Whole-Organism Chemical Screening Identifies Modulators of Pancreatic β-Cell Function

Functional pancreatic endocrine cells, located in the islets of Langerhans, are crucial for blood glucose homeostasis. In response to changes in blood glucose levels, islets secrete hormones that act on peripheral tissues to normalize glucose levels. An example of this response is insulin secretion by pancreatic β-cells when blood glucose concentration increases after a meal. Insulin stimulates glucose uptake in peripheral tissues and the conversion of glucose to glycogen in the liver. Thus, optimal control of blood glucose levels depends on the precise control of insulin production and secretion by the pancreatic β-cells. Loss and dysfunction of β-cells causes type 1 and type 2 diabetes, respectively. Therefore, identifying new molecules and/or molecular pathways that stimulate β-cell function will contribute to the discovery of novel targets for future therapies for diabetes.

Phenotypic chemical screens using zebrafish (Danio rerio) have emerged as a powerful tool for rapid dissection of biological processes underlying complex traits (1). Over 50 screens have been reported spanning behavioral, cardiac, metabolic, and regenerative phenotypes (2), and these in vivo screening strategies have become an efficient platform to identify drug candidates for clinical repurposing (3). Zebrafish is particularly well suited to study pancreas development and physiology (4–8). Several groups, including ours, have previously performed zebrafish-based small-molecule screens and reported novel molecules and signaling pathways that stimulated β-cell differentiation, regeneration, or proliferation (9–12). However, a screen targeting in vivo β-cell function has not yet been reported.

In this study, we performed a small-molecule screen using the zebrafish model to identify in vivo modulators of β-cell function and glucose metabolism. Using zebrafish larvae, we identified 84 insulin stimulators that reduced glucose levels and 7 insulin repressors that increased glucose levels. Among the stimulators, we discovered the Kv1.3 blocker Psora4, which also stimulated β-cell function and peripheral insulin response and reduced glucose levels in mouse. Additionally, our data indicate that cytoplasmic Kv1.3 modulates β-cell function, revealing a new mechanism to modulate β-cell function.

All zebrafish husbandry was performed under standard conditions in accordance with institutional and national ethical and animal welfare guidelines.

We used the following transgenic lines: Tg(ins:Luc2;cryaa:mCherry)^gi3 (13), abbreviated ins:Luc2; Tg(pck1:Luc2,cryaa:mCherry)^s952 (14), abbreviated pck1:Luc2; Tg(ins:H2BGFP;ins:dsRED)^s960 (15), abbreviated ins:H2BGFP; Tg(P0-pax6b:eGFP)^ulg515 (16), abbreviated pax6b:eGFP; TgBAC(neurod1:eGFP)^nl1 (17), abbreviated neurod1:eGFP; and Tg(−2.6mnx1:GFP)^ml59 (18), abbreviated mnx1:eGFP.

The ins:Luc2 zebrafish line was incrossed, the progeny raised and screened for homozygous transgenic animals, which were then outcrossed to AB wild-type zebrafish to collect large numbers of heterozygous transgenic animals for experiments. Healthy larvae (4 days postfertilization [dpf]) were selected and washed with egg water before distributing three animals per well in 200 μL 10 mmol/L HEPES-buffered egg water in 96-well plates. Drugs were diluted from a 1 mmol/L stock solution in DMSO to a treatment concentration of 10 μmol/L in 1% DMSO. After 48 h of treatment, the samples were incubated in Steady-Glo (Promega) as described previously (14). Bioluminescence signal was analyzed by FLUOstar Omega (BMG LABTECH).

For the small-molecule screen, three 4-dpf larvae were tested in duplicate for every compound. Differential modulation of insulin promoter activity was determined by normalizing to the average of the DMSO control in each plate. An average of 2.5-fold upregulation or 0.5-fold downregulation was assigned as the threshold for calling hits, which were further tested in validation experiments and glucose measurements.

After drug treatment, 15 larvae per group were homogenized, and free glucose levels were determined by using a glucose assay kit (BioVision) as described previously (14). Blood glucose measurements in adult zebrafish were performed as described previously (13).

Total RNA from control and drug-treated larvae was extracted by RNeasy Mini Kit (Qiagen). RT-PCR was performed by using SuperScript III First-Strand Synthesis System (Invitrogen) according to the manufacturer’s instructions with 500 ng total RNA as template.

