CRYs) in the negative limb and stabilizing factors (ROR/REV-ERB) in a cycle that repeats itself every 24 hours (1, 2). A transformation in our understanding of clock function emerged from the discovery of autonomous circadian oscillation within individual tissues, and even in fibroblasts, ex vivo (3). Molecular rhythms play a critical role in systemic health, as indicated by observations that disruption of central and peripheral clocks can alter body weight and glucose homeostasis. However, there has been a major gap in our understanding of how the molecular clock synchronizes transcription in distinct peripheral tissues to maintain overall physiological homeostasis (48).

Genome-wide analyses in liver indicate extensive rhythmicity of processed RNAs and noncoding enhancer RNAs (eRNAs) that are dependent on temporal binding of circadian transcription factors to both promoters and enhancers (911). Yet the circadian clock exerts different effects on glucose metabolism within liver and other peripheral tissues. Thus, we sought to examine the genomic mechanism of clock control of pancreatic β cell insulin secretion (5, 6). Here, we define the targets of clock transcriptional regulation within the β cell and investigate the impact of clock disruption on the temporal control of insulin secretion and glucose homeostasis.

The β cell clock produces rhythmic insulin secretion and secretory gene transcription

First, to determine whether transcriptional oscillations in pancreatic islets give rise to rhythmic islet physiology, we examined the phase dependence of pancreatic islet function by analyzing nutrient-induced insulin secretion in parallel with live-cell clock oscillation in islets from Per2Luc reporter mice (12). After synchronization with forskolin (6, 13), we assessed insulin secretion every 4 hours in individual groups of five islets at each time point over the ensuing 72-hour window (fig. S1A; materials and methods) and observed a striking self-sustained, time of day–dependent variation in the magnitude of response to stimulatory concentrations of both glucose and KCl, which triggers insulin exocytosis through direct depolarization of the β cell (Fig. 1A). Intracellular insulin content did not cycle (fig. S1B) despite rhythmic glucose-stimulated insulin secretion (GSIS) (Fig. 1A), consistent with circadian regulation at a step after translation of insulin. We further confirmed that GSIS rhythms were autonomous by monitoring insulin secretion after forskolin synchronization at times corresponding to the nadir (36 hours after forskolin shock) and zenith (48 hours after forskolin shock) of the wild-type GSIS rhythm in islets isolated from PdxCreER;Bmal1flx/flx mice (Fig. 1B and fig. S1C), which when treated with tamoxifen ex vivo displayed >60% reduction in Bmal1 expression (fig. S1D). Vehicle-treated islets displayed significantly higher GSIS at the zenith than at the nadir, whereas tamoxifen-treated islets exhibited constitutively low levels of insulin secretion (Fig. 1B). Together, these data suggest that the islet molecular clock gates the rhythmic secretory response downstream of membrane depolarization.

Fig. 1 Isolated pancreatic islets display rhythmic insulin secretion and transcription of secretory genes in mice and humans.

(A) Bottom: Glucose- and KCl-stimulated insulin secretion in synchronized wild-type mouse islets across 3 or 2 consecutive days, respectively (n = 3 replicate sets of islets pooled from 6 to 9 mice each). Top: Bioluminescence monitoring (counts/s) in islets from Per2Luc reporter mice was performed in parallel. (B) Glucose-stimulated insulin secretion in ethanol- or tamoxifen-treated islets from PdxCreER;Bmal1flx/flx mice at the nadir (36 hours after forskolin shock) and zenith (48 hours after forskolin shock) of cyclic insulin secretion in wild-type islets from Fig. 1A (n = 4 islet pools per time point, three replicates per islet pool). Ethanol-treated islets displayed significant difference in GSIS comparing 36 to 48 hours (P = 0.038), whereas tamoxifen-treated islets did not (P = 0.974). (C) Top: Bmal1 and Rev-erbα RNA expression. Middle: Heat map of all cycling genes identified by eJTK_CYCLE analysis. Bottom: Significantly enriched KEGG ontology pathways shown within the cycling gene set. (D) Left: Peak phase expression (hours after forskolin shock) of cycling genes in synchronized wild-type islets that were also altered in PdxCre;Bmal1flx/flx islets at ZT2. Right: Log2 change in expression in PdxCre;Bmal1flx/flx (KO) islets relative to Bmal1flx/flx (control) at ZT2 for subset of genes relevant to insulin secretion. (E) Heat map showing expression patterns of cycling trafficking and exocytosis genes in synchronized human islets. (F) Mapping of cycling RNAs in both human and mouse islets onto the “Insulin Secretion” KEGG pathway. All values in (A) and (B) represent mean ± SEM; *P < 0.05, ***P < 0.001.

We next sought to examine the genome-wide effect of rhythmic transcription on insulin secretory dynamics by performing RNA sequencing (RNA-seq) over two circadian cycles in RNA isolated from wild-type islets synchronized ex vivo (fig. S1A; materials and methods). We analyzed polyadenylated RNAs using eJTK_CYCLE (14), a modified nonparametric algorithm with increased sensitivity for detecting cycling transcripts. We detected a total of 3905 cycling transcripts (Bonferroni-corrected P < 0.05), which accounted for ~27% of all expressed transcripts within the islet that met a minimum mean expression threshold of 10 normalized counts (Fig. 1C). As expected, we observed high-amplitude rhythms for the core clock transcription factors, including Bmal1, Clock, Npas2, Per2, Cry1, Rev-Erbα, and Rorα, with Bmal1 and its repressor Rev-Erbα displaying antiphasic expression (Fig. 1C) (15).

