A cullin 4B-RING E3 ligase complex fine-tunes pancreatic δ cell paracrine interactions

The functions of different cell types within Langerhans islets are precisely regulated by paracrine interactions, and these coordinate the rate and level of reciprocal insulin and glucagon secretion, thereby maintaining glucose homeostasis in response to changes in nutrient levels in different physiological contexts (1–4). The disruption of this fine-tuned islet circuit by the aberrant regulation of paracrine islet interactions plays an important role in the development of diabetes (1–9). For example, the loss of the important paracrine factor urocortin 3 (UCN3) in pancreatic β cells caused dysregulation of a somatostatin-mediated negative feedback loop between islet δ and β cells, resulting in aberrant glucose-induced insulin secretion and contributing to the pathophysiology of diabetes (5, 10, 11). Recently, the functional importance of pancreatic δ cells in maintaining glucose homeostasis is increasingly being appreciated. More specifically, several studies have shown that somatostatin secretion from pancreatic δ cells is regulated by intracellular calcium and cyclic AMP, whereas somatostatin synthesis in pancreatic δ cells is regulated by the cAMP and β-arrestin1 pathways (12, 13). In spite of this progress, whether an epigenetic regulatory mechanism underlies the crosstalk between pancreatic δ and β cells has not been investigated.

Cullin family members accompanied by their partners in cullin-RING E3 ligase complexes (CRLs) play important roles in neuron activity, in cardiovascular function, and in the development of the reproductive system and pancreatic islets (14, 22). In particular, cullin 4B (CUL4B) mutations cause mental retardation, short stature, and obesity (23, 24). Unlike CUL4A and other cullins, CUL4B harbors an N-terminal nuclear localization sequence that directs it to nucleus to regulate cell functions in an epigenetic manner (18, 25–27). Recently, we have shown that cullin 4B-RING E3 ligase (CRL4B) functions as a transcriptional corepressor by catalyzing H2AK119 ubiquitination (H2A119ub1) and coordinating with polycomb repressive complex 2 (PRC2) to regulate histone H3 lysine 27 (H3K27) trimethylation (28, 29). Other studies have demonstrated that CRL4 ubiquitinates WD repeat-containing protein 5 (WDR5), a core subunit of H3K4 methyltransferase complexes, for degradation in the nucleus, thereby promoting increased H3K4 methylation levels (30). However, whether CUL4B or its E3 ligase complex CRL4B-PRC2 participates in the functioning or development of Langerhans islets has not been studied.

In this study, we found that CUL4B expression levels are decreased in db/db mice. Therefore, we crossed mice expressing Cre under the insulin II promotor (Ins2-Cre) and mice expressing Cre under the somatostatin promoter (Sst-Cre) with Cul4b^fl/fl mice to characterize CUL4B functions in specific cell types of the islet circuit. Although Ins2-Cre Cul4b^fl/Y male mice were found to exhibit normal plasma glucose levels, the constitutive ablation of CUL4B in pancreatic δ cells impaired glucose tolerance and reduced insulin secretion through enhanced somatostatin release. Moreover, mechanistic studies revealed that the CUL4B-PRC2 complex controls intracellular calcium and cAMP levels by epigenetically regulating the expression of ion channel CACNA1C (Cav1.2) and adenylyl cyclase 6 (AC6), thus adjusting somatostatin secretion from pancreatic δ cells.

CUL4B ablation in pancreatic δ cells other than β cells causes glucose intolerance. Previous studies have shown that mutations of CUL4B in patients lead to obesity (24, 31). To understand the potential roles of CUL4B in glucose homeostasis, we examined protein levels of CUL4B in the pancreatic islets of obese diabetic db/db mice. Total levels of islet CUL4B expression were decreased approximately 3-fold in 12-week-old db/db mice compared with their littermate db/+ controls, whereas the expression of its paralog CUL4A remained unaltered (Figure 1, A and B; see complete unedited blots in the supplemental material; supplemental material available online with this article; https://doi.org/10.1172/JCI91348DS1). Specifically, immunofluorescence results revealed that CUL4B expression decreased more drastically in pancreatic δ cells than in total islets in 12-week-old and 26-week-old db/db mice (Figure 1C and Supplemental Figure 1, A–C).

Figure 1

CUL4B deficiency in pancreatic δ cells impairs glucose metabolism. (A and B) Western blots and quantitative data for CUL4A and CUL4B protein levels in islets from 12-week-old diabetic db/db mice and their heterozygous littermates (db/+). n = 6 mice per group. Representative Western blots from at least 3 independent experiments are shown. (C) Immunostaining for CUL4B (green) and somatostatin (SST, red) in pancreatic sections from db/db and db/+ mice. Scale bar: 100 μm. n = 6 mice per group; 4–7 random areas were selected from each islet section, and 10 sections were randomly selected from each mouse. (D) Confirmation of pancreatic δ cell–specific CUL4B deficiency (Sst-Cre^+/– Cul4b^fl/Y) through immunofluorescence. The colocalization of somatostatin (red) and CUL4B (green) in δ cells of WT mice was absent in Sst-Cre^+/– Cul4b^fl/Y mice. Scale bar: 100 μm. (E) Immunostaining for insulin (red) and somatostatin (green) in WT and Sst-Cre^+/– Cul4b^fl/Y mice. Scale bar: 50 μm. (F) Quantitative data for islet density, pancreatic δ cell number, and β cell mass. n = 6 mice per group; 4–10 random areas were selected from each section, and 12 sections were randomly selected from each mouse. (G) The fasting and fed blood glucose levels of Sst-Cre^+/– Cul4b^fl/Y mice and their WT littermates. n = 8 mice per group. (H) Glucose tolerance test for Sst-Cre^+/– Cul4b^fl/Y mice and their WT littermates (n = 11–12). (I) Insulin tolerance test for Sst-Cre^+/– Cul4b^fl/Y mice and their WT littermates. Insulin-induced decreases in blood glucose levels were significantly lower in Sst-Cre^+/– Cul4b^fl/Y mice than in their WT littermates, and they did not return to baseline levels at the 2-hour time point, whereas the levels of their WT littermates did (n = 9–11). *P < 0.05; **P < 0.01; ***P < 0.001. db/db mice were compared with their db/+ littermates, and Sst-Cre^+/– Cul4b^fl/Y mice were compared with their WT littermates. Error bars in F represent mean ± SD; other bars represent mean ± SEM. All data were analyzed using 1-way ANOVA.

