Inhibition of TBK1/IKKε Promotes Regeneration of Pancreatic β-cells

β-cell proliferation induction is a promising therapeutic strategy to restore β-cell mass. By screening small molecules in a transgenic zebrafish model of type 1 diabetes, we identified inhibitors of non-canonical IκB kinases (IKKs), TANK-binding kinase 1 (TBK1) and IκB kinase ε (IKKε), as enhancers of β-cell regeneration. The most potent β-cell regeneration enhancer was a cinnamic acid derivative (E)-3-(3-phenylbenzo[c]isoxazol-5-yl)acrylic acid (PIAA), which, acting through the cAMP-dependent protein kinase A (PKA), stimulated β-cell-specific proliferation by increasing cyclic AMP (cAMP) levels and mechanistic target of rapamycin (mTOR) activity. A combination of PIAA and cilostamide, an inhibitor of β-cell-enriched cAMP hydrolyzing enzyme phosphodiesterase (PDE) 3, enhanced β-cell proliferation, whereas overexpression of PDE3 blunted the mitogenic effect of PIAA in zebrafish. PIAA augmented proliferation of INS-1β-cells and β-cells in mammalian islets including human islets with elevation in cAMP levels and insulin secretion. PIAA improved glycemic control in streptozotocin (STZ)-induced diabetic mice with increases in β-cell proliferation, β-cell area, and insulin content in the pancreas. Collectively, these data reveal an evolutionarily conserved and critical role of TBK1/IKKε suppression in expanding functional β-cell mass.

Non-canonical IκB kinases (IKKs), TANK-binding kinase 1 (TBK1) and IKKε, have high sequence homology with comparable phosphorylation profiling of substrate(s) 20 . These kinases regulate inflammatory reactions primarily through their action on the interferon regulatory factor (IRF) pathway 21,22 . Independent of their role in acute immune responses, TBK1 and IKKε were shown to be induced in response to obesity-dependent inflammation and directly phosphorylate phosphodiesterase (PDE) 3B 23 , a major cyclic AMP (cAMP) hydrolyzing PDE isoform in adipocytes 24 . Consequently, pharmacological inhibition of TBK1/IKKε with amlexanox, a small molecule inhibitor of these kinases, increased cAMP levels in adipocytes 23 . This led to the secretion of interleukin-6 (IL-6) and the activation of the hepatic Signal Transducer and Activator of Transcription 3 (STAT3) 25 , resulting in weight loss and reduced hepatic gluconeogenesis in obese mice 26 . In addition, IKKε was shown to be among putative targets of diarylamide WS6, a small molecule that promoted human β-cell proliferation in vitro 9,11 . Despite

Chemical screens identify TBK1/IKKε inhibitors as enhancers of β-cell regeneration in zebraf-
ish. To identify bioactive compounds that facilitate pancreatic β-cell regeneration, we screened a library of 75 small molecules with well-characterized biological and pharmaceutical activity in a transgenic zebrafish model of type 1 diabetes. We used the Tg(ins:CFP-NTR) s892 line, in which β-cells are eradicated by nitroreductase (NTR), an enzyme that converts the chemical metronidazole (MTZ) to a DNA interstrand cross-linking agent 47,48 . To easily follow the ablation and regeneration of β-cells, we used an additional transgenic line, Tg(ins: Kaede) jh6 , which expresses the brightly green fluorescent protein Kaede in β-cells 48 . [Tg(ins:CFP-NTR) s892 ; Tg(ins: Kaede) jh6 ] larvae were treated with MTZ at 3 days post-fertilization (dpf) for 24 hours to induce β-cell apoptosis, followed by washing out of MTZ (at 4 dpf, defined as 0 hours post-ablation (hpa)) and subsequent recovery in the presence or absence of chemical compounds for 48 hours (4-6 dpf, corresponding to 0-48 hpa). Using this system, we identified that the compound BX795 can approximately double the number of regenerated β-cells at 48 hpa (Fig. 1B, I). BX795 is a small molecule inhibitor that represses both 3-phosphoinositide-dependent kinase 1 (PDPK1) and non-canonical IKKs, TBK1 and IKKε 49 (Fig. 1C). To determine which pathway's suppression is primarily responsible for augmenting β-cell regeneration, we tested the potent PDPK1 inhibitor BX912 49 . BX912 caused minimal increase in β-cell regeneration (Fig. 1D, I).
