Salt-Inducible Kinase 1 Terminates cAMP Signaling by an Evolutionarily Conserved Negative-Feedback Loop in β-Cells

Glucose homeostasis is primarily regulated by the action of anabolic hormone insulin that is secreted from pancreatic β-cells. Impaired insulin secretion due to the β-cell dysfunction leads to the severe health problems in patients with diabetes that are associated with hyperglycemia-initiated symptoms including ketoacidosis, cardiovascular diseases, and kidney failure (1). Factors that regulate insulin secretion from β-cells include various hormonal and nutritional components including glucose itself, intracellular concentration of Ca²⁺, ATP, diacylglycerol, inositol 1,4,5-trisphosphate, and cAMP (2). cAMP signaling is shown to be critical in enhancing glucose-stimulated insulin secretion (GSIS) in β-cells. Incretins such as glucagon-like peptide 1 or gastric inhibitory peptide could potentiate this process by activation of G protein–coupled receptor–initiated increase in cAMP signaling in β-cells (3,4). In addition, glucose metabolism itself could control cAMP signaling, suggesting that incretin-independent activation of cAMP signaling by glucose could be also critical in insulin secretion and insulin granule exocytosis in pancreatic β-cells (2).

In addition to its role in insulin secretion, cAMP signaling was also shown to be critical in the β-cell function by a transcriptional mechanism. A previous report showed that CREB-dependent transcription is critical in the maintenance of β-cell mass by using RIP A-CREB transgenic mice, a β-cell–specific transgenic mice expressing CREB inhibitor (5). They showed that inhibition of CREB activity in β-cells resulted in the apoptosis primarily due to the lack of expression of insulin receptor substrate 2 and subsequent loss of the insulin signal transduction pathway in this cell type. The role of CREB coactivators such as CREB-binding protein (CBP) and CREB-regulated transcriptional coactivator 2 (CRTC2) were also shown to be critical in the maintenance of β-cell mass, accentuating the importance of a cAMP-dependent transcription pathway in the glucose homoeostasis via controlling the functions of pancreatic β-cells (6–8).

Salt-inducible kinases (SIKs), members of AMP-activated protein kinase (AMPK)-related kinases (AMPKRK), play critical roles in the regulation of cAMP-mediated transcription by enhancing inhibitory phosphorylation of CRTCs and glucose homeostasis (9–11). SIK1 is a founding member of the SIK family of kinases that comprise of SIK1, SIK2, and SIK3. In addition to its role as an inhibitory kinase for CRTCs and subsequent gluconeogenesis in hepatocytes, SIK1 is also shown to be involved in the regulation of myogenesis by direct phosphorylation of histone deacetylase 5 in muscle cells and in the regulation of lipogenesis in the liver by conferring an inhibitory phosphorylation mark on SREBP-1c, showing the importance of SIK1 as a regulator of metabolic pathways in insulin-sensitive tissues (12–14). In particular, as a direct target gene of CREB/CRTCs, SIK1 has been shown to be involved in the negative-feedback loop that would terminate cAMP-dependent transcriptional responses (12,13,15).

Here, we wanted to elucidate potential metabolic pathways that are regulated by SIK1. Insulin secretion from islets of SIK1-deficient mice was enhanced compared with that of wild-type mice, due to the elevation in cAMP concentration. We found that the evolutionarily conserved Ser¹³⁶ of phosphodiesterase (PDE)4D is a phospho-target of SIK1, a critical player in controlling the cAMP pathway. Ectopic expression of catalytically active SIK1 or PDE4D reduced cAMP concentration and insulin secretion from β-cells, and this effect was largely blunted by either the treatment of PDE4 inhibitor or PDE4D knockdown. These data suggest that SIK1 regulates insulin secretion by controlling a cAMP signaling pathway through a modulation of PDE4D in the pancreatic β-cells.

SIK1 knockout mice were generated from SIK1 KO embryonic stem cells (14108A-D8, C57BL/6NTac background; Velocigene, KOMP) by the Mouse Biology Program, University of California, Davis. The mice heterozygous for SIK1 were subsequently backcrossed with C57BL/6N 10 times. SIK2 knockout mice have previously been described (16). Fasting blood glucose levels were measured from mice that were fasted for 16 or 6 h with free access to water. For the glucose tolerance test, mice were fasted for 16 h and injected with glucose (1.5–2 g/kg body wt i.p.). For the insulin tolerance test, mice were fasted for 6 h and injected with insulin (0.75–1 unit/kg body wt i.p.). Blood glucose levels were measured from tail vein blood using an automatic glucose monitor (OneTouch; LifeScan, Inc., Milpitas, CA). All procedures were performed in a specific pathogen-free facility at the Korea University College of Medicine (Seoul, Korea), based on the protocols that were approved by the Korea University Institutional Animal Care and Use Committee.