To analyze CUL4B function in islet circuits, we crossed Cul4b^fl/fl mice with Ins2-Cre transgenic mice or Sst-Cre transgenic mice to generate cell type–specific Cul4b-knockout mice (Supplemental Figure 1, D–F). Both cell type–specific knockout mice were born at expected Mendelian frequencies and exhibited normal life expectancies. Compared with those in Ins2-Cre mice, CUL4B protein expression levels were decreased by 80% in the islets of Ins2-Cre^+/– Cul4b^fl/Y mice, as confirmed by Western blotting and immunofluorescence analysis (Supplemental Figure 1, G and H). Insulin and somatostatin immunofluorescence showed that Ins2-Cre^+/– Cul4b^fl/Y mice had the same number of islets and comparable β and δ cell numbers (Supplemental Figure 2, A and B). Furthermore, no significant differences in fasting and fed glucose levels or glucose tolerance were observed between the Ins2-Cre^+/–Cul4b^fl/Y mice and the Ins2-Cre^+/– mice (Supplemental Figure 2, C and D).

We next investigated the specific functions of CUL4B in somatostatin-secreting cells using Sst-Cre^+/– Cul4b^fl/Y mice (Supplemental Figure 1, D and E). CUL4B was not detected in the pancreatic δ cells of Sst-Cre^+/–Cul4b^fl/Y mice, compared with Sst-Cre^+/– mice (referred to as WT), by immunostaining (Figure 1D and Supplemental Figure 1E). Immunofluorescence also showed no significant differences in islet or β or δ cell numbers between Sst-Cre^+/–Cul4b^fl/Y and Sst-Cre^+/– mice (Figure 1, E and F). We then examined glucose homeostasis in Sst-Cre^+/– Cul4b^fl/Y mice. Although Sst-Cre^+/– Cul4b^fl/Y mice exhibited fasting glucose levels comparable to those of Sst-Cre^+/– mice, the knockout mice exhibited increased glucose levels after 2 hours of feeding following a 16-hour fast (Figure 1G). Accordingly, during a glucose tolerance test, the blood glucose levels of Sst-Cre^+/– Cul4b^fl/Y mice were 33% and 30% higher at 30 minutes and 120 minutes, respectively, than those of the control group consisting of both Sst-Cre^+/– and Cul4b^fl/Y mice (Figure 1H). In addition, during the insulin tolerance test, the blood glucose levels of Sst-Cre^+/– Cul4b^fl/Y mice unexpectedly decreased more dramatically and rapidly than those of the Sst-Cre^+/– control mice in response to insulin stimulation (Figure 1I). Insulin and somatostatin content as well as kidney, brain, heart, liver, and stomach weights were not found to be significantly different between the Sst-Cre^+/– Cul4b^fl/Y mice and their Sst-Cre^+/– littermates (Supplemental Figure 2, E–G). Together, these data indicate that ablation of CUL4B in mouse δ cells, but not β cells, impairs glucose tolerance.

Ablation of CUL4B in δ cells impairs glucose-mediated insulin secretion by increasing somatostatin secretion. The impaired glucose metabolism found in the Sst-Cre^+/– Cul4b^fl/Y mice was probably due to dysfunctional hormone secretion from the pancreatic islets. Therefore, we detected levels of plasma and secreted insulin and somatostatin in isolated islets from Sst-Cre^+/– Cul4b^fl/Y and Sst-Cre^+/– mice. After an overnight fast, the plasma insulin levels of Sst-Cre^+/– Cul4b^fl/Y mice showed no significant differences compared with those of Sst-Cre^+/– mice (Figure 2A). After a 2-hour feeding period, plasma insulin levels increased 3.7-fold in the Sst-Cre^+/– mice, whereas plasma insulin levels of Sst-Cre^+/– Cul4b^fl/Y mice increased by approximately 2-fold (Figure 2A). However, we found no significant differences in plasma somatostatin levels of fasted versus fed animals in both Sst-Cre^+/– Cul4b^fl/Y and Sst-Cre^+/– mice (Figure 2B). We then isolated pancreatic islets from Sst-Cre^+/– Cul4b^fl/Y and Sst-Cre^+/– mice and examined insulin and somatostatin secretion patterns in response to 1 mM, 5.5 mM, and 20 mM glucose or high KCl stimulation conditions. High glucose levels induced both insulin and somatostatin secretion from isolated islets (Figure 2, C–F). However, glucose-induced insulin secretion was 47% lower in islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice than in those from the control Sst-Cre^+/– mice (Figure 2, C and E). In contrast, glucose-induced somatostatin levels were found to be 140% higher than those in control mice (Figure 2, D and F). High potassium levels also induced insulin and somatostatin secretion from isolated islets, and islets from Sst-Cre^+/– Cul4b^fl/Y mice exhibited significantly decreased insulin secretion and corresponding increases in somatostatin (Figure 2, E and F). We then examined glucose-induced insulin and somatostatin secretion trends over different stimulation time points. The islets from Sst-Cre^+/– Cul4b^fl/Y mice compared with the islets from Sst-Cre mice exhibited decreased insulin secretion levels after both short- and long-term glucose exposure (Figure 2G). In contrast, glucose-induced somatostatin secretion from islets of Sst-Cre^+/– Cul4b^fl/Y mice was 140% and 40% higher than that of Sst-Cre^+/– mice at 5 minutes and 60 minutes, respectively (Figure 2H). In contrast to the results found for the Sst-Cre^+/–Cul4b^fl/Y mice, CUL4B deficiency in the pancreatic β cells of Ins2-Cre^+/– Cul4b^fl/Y mice did not significantly affect insulin and somatostatin secretion levels in response to high glucose relative to trends found for Ins2-Cre^+/– mice (Supplemental Figure 3, A and B).

Figure 2

Impaired glucose tolerance in Sst-Cre^+/– Cul4b^fl/Y mice is due to increased somatostatin paracrine signaling. (A and B) Plasma insulin levels (A) and somatostatin levels (B) of WT and Sst-Cre^+/– Cul4b^fl/Y mice in the fasted and fed states. n = 6–7 mice per group. (C and D) 1 mM, 5.5 mM, and 20 mM glucose-induced insulin (C) and somatostatin (D) secretion in islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice and WT littermates after 5 minutes. n = 5 mice per group. (E and F) Glucose- and high KCl–induced insulin secretion (E) or somatostatin secretion (F) of islets isolated from WT and Sst-Cre^+/– Cul4b^fl/Y mice after 5 minutes. n = 4 mice per group. (G and H) Short- and long-term glucose-induced insulin secretion (G) and somatostatin secretion (H) of islets isolated from WT mice and Sst-Cre^+/– Cul4b^fl/Y mice. n = 4 mice per group. (I) Effect of the somatostatin receptor antagonist cyclosomatostatin (12 μM) on glucose-induced insulin secretion of islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice and WT littermates after 5 minutes. n = 4 mice per group. (J) Glucose- or high KCl–induced somatostatin secretion from CUL4B knockdown and control TGP52 cells after 1 hour (n = 5). A–I, *P < 0.05; **P < 0.01. Sst-Cre^+/– Cul4b^fl/Y mice were compared with their WT littermates. J, *P < 0.05. CUL4B knockdown cells were compared with control shRNA knockdown cells. Bars represent mean ± SEM. All data were analyzed using 1-way ANOVA.

