Pharmacologic rescue of circadian β-cell failure through P2Y1 purinergic receptor identified by small-molecule screen

The mammalian circadian clock drives daily oscillations in physiology and behavior through an autoregulatory transcription feedback loop present in central and peripheral cells. Ablation of the core clock within the endocrine pancreas of adult animals impairs the transcription and splicing of genes involved in hormone exocytosis and causes hypoinsulinemic diabetes. However, identification of druggable proteins and pathways to ameliorate the burden of circadian metabolic disease remains a challenge. Here, we generated β cells expressing a nano-luciferase reporter within the proinsulin polypeptide to screen 2,640 pharmacologically-active compounds and identify insulinotropic molecules that bypass the secretory defect in clock mutant β cells. We validated lead compounds in primary mouse islets and identified known modulators of ligandgated ion channels and G-protein coupled receptors, including the antihelmintic ivermectin. Single-cell electrophysiology in circadian mutant mouse and human cadaveric islets validated ivermectin as a glucose-dependent secretagogue. Genetic, genomic, and pharmacologic analyses established that the molecular clock controls the expression of the purinergic P2Y1 receptor to mediate the insulinotropic activity of ivermectin. These findings identify the P2Y1 purinergic receptor as a target to rescue circadian β-cell failure and establish a chemical genetic screen for endocrine therapeutics.


Summary:
The mammalian circadian clock drives daily oscillations in physiology and behavior through an autoregulatory transcription feedback loop present in central and peripheral cells. Ablation of the core clock within the endocrine pancreas of adult animals impairs the transcription and splicing of genes involved in hormone exocytosis and causes hypoinsulinemic diabetes. However, identification of druggable proteins and pathways to ameliorate the burden of circadian metabolic disease remains a challenge. Here, we generated b cells expressing a nano-luciferase reporter within the proinsulin polypeptide to screen 2,640 pharmacologically-active compounds and identify insulinotropic molecules that bypass the secretory defect in clock mutant b cells. We validated lead compounds in primary mouse islets and identified known modulators of ligandgated ion channels and G-protein coupled receptors, including the antihelmintic ivermectin.
Single-cell electrophysiology in circadian mutant mouse and human cadaveric islets validated ivermectin as a glucose-dependent secretagogue. Genetic, genomic, and pharmacologic analyses established that the molecular clock controls the expression of the purinergic P2Y1 receptor to mediate the insulinotropic activity of ivermectin. These findings identify the P2Y1 purinergic receptor as a target to rescue circadian b-cell failure and establish a chemical genetic screen for endocrine therapeutics. 5 PMA (positive control known to enhance insulin secretion in both Bmal1 -/mouse islets and Beta-62 TC-6 cells) (10). Luciferase intensity from the supernatant was measured following exposure to 63 NanoGlo Luciferase Assay Substrate (Fig 1E). 64 65 We initially identified 19 hit compounds that both significantly enhanced insulin secretion and 66 elicited a response of greater than 3 standard deviations from the mean (Z score > 3) with more 67 than a 1.25-fold increase, exceeding the upper 99% confidence interval of the negative control 68 (Fig 2A, Fig S1A, Table S1). Of these, seven were excluded from further analysis because of 69 reported toxic effects or lack of availability of the compound (Fig S1A). The remaining 12 hit 70 compounds mediate activity of ligand-gated cell surface receptors and ion channels that stimulate 71 second messenger signaling cascades (Fig 2B-C) (12,13). Of these, four target ion channels 72 (tacrine hydrochloride, suloctidil, dyclonine hydrochloride, and ivermectin) (Figs 2B-C) (14-23). 