Hepatic triacylglycerol (TG), plasma TG, and nonesterified fatty acid (NEFA) were measured by colorimetric assay kits (Wako). Plasma insulin was measured by an ELISA kit (Alpco, Salem, NH).

Full-length sequence of PDE4D was PCR amplified from mouse brain and was subcloned into pcDNA3-hemagglutinin (HA). PDE4D mutants (S136A, S141A, and S136/141A) were generated using site-directed mutagenesis. For generation of pU6-PDE4D RNA interference (RNAi), palindromic sequences corresponding to nucleotides 444–464 from mouse PDE4D coding sequence (5′-AACTCCTCGATTGCCAGTGAT-3′) were linked to mouse U6 promoter in the pBluescript KS vector (Stratagene).

Ad-GFP, Ad-US (nonspecific RNAi control), Ad-SIK1, and Ad-SIK1 RNAi adenoviruses were previously described (13). Adenovirus expressing Ad-SIK1 T182A, Ad-PDE4D, Ad-PDE4D mutants, and Ad-PDE4D RNAi was generated by homologous recombination between adenovirus backbone vector pAD-Easy and linearized transfer vector pAD-Track as previously described (17).

INS-1 cells were grown in RPMI-1640 medium (Gibco) supplemented with 10% FBS, 11 mmol/L glucose, 10 mmol/L HEPES, 50 μmol/L β-mercaptoethanol, 100 units/mL penicillin, and 100 μg/mL streptomycin as previously described (18). Islets were isolated by collagenase (sigma) digestion of the wild-type (WT), SIK1^+/−, SIK1^−/−, and SIK2^−/− mouse pancreas. Pancreata were digested at 37°C for 10–15 min, and islets were handpicked on Hanks’ balanced salt solution. Islets were cultured in RPMI-1640 media supplemented with 10% FBS, 11 mmol/L glucose, 100 units/mL penicillin, and 100 μg/mL streptomycin overnight at 37°C prior to experiment for recovery.

For insulin secretion, INS-1 cells or islets were preincubated for 2 h in glucose-free Krebs-Ringer bicarbonate buffer (KRBB) and then switched to KRBB with either 2.8 mmol/L glucose or 16.7 mmol/L glucose for 30 min unless otherwise indicated in the figure legends. Secreted insulin concentration in the media was determined by using an ELISA kit (Alpco, Salem, NH) and was normalized to insulin contents.

Dynamic insulin secretion in the perifusate was measured via an ultrasensitive enzyme immunoassay (Mercodia, Uppsala, Sweden) using an automated perifusion system. Briefly, a low-pulsatility peristaltic pump pushed Krebs-Ringer Phosphate-HEPES buffer containing 0.2% BSA and 2.8 mmol/L glucose at a perifusion rate of 100 mL/min through a perifusion chamber (Biorep Technologies, Miami, FL) containing 100 pancreatic islets immobilized in Bio-Gel P-4 Gel (Bio-Rad, Hercules, CA). After a 30-min stabilization period, the groups of islets were successively stimulated with 16.7 mmol/L glucose. The perifusate was collected in an automatic fraction collector designed for a 96-well plate format. The perifusion chamber containing the islets were kept at 37°C, and the perifusate in the collecting plate was kept at <4°C. The perifusate solutions were gassed with 95% O₂ and 5% CO₂ and maintained at 37°C. Perifusates were collected every minute.