It was reported that somatostatin inhibits glucose-induced insulin secretion from islet β cells by activating G protein–coupled somatostatin receptors (32, 35). In agreement with previous results, our data showed that somatostatin behaved as a paracrine regulator of glucose metabolism and insulin secretion in Sst-Cre^+/– Cul4b^fl/Y mice. Downregulation of insulin secretion from pancreatic islets of Sst-Cre^+/– Cul4b^fl/Y mice may have resulted from the increased levels of somatostatin release. To this end, we pretreated islets with cyclosomatostatin, a general somatostatin receptor antagonist, and we then stimulated them under high glucose levels. The administration of cyclosomatostatin abolished differential glucose-induced insulin secretion observed between Sst-Cre^+/– Cul4b^fl/Y and Sst-Cre^+/– pancreatic islets (Figure 2I). To further confirm the direct regulatory role of CUL4B in pancreatic δ cell functioning, we used the in vitro TGP52 cell system and knocked down CUL4B by shRNA. TGP52 is a pancreatic epithelial cell line that secretes somatostatin only rather than insulin in response to high glucose or high potassium stimulation (Supplemental Figure 4, A and B) (36). Two TGP52 stable cell lines treated with different shRNAs showed significantly decreased expression levels of CUL4B (Supplemental Figure 4, C and D). In agreement with our in vivo results, TGP52 cells with lower CUL4B expression levels secreted more somatostatin after high glucose or high potassium stimulation (Figure 2J).

Elevated calcium and cAMP levels in δ cells contribute to enhanced somatostatin secretion. ATP-sensitive potassium (KATP) channels are key players in the regulation of insulin and somatostatin secretion from pancreatic β and δ cells, respectively, because activation of KATP channels completely abolishes glucose-induced insulin and somatostatin secretion (5, 37, 38). To investigate the mechanisms underlying increased somatostatin secretion levels found in Sst-Cre^+/– Cul4b^fl/Y mice, we first examined the effects of diazoxide, a specific KATP channel opener, on glucose-induced somatostatin secretion in islets isolated from Sst-Cre^+/–Cul4b^fl/Y mice or Sst-Cre^+/– littermates. Treatment with diazoxide inhibited somatostatin secretion in islets isolated from both Sst-Cre^+/– Cul4b^fl/Y and control mice to a similar extent, thus indicating that increased somatostatin secretion was not due to changes in KATP channel activity in pancreatic δ cells (Figure 3A). In contrast, diazoxide inhibited glucose-induced insulin secretion to the same extent in islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice and from their WT littermates, possibly because of the common essential downstream pathway mediated by KATP channels governing insulin secretion in pancreatic β cells in mice of both genotypes (Figure 3B).

Figure 3

L-type calcium channels play key roles in enhanced somatostatin secretion from CUL4B-deficient δ cells. (A) Effect of diazoxide (250 μM), a KATP channel opener, on glucose-induced somatostatin secretion in islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice and WT littermates after 5 minutes. (n = 4). (B) Effect of diazoxide on glucose-induced insulin secretion in islets isolated from Sst-Cre^+/– Cul4b^fl/Y or WT littermates after 5 minutes (n = 4). (C) Glucose-induced somatostatin secretion from islets in the absence or presence of the R-type channel blocker SNX482 (100 nM) and the L-type channel blockers isradipine (10 μM) and nicardipine (10 μM) after 5 minutes (n = 4). (D) Glucose-induced insulin secretion from islets in the absence or presence of an R-type channel blocker or L-type channel blockers after 5 minutes (n = 4). (E and F) Representative curves and statistical analysis of a high glucose–induced calcium signal in isolated pancreatic δ cells from Sst-Cre^+/– GFP^fl/+ Cul4b^fl/Y mice (E, lower panel) and WT littermates (E, upper panel) (n = 8). (G and H) Representative curves and statistical analysis of high KCl–induced calcium signal in δ cells from Sst-Cre^+/– GFP^fl/+ Cul4b^fl/Y mice (G, lower panel) or WT littermates (G, upper panel) (n = 8). *P < 0.05; **P < 0.01; ***P < 0.001. Sst-Cre^+/– GFP^fl/+ Cul4b^fl/Y mice were compared with their WT littermates. Bars represent mean ± SEM. All data were analyzed using 1-way ANOVA.

Recent studies have shown that L-type and R-type calcium channels together with the ryanodine receptor pathway are involved in glucose-induced somatostatin secretion (5, 12). In particular, the application of the L-type channel blocker isradipine abolished the potentiation effects of UCN3 on glucose-stimulated somatostatin secretion from pancreatic δ cells (5). Therefore, we examined glucose-induced islet somatostatin secretion in the absence or presence of the R-type channel blocker SNX482, the L-type channel blockers nicardipine and isradipine, and the endoplasmic reticulum Ca²⁺ ATPase inhibitor thapsigargin (Figure 3C). Similarly to what was recently reported (12), glucose-induced somatostatin secretion was significantly blocked by R-type channel antagonist SNX482 (Figure 3C). However, even with SNX482 pretreatment, islets from Sst-Cre^+/– Cul4b^fl/Y mice still released somatostatin levels twice as high as those of their WT littermates, indicating that the increased somatostatin levels found in the knockout islets did not form because of changes in the opening of R-type calcium channels. Interestingly, when we blocked L-type calcium channels with nicardipine or isradipine, we found that islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice released the same amount of somatostatin as those from control mice (Figure 3C). These results showed that L-type calcium channels play an important role in increased somatostatin secretion in Sst-Cre^+/– Cul4b^fl/Y mice. We then used TGP52 pancreatic δ cells to identify a direct interaction between CUL4B and L-type calcium channels in somatostatin secretion of pancreatic δ cells. Indeed, isradipine blocked enhanced somatostatin secretion in CUL4B knockdown TGP52 cells, whereas treatment with SNX482, thapsigagin, or diazoxide elicited insignificant effects (Supplemental Figure 5A). We next measured insulin secretion after treatment with the same panel of Ca²⁺ channel and ATPase inhibitors. Treatment with both the R-type calcium channel blocker SNX482 and the L-type channel inhibitor nicardipine markedly inhibited glucose-induced insulin secretion from islets. However, although the insulin secretion of islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice was still much lower than that of their WT littermates after SNX482 treatment, this difference was eliminated after nicardipine treatment (Figure 3D). These data are in agreement with the observed somatostatin secretion patterns and reinforce the conclusion that L-type calcium channels in pancreatic δ cells are key components mediating the increased somatostatin secretion and impaired glucose metabolism in Sst-Cre^+/– Cul4b^fl/Y mice.