73 Five target seven-transmembrane G-protein coupled receptors (GPCRs) that signal through 74 phospholipase C (PLC) and diacylglycerol (DAG) to activate insulin secretion and β-cell gene 75 transcription (benzalkonium chloride, carbachol, isoetharine mesylate, pipamperone, and 76 ivermectin) (Figs 2B-C) (17,(24)(25)(26)(27)(28)(29)(30). Similar to the hit compounds of our screen, our previous 77 results showed that carbachol, a muscarinic Gq-coupled receptor agonist, and the DAG mimetic 78 PMA rescue insulin secretion in Bmal1 -/islets (10). Four additional hit compounds act as 79 acetylcholinesterase inhibitors, promoting enhanced glucose-dependent insulin secretion in 80 response to acetylcholine through the muscarinic GPCRs, as well as the ionotropic nicotinic 81 acetylcholine receptors (tyrothricin, tomatine, carbachol, and tacrine hydrochloride) (Figs 2B-C) 82 (31-36). One compound has been shown to promote insulin secretion by inhibition of the 83 mitochondrial protein tyrosine phosphatase PTPM1 (alexidine hydrochloride) (Figs 2B-C) (37,84 38), and another likely affects b-cell function by signaling through the mineralocorticoid receptor 85 (deoxycorticosterone) (Figs 2B-C) (39). Finally, in addition to ion channels and GPCRs, the 86 macrolide ivermectin has also been shown to signal in micromolar concentrations though several 87 ionotropic receptors, including purinergic, GABAergic, and glycine receptors, as well as through 88 the farnesoid X nuclear receptor (17,40,41). 89 90 Ten of these twelve hit compounds were not considered for further analysis because of either the 91 high dose required to achieve insulin secretion (Fig S1B) or because they augmented insulin 92 6 release in low basal glucose (2 mM) in intact WT mouse primary islets (Fig 2D). One of the 93 remaining compounds induces hepatotoxicity after prolonged use (tacrine hydrochloride) (42). We 94 therefore focused our attention on ivermectin (IVM) due to its dose-dependent enhancement of 95 glucose-stimulated insulin secretion in insulin-NanoLuc-expressing Beta-TC-6 cells, as well as its 96 robust rescue of insulin secretion in Bmal1 -/islets (Fig 2D-E). with our initial bioluminescence assay, we observed that IVM enhanced insulin secretion in a 103 glucose-dependent manner following both 1-hr IVM exposure and 24-hr pre-treatment with IVM 104 in β-cell lines and WT mouse islets, suggesting both acute and longer-term exposure to IVM 105 enhances b-cell function (Figs 3A, S2A). Since there was not a significant increase in insulin 106 secretion with overnight (~2 fold) compared to acute (~1.5-1.6 fold) IVM exposure, further 107 analysis of IVM as a potentiator of insulin secretion was performed only with acute treatment. 108 109 Chemical energy from ATP generated by glucose metabolism within the β cell triggers closure of 110 the sulfonylurea-linked potassium channel, depolarization of the plasma membrane, and opening 111 of voltage-gated calcium channels leading to stimulus-secretion coupling. To assess the 112 mechanism of IVM-induced insulin secretion, we next monitored real-time calcium influx using 113 ratiometric fluorescence imaging in WT b cells in the presence of both glucose and IVM. We 114 observed an immediate and robust glucose-stimulated intracellular calcium response within 2 115 minutes of IVM stimulation (p<0.05) (Fig 3B). Importantly, this effect was only observed in the 116 presence of high glucose, consistent with results of our initial NanoLuc 384-well plate screening 117 and subsequent ELISA-based analyses of glucose-stimulated insulin secretion. In contrast, the 118 Ca 2+ channel inhibitor isradipine completely suppressed Ca 2+ influx and insulin secretion (Fig  119   S2D-E) (43). To determine whether increased calcium influx corresponded with productive insulin 120 release following IVM treatment, we used a perifusion system to directly measure NanoLuc 121 activity in eluates harvested from IVM-treated β cells during both the first and second phase of 122 insulin secretion (Fig 3C). IVM significantly increased insulin release by 12 minutes post-123 7 stimulation (p<0.05) and throughout most of the second phase of insulin secretion (>15 min), 124 consistent with continuous release of reserve insulin granules (44) (Fig 3C). 