Quantitative RT-PCR was performed with 2× Dynamo Color Flash SYBR Green Master Mix (Thermo Fisher Scientific), and gene expression levels were quantified on a CFX Connect Real-time System (Bio-Rad) with gene-specific primers (Supplementary Table 5). Each sample was normalized to a housekeeping gene (actb).

Plasma insulin detection was performed as described previously (13).

Fluorescence images were acquired with an LSM 700 laser scanning confocal microscope (Carl Zeiss), and fluorescence intensity or each transgenic line was calculated using the ZEN software. Cell numbers were counted manually for experiments using the ins:H2BGFP transgenic line.

Immunostaining was performed as described previously (19). The following primary antibodies were used at the indicated dilutions in 1% FBS in PBS containing 0.1% Triton X-100: guinea pig anti-insulin (Abcam) at 1:100 and rabbit anti-Kv1.3 (Anemone Laboratories) at 1:100 (20,21). Alexa Fluor–conjugated secondary antibodies (Invitrogen) were used at 1:500–1:1,000 dilution with 1% FBS in PBS containing 0.1% Triton X-100.

Pancreatic islets were isolated from 16–20-week-old C57BL/6J mice following LiberaseTL (Roche) infusion via the common bile duct to disrupt the pancreatic exocrine tissue and filtration to recover intact islets. Islets were handpicked and cultured overnight in RPMI 1640 supplemented with 10% FBS, 100 units/mL penicillin, and 100 µg/mL streptomycin. Islets were treated with 10 μmol/L of the hit compounds for 24 h in RPMI 1640 containing 0.2% BSA, 100 units/mL penicillin, and 100 µg/mL streptomycin. Glucose-stimulated insulin secretion (GSIS) was conducted in Krebs-Ringer bicarbonate (KRB) buffer containing 4.6 mmol/L KCl, 2.5 mmol/L CaCl₂, 1.2 mmol/L MgSO₄, 1.2 mmol/L KH₂PO₄, 17.7 mmol/L NaHCO₃, 10 mmol/L HEPES, 117 mmol/L NaCl, and 0.2% BSA. Ten size-matched islets were preincubated for 1 h in KRB containing 3 mmol/L glucose prior to collection of the supernatant from incubations with fresh KRB containing 3 or 15 mmol/L glucose. Insulin secretion was assessed with a mouse insulin ELISA kit (Alpco) following the manufacturer’s protocol.

Hyperglycemia was induced in 8-week-old C57BL/6N male mice by intraperitoneal (IP) injection of 150 mg/kg streptozotocin (STZ) (22). Mice with nonfasting blood glucose levels >350 mg/dL 1 week after STZ administration were used. Psora4 was dissolved in peanut oil at 10 mg/mL with stirring at 70°C and IP injected to diabetic mice at 40 mg/kg. A total of 100 µL peanut oil was injected as a control. After 6 h of fasting, blood glucose levels were measured by MS-GT102 (Terumo Corporation) from peripheral blood obtained from the tail vein every 24 h starting at 1 day before the injection. A glucose tolerance test was performed 5 days after Psora4 injection. Blood glucose levels were measured at 0, 15, 30, 60, 90, and 120 min after IP injection of 2 g/kg glucose after 6 h of fasting. Insulin tolerance test was performed 5 days after Psora4 injection. Blood glucose levels were measured after IP injection of 0.5 units/kg Humulin R (Eli Lilly and Company) after 6 h of fasting. All mouse experiments were approved by the Institutional Animal Care and Use Committee and performed in accordance with the guidelines of the University of Tokyo.

To enable high-throughput small-molecule screening for modulators of β-cell function, we first optimized the quantification of insulin expression using an in vivo luciferase transgenic reporter (ins:Luc2) (13). We identified the optimal treatment window for screening using known modulators of β-cell differentiation (N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester [DAPT]) (23), proliferation (retinoic acid, glucosterone, trazodone, and prednisolone) (11), and maturation (triiodothyronine) (13) (Fig. 1A and B and Supplementary Fig. 1). We determined that treatments from 4–6 dpf generated the greatest dynamic range in the assay. Therefore, we performed a primary screen of 4,640 small molecules from 4–6 dpf at a fixed concentration of 10 μmol/L. We identified 229 small molecules that enhanced insulin reporter activity by at least 2.5-fold and 31 small molecules that repressed insulin reporter activity to less than half of the DMSO control samples (Fig. 1C).