L-type calcium channels are responsible for the influx of extracellular calcium during hormone secretion. We therefore measured high glucose– and high potassium–induced calcium signals in isolated primary pancreatic δ cells from Sst-Cre^+/– GFP ^fl/+ Cul4b^fl/Y mice and their WT littermates. The pancreatic δ cells derived from Sst-Cre^+/– GFP ^fl/+ Cul4b^fl/Y mice showed significantly higher frequencies of high glucose–induced calcium signals than the WT primary pancreatic δ cells (Figure 3, E and F). Similarly, in pancreatic δ cells derived from Sst-Cre^+/– GFP ^fl/+ Cul4b^fl/Y mice, potassium-induced calcium increases were significantly greater than those in the WT littermates (Figure 3, G and H). However, after isradipine pretreatment, intracellular calcium signals of the Sst-Cre^+/– GFP ^fl/+ Cul4b^fl/Y mice were decreased to 50%, reflecting patterns found for the WT (Figure 3, G and H). Therefore, intracellular calcium increases contribute to increased somatostatin secretion in Sst-Cre^+/– GFP ^fl/+ Cul4b^fl/Y mice and are dependent on L-type calcium channel activity.

In addition to intracellular calcium, the second messenger cAMP, the G_q-PLC pathway, and kinase signaling are actively involved in hormone secretion from pancreatic islets (5, 13, 39–41). Therefore, we preincubated the islets with several inhibitors, including cAMP-PKA signaling inhibitors Rp-CAMPs and H89, G_q-PLC inhibitor U73122, MEK inhibitor U0126, and adenyl cyclase (AC) inhibitor 2′,5′-dideoxyadenosine (DDA), and we examined glucose-induced somatostatin secretion patterns (Figure 4A). Interestingly, cAMP antagonist Rp-cAMPs and AC inhibitor DDA completely blocked enhanced somatostatin secretion from the islets of Sst-Cre^+/– Cul4b^fl/Y mice compared with their WT littermates whereas no significant effects were found for the other inhibitors. Moreover, treatment with cAMP antagonist Rp-cAMPs completely eliminated enhanced somatostatin secretion in Cul4b knockdown TGP52 cells compared with control cells (Supplemental Figure 5B). These results suggest that the cAMP pathway actively participates in the increased somatostatin secretion observed in Sst-Cre^+/– Cul4b^fl/Y mice. We therefore measured intracellular cAMP levels in CUL4B knockdown TGP52 and control cells. High glucose levels elicited time-dependent intracellular cAMP accumulation; intracellular cAMP levels in the CUL4B knockdown TGP52 cells increased 4-fold, 3-fold, and 3-fold over 2 minutes, 5 minutes, and 15 minutes, respectively (Figure 4B).

Figure 4

CRHR2 activity and the cAMP signaling pathway are important in increased somatostatin secretion. (A) Glucose-induced somatostatin secretion in islets isolated from WT and Sst-Cre^+/– Cul4b^fl/Y mice treated with inhibitors for PKA (H89: 10 μM), cAMP (Rp-cAMP: 100 μM), ERK (U0126: 10 μM), PLC (U73122: 10 μM), or AC (DDA: 100 μM) after 5 minutes (n = 3). (B) Glucose-induced cAMP signals in CUL4B knockdown and control TGP52 cells (n = 4). (C–D) Effects of Ast2B on glucose-induced insulin (D) and somatostatin (C) secretion in islets isolated from Sst-Cre^+/– Cul4b^fl/Y or WT littermates after 5 minutes (n = 4). A, C, D, *P < 0.05; ***P < 0.001. Sst-Cre^+/– Cul4b^fl/Y mice were compared with their WT littermates. B, *P < 0.05; ***P < 0.001. CUL4B knockdown cells were compared with control shRNA knockdown cells. Bars represent mean ± SEM. All data were analyzed using 1-way ANOVA.

Pancreatic islet δ cells receive the UCN3, a paracrine signal from pancreatic β cells, which consequently regulates somatostatin secretion, potentially through Gs-cAMP signaling (5). We therefore explored whether UCN3 and its receptor CRHR2 participate in the hypersecretion of somatostatin in Sst-Cre^+/– Cul4b^fl/Y mice. Interestingly, CRHR2 antagonist astressin 2B (Ast2B) significantly decreased glucose-stimulated somatostatin secretion from islets isolated from Sst-Cre^+/– Cul4b^fl/Y mice, whereas it was not found to have a significant effect on islets isolated from their WT littermates (Figure 4C). Correspondingly, high levels of glucose-induced insulin secretion from Sst-Cre^+/– Cul4b^fl/Y mice were restored to levels similar to those of the control WT mice via the application of Ast2B (Figure 4D).

Together, these results show that increased intracellular calcium and cAMP levels in pancreatic δ cells contribute to enhanced somatostatin secretion and impaired glucose homeostasis in Sst-Cre^+/– Cul4b^fl/Y mice. Specifically, the UCN3 receptor CRHR2 and L-type calcium channels play important roles in the increased levels of somatostatin secretion in Sst-Cre^+/– Cul4b^fl/Y mice.

Transcriptional upregulation of Cav1.2 and AC6 contributes to enhanced somatostatin secretion in CUL4B-deficient pancreatic δ cells. To investigate the molecular mechanisms underlying the increased intracellular calcium and cAMP levels in response to high glucose and high potassium levels in CUL4B-deficient mice, we measured the mRNA levels of calcium channels, adenylyl cyclases, G proteins, PKA, and transcription factors in CUL4B knockdown and control TGP52 cells (Figure 5A). Specifically, mRNA levels of Cav1.2, Cav1.3, AC6, and AC7 increased in CUL4B knockdown cells, and the mRNA level of hematopoietically expressed homeobox (Hhex) decreased. Consistent with upregulation of transcription, the protein levels of Cav1.2, Cav1.3, and AC6 in the CUL4B knockdown cells were higher than those in the control cells (Figure 5, B–E). To confirm that the expression of these specific calcium channels and adenylyl cyclases was regulated by CUL4B in a physiological context, we measured mRNA levels of the α subunits of L-type calcium channels and different adenylyl cyclases in primary δ cells isolated from Sst-Cre^+/– GFP^fl/+ Cul4b^fl/Y mice and their WT littermates (Figure 5F and Supplemental Figure 12). These results showed that the expression of key regulators of extracellular calcium influx and intracellular cAMP levels, including those of Cav1.2, Cav1.3, AC6, and AC7, were increased in CUL4B-deficient cells.