To determine the identity of functional circadian gene networks in the islet, we tested for overrepresentation of defined KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways among rhythmic RNAs. We observed enrichment of factors mediating vesicle exocytosis, which suggests that daily variation in insulin secretory capacity arises from genomic regulation of the transport and release of peptidergic hormone (Fig. 1C and table S1). Overrepresented pathways in the circadian transcriptome included factors involved in (i) vesicle budding, including genes encoding the COPII coat proteins (Sec24a and Sec31a), which mediate vesicle budding from the endoplasmic reticulum (16, 17); (ii) cargo trafficking, specifically the motor proteins (Kif1b, Myo9a, and Dync2h1) that enable vesicle transport along cytoskeletal filaments (18); and (iii) vesicle tethering and fusion to the plasma membrane, including v- and t-SNAREs such as Vamp1, Vamp5, Vamp8, Stx1a, Stx4a, and Stx8 (19, 20). In addition to the cycling of RNAs that encode factors involved in insulin exocytosis, we also identified rhythmic RNA expression of insulinotropic signals involved in vesicle movement and membrane fusion, including (i) targets of cAMP/EPAC (cyclic adenosine monophosphate/exchange protein activated by cAMP) signaling (Pclo, Rims2, Rab3b, Rap1a, Rap1b, Rapgef2, Rapgef6), which mediate vesicle docking and fusion to the plasma membrane (21, 22); (ii) Ca2+-sensing synaptotagmins (Syt11, Syt14, Syt16), which stimulate membrane fusion of synaptic vesicles (23, 24); and (iii) calmodulin-dependent protein kinases (Camk1, Camk4, Camkk2, Camk2g), which regulate vesicle exocytosis and recycling (25). Lastly, we detected significant oscillation in targets of phosphoinositide signaling, including protein kinase C (Prkca, Prkcb) (26), exocyst actin-interacting factors such as Exoc1/Sec3 (27), and the cytoskeletal filament–remodeling Rho guanosine triphosphatases (GTPases) Rho, RhoA, RhoB, and RhoC (18). Collectively, cycling of RNAs that encode factors involved in insulin exocytosis and signaling components reveals a genomic basis for circadian variation in insulin secretion.

To further understand the physiologic function of tissue-specific rhythmic gene transcription, we compared genome-wide rhythms of RNA expression in wild-type islets to those in pancreas-specific clock mutant mice (PdxCre;Bmal1flx/flx), which exhibit severe hypoinsulinemic diabetes due to defects downstream of glucose metabolism and mitochondrial respiration (fig. S2) (6). We performed RNA-seq using RNA isolated from PdxCre;Bmal1flx/flx and control littermate islets at the start of the light phase [zeitgeber time 2 (ZT2), the time of maximal GSIS impairment] (fig. S3A) (6). We identified changes in the expression of 1757 genes in islets isolated from clock mutant animals relative to littermate controls (Bmal1flx/flx), including transcripts that were both decreased (1074) and increased (683) in expression, consistent with actions of the clock as both an activator and repressor of gene expression [false discovery rate (FDR)–adjusted P < 0.05] (fig. S3, B and C).

Many of the RNAs that were altered in islet clock knockout mice were identified as cycling RNAs in wild-type islets; overall, a total of 720 oscillating genes exhibited altered expression in animals with disrupted pancreatic clock function (fig. S3C), indicating an autonomous role of the islet clock in the rhythmic transcriptional regulation of insulin secretion. Among the most significantly changed RNAs were factors in the negative limb of the core clock containing the canonical E-box transcription motif, in addition to circadian PAR bZip transcription factors including Per2, Rev-Erbα (Nr1d1), Tef, and E4bp4 (Nfil3) (Fig. 1D). We also found a broad range of alterations in cycling genes that are circadian outputs and grouped by KEGG annotation into exocytosis networks similar to those described for the wild-type islets, including genes encoding factors involved in trafficking, such as the vacuolar protein sorting factors Vps13b and Vps13c; Myo9, the motor protein involved in vesicular transport; the kinesin transport factor Kif21; and the small GTPase Rab11, a factor in trans-Golgi vesicular biogenesis (28) (Fig. 1D, fig. S3, E and F, and table S1). Ontology analysis also identified genes related to vesicle tethering and fusion as altered in clock-deficient islets, including the conserved exocyst component Exoc1/Sec3, cAMP/EPAC-controlled Rims2 and Pclo, and the synaptotagmin Syt14 (Fig. 1D); islet genes involved in glucose sensing were unchanged (table S2). Whereas the complete set of cycling RNAs displayed broadly distributed peak phases (fig. S3D), the majority of exocytosis-related RNAs that were differentially expressed in clock mutants exhibited peak expression at two distinct phases (48 and 60 hours after forskolin shock) (Fig. 1D). Although this suggests that these genes may represent direct targets of CLOCK/BMAL1 and/or a clock repressor (REVERBα/β or E4BP4), nascent RNA-seq studies indicate that peak circadian mRNA phases are not directly correlated with nascent transcription (11). Collectively, sequencing results indicate that secretory pathway genes represent a major output of the islet clock.

To determine whether the rhythmic islet transcriptome is conserved from mouse to humans, we performed RNA-seq in RNA isolated from synchronized human islets (fig. S4A). Human islets displayed characteristic circadian patterns in the expression of core clock components BMAL1 and REV-ERBα (fig. S4B) (29) and genome-wide rhythmic patterns in the transcriptome with 1800 cycling RNAs (Bonferroni-corrected P < 0.05) (fig. S4B). Although striking differences have been described between mouse and human islet cell composition and cytoarchitecture (30), the expression of key genes involved in insulin release is conserved between species (30). Remarkably, 481 of the rhythmic human islet genes were orthologous to those in mouse islets (fig. S4C), including factors involved in exocytosis, trafficking, and fusion (Fig. 1, E and F, and fig. S4C). Mapping cycling human islet RNAs onto KEGG-curated human insulin secretion pathways revealed regulation of heterotrimeric G protein–coupled receptor (GPCR), cAMP, Ca2+, and phosphoinositide-responsive signaling molecules important in nutrient response and hormone release (Fig. 1, E and F). Specifically, these included GNAQ (Gq protein), RIMS2 and PCLO (insulin vesicle–associated), CAMK2G (calmodulin-activated protein kinase), and PLCB4 (phospholipase C), all of which were also rhythmic in mouse islets (Fig. 1, E and F, and fig. S4D). Circadian gene regulation in the endocrine pancreas of both mice and humans thus converges on the late secretory pathway, demonstrating conservation of clock control of rhythmic tissue function across species.