Exocytosis was detected as changes in cell capacitance as described with a modification (19). Briefly, secretion was elicited by trains of nine 500-ms depolarizations (1-Hz stimulation frequency) from −70 mV to 0 mV. Patch electrodes were made from borosilicate glass capillaries (Clark Electromedical Instruments, Pangbourne, U.K.) using a puller (PP-830; Narishige, Tokyo, Japan). The pipette resistance ranged between 2 and 4 MΩ when the pipettes were filled with the intracellular solutions specified below. The zero-current potential of the pipette was adjusted in the bath before giga-seal formation. We used a perforated whole-cell patch-clamp technique, in which the cells remain metabolically intact. All measurements were conducted using an EPC-10 patch-clamp amplifier (HEKA Elektronik, Lambrecht/Pfalz, Germany) and Patchmaster software. Normal Tyrode was used as the bathing solution and contained 123 mmol/L NaCl, 5.4 mmol/L KCl, 0.33 mmol/L NaH₂PO₄, 1.8 mmol/L CaCl₂, 0.5 mmol/L MgCl₂, 5 mmol/L HEPES, 5 mmol/L glucose, and 20 mmol/L tetraethylammonium-Cl adjusted to pH 7.4 with NaOH. The pipette solution contained 140 mmol/L CsCl, 1 mmol/L MgCl₂, 10 mmol/L HEPES, 5 mmol/L EGTA, and 200 μg/mL nystatin. The pH was adjusted to 7.2 with CsOH. All experiments were conducted at 35 ± 1°C.

For the measurement of cAMP concentration in INS-1 cells, cells were seeded in sixwell plates (2 × 10⁶ cells) and were treated with 10–30 μmol/L forskolin for 16 h unless otherwise noted in the figure legends. For the measurement of cAMP concentration in primary islets, after 36–48 h recovery, 35–50 islets were incubated for 1–4 h in KRBB with 2.5 mmol/L glucose as stated in the figure legends. Cells were lysed and clarified by centrifugation, and concentration of cAMP was detected by Chemiluminescent Immunoassay system (Applied Biosystems) according to the manufacturer’s instructions. Concentrations of cAMP were normalized to protein content.

Western blot analyses of whole-cell extracts were performed as previously described (20). Antibodies for AKT, phospho-Ser⁴⁷³ AKT, CREB, and phospho-AMPK substrate antibody were from Cell Signaling Technology; antibodies for HA-horseradish peroxidase and flag-M2 were from Sigma-Aldrich; antibody for CRTC2 was from Calbiochem; antibody for phospho-Ser¹³³ CREB was from Rockland; and antibody for PDE4D was from Proteintech. Antibodies to heat shock protein-90 (HSP90; Santa Cruz) or α-tubulin (Sigma-Aldrich) were used as loading controls.

Total RNA from either cultured cells or tissues was extracted using RNeasy mini-kit (Qiagen). cDNA generated by Superscript II enzyme (Invitrogen) was analyzed by quantitative PCR (Q-PCR) using an SYBR green PCR kit and CFX Connect Real-Time System (Bio-Rad). All data were normalized to ribosomal L32 expression.

Induction and purification of glutathione S-transferase (GST) fusion proteins in Escherichia coli were performed according to the manufacturer’s protocol (Amersham Pharmacia Biotech). SIK1 protein was expressed by transfection in 293T cells and immunoprecipitated with anti-Flag agarose (Sigma-Aldrich) and then eluted using 3X Flag peptide. The kinase assay was performed using GST-fused substrates with the eluted SIK1 in the presence of 2.5 μCi ATP for 1 h at 30°C. Results were assessed by autoradiography (GE Healthcare) (21).

For the analysis of phosphorylated Ser/Thr residues on PDE4D, pcDNA3-Flag-SIK1 and pcDNA3-HA-PDE4D constructs were transiently transfected into 293T cells for 48 h. HA-PDE4D was isolated by immunoprecipitation with anti-HA agarose and analyzed by SDS-PAGE gel. The protein band corresponding to PDE4D was excised and subjected to in-gel digestion with trypsin. Nano liquid chromatography–mass/mass analysis was performed using an LTQ ABI Q-STAR mass spectrometer (Agilent Technologies). The modified peptides were analyzed by using the mascot algorithm (Matrix Science). The experiment was performed in Yonsei Proteome Research Center, Seoul, Korea.

The motif search program at the ScanProsite webpage (http://prosite.expasy.org/scanprosite) was utilized to identify proteins that contain the consensus motif for SIK1 substrate ([LIV]-[R/K]-X-[S/T]-X-X-X-[L/I/V]) in the SwissPort database.

Results are represented as either the mean ± SEM (for metabolites) or the mean ± SD (for Q-PCR, insulin, and cAMP assays). The comparison of two groups was carried out by using a two-tailed unpaired Student t test. The comparison of multiple groups was carried out by using an ANOVA with a Dunnett C post hoc test. In all statistical comparisons, P values <0.05 were considered significant and reported as in the figure legends.