Figure 5

Transcriptional upregulation of Cav1.2 and AC6 contributes to the enhanced somatostatin secretion in Cul4b-deficient pancreatic δ cells. (A) The mRNA levels of L-type calcium channels, adenylyl cyclases, G proteins, PKA and transcription factors in CUL4B knockdown and control TGP52 cells, as determined by quantitative reverse-transcriptase PCR (qRT-PCR) (n = 3). (B–E) Cav1.1, Cav1.2, Cav1.3, AC6, AC7 and Gs protein levels in CUL4B knockdown and control TGP52 cells. B and D, representative Western blots from at least 3 experiments are shown. C and E, bar graph representation and statistical analyses of B and D, respectively. (F) qRT-PCR was used to analyze the mRNA levels of L-type calcium channel α subunits and adenylyl cyclases in primary δ cells from Sst-Cre^+/– GFP^fl/+ Cul4b^fl/Y and WT littermates (n = 7). (G) Effects of Cav1.2 or Cav1.3 knockdown on somatostatin secretion from CUL4B knockdown and control TGP52 cells (n = 4). (H) Effects of AC6 or AC7 knockdown on somatostatin secretion from CUL4B knockdown and control TGP52 cells. (I) Glucose-induced cAMP levels in AC6 or AC7 knockdown and control TGP52 cells, with or without CUL4B knockdown. Cells were incubated with 20 mM glucose for 15 minutes.*P < 0.05; **P < 0.01; ***P < 0.001. Sst-Cre^+/– GFP^fl/+ Cul4b^fl/Y mice were compared with their WT littermates, and CUL4B knockdown cells were compared with control shRNA-treated cells. The bars represent mean ± SEM. All data were analyzed using 1-way ANOVA.

To characterize the functional importance of the increased expression of specific L-type calcium channels, we knocked down Cav1.2 and Cav1.3 in CUL4B knockdown and control TGP52 cells. We found that Cav1.2, but not Cav1.3, knockdown completely eliminated enhanced somatostatin secretion (Figure 5G and Supplemental Figure 7, A–C). We next measured the effects of AC6 and AC7 knockdown on somatostatin secretion in CUL4B knockdown and control TGP52 cells. Both AC6 and AC7 knockdown decreased somatostatin secretion levels in CUL4B knockdown and control TGP52 cells (Figure 5H and Supplemental Figure 7, D–F). In particular, somatostatin secretion in CUL4B knockdown TGP52 cells with siAC6 was decreased to control levels whereas the somatostatin secretion of CUL4B knockdown TGP52 cells with siAC7 was still higher than that of control cells (Figure 5H). Correspondingly, only the knockdown of AC6 completely eliminated high glucose–induced cAMP elevation in CUL4B knockdown cells compared with the control cells (Figure 5I). These results showed that specific key second messenger regulators, including Cav1.2 and AC6, promote enhanced somatostatin secretion in Cul4b-deficient pancreatic δ cells.

Epigenetic mechanisms and HHEX regulation driving Cav1.2 and AC6 expression in CUL4B-deficient pancreatic δ cells. CUL4B is a known E3 ligase that directly regulates protein levels through ubiquitination (30, 42). To delineate how CUL4B regulates expression of Cav1.2 and AC6, we examined Cav1.2 and AC6 ubiquitination in CUL4B-overexpressing and control TGP52 cells (Figure 6A). In the presence or absence of protease inhibitor MG132, ubiquitination levels of Cav1.2 and AC6 were not increased in CUL4B-overexpressing TGP52 cells compared with control cells (Figure 6, A–C). These results demonstrate that CUL4B does not promote Cav1.2 and AC6 ubiquitination in pancreatic δ cells.

Figure 6

CUL4B negatively regulates Cav1.2 and AC6 expression in pancreatic δ cells through an epigenetic mechanism. (A–C) Levels of ubiquitinated Cav1.2 and AC6 in CUL4B overexpression and control TGP52 cells. HA-tagged ubiquitin was transiently transfected into WT and CUL4B overexpression TGP52 cell lines in the presence or absence of 10 nM MG-132. The ubiquitinated proteins were pulled down with anti-HA beads, and ubiquitinated Cav1.2 and AC6 were detected with specific antibodies. Representative Western blots from at least 3 independent experiments are shown. (D–I) qChIP analysis of the recruitment of the indicated proteins on specific regions of the Cav1.2 promoter in CUL4B knockdown and control TGP52 cells. Whereas CUL4B, DDB1, K119-ubiquitinated-H2A, EZH2, and K27-3–methylated H3 protein bound to the –1500 to –1219 region of the Cav1.2 promoter, K4-3–methylated H3 bound to the –1060 to –900 region (n = 3). (J) qChIP analysis of the recruitment of CUL4B on specific regions of the Cav1.2 and AC6 promoter in HHEX knockdown and control TGP52 cells (n = 4). (K and L) qChIP analysis of HHEX binding patterns of the Cav1.2 and AC6 promoters in TGP52 cells (n = 4). (M) Coimmunoprecipitation of HHEX and the CRL4B-PRC2 component in TGP52 cell lines. TGP52 cells were lysed and incubated with HHEX antibody or normal rabbit/mouse IgG on a rotator at 4°C overnight, followed by the addition of protein A/G beads for 2 hours at 4°C. Beads were then washed 5 times with lysis buffer. *P < 0.05; **P < 0.01; ***P < 0.001. CUL4B knockdown cells were compared with control shRNA-treated cells. Bars represent mean ± SEM. All data were analyzed using 1-way ANOVA.

CRL4B has been shown to repress transcription in cooperation with the PRC2 complex (28). We then investigated whether CUL4B regulates the gene expression of Cav1.2 and AC6 through an epigenetic mechanism. AC6 has only 1 mRNA isoform, whereas Cav1.2 has 3 mRNA-splicing isoforms at the initial transcriptional site (43) (Supplemental Figure 8A). Pancreatic δ cells primarily express Cav1.2b and Cav1.2c isoforms, which share the same promoter (Supplemental Figure 8B). The mRNA levels of these isoforms increased in CUL4B knockdown cells (Supplemental Figure 8, B–D). We therefore performed a quantitative ChIP (qChIP) profiling analysis of the binding patterns of CUL4B, DDB1, and the histone-lysine N-methyltransferase EZH2 for a region located approximately 200 bp upstream to 1,500 bp downstream of the Cav1.2 and AC6 transcription start sites (Figure 6, D–I, and Supplemental Figure 8, E–J). The qChIP assay revealed that the occupancy sites of these proteins overlap and that the binding of CRL4B (CUL4B and DDB1) and PRC2 (EZH2) both peaked in the region at around –1,500 bp to –1,219 bp and –740 bp to –423 bp of the Cav1.2 and AC6 promoters, respectively (Figure 6, D–F, and Supplemental Figure 8, E–G). Notably, H2AK119ub1 and H3K27me3, 2 histone marks of transcription repression, were also enriched in the same region (Figure 6, G and H and Supplemental Figure 8, H and I).