BMAL1 and CLOCK bind near cell type–specific enhancers in pancreatic β cells

Because our genome-wide RNA sequencing studies indicate that genomic regulation by the clock gives rise to rhythmic insulin secretion, we next sought to analyze how core circadian transcription factors regulate this process by analyzing the extent of binding by BMAL1 and CLOCK to rhythmically expressed genes. In this context, cistrome studies have recently characterized β cell transcriptional hubs encoding genes that program both development and function (31), revealing colocalization within regions of accessible chromatin (H2A.Z) and active enhancers [monomethylated histone 3 Lys4 (H3K4Me1) colocalized with acetylated histone 3 Lys27 (H3K27Ac)] containing binding sites for lineage-determining transcription factors (PDX1, MAFB, FOXA2, NKX6-2, and NKX2-2) (3133). To determine the intersection between circadian transcription factor regulation and genomic binding at regulatory loci, we performed chromatin immunoprecipitation sequencing (ChIP-seq) in the mouse β cell line Beta-TC6 (Fig. 2A). As expected, we found that both BMAL1 and CLOCK physically bound to sites at core clock and other gene targets in β cells that were enriched for the canonical CACGTG E-box motif, often occurring in tandem, as previously reported at BMAL1 binding sites in liver (fig. S5A) (P = 10−38 and P = 10−91, respectively) (9, 34). Moreover, we also observed a correlation between the genome-wide binding of BMAL1 and CLOCK (fig. S5B). A representative UCSC Genome Browser track at the Rev-erbα (Nr1d1) locus is shown in Fig. 2A, revealing colocalization of BMAL1 and CLOCK binding sites at three distinct regulatory regions at the Nr1d1 locus, including within the promoter region [shaded light orange and defined as within 2 kb of the transcription start site (TSS)] and within intragenic and intergenic distal enhancer regions (shaded light green and defined as binding regions greater than 2 kb from the TSS). Histone markers representing active and accessible chromatin (H3K27Ac and H2A.Z, respectively) localized to the same promoter and enhancer regions within the Nr1d1 locus, indicating active transcriptional regulation by BMAL1 and CLOCK (Fig. 2A).

Fig. 2 BMAL1 and CLOCK bind to cycling genes at distal regulatory sites.

(A) Top: Model of transcriptional targets and chromatin modifications for ChIP-seq experiments. Bottom: UCSC Genome Browser tracks at Nr1d1 (Rev-erbα) locus in β cells. Maximum track heights within viewable window are indicated to the right of each factor. Shaded columns are described in text. (B) Distribution of BMAL1 and CLOCK peaks at cycling and noncycling gene targets expressed in islets. Binding sites at cycling genes are separated into promoter proximal and distal sites (<2 kb and >2 kb from TSS of nearest gene, respectively). (C) KEGG ontology terms enriched in cycling and noncycling BMAL1 and CLOCK target genes.

To determine whether BMAL1 and CLOCK directly regulate the oscillating transcripts identified in the synchronized wild-type islets (Fig. 1C), we evaluated the overlap between the BMAL1 and CLOCK cistromes with genes oscillating in the wild-type islets. Among binding sites localized to expressed RNAs, 30% (862 binding sites) and 29% (330 binding sites) of the BMAL1 and CLOCK targets, respectively, exhibited rhythmic transcription in synchronized wild-type islets (Fig. 2B), which collectively accounted for 742 cycling direct target genes, of which 165 were differentially expressed in Bmal1 knockouts (fig. S5C). These findings suggest direct BMAL1 and CLOCK regulation. Moreover, KEGG analysis of the direct gene targets in mouse islets that were present in BMAL1 and CLOCK cistromes revealed enrichment in pathways related to protein export, COPII-mediated vesicle budding from the endoplasmic reticulum, and SNARE vesicular transport and membrane fusion, in the cycling set relative to the noncycling set of BMAL1- and CLOCK-controlled transcripts (KEGG pathways listed in order of descending –log10 P values in Fig. 2C and table S1). Together, these findings identify direct transcriptional targets of CLOCK/BMAL1 that mediate rhythmic islet physiology.

BMAL1 binds to distinct enhancers in liver and pancreatic β cells

Given evidence for tissue-specific regulation at enhancers as a predominant mode of circadian regulation in liver (10), we next analyzed the binding position of BMAL1 and CLOCK in relation to the TSS of rhythmic genes in β cells. Because genome-wide promoter activity studies and epigenetic characterization of mammalian regulatory regions have indicated that the majority of core promoter activity is localized within 2 kb of the TSS (3537), we classified binding events occurring within 2 kb of the nearest annotated gene TSS as promoter-proximal. Surprisingly, we found that BMAL1 and CLOCK bind predominantly at distal sites (defined as greater than 2 kb from the TSS) rather than at proximal promoter sites (defined as less than 2 kb from the TSS) of rhythmically regulated genes (Fig. 2B and fig. S5D). This finding suggests that the islet clock transcription factors affect rhythmic physiology through binding to distal regulatory sites, an observation concordant with the general finding that transcription factors exert physiologic effects through regulation within tissue-specific enhancers (31).

Although clock factors have been shown to exert distinct physiologic functions across tissues, a major gap remains in understanding the underlying genomic mechanisms accounting for these tissue-specific functions. To determine whether BMAL1 regulates rhythmic genes through unique sites in the β cell compared to liver, the tissue in which the circadian cistrome has been best characterized (911, 34, 38), we compared sites of BMAL1 occupancy in the β cells to a published set of liver BMAL1 peaks (9). Unexpectedly, although there was a considerable overlap of genes identified as direct BMAL1 binding targets in β cells and liver (40%, 1063 genes out of 2660 total β cell target genes) (Fig. 3A), BMAL1 binding at the regulatory regions of those shared gene sets localized to distinct sites (Fig. 3A). Remarkably, in comparing genome-wide binding patterns, we observed common locations of binding in only 4% of these instances; thus, BMAL1 binding at all β cell–defined sites is uncorrelated with BMAL1 binding at all liver-defined sites (Fig. 3A and fig. S5E) (R2 = 0.01874 and 0.03286 for BMAL1 binding at β cell and liver sites, respectively), whereas binding at canonical E-box sites in Per2, Cry1, and Dbp was similar between tissues (fig. S5F). Furthermore, when we compared the shared set of BMAL1 target genes that were rhythmic in islets and also reported to be rhythmic at the mRNA level in liver, BMAL1 likewise bound to unique sites (9) (Fig. 3B). These data suggest convergent regulation of BMAL1 targets in β cell and liver sites through divergent regulatory elements.