125 126 Since our cell-based studies indicated that IVM stimulates GSIS within immortalized b-cell lines, 127 we next sought to determine whether IVM restores insulin secretion in the context of circadian 128 disruption within primary islets, which are composed of multiple hormone-releasing cell types 129 (45). To test this idea, we administered IVM to mouse islets isolated from pancreas-specific Bmal1 -130 /mice, revealing a 3.3-fold elevation of GSIS following exposure to the drug in the mutant islets 131 (Fig 3D). To determine if IVM can improve glucose homeostasis in diabetic animals, we next 132 tested the effects of chronic IVM administration in the well-characterized Akita model of β-cell 133 failure (46). Daily intraperitoneal IVM (1.3 mg/kg body weight) was administered to Akita mice 134 over a 14-day period (47), terminating in assessment of glucose tolerance and ex vivo GSIS. 135 Treatment with IVM significantly improved glucose tolerance and augmented glucose-stimulated 136 insulin release from islets isolated from these mice (Fig S2B-C). Given that our prior genomic and 137 cell physiologic studies have localized the β-cell defect in circadian mutant mice to impaired 138 insulin exocytosis (11), and as IVM augmented insulin secretion in Bmal1 mutant islets, we next 139 sought to determine whether IVM might enhance depolarization-induced exocytosis using 140 electrophysiologic analyses (48). We assessed cumulative capacitance, a measure of increased cell 141 surface area as insulin granules fuse to the plasma membrane, in β cells from islets of control and 142 pancreas-specific Bmal1 mutant mice, as well as from human cadaveric islets. While Bmal1 mutant 143 cells displayed reduced rates of exocytosis following direct depolarization (as indicated by reduced 144 capacitance), 10 µM IVM treatment rescued the defect in Bmal1 mutant cells, increasing 145 cumulative capacitance from 11.0 to 20.7 fF/pF after 10 consecutive depolarization steps (Fig 3E). 146 IVM treatment also enhanced cumulative capacitance in human β cells from 17.9 to 39.7 fF/pF 147 (Fig 3F). Together these data show that IVM augments β-cell early calcium influx in a glucose-148 dependent manner to promote increased vesicle fusion and release. 149 150 Purinergic receptor P2Y1 mediates IVM-induced insulin exocytosis. In addition to IVM, 151 several of the predicted targets of the insulinotropic compounds from our screen involve second-152 messenger signaling, raising the possibility that circadian disruption may be overcome by 153 augmenting hormonal or metabolic factors that promote peptide exocytosis. IVM is a readily-154 8 absorbable and potent derivative of avermectin B1 that acts to allosterically regulate several 155 different types of cell surface receptors, including the purinergic and GABA receptors, as well as 156 nuclear transcription factors such as the farnesoid X receptor (FXR) (47, 49-51). Since IVM 157 augments insulin secretion in Bmal1 -/cells, we hypothesized that the expression of putative IVM 158 targets may be reduced during circadian disruption. First, through RNA-sequencing we observed 159 significantly higher levels of expression of the transcript encoding the purinergic receptor P2Y1 160 (P2ry1) in WT b cells compared to transcripts encoding FXR or GABA components (Fig S3A). 161 We further observed enrichment of BMAL1 chromatin binding within enhancer regions 266 -41 162 kb upstream of the P2ry1 gene transcription start site by chromatin immunoprecipitation-163 sequencing (GSE69889) (Fig 4A), as well as rhythmic expression of P2ry1 in wild-type Beta-TC-164 6 pseudoislets (Fig S3B). Pharmacologic inhibition of P2Y1 using the subtype-specific inhibitor MRS2179 in the presence 173 of both high glucose and 10 µM IVM resulted in a 52% reduction in insulin secretion by 174 bioluminescence and a reduction in calcium influx to levels similar to those observed during high 175 glucose alone, as assessed by Fura2-AM ratiometric determination of intracellular calcium (Figs 176 4B-C). In addition to evidence that pharmacological blockade of P2Y1 receptor signaling 177 abrogates IVM activity, we also tested the requirement of P2Y1 receptor signaling following 178 CRISPR-Cas9-mediated knockout of the P2Y1 receptor in both WT and Bmal1 -/b cells (Fig S4A). 179 While IVM enhanced glucose-stimulated insulin secretion in WT and Bmal1 -/b cells by 60% and 180 80%, respectively, IVM did not significantly enhance glucose-stimulated insulin secretion in cells 181 lacking the P2Y1 receptor (Fig 4D). Similar to the pharmacologic findings with the P2Y1 182 antagonist MRS2179, these results demonstrate a requirement for P2Y1 in IVM-induced GSIS. in β cells (Fig 4E). In WT cells, IVM induces differential expression of 65 transcripts (1.5-fold 194 change, Adj. P value < 0.05), including up-regulation of the immediate early gene Fos (58) and 195 down-regulation of Aldolase B, whose expression has been linked to reduced insulin secretion in 196 human islets (59) (Figs 4F, S4B). Strikingly, none of these transcripts were significantly altered 197 by IVM in the P2ry1 -/β cells (all adjusted P value > 0.05) (Figs 