Next, we applied a functional filter to identify compounds that were likely to modulate endogenous insulin expression. In the absence of compensatory responses, we predicted that insulin expression and glucose levels would be inversely correlated. Zebrafish larvae were treated with the hit compounds from 4–6 dpf, and free glucose levels were analyzed. In these experiments, 84 of the 229 insulin stimulators showed >20% glucose reduction, and 7 of the 31 insulin repressors showed >20% glucose elevation (Fig. 1D). Thus, these 91 compounds (1.96% of the library), affecting both insulin reporter expression and glucose metabolism, were selected for subsequent analyses.

The 48-h small-molecule treatments used in the primary screen may have had direct effects on insulin transcription and/or indirect effects on β-cell function because of other physiological changes in metabolism. In order to identify small molecules that were most likely to directly modulate insulin expression, shorter drug treatment windows were tested using the ins:Luc2 line. After 12-h drug treatments, 32 of the 84 stimulators from the primary screen enhanced the activity of the insulin expression reporter (12-h insulin stimulators) (Fig. 1E, Table 1, and Supplementary Table 1). An additional 14 compounds required 24 h of exposure to enhance insulin expression reporter activity (24-h insulin stimulators) (Fig. 1E and Supplementary Table 2). The remaining 38 insulin stimulators were only active under the original screening conditions (48-h insulin stimulators) (Fig. 1E and Supplementary Table 3), suggesting that this class indirectly induced insulin reporter expression. Surprisingly, two of the seven insulin repressors from the primary screen stimulated insulin promoter activity in 12- or 24-h assays, suggesting that they induce a compensatory response at 48 h. The five remaining insulin repressors required 48 h to repress ins:Luc2 in the assay (Fig. 1E and Supplementary Table 4).

The glucoregulatory effects of the insulin reporter modulators could be due to effects on gluconeogenesis itself. To investigate this possibility, we used a validated pck1:Luc2 reporter (14) to quantify the effects of the primary hits on the expression of a key, transcriptionally regulated, gluconeogenic enzyme. We classified small molecules that increased reporter activity by >3.5-fold as “strong pck1 inducers,” compounds that increased reporter activity from 1–3.5-fold as “weak pck1 inducers,” and drugs that led to reporter activity to <1-fold as “pck1 repressors” (Fig. 1F).

We prioritized the 12-h insulin stimulators for additional analyses because they were most likely to act directly on β-cells. Furthermore, we filtered the 12-h insulin stimulators to identify those that lowered glucose levels despite elevated pck1 expression, because enhanced gluconeogenesis could represent a homeostatic response to increased Insulin action. Based on these criteria, we selected 13 out of 32 of the 12-h insulin stimulators for further investigation (Fig. 1F and Table 1). Compensatory gluconeogenesis could reflect increased β-cell function and/or insulin sensitivity in target tissues. Interestingly, the 13 compounds included pioglitazone, a frequently prescribed antidiabetic drug of the thiazolidinedione family, which stimulates both β-cell function and insulin sensitivity in peripheral tissues (24,25). To identify novel β-cell stimulators, we selected 6 compounds among the remaining 12 that had not been previously characterized in detail: 2,4-dichlorophenoxybutyric acid (2,4-DB; auxin), guanidinyl-naltrindole di-trifluoroacetate (GNTI; κ-opioid receptor antagonist), Psora4 (Kv1.3 blocker), 2′,4′-dihydroxy-4-methoxychalcone (2′,4′-D-4-M; chalcone), SR-2640 (LTD4/LTE4), and Karanjin (flavonoid). Because the primary screen was performed at a fixed dose of 10 μmol/L, we performed dose-response assays (0–100 μmol/L) to determine the most effective concentration for each compound. Using acute 12-h treatments, we determined the concentration of each compound that elicited maximal induction of the ins:Luc2 reporter (Supplementary Fig. 2), which was then used in subsequent experiments. For compounds that exhibited toxicity at higher doses, we determined dose-dependent survival (Supplementary Fig. 3) and excluded 2′,4′-D-4-M from further analysis due to complete lethality at 20 μmol/L (double the primary screening dose).