Fig. 3 β cell circadian cistrome is determined by tissue-specific enhancer repertoire.

(A) Top: Overlap of genes identified at BMAL1 binding sites in β cells and liver. Middle: Scatterplots show BMAL1 binding in liver (y axis) versus β cells (x axis) within 500-bp windows surrounding peaks identified in each tissue. Bottom: Browser track view of BMAL1 binding in β cells and liver at the Gpr137 locus. (B) Top: Overlap of cycling and direct BMAL1 target genes in β cells that have been reported to cycle in liver. Bottom: Cycling BMAL1 direct target genes containing shared or unique binding sites in β cells and liver. (C) Top: Heat maps comparing binding of indicated factors within 1-kb windows surrounding promoter (3492) and enhancer (5771) localized H3K4Me2 peaks annotating to genes containing cycling RNAs in wild-type islets. Bottom: Histograms summarizing normalized tag counts for H3K27Ac (in β cells and liver) and PDX1 (in β cells) across 6-kb span centered at all β cell H3K4Me2 peaks. (D) Box-and-whisker plots (whiskers represent interquartile range 1.5) comparing BMAL1 binding in β cells and liver at loci corresponding to H3K4Me2 peaks defined in heat maps. Poised enhancers refer to H3K4Me2 sites that do not colocalize with H3K27Ac, whereas active enhancers are defined as H3K4Me2 sites colocalized with H3K27Ac. ***P < 0.0001 by Mann-Whitney nonparametric unpaired t test. All reported ChIP-seq tag counts were normalized per 107 reads.

Because BMAL1 predominantly bound at distal regulatory regions in islets that were divergent from liver, we next sought to examine the chromatin regulatory context at all cycling genes in β cells. To do so, we defined all regulatory regions at cycling loci using dimethylated histone 3 Lys4 (H3K4Me2) peaks within 2 kb of the TSS (promoter) and more than 2 kb from the TSS (enhancer) (Fig. 3C). The binding patterns of the histone marks H3K4Me2, H2A.Z, and H3K27Ac (which represent promoter/enhancer regulatory regions, chromatin accessibility, and enhancer activity, respectively), as well as binding of the lineage-determining transcription factor for β cells PDX1 (39) at promoters and enhancer regions, are displayed in heat maps in Fig. 3C. Hierarchical clustering revealed that all epigenetic and PDX1 signals at promoter and distal enhancer regions at cycling genes more frequently displayed correlated binding than did H3K27Ac at these loci in liver, as indicated by the clustering dendrogram (Fig. 3C). Accordingly, the genomic coordinates in liver corresponding to enhancers defined in β cells displayed markedly reduced H3K27Ac, indicating that these enhancers defined specific loci of β cell regulation (Fig. 3C). Frequent binding of PDX1 at distal enhancer loci suggested that tissue specificity arose from early events in islet cell development (Fig. 3C) (40). Consistent with tissue-specific clock transcription factor regulation at β cell regulatory regions, BMAL1 displayed a greater degree of binding to promoter and enhancer regions at cycling genes in β cells than in liver, particularly at active enhancers containing both H3K4Me2 and H3K27Ac (Fig. 3D). These results indicate that clock transcription factors generate unique patterns of rhythmic RNA expression across tissues according to the pattern of cell-specific enhancer repertoires and provide a molecular basis for the distinct and opposing effects of the clock in pancreas and liver, which primarily affect postprandial and fasting glucose metabolism, respectively (5, 6).

β cell clock disruption during adulthood impairs insulin secretion and causes diabetes

To test the hypothesis that clock genes modulate genome-wide transcription on a daily basis throughout adult life, we examined the impact of acute clock inhibition on glucose metabolism in PdxCreER;Bmal1flx/flx mice at 2 to 3 months of age after administration of tamoxifen, which abrogates BMAL1 expression exclusively within the β cell (fig. S6) (41). Although these mice displayed normal wheel-running rhythms, period length, food intake, and body weight (fig. S7) relative to littermate tamoxifen-treated PdxCreER and Bmal1flx/flx animals, they developed significant hyperglycemia, impaired glucose tolerance, and hypoinsulinemia within 10 to 14 days after tamoxifen administration during both the day (ZT2) and night (ZT14) (Fig. 4, A and B, and fig. S8), despite no differences in islet mass (fig. S9A). These results establish that circadian disruption in fully differentiated cells is sufficient to induce metabolic disease, independent of effects on early development.

Fig. 4 Clock disruption in β cells during adulthood causes acute hypoinsulinemic diabetes in mice.

(A) Blood glucose levels in ad libitum–fed PdxCreER;Bmal1flx/flx mice and littermate controls before and after tamoxifen administration (n = 6 to 12 mice per genotype). (B) Glucose tolerance and insulin secretion at ZT2 after intraperitoneal glucose administration in PdxCreER;Bmal1flx/flx mice and littermate controls before and after tamoxifen treatment (n = 4 to 11 mice per genotype). Inset represents area under the curve for glucose (104 mg/dl per 120 min). (C) Model of intersecting pathways driving insulin exocytosis. Nutrient, Gs, and Gq receptor signaling that are used to stimulate insulin secretion in (D) to (F) are highlighted. (D to F) Glucose- and nutrient-stimulated (D), cyclase pathway–stimulated (E), and catecholamine-stimulated (F) insulin secretion in islets isolated from tamoxifen-treated PdxCreER;Bmal1flx/flx and control mice (n = 3 to 8 mice per genotype, three repeats per mouse). Inset of (F) shows ratiometric determination of intracellular Ca2+ using Fura2-AM dye in Beta-TC6 cells in response to insulin secretagogs (n = 3 replicates per condition). All values represent mean ± SEM. For (B), asterisks denote significance between Bmal1flx/flx and PdxCreER;Bmal1flx/flx; plus symbols denote significance between PdxCreER and PdxCreER;Bmal1flx/flx. */+P < 0.05, **/++P < 0.01, +++P < 0.001.