4F, S4B). Taken together, these 198 data suggest that the circadian clock program controls P2Y1 expression to modulate glucose-199 Little is known about P2Y1 targeting in disease states, such as circadian disruption and/or type 2 210 diabetes, or whether P2Y1 is controlled at a transcriptional level. Our evidence that P2Y1 is 211 expressed under control of the circadian clock derives from analyses at the level of both chromatin 212 binding by the core clock factor BMAL1 and genome-wide differential RNA expression analysis 213 in circadian mutants. Intriguingly, P2X and P2Y receptors are required for Ca 2+ signaling in the 214 suprachiasmatic nucleus (61, 62), yet their role in circadian regulation of peripheral tissues has not 215 been well studied. Our pharmacological and genetic analyses are the first to reveal that 216 enhancement of P2Y1 receptor activity can bypass the transcriptional deficits exhibited in 217 circadian mutant b cells and restore insulin secretion. Future studies will be required to determine 218 the precise mechanism by which IVM modulates P2Y1 activity. One possibility is that IVM may 219 augment P2X-P2Y1 crosstalk to drive insulin secretion, which has been shown to drive Ca 2+  The study of transcriptional rhythms across the 24-hr circadian cycle has previously revealed a 237 diverse landscape of clock-controlled genes and pathways (67). Despite the identification of 238 thousands of tissue-specific and clock-controlled transcripts, limited advances have been made in 239 utilizing this information to treat diseases associated with circadian disruption, including type 2 240 diabetes. One approach to this challenge has been to intervene and restore the molecular clock 241 program using pharmacology (Nobiletin) (8), micronutrient supplementation (NAD + precursors) 242 (68, 69), or enforced behavioral rhythms (such as time restricted feeding) (70). However, it remains 243 unclear how altering the whole-body clock will affect nutritional and hormonal dynamics at a 244 cellular level. Another approach has been to directly target clock-controlled genes with known 245 function in health and disease (71), or to look at gain/loss of circadian control in health versus 246 11 disease (72). This approach requires an understanding of gene function within a given tissue, and 247 thus limits the identification of novel therapeutic targets. In the studies performed here we sought 248 to address the challenge of connecting clock control of transcription with druggable targets by 249 using an unbiased small molecule drug screen, in tandem with functional genomics, to elucidate 250 mechanisms of insulin secretory dynamics. Since the circadian timing system has been shown to 251 not only regulate the function of mature b cells, but also the regenerative capacity of islets in both 252 the context of the mouse (73) and in human embryonic stem cell differentiation (74), molecules 253 identified in cell-based genetic screens may provide broad applicability as therapeutics. at Northwestern University. Screen feasibility was determined by calculating Z'-factor using the 317 following formula: Z'-factor = 1-3(σp + σn) / (µp -µn) (where σp is the standard deviation of positive 318 control, σn is the standard deviation of negative control, µp is the mean intensity of positive control, 319 and µn is the mean intensity of the negative control). 320 321 Determination of hit compounds. Z scores for luciferase intensities produced by screened 322 compounds were calculated from the following formula: z = (X -µ) / σ (where z is the Z score, X 323 is the intensity of the compounds, µ is the intensity of negative control (20mM glucose), and σ is 324 the standard deviation of negative control). A row-based correction factor was applied to all 325 luciferase readings to adjust for logarithmic signal decay. Hit compounds were defined as those 326 that elicited a response of greater than 3 standard deviations from the mean (Z score > 3) and more 327 than 1.25-fold increase compared to negative control, which is the cut-off for ~10% chance of the 328 observation occurring by random chance. Validated hit compounds that augmented insulin 329 secretion at low drug dose were considered lead compounds.

Compounds excluded
• inability to validate initial secretion results (n=3) • compound effect only at high dose (n=5)

Compounds validated
in WT islets for increased insulin secretion (n=4)

Compounds selected
for further analysis (n=1)