To elucidate the interactions between TBK1/IKKε and inhibitors, we performed molecular docking simulations using the well-characterized TBK1 crystal structure in complex with BX795 54 (PDB entry 4EUT). We observed that amlexanox and PIAA adapt docking poses that closely mimic the crystallographically obtained structure of BX795 by forming multiple hydrogen bonds (H-bonds) with key residues within the kinase domain of TBK1 (Fig. 2E-G). The core aromatic moieties, the carboxylate of amlexanox and PIAA, and the thiophene amide of BX795 all bound within an unvaried region of the kinase domain ( Fig. 2E-G), while BX795, being a longer molecule, used its urea moiety to extend toward the outer rim of the kinase domain through interaction with the carbonyl group of Pro-90 (Fig. 2F, G). To corroborate the docking simulations, we further performed structure-activity relationship (SAR) studies of PIAA. We synthesized four analogs: PIAA-1 (an analog lacking the carboxylate group), PIAA-2 (a methyl ester analog of PIAA), PIAA-3 (an analog with an open isoxazolyl ring), and PIAA-4 (an amide analogue of PIAA) (Figs 2H and S1A). The carboxylate group and the intact isoxazolyl ring were both required in enhancing β-cell regeneration, since analogs lacking either group were inactive (Fig. 2I). Thus, the SAR data and modeling of TBK1 and inhibitor interactions delineate the molecular basis of the selectivity of the TBK1/IKKε-Is used in in vivo chemical screens. Taken together, these results indicate that suppression of TBK1/IKKε augments β-cell regeneration in the zebrafish model of type 1 diabetes.
Repression of TBK1/IKKε increases β-cell regeneration by primarily promoting their proliferation. To exclude a substantial contribution of pre-existing β-cells to regeneration of β-cells, we converted the fluorescence of the Kaede protein from green to red by exposing the [Tg(ins:CFP-NTR) s892 ; Tg(ins:Kaede) jh6 ] larvae to UV light at 3 dpf immediately after MTZ treatment (Fig. S3A). We found that 48 hpa islets contained only unconverted green-only β-cells both in DMSO-or TBK1/IKKε-I-treated recovering larvae with a greater number of green-only β-cells in TBK1/IKKε-I-treated larvae (Fig. S3B). These results demonstrate that essentially all β-cells were ablated by MTZ treatment and that TBK1/IKKε-Is augment the number of the newly formed β-cells.
A previous study showed that β-cell neogenesis occurs through both α-to-β-cell transdifferentiation and ductal progenitor-to-β-cell conversion during the initial stages of regeneration in zebrafish 59 . Administering TBK1/IKKε-Is, specifically amlexanox and PIAA, for 24 hours (from 4-5 dpf) rather than for 48 hours (from 4-6 dpf), caused minimal enhancement of the number of β-cells in/adjacent to the hepatopancreatic ductal (HPD) system and that of cells co-expressing Insulin and Somatostatin (Fig. 4A-D and data not shown). There was a slight increase in the number of cells that co-express Insulin and Glucagon (Fig. 4E-H). These results indicate that TBK1/IKKε repression does not primarily affect β-cell neogenesis. To further test whether these compounds increase proliferation of newly formed β-cells, we introduced a transitional day between β-cell ablation and TBK1/IKKε-I treatment. This time lag allowed MTZ-induced ablation to conclude and the default neogenesis to begin before the compounds were added. Amlexanox and PIAA both potently increased the number of EdU incorporated β-cells with the transitional day ( Fig. 4I-L). Taken together, these data suggest that TBK1/IKKε suppression primarily enhances β-cell proliferation to promote regeneration of β-cells.