We further found that islets isolated from tamoxifen-treated PdxCreER;Bmal1flx/flx mice secreted significantly less insulin relative to littermate controls when exposed to (i) 20 mM glucose; (ii) 10 mM leucine combined with 2 mM glutamine, which bypasses glycolysis to trigger mitochondrial adenosine triphosphate (ATP) production; or (iii) 30 mM KCl, which chemically closes the KATP channel, thus inducing membrane depolarization distal to glucose metabolism and an increase in cytosolic calcium (Fig. 4, C and D); glucose-stimulated calcium influx was unchanged between the two groups (fig. S9B). Remarkably, these data are consistent with our observation that circadian oscillation in insulin secretory capacity is regulated downstream of KATP channel closure. Consistent with impaired Gs-coupled GPCR signaling, PdxCreER;Bmal1flx/flx islets also secreted significantly less insulin than controls in response to glucose together with the cyclase agonist forskolin and the nonhydrolyzable cAMP analog 8-br-cAMP (Fig. 4, C and E).

Finally, we tested the response to Gq-type GPCR signaling by stimulating islets with the muscarinic agonist carbachol, the diacylglycerol (DAG) mimetic phorbol 12-myristate 13-acetate (PMA), and the Ca2+ ionophore ionomycin. Surprisingly, carbachol and PMA restored insulin secretion in PdxCreER;Bmal1flx/flx islets (Fig. 4, C and F), whereas the response to ionomycin, which raised intracellular Ca2+ in β cells (Fig. 4F), was significantly reduced in mutants, indicating that the DAG arm of the Gq pathway restored second messenger signaling. DAG regulates exocytosis in β cells and other neurosecretory cells by acting as a ligand for the vesicle priming protein Munc13-1 (42) and protein kinase C (PKC), which phosphorylates and activates SNAP25 and MUNC18-1 to initiate vesicle fusion (43). We observed rhythmic RNA expression of the PKC-activating Rho and Rap GTPases Rho, Rhoa, Rhob, and Rap1a in wild-type islets, which raises the intriguing possibility that elevated DAG concentrations in carbachol- or PMA-treated islets pharmacologically bypass a deficiency in Rho- and Rap-mediated signaling. Together, these results demonstrate that pharmacologic Gq agonism reverses the insulin secretory blockade induced by clock disruption, indicating convergence of cholinergic and phosphoinositol signaling within the β cell in temporal homeostasis.

Discussion

We have established the genome-wide basis of coordinated cross-tissue circadian oscillation through integrated studies of β cell physiology and cistrome regulation. We focused on pancreatic β cells as a paradigm of peripheral clock regulation of metabolism because clock disruption in the islet leads to severe hypoinsulinemic diabetes and has direct application to understanding human tissue rhythms and disease. Although the circadian system functions as a hierarchy in the intact animal, our results reveal organ-autonomous cycles of nutrient-coupled insulin secretion in isolated islets ex vivo that result in a high amplitude in maximal glucose responsiveness, suggesting that the clock primes insulin secretion within limited windows each day. We further find that circadian-driven transcriptional oscillation within pancreas drives daily waves of expression of genes involved in the biogenesis, transport, and signal-induced activation of peptide exocytosis, indicating that genomic rhythmicity gives rise to tissue-specific function of the clock. Our observations suggest that autonomous transcription cycles enable islet cells to anticipate diurnal changes in the demand for insulin.

Cistromic profiling within the β cell provides further insight into the regulation of tissue-specific genome oscillation. We found that CLOCK and BMAL1 bind predominantly within distal tissue-specific enhancers rather than the promoters of cycling genes in proximity to H3K4Me2-, H2A.Z-, and H3K27Ac-modified nucleosomes that are co-occupied by PDX1. Consistent with tissue specificity in enhancer selection across cell types, BMAL1 binding in islet cells was highly divergent from liver, even within shared cycling target genes across the two tissues. These findings suggest that the establishment of accessible chromatin domains during development is a critical determinant of the available regulatory sites for clock-mediated transcription across distinct cell types. Further studies will be needed to define the underlying mechanisms through which divergent tissue-specific regulation gives rise to convergent oscillation of rhythmic genes.

Finally, our studies using chemically inducible genetic clock inactivation demonstrate that inhibition of circadian signaling in differentiated β cells acutely blocks peptide exocytosis and leads to hypoinsulinemic diabetes, providing evidence that clock function throughout adult life is necessary for glucose constancy. An intriguing possibility is that cell-autonomous genomic rhythms may regulate peptidergic secretion across diverse secretory and neuronal cell types, coordinating the availability of signaling molecules with the sleep/wake cycle each day. Furthermore, given the association between circadian and sleep disruption with human metabolic disease in both clinical (44, 45) and genetic (46) studies, the finding that circadian transcription is conserved in human islets suggests that clock dysregulation in β cells may contribute to the pathogenesis of human diabetes. The demonstration of coordinated circadian genomic and physiologic rhythms in pancreatic β cell insulin exocytosis and its control by enhancers provides a new window to understanding how geophysical and physiologic time are transcriptionally coupled, and how errors in this process may contribute to diabetes and other metabolic disorders.