The effects of the remaining five compounds (2,4-DB, GNTI, Psora4, SR-2640, and Karanjin) (Fig. 2A) on the expression level of endogenous insulin were quantified in wild-type zebrafish (Supplementary Fig. 4A and B). All of the compounds stimulated endogenous insulin expression after acute (12-h) as well as 48-h treatments (Supplementary Fig. 4A and B), further validating the primary screen. To determine whether the drug treatments impacted β-cell mass as a result of increased proliferation or differentiation, we treated ins:H2BGFP reporter fish, which express a stabilized nuclear-localized GFP that facilitates cell number quantification, starting at 4 dpf (Fig. 2B–G). As expected, based on the effects on endogenous insulin expression levels, ins:H2BGFP intensity was increased by all compounds by 6 dpf (Fig. 2E–G). Quantification of β-cell numbers revealed a small but significant increase in β-cell mass by 7 dpf with Psora4, SR-2640, and Karanjin treatments and no effect with 2,4-DB or GNTI treatments (Fig. 2E–G), further suggesting that Psora4, SR-2640, and Karanjin have effects on β-cell differentiation and/or proliferation.

To understand how these compounds regulate insulin expression, we investigated the effects of these compounds on upstream endocrine transcription factor genes. We used transcriptional eGFP reporters for pax6b, neurod, and mnx1, which encode known regulators of insulin expression and endocrine differentiation, and then analyzed eGFP expression in the principal islet (Fig. 3 and Supplementary Fig. 6). Interestingly, all compounds enhanced mnx1:eGFP expression by 2 days of treatment (Fig. 3F and G) without altering mnx1:eGFP expression after 24 h of treatment (Supplementary Fig. 6C). These data suggest that mechanisms used for enhancement of insulin expression at 12 h of treatment are different from those used for maintenance of insulin expression at the later time points (24 or 48 h). Furthermore, quantitative PCR analysis revealed that SR-2640 and Psora4 enhanced pdx1 expression at 2 days of treatment, but not at 12 h of treatment (Supplementary Fig. 4A and B). Altogether, these data suggest that five compounds (2,4-DB, GNTI, Psora4, SR-2640, and Karanjin) affect the expression of developmental endocrine regulators, as well as that of insulin.

Adults possess glucoregulatory mechanisms that are not present during development. To determine whether our prioritized insulin activators induced hypoglycemia in adult zebrafish, we injected adult animals with each compound daily and measured blood glucose levels after the third injection (Fig. 4A). Encouragingly, none of the compounds caused hypoglycemia under basal conditions (Fig. 4B); and in an IP glucose tolerance test, four of the compounds (GNTI, Psora4, SR-2640, and Karanjin) improved blood glucose clearance (Fig. 4B). Furthermore, and consistent with the effects on larval insulin expression, all of the compounds increased blood Insulin levels after glucose challenge (Fig. 4C and D), indicating that the drug treatments potentiated Insulin secretion from β-cells.

To determine whether the insulin-stimulating effects were conserved in a mammalian model, we measured GSIS from isolated mouse islets that were treated for 24 h with each compound (Fig. 5A). Only Psora4 potentiated GSIS in mouse islets (Fig. 5A). To establish whether the effects of Psora4 on insulin expression and β-cell function are mediated by on-target Kv1.3 inhibition, we tested two other Kv1.3 inhibitors, PAP-1 and clofazimine (Supplementary Fig. 7). Both PAP-1 and clofazimine stimulated insulin expression upon 12 h of treatment (Supplementary Fig. 7A and B), although both PAP-1 and clofazimine exhibited higher toxicity than Psora4 (Supplementary Figs. 7A and 8A–C). Furthermore, both PAP-1 and clofazimine lowered blood glucose levels (Supplementary Fig. 7F) and enhanced blood insulin secretion (Supplementary Fig. 7G and H) in adult zebrafish. Additionally PAP-1 increased β-cell numbers and ins:eGFP expression in larval zebrafish (Supplementary Fig. 7C and D). We conclude that the metabolic effects of Psora4 are likely through on-target Kv1.3 inhibition. Next, to confirm that the effects of Kv1.3 inhibition were glucose dependent, we sequentially incubated Psora4, PAP-1, or vehicle-treated mouse islets in low (3 mmol/L), high (15 mmol/L), and low (3 mmol/L) glucose media, respectively. Psora4- and PAP-1–treated islets exhibited enhanced glucose-dependent insulin secretion that returned to baseline (Fig. 5B and C). Importantly, insulin secretion returned to basal levels when the islets were re-exposed to 3 mmol/L glucose, suggesting that β-cell repolarization was not affected by Kv1.3 inhibition. Psora4 and PAP-1 had modest effects on Ins1 and Ins2 expression (Supplementary Fig. 9), suggesting that in mouse, these compounds stimulate insulin secretion mostly through a posttranscriptional mechanism.