TBK1/IKKε inhibition selectively accelerates proliferation of β-cells. To determine whether TBK1/
IKKε-Is increase proliferation of β-cells specifically or whether they trigger a general increase in cell proliferation, we assessed the replication rate of other pancreatic endocrine cells, specifically Somatostatin-producing δ-cells, Glucagon-producing α-cells, and liver cells. Inhibition of TBK1/IKKε led to minimal increase of EdU incorporation in δ-cells, whereas it enhanced proliferation of β-cells in the regenerating pancreas ( Fig. 5A-D). TBK1/ IKKε-Is also displayed no significant effects on α-cell replication ( Fig. 5E-H). Furthermore, there was no considerable difference between DMSO-and TBK1/IKKε-I-treated larvae on the number of liver cells that incorporated EdU ( Fig. S4A-D). A longer treatment with TBK1/IKKε-Is, for 96 hours after β-cell ablation, showed that TBK1/ IKKε-Is did not lead to an overshoot in β-cell number (Fig. S4E-J and data not shown). These results suggest that suppression of TBK1/IKKε enhances β-cell proliferation during the most dynamic period of β-cell regeneration without inducing a general increase in proliferation of other cell types.
Next, we examined the ability of TBK1/IKKε-Is to restore normoglycemia. Free glucose levels were elevated after β-cell ablation but declined from 24-72 hpa (corresponding to 5-7 dpf) in DMSO-and TBK1/ IKKε-I-treated larvae (Fig. 5I). Importantly, normal levels of free glucose were recovered significantly faster in TBK1/IKKε-I-treated, especially PIAA-treated, larvae than in DMSO-treated larvae (Fig. 5I). Altogether, these data suggest that inhibition of TBK1/IKKε induces selective increase in β-cell number that accelerates restoration of sufficient overall β-cell function.

Repression of TBK1/IKKε enhances β-cell replication via cAMP-PKA-mTOR activation.
To explore the mechanisms of how TBK1/IKKε suppression stimulates β-cell replication, we measured the regeneration efficiency of β-cells simultaneously treated with several β-cell replication-pathway inhibitors 8 in the presence  (Fig. S5I, J and S5N). We observed the same effect when each of LY294002 and MK2206 was tested alone (Fig. S5E, F and S5M). These results implicate that cAMP-PKA and mTOR signaling mediate the β-cell regeneration response to TBK1/IKKε-Is (Fig. S5O) and that PI3K-Akt signaling, cooperatively with TBK1/ IKKε signaling, may play a repressive role in β-cell regeneration (Fig. S5O).
Given the previous studies showing PKA directly phosphorylating mTOR and RAPTOR in white adipose browning via βAR-cAMP-PKA-mTORC1 signaling pathway 42 and TBK1 and IKKε direct phosphorylating and activating PDE3B in vitro 23 , and our replication-pathway inhibitor data ( Fig. S5I-J and S5N), we hypothesized that TBK1/IKKε-Is induce the activity of the cAMP-PKA-mTOR for β-cell regeneration via suppression of the TBK1/IKKε-PDE3 signaling axis (Fig. 6A). To test this hypothesis, we first measured the cellular cAMP levels and determined the phosphorylation status of the S6 kinase 1 (S6K1), a downstream target of the mTOR signaling cascade 43 , in the PIAA-treated regenerating larvae. Treatment with PIAA led to pronounced increases in both cAMP levels and S6K1 protein phosphorylation (Figs 6B, C, S6A and S7). Furthermore, a combination of PIAA and PDE3 inhibitor cilostamide enhanced β-cell regeneration with increases in cAMP levels, number of β-cells that were EdU incorporated, and immunoreactive for phosphorylated ribosomal protein S6 (pRPS6), another downstream target of the mTOR signaling cascade 43 , compared to individual compound-treated larvae (Figs 6E-6G' , 6J-6K, S6B, and S8A-E). These effects were suppressed by rapamycin treatment (Figs 6H-6I' , 6J, K, and data not shown). In a converse experiment, we assessed the effects of ectopic expression of pde3a on mitogenic potential of TBK1/IKKε-Is using a heat-inducible transgene Tg(hsp:pde3a; hsp:GFP) gt4 . Pde3a is the only PDE3 isoform in zebrafish. When pde3a expression was induced during recovery period in the presence of PIAA, the proportion of new β-cells, which were EdU incorporated and pRPS6-positive, was decreased compared to PIAA-only-treated larvae (Fig. S8F-K). These data suggest that suppression of TBK1/IKKε bestows an increase in β-cell number by regulating cAMP and mTOR activity through PDE3 in the zebrafish model of type 1 diabetes (Fig. S8L).