Materials and methods

Animals

Male wild-type C57BL/6J mice were purchased from the Jackson Laboratory. Per2Luc (12) and PdxCre;Bmal1flx/flx (6) mice were produced and maintained on a C57BL/6J background at the Northwestern University Center for Comparative Medicine. Bmal1flx/flx mice (47) were crossed with PdxCreER transgenic mice (kindly provided by D. Melton, Harvard University) (48) to generate PdxCreER;Bmal1flx/flx mice, as well as Bmal1flx/flx and PdxCreER littermate controls. Unless otherwise stated, mice were maintained on a 12:12 light:dark (LD) cycle with free access to regular chow and water. All animal care and use procedures were conducted in accordance with regulations of the Institutional Animal Care and Use Committee at Northwestern University.

Islet isolation, insulin secretion assays, and in vitro islet synchronization

Mouse pancreatic islets were isolated via bile duct collagenase digestion (Collagenase P, Sigma) and Ficoll gradient separation and left to recover overnight (16 hours) at 37°C in RPMI 1640 with 10% fetal bovine serum (FBS), 1% l-glutamine, and 1% penicillin/streptomycin. For standard insulin release assays, five islets were statically incubated in Krebs-Ringer buffer (KRB) and stimulated for 1 hour at 37°C with various glucose concentrations, 30 mM KCl, 2.5 μM forskolin, 1 mM 8-Br-cAMP, 10 mM l-leucine + 2 mM l-glutamine, 1 mM carbachol, 10 μM PMA, or 10 μM ionomycin. Supernatant was collected and assayed for insulin content by enzyme-linked immunosorbent assay (ELISA; Crystal Chem Inc.). Islets were then sonicated in acid-ethanol solution and solubilized overnight at 4°C before assaying total insulin content by ELISA. For rhythmic insulin release assays, islets were first synchronized with 10 μM forskolin (Sigma) for 1 hour and allowed to recover for 16 hours. Insulin secretion assays were then performed as above in individual groups of five islets every 4 hours for 72 hours (fig. S1A). Human islets (obtained from IIDP) were cultured in RPMI 1640 with 10% human AB serum, 1% l-glutamine, and 1% penicillin/streptomycin (see table in fig. S4A for details of sex, age, BMI, and IIDP ID numbers of the three donors). For the rhythmic analysis of RNAs in murine and human islets, RNA was isolated (described below) in groups of 200 islets every 4 hours for 48 or 24 hours, respectively, starting 40 hours after forskolin synchronization (fig. S1A).

LumiCycle analysis

Approximately 2 hours before the start of the dark period (i.e., lights off), ~100 to 150 pancreatic islets were isolated from Per2Luc mice as described above. Islets were cultured on tissue culture membranes (Millipore) in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, 1.2 ml) containing sodium bicarbonate (352.5 μg/ml), 10 mM HEPES (Gibco), 2 mM l-glutamine, 2% B-27 serum-free supplement (Invitrogen), penicillin (25 U/ml), streptomycin (Gibco, 20 μg/ml), and 0.1 mM luciferin sodium salt (Biosynth AG). Sealed cultures were placed at 37°C in a LumiCycle luminometer (Actimetrics) and bioluminescence from tissues was recorded continuously. After several days in culture, islets were synchronized by 10 μM forskolin (Sigma) treatment for 1 hour followed by incubation in fresh media. Period was calculated via a modified best-fit sine wave analysis using LumiCycle analysis software (Actimetrics).

Measurement of islet oxygen consumption

After bile duct collagenase digestion, 40 purified pancreatic islets were plated in wells of a 96-well respirometry plate (Seahorse Bioscience) and cultured overnight in complete medium. The next day, culture medium was replaced with assay buffer containing 3 mM glucose, 0.8 mM Mg2+, 1.8 mM Ca2+, 143 mM NaCl, 5.4 mM KCl, 0.91 mM NaH2PO4, and phenol red (Seahorse Bioscience; 15 mg/ml) and allowed to equilibrate at 37°C in a CO2-free incubator for 1 to 2 hours. The plate was then loaded into a Seahorse XF96 instrument, and the oxygen consumption rate (OCR) was measured for four sequential 3-min intervals at basal conditions and after injection of glucose (20 mM final concentration), oligomycin (F1FO ATP synthase inhibitor) (5 μM final concentration), and antimycin A (complex III inhibitor) (5 μM final concentration). OCR values given represent the average of four sequential measurements. Mitochondrial oxygen consumption was calculated by subtracting OCR values after antimycin A treatment (representing nonmitochondrial oxygen consumption).

RNA isolation and qPCR mRNA quantification

Islets were added to microfuge tubes containing Tri Reagent (Molecular Research Center Inc.) and frozen at –80°C. RNA was isolated according to the manufacturer’s protocol and purified using RNeasy columns (Qiagen). cDNAs were then synthesized using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time quantitative polymerase chain reaction (qPCR) analysis was performed with SYBR Green Master Mix (Applied Biosystems) and analyzed using an Applied Biosystems 7900 Fast Real-Time PCR System. Relative expression levels were determined using the comparative CT method to normalize target gene mRNA to Gapdh. Exon-specific primer sequences for qPCR were as follows: Bmal1 exons 5 to 7, forward, 5′-ATCGCAAGAGGAAAGGCAGT-3′; reverse, 5′-ATCCTTCCTTGGTGTTCTGCAT-3′. Bmal1 exons 7 to 9, forward, 5′- AGGCCCACAGTCAGATTGAA-3′; reverse, 5′-TGGTACCAAAGAAGCCAATTCAT-3′. Bmal1 exon 8, forward, 5′-GGCGTCGGGACAAAATGAAC-3′; reverse, 5′-TCTAACTTCCTGGACATTGCAT-3′. Bmal1 exons 8 and 9, forward, 5′-TGCAATGTCCAGGAAGTTAGAT-3′; reverse, 5′-TGGTGGCACCTCTCAAAGTT-3′. Bmal1 exons 10 to 12, forward, 5′-TAGGATGTGACCGAGGGAAG-3′; reverse, 5′- AGCTCTGGCCAATAAGGTCA-3′.