We next examined the effect of CUL4B depletion on the recruitment of the CRL4B-PRC2 complex to the Cav1.2 and AC6 promoters in pancreatic δ cells. Knockdown of CUL4B resulted in a marked reduction in the levels of CUL4B, DDB1, and EZH2 bound to the Cav1.2 and AC6 promoters (Figure 6, D–F, and Supplemental Figure 8, E–G). Consistently, the levels of H2AK119ub1 and H3K27me3 were also markedly decreased at the Cav1.2 and AC6 promoters (Figure 6, G and H, and Supplemental Figure 8, H and I). In particular, H3K4me3 was enriched in the 1,060-bp to –900-bp region of the Cav1.2 promoter (Figure 6I). Increased H3K4me3 levels in the promoter regions of Cav1.2 and AC6 were also observed in CUL4B knockdown pancreatic δ cells, potentially contributing to elevated Cav1.2 and AC6 expression levels observed in these cells (Figure 6I and Supplemental Figure 8J). Together, these data show that CUL4B, a key component of CRL4B, represses the expression of Cav1.2 and AC6 by colocalizing with PRC2 at specific promoter regions of Cav1.2 and AC6, thus epigenetically regulating the ubiquitination and methylation of their associated histones.

Interestingly, we noticed that upregulation of Cav1.2 and AC6 are specific to CUL4B-deficient pancreatic δ cells, but not β cells (Supplemental Figure 6, A–E). We therefore tested the connection between HHEX, a pancreatic δ cell–specific transcriptional factor essential for δ cell differentiation (11), and CUL4B-regulated Cav1.2 and AC6 expression. qChIP showed that knockdown of HHEX in TGP52 cells significantly decreased CUL4B recruitment to the promoters of Cav1.2 and AC6 (Figure 6J), suggesting that HHEX promotes the recruitment and/or retention of the CUL4B complex at the promoters of targeted genes. We next performed a qChIP assay to profile the binding patterns of HHEX on a region of approximately 5-kb upstream and downstream around the Cav1.2 and AC6 transcription start sites (Figure 6, K and L, and Supplemental Figure 10, A–C). The results indicated that HHEX directly binds to the DNA region of the 4,568–4,918 downstream of the transcriptional initiation site of Cav1.2 and the DNA region of 3,981 to 4,417 downstream of the initiation site of AC6 (Figure 6, K and L). Moreover, HHEX physically associated with the CRL4B-PRC2 complex in TGP52 cells, as demonstrated by coimmunoprecipitation assays (Figure 6M and Supplemental Figure 10D). Together, these results show that HHEX participates in the epigenetic regulation of Cav1.2 and AC6 by CUL4B in pancreatic δ cells.

Constitution of the CRL4B-PRC2 complex regulating Cav1.2 and AC6 expression. To further define the functional interplay between the CRL4B and PRC2 complexes in the transcriptional repression of Cav1.2 and AC6, we performed ChIP – Re-ChIP experiments. Soluble chromatin from TGP52 cells was first immunoprecipitated with antibodies against CUL4B, DDB1, or EZH2 and subsequently reimmunoprecipitated with the other antibodies. The results indicated that the Cav1.2 and AC6 promoters that immunoprecipitated with CUL4B antibody could be reimmunoprecipitated with antibodies against DDB1 and EZH2 (Figure 7, A and B). Similar results were acquired in sequential ChIP using DDB1 or EZH2 as the first antibody (Figure 7, A and B). These data support the notion that CUL4B, DDB1, and EZH2 constitute 1 multi-subunit complex at the promoters of targeted genes.

Figure 7

CUL4B and DDB1 regulate somatostatin secretion by recruiting EZH2 to constitute a functional CRL4B-PRC2 complex at the promoter regions of Cav1.2 and AC6. (A and B) ChIP and Re-ChIP (sequential ChIP) experiments revealed that CUL4B, DDB1, and EZH2 form multiple protein complexes in the promoter regions of Cav1.2 (A) and AC6 (B) in TGP52 pancreatic δ cells. (C and D) qChIP analysis of the Cav1.2 (C) and AC6 (D) promoters in CUL4B, DDB1, or EZH2 knockdown TGP52 cells using the indicated antibodies. The results indicate that CRL4B promotes the recruitment of PRC2 to the promoters of Cav1.2 and AC6 (n = 3). (E and F) Western blot analysis demonstrated increased expression of Cav1.2 and AC6 in DDB1 (E) and EZH2 (F) knockdown TGP52 cells. A representative Western blot from at least 3 experiments is shown. (G) High glucose–induced somatostatin secretion in CUL4B, DDB1, and EZH2 knockdown TGP52 cells after 15 minutes of stimulation (n = 3). (H) Schematic of the sequential IP strategy and a model of CUL4B/DDB1/EZH2 assembly at the promoters of Cav1.2 and A6. *P < 0.05; **P < 0.01. CUL4B, DDB1, and EZH2 knockdown cells were compared with control shRNA- or siRNA-treated cells. Bars represent mean ± SEM. All data were analyzed using 1-way ANOVA.

We next knocked down CUL4B, DDB1, and EZH2 individually to determine how the CRL4B-PRC2 complex is assembled at promoter regions of Cav1.2 and AC6. CUL4B and DDB1 knockdown drastically reduced the recruitment of other components of the CRL4B-PRC2 complex as well as H3K27me3 levels of promoter regions of Cav1.2 and AC6 (Figure 7, C and D, and Supplemental Figure 9, A and B). In contrast, knockdown of EZH2 did not affect the recruitment of CUL4B and DDB1 to the promoter region of Cav1.2 and AC6 (Figure 7, C and D, and Supplemental Figure 9, A and B). These data suggest that CRL4B promotes the recruitment and/or retention of PRC2 at the promoter regions of Cav1.2 and AC6. Critically, when we knocked down EZH2 or DDB1 in TGP52 cells, we observed significant increases in protein levels of Cav1.2 and AC6 (Figure 7, E and F, and Supplemental Figure 9, C and D). Moreover, high glucose–induced somatostatin secretion levels were significantly increased in EZH2 or DDB1 knockdown TGP52 cells and to an extent similar to that observed in CUL4B knockdown cells (Figure 7G). Together, these data demonstrate that the integrity of the CRL4B-PRC2 complex is required for the transcriptional repression of Cav1.2 and AC6 in pancreatic δ cells (Figure 7H).

Glucose and UCN3 inversely regulate Cav1.2 and AC6 expression by modulating the CRL4B-PRC2 complex. We monitored how protein levels of key second messenger regulators including Cav1.2 and AC6 are regulated by the CRL4B-PRC2 complex in TGP52 cells in response to different stimulations (e.g., glucose stimulation or UCN3 action). Interestingly, a significant increase in CUL4B protein levels was observed in high glucose–treated islets and TGP52 cells in a time-dependent manner (Figure 8, A and B, and Supplemental Figure 11, A and B). Similarly, EZH2 protein levels increased after persistently high glucose application whereas protein levels of Cav1.2 and AC6 decreased (Figure 8, C and D). In contrast, UCN3 induced significant decreases in the protein levels of CUL4B and EZH2 in TGP52 cells whereas increases in Cav1.2 and AC6 expression were observed with continued incubation (Figure 8, E and F). A similar inverse trend in Cul4b, Cav1.2, and AC6 mRNA levels was observed in TGP52 cells and primary pancreatic δ cells (Figure 8, G and H, and Supplemental Figure 11, C–E). Importantly, CUL4B knockdown and Ast2B application in TGP52 cells eliminated protein- and mRNA-level changes in Cav1.2 and AC6 in response to high glucose levels and UCN3 stimulation, respectively (Figure 8, I–L, and Supplemental Figure 11F). These data showed that CRHR2 activity and CUL4B may regulate Cav1.2 and AC6 expression in pancreatic δ cells in response to high glucose or UCN3 levels and thus may play a role in the regulation of somatostatin secretion.