We administered Psora4 to the STZ-induced mouse diabetes model to determine whether Psora4 can improve glucose metabolism in vivo. Following confirmation of STZ-induced hyperglycemia (>350 mg/dL), Psora4 was delivered by IP injection, and blood glucose levels were monitored daily after 6 h of fasting. Compared with vehicle-injected mice, blood glucose levels in Psora4-injected mice were markedly reduced by 1 day after the first Psora4 injection and remained lower with every-other-day treatments (Fig. 6A and B). A glucose tolerance test at 5 days after the first Psora4 injection showed improved glucose tolerance in the Psora4-treated animals (Fig. 6C and D), accompanied by increased insulin secretion (Fig. 6E). Finally, we performed an insulin challenge after 7 days of Psora4 treatment. Blood glucose levels were modestly, yet significantly, lower in Psora4-treated animals (Fig. 6F and G). Together, our data indicate that Psora4 exhibits conserved effects on glucose metabolism in vivo, mainly by increasing insulin secretion.

There are no previous reports on Kv1.3 expression in pancreatic β-cells, although Kv1.3 has been detected in islets by RT-PCR (26). Therefore, we investigated Kv1.3 expression in mouse (Fig. 7A) and zebrafish (Fig. 7B) islets using an antibody, which recognized a COOH-terminal region of mammalian Kv1.3 (Supplementary Fig. 10) and has been used on rat brain sections (21) as well as on Western blots of mouse samples (20). Surprisingly, Kv1.3 immunostaining was mostly localized in the cytoplasm, but not at the cell membrane, of mouse pancreatic β-cells (Fig. 7A). In these cells, Kv1.3 immunostaining was observed in a subset of insulin granules (Fig. 7A). Using this antibody, we also observed immunostaining in the cytoplasm and nucleus of zebrafish β-cells at 5 dpf (Fig. 7B). Altogether, these results suggest a putative novel role for cytoplasmic Kv1.3 in GSIS.

In this study, we screened 4,640 small molecules using zebrafish and identified 91 compounds that affected insulin expression and glucose homeostasis. We found that a voltage-gated potassium (Kv) channel 1.3 inhibitor, Psora4, stimulated insulin secretion and reduced blood glucose levels in zebrafish and mice. Furthermore, our results suggest that Psora4 might act as a direct regulator of insulin expression, because Psora4 induced expression of insulin within 12 h without upregulating other endocrine differentiation genes. Although Kv1.3 expression was previously detected in rat islets (26), it has been unclear how Kv1.3 functions in pancreatic β-cells. In contrast, other members of the Kv channel subfamily are expressed in pancreatic β-cells, and their functions in β-cells have been reported (26–33). During GSIS, high glucose causes the closure of K_ATP channels, thereby causing β-cell depolarization that leads to the activation of voltage-dependent Ca²⁺ channels and then stimulation of insulin secretion. It has been proposed that Kv channels control membrane repolarization and thereby suppress prolonged insulin secretion (27). This model assumes that Kv channels are localized to the plasma membrane. However, Kv1.3 appears to be localized mostly to the cytoplasm of zebrafish and mouse β-cells, and, to the best of our knowledge, it has not been clearly determined where other Kv channels are localized in pancreatic β-cells. However, additional reagents, including Kv1.3-null mutant and GFP-tagged Kv1.3, will be required to determine the precise localization of Kv1.3 and related channels, to further explore this model. It is also unclear how Kv channels are trafficked, how they function, and whether Kv channels have cytoplasmic roles in other tissues. A deeper understanding of cytoplasmic Kv channel regulation may reveal new regulatory mechanisms for insulin secretion in pancreatic β-cells. Kv1.3 blockers have been shown to improve peripheral insulin sensitivity (34,35), enhance metabolic rate and weight loss (36), and decrease the incidence of autoimmune diabetes in a type 1 diabetic rat model by resetting T-cell memory (37). We propose that the reduced blood glucose levels in STZ-treated mice injected with Psora4 are due to the combined effects of Kv1.3 inhibition on insulin production and peripheral insulin sensitivity. Therefore, the Kv family, including Kv1.3, are promising targets for antidiabetic drug discovery, and Psora4 is an interesting candidate to investigate a new class of antidiabetic drug.