RNA sequencing and analysis

After RNA isolation (described above), RNA quality was assessed using a Bioanalyzer (Agilent), and sequencing libraries were constructed using an Illumina TruSeq Stranded mRNA sample prep kit LT (Illumina, RS-122-2101) according to the manufacturer’s instructions. Libraries were quantified using both a Bioanalyzer (Agilent) and qPCR-based quantification (Kapa Biosystems) and sequenced on either an Illumina HiSEq 2000 or NextSEq 500 instrument to a depth of at least 30 million reads using 100–base pair (bp) or 75-bp paired-end reads, respectively.

For differential expression comparison between PdxCre;Bmal1flx/flx and Bmal1flx/flx islets, RNA raw sequence reads were aligned to the reference genome (mm10) using STAR version 2.3.1s_r366 (49). Differentially expressed RNAs were identified using DESEq 2 version 1.6.3 (50) (FDR-adjusted P < 0.05).

For cycling RNAs, raw sequence reads were similarly aligned using STAR (mm10 index for mouse and hg19 for human), and uniquely mapped reads (tags) were normalized using the algorithm used in DESEq 2 (50). The geometric mean of the raw read counts was calculated for each gene. A normalization factor was calculated for each sample using the median of the raw read counts of each gene divided by the geometric mean of the gene. The normalized read counts were computed by dividing the raw read counts by the normalization factor. The normalized tags for the mouse and human time series were separately concatenated and z-scored within each gene (14). Rhythm detection of the z-scored and normalized counts was performed with empirical JTK_CYCLE with asymmetry search, which increases sensitivity of detecting cycling transcripts by extending comparisons to reference waveforms beyond cosines, including arbitrary asymmetric waveforms that better represent expression patterns seen in biological data. Rhythmic time series were examined with reference waveforms with a period of 24 hours; a phase of 0, 4, 8, 12, 16, or 20; and an asymmetry of 4, 12, or 20. Because of the small number of waveforms compared, the Bonferroni correction was used instead of the empirical P values. Genes with a Bonferroni-adjusted P value below 0.05 were considered to be rhythmic.

For KEGG ontology term enrichment (51, 52), Ensembl gene IDs were supplied and analyzed using Homer (version 4.7.2) command “findGO” (53).

Genes exhibiting rhythmic mRNA accumulation in vivo in liver were derived from reported “exon cycling” transcripts (9).

β cell culture

Beta-TC6 cells were purchased from ATCC (CRL-11506) and cultured in DMEM supplemented with 15% FBS, 1% l-glutamine, and 1% penicillin/streptomycin. All cells used in experiments were at fewer than 15 passages.

Mouse BMAL1 and CLOCK polyclonal antibody generation

Guinea pig anti-mouse BMAL1 and CLOCK polyclonal antibodies were generated using a 37– and 39–amino acid peptide fragment of the mouse BMAL1 and CLOCK proteins, respectively (RS synthesis). Guinea pigs were immunized with KLH-conjugated peptides (Pocono Farms), and BMAL1- and CLOCK-specific antibodies were affinity-purified from whole serum using resin cross-linked with antigen peptides (Pierce).

Chromatin immunoprecipitation (ChIP)

Beta-TC6 cells (~40 to 160 million) were fixed for 30 min in 2 mM disuccinimidyl glutarate and for 10 min in 1% formaldehyde and then either frozen at –80°C or processed immediately. Nuclei were isolated in buffer containing 1% SDS, 10 mM EDTA, 50 mM Tris-HCl (pH 8.0), and protease inhibitors and sonicated using a Diagenode Bioruptor to shear chromatin into 200- to 1000-bp fragments. Protein-DNA complexes were incubated with antibodies against BMAL1 and CLOCK (affinity-purified guinea pig IgGs as described above), H3K4Me2 (Abcam), H3K27Ac (Active Motif), H2AZ (Active Motif), or PDX1 (Novus Biologicals) and immunoprecipitated with IgG paramagnetic beads (Invitrogen). Eluted chromatin was isolated using MinElute PCR purification columns (Qiagen).

ChIP sequencing and analysis

Sequencing libraries were generated using KAPA DNA Library Preparation kits (Kapa Biosystems, KK8504) according to manufacturer’s instructions. Library concentrations were assessed by both a Bioanalyzer (Agilent) and qPCR-based quantification (Kapa Biosystems). Libraries were sequenced using 75-bp single-end reads on an Illumina Next-SEq 500 instrument to a depth of >10 million mapped reads.

Raw sequence reads were aligned to the mm10 reference genome and displayed using UCSC annotated genes using bowtie version 1.1.1 (54) with parameters “-best” and “-m 1” to ensure reporting of uniquely mapped reads (tags). ChIP-seq peaks were designated as regions with a factor of 4 enrichment over both the input sample and the local background and were normalized to 10 million reads using default parameters for the Homer “findPeaks” command (53) and specifying “-style factor” for BMAL1, CLOCK, and PDX1 and “-style histone” for H2A.Z, H3K4Me2, and H3K27Ac. For BMAL1 and CLOCK peaks, promoter binding was defined as peaks occurring within 2 kb of the nearest gene TSS, and distal binding was defined as those occurring greater than 2 kb from a nearest TSS.

To identify consensus motifs for BMAL1 and CLOCK, we scanned 50-bp windows surrounding transcription factor peaks using “findMotifsGenome.pl” with standard background (random genomic sequences sampled according to GC content of peak sequences). We determined the occurrence of tandem E-boxes with variable-length spacing by generating synthetic canonical E-box motifs separated by the indicated number of random spacers (i.e., CACGTGNNNCACGTG = 3 spacers) using “seq2profile.pl” allowing for two mismatches and testing for their occurrence at BMAL1 and CLOCK peaks using “annotatePeaks.pl”.

Fastq files for all BMAL1 and H3K27Ac ChIP-seq were downloaded from the ENA server (study accession number SRP014752) and raw sequence reads for 12 sequential time points were concatenated into a single file. Alignments and peak calling were performed using bowtie and Homer as described above. Shared BMAL1 binding sites were identified by comparing binding locations between β cells and liver using the Homer command “mergePeaks” and specifying “-d 200,” which identified peaks occurring within 200 bp as shared across tissues.