TBK1/IKKε inhibition augments β-cell function and proliferation in mammalian systems.
To determine whether the effects of TBK1/IKKε suppression on β-cells are conserved across species, we first performed glucose-stimulated insulin secretion (GSIS) assay in primary rat and human islets as elevation of cAMP levels has shown to lead to enhanced β-cell replication, survival, and insulin secretion 10,28,29 . PIAA treatment significantly increased glucose stimulation indices in rat and human islets (Fig. 7I, M). Next, we investigated the mitogenic effects of TBK1/IKKε inhibition by analyzing the ability of PIAA to increase β-cell proliferation in INS-1 cells 60 Table 1) and levels of cAMP (Fig. 7D). Furthermore, PIAA augmented phosphorylation of PKA substrate and S6K1, RPS6, and Grb10 proteins, which are established targets of mTOR 61 , as well as ERK1/2, whereas it triggered minimal phosphorylation of Akt (Figs 7E, S9A, B, and S10) in INS-1 and INS-1-derived 832/13 cells 62 . These data suggest that suppression of TBK1/IKKε promotes replication of β-cells via activation of the cAMP-mTOR signaling axis in mammalian systems. PIAA also increased β-cell proliferation in both dispersed and whole rat islets in a dose-dependent manner (Fig. 7F-H, Supplementary Table 2, and data not shown), consistent with its mitogenic effect on β-cell formation even in the absence of injury in zebrafish (Fig. S11A-D). Importantly, treatment of PIAA on primary human β-cells using islets obtained from 3 cadaveric organ donors caused a notable, dose-dependent induction of β-cell proliferation (Fig. 7J-L and Supplementary Table 3).
We further investigated whether PIAA could increase β-cell regeneration in the streptozotocin (STZ)-induced mouse model of type 1 diabetes. PIAA administration started causing a substantial reduction of non-fasting blood glucose levels after 4-5 days of intraperitoneal injection (Fig. 8A). Significant improvement in glucose and insulin tolerance was observed compared with vehicle treatment (Fig. 8B, C). Morphometric analysis of pancreas sections showed that the β-cells, not α-cells, in PIAA-treated mice were more likely to be Ki67 + , indicating that they were proliferating at a higher rate (Fig. 8D-F and Supplementary Table 4). Moreover, β-cell area and insulin content were increased in PIAA-treated compared with vehicle-treated diabetic mice (Fig. 8G, H). There was no difference in the weight of the mice based on treatment, neither at the start nor at the end of the experiments, indicating that the mice were not generally affected by PIAA treatment (data not shown). Taken together, these

Discussion
In this study, we identified TBK1/IKKε-Is as selective enhancers of β-cell regeneration in a transgenic zebrafish model of type 1 diabetes. We further demonstrated that inhibition of TBK1/IKKε promotes amplification of β-cells in mammalian systems including primary rat and human islets as well as STZ-induced diabetic mice. The proliferative effects of TBK1/IKKε-Is are likely to be mediated by the cAMP-PKA-mTOR signaling axis via PDE3, indicating that TBK1/IKKε play a previously unappreciated role in modulating β-cell mass.