Our investigation of four additional compounds—2,4-DB, GNTI, SR-2640, and Karanjin—showed that they stimulated insulin expression, reduced glucose levels in larval zebrafish, and increased insulin secretion in adult zebrafish. Furthermore, GNTI, SR-2640, and Karanjin reduced blood glucose levels in adult zebrafish, and SR-2640 and Karanjin increased β-cell mass in larval zebrafish. However, none of these compounds could stimulate insulin secretion in mouse islets. These results suggest the possibility that these compounds do not act on β-cells directly or that the mechanism of action for these drugs differs between species. However, it is also possible that differences in results arise from the in vitro and in vivo nature of the experiments. For example, SR-2640 and GNTI inhibit the function of receptors by competing with their ligands. Therefore, these compounds cannot inhibit the functions of these receptors under conditions without ligands, as is possibly the case with in vitro mouse islet culture experiments. Furthermore, it has been reported that the action of thyroid hormone is different between isolated islets and in vivo islets (38). Thus, isolated islets might not perfectly mimic in vivo islet behavior. In addition, metabolites of these compounds, but not the compounds themselves, might be the active agents. As another example, it has been reported that Karanjin can reduce blood glucose levels in STZ-treated and db/db mice (39), although it is unclear whether it acts directly on pancreatic β-cells. Thus, it will be interesting to investigate how these drugs affect mammalian β-cells in vivo and consequently organismal glucose metabolism.

In our study, many interesting compounds were identified. For example, levothyroxine (thyroid hormone receptor agonist), harmine (DYRK1A inhibitor), and Bay K8644 (calcium channel activator), three known stimulators of Mafa expression and/or β-cell maturation (38,40–43), turned up in the list of 84 insulin stimulators. These 84 compounds are likely to include new drugs that stimulate β-cell maturation as well as β-cell function. In addition, 4 out of 84 compounds caused a decrease in pck1 activity. These compounds hold high potential in antidiabetic therapy, because they not only stimulate insulin production, but also reduce gluconeogenesis. In addition, we reported that 6 other compounds reduced glucose levels and that 43 compounds, including 7 insulin repressors, elevated glucose levels (Fig. 1D). Investigating downstream mechanisms of these molecules in detail should provide new insights into glucose metabolism. Furthermore, another 127 compounds from the screen modulated insulin promoter activity without changing organismal glucose levels. Therefore, it will be interesting to investigate the effects of these compounds on β-cell differentiation and proliferation, as they could aid in vitro differentiation protocols. Thus, our screen has identified several classes of drugs of interest to the fields of β-cell biology, glucose homeostasis, and antidiabetic therapies.

Acknowledgments. The authors thank Viola Graef (Max Planck Institute for Heart and Lung Research) for critical reading of the manuscript, Sabine Fischer (Max Planck Institute for Heart and Lung Research) for help with the animal protocols, all members of the Stainier laboratory for helpful discussion, and the fish facility staff at the Max Planck Institute for Heart and Lung Research for fish care.

Funding. This work was supported in part by funds from the Max Planck Society and the EU (HumEn) to D.Y.R.S.

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

Author Contributions. H.Mat., S.T.M., Y.H.C.Y., and H.Mas. generated data. H.Mat., S.T.M., Y.H.C.Y., H.Mas., D.H., and D.Y.R.S. contributed to discussion of data and editing and writing the manuscript. H.Mat. and D.Y.R.S. conceived the project. H.Mat. and D.Y.R.S. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.