Tamoxifen treatment

For in vivo delivery of tamoxifen (Sigma, dissolved in corn oil), mice received three intraperitoneal (i.p.) injections of 200 μg tamoxifen/g body weight, administered every other day. Subsequent experiments were conducted 10 to 14 days after tamoxifen treatment. For in vitro administration of tamoxifen, isolated islets were incubated for 24 hours with 1 mM tamoxifen (dissolved in ethanol) prior to transfer to complete media for 24 hours to recover. Islets were then synchronized with forskolin prior to insulin secretion assays as described above.

Immunohistochemical analysis

Mice were anesthetized with i.p. injection of phenobarbital (Nembutal, 50 mg/ml) and perfused with heparinized saline, followed by 4% paraformaldehyde (PFA) (Sigma) in PBS. Brain and pancreas were removed and post-fixed with 4% PFA overnight at 4°C. Brain tissues were then cryoprotected in 30% sucrose (Sigma), frozen in O.T.C. (Tissue Tek), and 30-μm brain sections collected for antibody staining. Pancreata were embedded in paraffin, and blocks of 6-μm sections were mounted on slides. The following primary antibodies were used for staining: guinea pig anti-insulin (1:500, DAKO), mouse anti-glucagon (1:500, Sigma), and rabbit anti-BMAL1 (1:500, Novus Biological). Triple staining was visualized with the following secondary antibodies: AMCA goat anti–guinea pig (1:400, Jackson ImmunoResearch), Alexa Fluor 488–conjugated goat anti-mouse (1:400, Invitrogen), and Alexa Fluor 546–conjugated goat anti-rabbit (1:400, Life Technologies). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) as indicated. Images were acquired with PictureFrame 1.0 using a Zeiss Axioskop 50. β cell mass was assessed by morphometric analysis of insulin immunostained pancreatic sections (DAKO; Histomouse Plus kit, Life Technologies). Four pancreatic sections, spaced 50 μm apart, were stained for each animal, and endocrine versus total pancreas area was calculated using Image-Pro Premier software (Media Cybernetics) using the smart segmentation feature.

Glucose and insulin measurements and glucose tolerance tests

Blood glucose and plasma insulin levels in ad libitum–fed mice were assessed at ZT2 and ZT14 from tail vein bleeds. Glucose tolerance tests were performed in mice after a 14-hour fast, and blood glucose and plasma insulin levels were measured at the indicated times after i.p. glucose injection of either 2 or 3 g/kg body weight, respectively. Plasma insulin levels were measured by ELISA.

Behavioral analysis

Locomotor activity was analyzed in 2- to 4-month-old pancreas-specific Bmal1 knockout mice and their respective littermate controls after tamoxifen treatment. All animals were individually housed in standard mouse cages equipped with running wheels and allowed free access to food and water. Mice were placed in a 12:12 LD cycle for 14 days, followed by 14 days in constant darkness (DD). Total activity data was recorded and analyzed in 6-min bouts using ClockLab software (Actimetrics). The free-running period was determined as the duration of time between the major activity periods on consecutive days in DD. Period was calculated using a χ2 periodogram for days 7 to 14 in DD. Food consumption was analyzed in pancreas-specific Bmal1 knockout mice and their littermate controls before and after tamoxifen treatment. All animals were individually housed with free access to water and regular chow. Daytime and nighttime food consumption was determined by manual measurement of food at both ZT0 and ZT12 for three consecutive days.

Intracellular calcium determination

BetaTC-6 cells were plated at a density of 100,000 cells per well in black 96-well plates with clear bottoms and cultured overnight at 37°C and 5% CO2. Islets were dispersed to single cells by incubating in 0.05% Trypsin-EDTA at 37°C for 3 min and plated at a density of 100 islets per well in laminin-treated black 96-well plates with clear bottoms and cultured in complete media for 48 hours at 37°C and 5% CO2. Cells were then washed with BSA-free KRB buffer with no glucose and loaded with 5 μM Fura-2 (Invitrogen) and 0.04% Pluronic F-127 (Invitrogen) for 30 min at 37°C. Following a wash with BSA-free KRB, Fura-2 intensity was measured after injection of either glucose or ionomycin (Sigma) to final concentrations of 20 mM or 10 μM, respectively. Cells were alternately excited with 340- and 380-nm light, and the emitted light was detected at 510 nm using a Cytation 3 Cell Imaging Multi-Mode Reader (Bio Tek) at sequential 30-s intervals. Raw fluorescence data were exported to Microsoft Excel and expressed as the 340/380 ratio for each well.

Supplementary Materials

References and Notes

  1. Acknowledgments: Supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant R01DK090625, National Institute on Aging grant P01AG011412, Chicago Biomedical Consortium S-007, Juvenile Diabetes Research Foundation grants 17-2013-511, 1-INO-2014-178-A-V, and 1-INO-2015-23-A-V, and University of Chicago Diabetes Research and Training Center grant P60DK020595 (J.B.); NIDDK T32 grant DK007169 (B.M.); National Heart, Lung, and Blood Institute T32 grant HL007909 (M.P.); Defense Advanced Research Projects Agency grant D12AP00023 (A.R.D.); and National Institute of Environmental Health Science grant ES05703 (C.A.B.). A.L.H. is a trainee of the NIH Medical Scientist Training program at the University of Chicago (National Institute of General Medical Sciences grant T32GM07281). We thank E. Rosenzweig and The Next Generation Sequencing Core Facility at Northwestern University for assistance with library generation, sequencing, and analysis, as well as all members of the Bass and Barish laboratories for helpful discussions. J.B. has financial interest in and serves as an advisor to Reset Therapeutics. RNA-seq data sets, including lists of cycling genes in mouse and human islets, have been deposited in the NCBI’s Gene Expression Omnibus accession GSE69889, and ChIP-seq data sets have been deposited in accession GSE70960.
View Abstract

Navigate This Article