Figure 8

Reciprocal regulation of the CUL4B/PRC2 complex, Cav1.2, and AC6 by glucose and UCN3 in pancreatic δ cells. (A and B) Protein-level changes in CUL4B, EZH2, and DDB1 expression in TGP52 cells stimulated with 20 mM glucose for the indicated time intervals. (C, D, I, and J) Protein-level changes in Cav1.2 and AC6 expression in CUL4B knockdown and control TGP52 cells stimulated with 20 mM glucose for the indicated time intervals. (E and F) Protein level changes in CUL4B, EZH2, DDB1, Cav1.2, and AC6 expression in TGP52 cells stimulated with 100 nM UCN3. (G and H). The mRNA levels of Cul4b, Cav1.2, and AC6 in primary isolated islet δ cells treated with 20 mM glucose (G) or 100 nM UCN3 (H) for the indicated number of hours. (K and L) Protein level changes in CUL4B, EZH2, DDB1, Cav1.2, and AC6 expression in TGP52 cells stimulated with 100 nM UCN3 and CRHR2-selective antagonist Ast2B for the indicated time intervals. (A, C, E, I, and K) Representative Western blots from at least 3 independent experiments are shown. (B, D, F, J, and L) Statistical analyses of A, C, E, I and K, respectively. (M and N) AC inhibitor DDA (100 μM) and L-type channel blocker isradipine (10 μM) completely blocked the enhanced somatostatin secretion from human islets incubated with UCN3 (100 nM) (M) or MLN4924 (200 nM) (N) compared with their control islets. *P < 0.05; **P < 0.01; ***P < 0.001. Glucose- and UCN3-stimulated cells were compared with nonstimulated cells, and human islets incubated with UCN3 were compared with control islets. Bars represent mean ± SEM. All data were analyzed through 1-way ANOVA.

We therefore activated CRHR2 in human pancreatic islets via incubation with UCN3 for 4 hours. Previous studies have demonstrated that UCN3 potentiates high glucose–induced somatostatin secretion, which was inhibited by the L-type calcium channel blocker isradipine (5). Confirming previous findings, we observed that 4 hours of incubation with UCN3 caused an adaptive increase in somatostatin secretion in human islets in response to high glucose, which was eliminated by incubation with the AC inhibitor DDA or the Cav1.2 antagonist isradipine (Figure 8M). Moreover, when we used selective NEDD8-activating enzyme (NAE) inhibitors (MLN4924) to block the CRL4B function in human islets, significantly higher levels of somatostatin secretion in response to high glucose levels were observed (Figure 8N). The effects of MLN4924 on human islet δ cells were eliminated through the incubation of either the AC inhibitor DDA or the Cav1.2 antagonist isradipine (Figure 8N). Therefore, the epigenetic mechanism modulates the extent of glucose-induced somatostatin secretion, which requires both AC and L-type calcium channel activities.

The pancreatic islets of Langerhans are composed of at least 5 different cell types, including insulin-secreting β cells, glucagon-secreting α cells, somatostatin-secreting δ cells, ghrelin-secreting ε cells, and pancreatic polypeptide-secreting pp cells (44). Specific paracrine interactions among these islet cells are critical in mediating appropriate responses to different physiological contexts that cannot be achieved by a single β or α cell (2, 4, 5, 45). For example, recent studies have shown that the loss of UCN3 stimulation by pancreatic β cells decreases somatostatin-dependent negative feedback and disrupts glucose homeostasis (5). As an important paracrine factor, UCN3 potentiates high glucose–induced somatostatin secretion, which requires L-type calcium channel activity (5). Here, we confirmed these observations in human islets (Figure 8M). Moreover, our results demonstrate that somatostatin secretion is controlled by an epigenetic mechanism through modulation of the expression of CUL4B and EZH2, 2 core components of the CRL4B-PRC2 supercomplex. High concentrations of glucose induce somatostatin secretion, and this is immediately followed by gradually increasing levels of CUL4B and EZH2 expression (Figure 8, A, B, and G, and Figure 9). When high glucose levels persist, these signals may progressively decrease somatostatin secretion and impede the somatostatin negative feedback loop. In contrast, UCN3 inhibits the expression of CUL4B and EZH2, thereby increasing somatostatin secretion in a time-dependent manner and ensuring decreased insulin release after glucose levels return to baseline levels (Figure 8, E, F, and H, and Figure 9). Therefore, we found that the CRL4B-PRC2 complex plays an important role in the modulation of paracrine interactions between pancreatic δ and β cells. Moreover, we identified that CUL4B dysfunction in pancreatic δ cells results in insufficient insulin secretion and in glucose homeostasis disturbance.

Figure 9

Graphic model of the CRL4B-PRC2 complex–mediated regulation of pancreatic somatostatin paracrine signaling and of adjustments in glucose homeostasis in response to high glucose or UCN3 stimulation. As demonstrated by previous studies, paracrine factor UCN3 and insulin were coreleased from the same vesicles from pancreatic β cells (5). The secreted UCN3 increases somatostatin secretion in response to high glucose stimulation in pancreatic islets, and L-type channel activity is required for the potentiation effects of UCN3 (5). Here, we found that in response to sustained UCN3 stimulation, the expression levels of CUL4B and EZH2, the core components of CRL4B-PRC2, decreased in pancreatic δ cells. Under the control of HHEX, the CRL4B-PRC2 complex epigenetically regulates the expression of calcium channel Cav1.2 and adenylyl cyclase AC6 through the modulation of methylation and ubiquitination states of the histone proteins that bind to the promoters of these 2 proteins in pancreatic δ cells. The decreased expression of CUL4B and EZH2 releases the repressive role of the CRL4B-PRC2 complex in the expression of Cav1.2 and AC6, thereby increasing levels for second messengers Ca²⁺ and cAMP in pancreatic δ cells and ultimately leading to increased somatostatin secretion levels. Consistent with a previous study (5), this feedback loop is essential for avoiding excessive sustained insulin secretion in a timely manner and for coordinating normal glucose homeostasis. In contrast, the protein levels of CUL4B and EZH2 increase in response to sustained high glucose stimulation. The upregulated CRL4B-PRC2 complex function decreases the expression of Cav1.2 and AC6, which in turn decreases somatostatin secretion and thus ensures sufficient insulin secretion for maintaining glucose homeostasis. The epigenetic regulation of somatostatin secretion by the CRL4B-PRC2 complex is important for maintaining glucose homeostasis.