Utilizing small molecule inducers of β-cell proliferation is one of the most tangible methods to restore functional β-cell mass. We have pinpointed that inhibition of TBK1/IKKε promotes β-cell proliferation in multiple species including primary human β-cells. Of essential importance is that suppression of TBK1/IKKε using amlexanox and PIAA, which exhibited the highest potency among tested, can increase β-cell proliferation selectively without inducing a general increase in proliferation of other cell types/tissues. Furthermore, a longer treatment of β-cell ablated zebrafish with these small molecules did not lead to over-proliferation of β-cells once normoglycemia was approached. Since oncogenicity can arise as a result of modulating mitogenic or regenerative pathways 4 , specific TBK1/IKKε-Is' selectivity to β-cells and ability to increase β-cell proliferation primarily during the most active period of β-cell regeneration present valid strategies for expanding β-cell mass. Furthermore, PIAA exhibited substantial efficiency in β-cell regeneration with minimum toxicity. Biologically less characterized azabenzimidazole (AZ) derivatives 5c and 5e have IC50 of 0.032 μM and 0.102 μM (AZ-5c) as well as 0.038 μM and 0.204 μM (AZ-5e) against TBK1 and IKKε, respectively 51 , compared to that of 0.4 μM and 1.07 μM (PIAA). Another AZ derivative AZ13102909 has an IC50 of 0.005 μM against TBK1, promoting apoptosis in melanoma cells 63 . It is noteworthy that AZ-5c and AZ-5e demonstrated significant toxicity when testing in the β-cell ablated zebrafish at nM range with less efficiency of β-cell regeneration than PIAA (data not shown). Diarylamide WS6, which was previously suggested to target IKKε with increasing β-cell proliferation potency in primary human islet culture 9,11 , increased β-cell regeneration much less efficiently than PIAA in the zebrafish model of type 1 diabetes (data not shown). Our results support a separate study showing that WS6 had modest effects on human β-cell proliferation 12 . Therefore, further design and validation of new molecular structures with potent TBK1 and/ or IKKε inhibition activities and minimal toxicity using the PIAA as a scaffold will allow us to identify legitimate strategies for developing human β-cell-specific proliferogens.
Previous studies have suggested that modulation of cAMP levels via GPCR in β-cells is essential for β-cell replication, survival, and insulin secretion 10,28,29 . Our results provide compelling evidence that inhibition of TBK1/IKKε enhances selective β-cell proliferation by increasing cAMP levels via PDE3. Time-course analysis in zebrafish treating PDE3 inhibitor cilostamide, PIAA, and a combination of PIAA and cilostamide demonstrated the correlation between the levels of cAMP and expansion of β-cell mass: the most dynamic period of  β-cell regeneration corresponds to high levels of cAMP, whereas the period reaching to normoglycemia without over-proliferation of β-cells links to low levels of cAMP. Moreover, PIAA treatment of INS-1 rat β-cells resulted in increased percentage of proliferating Insulin-positive cells (co-expressed Ki-67) with concurrent augmentation of cAMP levels. It is likely that PKA stimulation is responsible for linking PIAA-enhanced cAMP levels and subsequent β-cell proliferation. In INS-1 and INS-1-derived 832/13 cells, PIAA increased phosphorylation of PKA substrate and ERK1/2, which was shown to be driven β-cell proliferation via cAMP-PKA signaling cascade upon GIP/GIPR stimulation 35 . Phosphorylation of mTOR targets including S6K1, RPS6, and Grb10, was also greatly augmented with PIAA treatment. Both mammalian and zebrafish mTOR proteins contain 3 conserved PKA target RR/KXS motifs and RAPTOR proteins have 1 RRXS motif (JX and CHS, unpublished observation) that were directly phosphorylated by PKA upon treatment with βAR agonists in vitro 42 . PIAA did not trigger phosphorylation of Akt T308 and Akt S473 , suggesting that the effect of TBK1/IKKε repression is at least in part through a cAMP-PKA-mTOR signaling, not via a well-established Akt-mTOR signaling. The suggested suppressive effect of PI3K-Akt signaling in β-cell regeneration in our studies is consistent with a report showing the inhibitory role of PI3K signaling in β-cell formation in zebrafish 64 and may act through PDE3. Akt has been demonstrated to phosphorylate PDE3B in vitro 23,65 and PI3K has shown to directly activate Akt1 and stimulate Akt2 in a PDPK1-dependent manner 66 . Consistently, treatment of cilostamide with PI3K and Akt inhibitors led to increase in β-cell regeneration in zebrafish (data not shown). In this regard, PDPK1 triggering activation of both Akt and S6K1 67 may be the reason for minimal increase in β-cell number with PDPK1 inhibition. While α 2 -adrenergic receptor antagonist mirtazapine and several PDE inhibitors including a PDE3 inhibitor cilostamide have displayed their potency to stimulate β-cell replication in a cAMP-dependent manner 10 , enhanced insulin secretion was observed in Pde3b knockout (KO) mice 68 . However, Pde3b KO mice fail to suppress hepatic glucose production and display insulin resistance with a number of cAMP-signal transduction components being altered in Pde3b-deficient livers 68 . Contrarily, genetic deletion of IKKε and pharmacological inhibition of TBK1/IKKε led to improved insulin sensitivity through the inhibition of hepatic glucose production with decrease in PDE3B activity and increase in cAMP levels in adipocytes, not in livers, in obese mice 50,69 . Thus, despite the ameliorated insulin tolerance in PIAA-treated STZ diabetic mice may be a secondary effect of hyperglycemic reversal by augmenting β-cell mass, our findings of modulation of PDE3 activity and cAMP levels via suppression of TBK1/ IKKε is likely to enable us to pinpoint bona fide therapeutic approaches that increase the number of functionally adequate β-cells with direct or indirect improvement of insulin sensitivity.
In progression of T1 and T2DM, decreasing β-cell mass by cytokine-and/or glucolipotoxicity-induced apoptosis is a common feature. Thus, prevention of β-cell loss can be an alternative approach for increasing β-cell mass in diabetes. Considering our studies demonstrating that repression of TBK1/IKKε increases cAMP levels and the previous studies showing that anti-apoptotic gene bcl2 is induced by the cAMP-PKA-cAMP response element binding protein (CREB) signaling axis 38 , it is plausible to speculate that suppression of TBK1/IKKε can preserve β-cell mass. Intriguingly, adipose-specific genetic ablation of TBK1 attenuates diet-induced obesity with exaggeration in glucose intolerance/insulin resistance, while genetic deletion of IKKε increases energy expenditure with improvement in insulin sensitivity on a high fat diet 70 . Thus, a careful dissection and elucidation of TBK1-and/ or IKKε-controlled signaling networks will shed light on modulating β-cell survival with concomitant increase in functional β-cell mass, opening up new avenues of therapies for mitigating diabetes.

Experimental Procedures
Zebrafish strains. Adult fish and embryos/larvae were raised and maintained under standard laboratory conditions 71 . We used the following published transgenic lines: Tg(ins:CFP-NTR) s89247 and Tg(ins:Kaede) jh6 48 . Zebrafish studies conducted and protocols used were approved by the Institutional Animal Care and Use Committees of Mayo Clinic and Georgia Institute of Technology, and were in accordance with National Institutes of Health guidelines. β-cell ablation, chemical treatment, and photoconversion. To ablate β-cells, [Tg(ins:CFP-NTR) s892 ; Tg(ins:Kaede) jh6 ] larvae were treated with freshly prepared 5 mM metronidazole (MTZ) (Sigma) from 3 dpf to 4 dpf in the dark, followed by washing out the MTZ, and subsequent 48 hours recovery in the presence of DMSO, individual, or combination of the chemical compounds. The names and the concentrations of the chemical compounds are described in the Supplemental Experimental Procedures. To demonstrate that all β-cells are ablated by MTZ treatment, Tg(ins:Kaede) jh6 -expressing β-cells were converted from green to red by exposing them to UV light at 3 dpf immediately after MTZ treatment.

Chemical screening.
To perform a chemical screen for compounds enhancing β-cell regeneration, we tested 75 compounds from the Stem Cell Signaling Compound Library (Selleckchem). Further details are described in the Supplemental Experimental Procedures.

SCIEnTIfIC
Mammalian ex vivo islet culture and mice experiments. Male Lewis rat pancreatic donors were purchased from Charles River (Wilmington, MA).
Human islets from healthy donors were purchased from Prodo Laboratories (Irvine, CA). 6-8-wk old C57BL/6 male mice (Jackson Laboratory) were used. Studies conducted and protocols used were approved by the Institutional Animal Care and Use Committee of Mayo Clinic and Georgia Institute of Technology and were in accordance with National Institutes of Health guidelines. See Supplemental Experimental Procedures for additional information.