Recent studies have demonstrated that both the R-type calcium channel and the L-type calcium channel play important roles in the modulation of somatostatin secretion (5, 12). Here, we demonstrate that the expression levels of L-type calcium channels rather than R-type calcium channels are closely regulated by the CRL4B-PRC2 complex. Cav1.2 mediates excitation-contraction coupling in the heart (43), excitation-transcription coupling in the brain, and excitation-secretion coupling in many endocrine cells (46). Mutations in the Cav1.2 gene cause psychiatric disorders, such as bipolar disorder and schizophrenia (47). However, little is known of the transcriptional regulation of Cav1.2 (48, 49), and specific epigenetic mechanisms that regulate Cav1.2 have not been identified. Here, we found that CUL4B epigenetically modifies Cav1.2 expression levels in the CRL4B-PRC2 complex. Our data demonstrate that the CRL4B-PRC2 complex binds to the Cav1.2 promoter, thus repressing its expression by catalyzing H2K119 monoubiquitin and H3K27 methylation (Figure 6, D–I; Supplemental Figure 8, E–J, and Figure 9). Moreover, we observed that Cav1.2 expression was specifically regulated in pancreatic δ cells, but not in pancreatic β cells, owing to selective HHEX interactions (Figure 6, J–M). The diabetes gene Hhex is specifically expressed in pancreatic δ cells and is important for δ cell differentiation (11). In TGP52 cells, HHEX associates with the CRL4B-PRC2 complex to control the expression of Cav1.2 and AC6 (Figure 6, J–M). These results further indicate that the epigenetic control of key regulators of hormone secretion is cell-type specific.

Notably, Ins2-Cre and Sst-Cre mice may exhibit induced recombination in nonislet cells, such as in the brain and in the stomach (50, 51). Although we demonstrated that the hyposecretion of insulin and the hypersecretion of somatostatin from pancreatic islets aid in maintaining normal glucose homeostasis (Figure 1H and Figure 2), the deficiencies of CUL4B in SST-synthesized cells other than islets may also contribute to impaired glucose tolerance in Sst-Cre^+/– Cul4b^fl/Y mice, a possibility that requires further study. In addition, many of our biochemical studies used the pancreatic epithelial-like TGP52 cell line, an imperfect model for monitoring epigenetic interactions of islets. Therefore, in vivo chromatin characterizations and their relationships to the CRL4B-PRC2 complex in islets may provide important information on the epigenetic regulation of islet paracrine interactions. In the current study, to further extend the identified functional importance of the epigenetic mechanism for the regulation of somatostatin secretion in human islets, we blocked CUL4B activity using the NEDD8-activating enzyme inhibitor MLN4924 (Figure 8N), which adaptively potentiates high glucose–induced somatostatin secretion, phenocopying the effects of UCN3 (Figure 8M), as shown in previously published results (5). Moreover, DDA and isradipine application eliminated the effects of long-term MLN4924 treatment on facilitating somatostatin secretion, suggesting that our hypothesis derived from mouse islets and TGP52 cells may also be applied to human tissues (Figure 8N). Therefore, the dysfunction of epigenetic regulation in pancreatic δ cells may lead to somatostatin hypersecretion. Given that antagonizing CRHR2 activity rescued somatostatin hypersecretion in a mouse model of Cul4b deficiency specifically in δ cells (Figure 4, C and D) and that CUL4B mutations are related to obesity (24), a CRHR2 antagonist, such as Ast2B (5), might be developed as a more specific therapeutic method for malfunctions of CRL4B-PRC2 complexes.

Together, our results reveal an important epigenetic mechanism that modulates paracrine interactions between pancreatic δ and β cells during the maintenance of glucose homeostasis, the functional importance of the CRL4B-PRC2 complex in this system of regulation, and the mechanism by which the expression of a calcium channel and adenylyl cyclase are specifically controlled in pancreatic δ cells.

Animals. The cre driver mice Sst-Cre (Ssttm 2.1[cre], stock number 013044) and Ins2-Cre (B6.Cg-Tg[Ins2-cre]25Mgn/J, stock number 003573) and GFP mice (stock number 007676) were purchased from Jackson Laboratory. Ssttm 2.1 (cre) mice with a congenic C57BL/6 background were obtained by crossing Ssttm 2.1 (cre) mice with a C57BL/6 background. Cul4b^fl/fl mice were generated at the National Resource Center of Mutant Mice/Model Animal Research Center of Nanjing University, Nanjing, Jiangsu, China) (52). Tdt-tomato mice were obtained from Tian Xue (53). All experiments were performed with male mice.

Glucose tolerance test. Male mice 12 to 14 weeks of age were fasted for 16 hours and given free access to water. Blood glucose levels were measured in tail blood by using a FreeStyle Lite Glucose Meter (Roche) before (baseline, 0 minutes) as well as 15, 30, 60, 90, and 120 minutes after mice were intraperitoneally challenged with 2 g/kg body weight glucose.

ChIP. We analyzed the enrichment of the DNA template by conventional PCR using primers specific for Cav1.1, Cav1.2, Cav1.3, Cav1.4, AC6, or AC7 promoters. All primers used for ChIP assay of CRL4B or PRC2 complexes are listed in Supplemental Table 1. Detailed information is available in Supplemental Materials and Methods.

RNA extraction and quantitative reverse-transcriptase PCR. RNA from CUL4B knockdown and control TGP52 cells as well as islets derived from Sst-Cre^+/– GFP ^fl/+ Cul4b^fl/Y mice and their WT littermates (Sst-Cre^+/– GFP ^fl/+ Cul4b^+/Y mice) were extracted with TRIzol reagent (Invitrogen). The single δ cell was picked by hand and incubated with buffer containing the RNase inhibitor (N8080119, Thermo Fisher Scientific). We performed cDNA synthesis using the qRT-PCR Kit (Toyobo, FSQ-101). We conducted quantitative reverse-transcriptase PCR (qRT-PCR) in the LightCycler qPCR apparatus (Bio-Rad) with the FastStart SYBR Green Master (Roche). All primers used for qRT-PCR assay are listed in Supplemental Table 2.

Statistics. All data are presented as mean ± SD or mean ± SEM. Differences between different groups were determined using 1-way ANOVA in GraphPad Prism 5 software. A P value of less than 0.05 was considered significant.

Study approval. The use of isolated human islets was approved by the ethics committee of The Central South University Second Xiangya Hospital (protocol MSRC2016LF). All animal care and experiments were reviewed and approved by the Animal Use Committee of Shandong University School of Medicine.