To delineate the metabolic function of SIK1 in vivo, we utilized a genetic mouse model for SIK1 deficiency. Mice homozygous for SIK1-null allele did not display embryonic lethality but were born at less than the Mendelian ratio with a much smaller body weight (Fig. 1A and Supplementary Fig. 1A). Compared with the WT control, SIK1 homozygous mice (SIK1^−/−) showed improved glucose tolerance with a lower fasting glucose level (Supplementary Fig. 1B and C). Since mice heterozygous for SIK1 were born in a normal Mendelian ratio and did not display any apparent growth phenotype, we utilized SIK1 heterozygous mice (SIK1^+/−) for most of our study. Interestingly, SIK1^+/− mice also showed lower blood glucose levels than their WT littermate both under normal diet and under high-fat-diet conditions without changes in body weight or plasma TG and NEFA levels (Fig. 1B and Supplementary Fig. 2A and B). In addition, we observed an improved glucose tolerance in SIK1^+/− mice compared with their control mice (Fig. 1C and Supplementary Figs. 1C and 2C). Interestingly, plasma insulin levels were elevated in SIK1^+/− mice during the glucose tolerance test, suggesting indeed that depletion of SIK1 could promote increased insulin secretion from β-cells in vivo (Fig. 1D and Supplementary Fig. 2D). Subsequent insulin tolerance test revealed no difference between mice with two genotypes, showing that insulin response itself was not disturbed upon haploinsufficiency of SIK1 (Supplementary Fig. 2E and F). In line with the latter finding, we did not observe any changes in insulin signaling in insulin-sensitive tissues, as evidenced by the lack of changes in Ser⁴⁷³ phosphorylation level of Akt in the liver, muscle, or white adipose tissue (Supplementary Fig. 3A and B).

To corroborate the hypothesis that SIK1 deficiency may promote increased insulin secretion from β-cells, we utilized primary islets in culture that were prepared from either WT or SIK1^+/− mice. Indeed, we observed that the insulin secretion from SIK1^+/− islets was higher than those from WT islets without changes in total insulin contents (Fig. 2A). Perifusion analysis also showed increased GSIS in SIK1^+/− islets compared with the control both in the 1st phase and in the 2nd phase of GSIS (Fig. 2B). In addition, we also observed that the cAMP levels from SIK1^+/− islets were higher than those from WT islets (Fig. 2C). cAMP signaling was previously shown to be involved in the exocytosis of insulin granules. We presumed that the increased cAMP concentration in SIK1^+/− islets could also lead to the enhanced exocytosis of insulin granules from these cells. Thus, we measured the membrane capacitance of single β-cells from either WT or SIK1^+/− mice and found that SIK1^+/− β-cells displayed significantly elevated secretory responses compared with that of the control (Fig. 2D). These data collectively suggest that increased cAMP concentration upon SIK1 depletion in the β-cells could lead to the increased insulin secretion via the enhanced exocytosis of insulin granules.

Based on the presence of broader substrates for AMPK, a close paralogue of SIK kinases, we speculated that SIK1 could also have still unidentified substrates that might play roles in various signaling cascades including cAMP signaling. To this end, we used amino acids 164–176 from mouse CRTC2 to determine the consensus sequence for SIK1 substrates. We randomly substituted each amino acid except the Ser¹⁷¹ residue, a site that is phosphorylated by SIK1, and performed in vitro kinase assay. The assay revealed that SIK1-dependent phosphorylation requires the presence of hydrophobic residues (Leu, Ile, and Val) in the −5 and +4 positions relative to the phospho-acceptor Ser residue (Fig. 3A). In addition, the presence of basic residues (Arg, Lys) is preferred in the −3 position, in line with the original observation (22). We thus defined [LIV]-X-[RK]-X-X-[ST]-X-X-X-[LIV] as a consensus site of SIK1 and screened for the additional putative substrates. The motif search program identified candidate proteins that matched with the consensus sequences, and 56 candidates were subsequently selected that could be potentially relevant to the function of SIK1 (Supplementary Table 1). For confirmation of whether the selected proteins could serve as substrates for SIK1, GST-tagged peptides containing 13 amino acids from the candidates were generated and were subsequently tested by in vitro kinase assay. Among 56 candidates, 10 peptides were shown to be phosphorylated by SIK1 in vitro (Fig. 3B), including proteins involved in the energy metabolism (p300 and AAKG3), cardiac muscle differentiation and maturation (SRF), cell-cycle checkpoint (MLTK), or other cellular signaling cascades (PDE4D, TAB2, RAF1, JKIP, and BRAF), which were further confirmed through in vitro kinase assay by using truncated proteins (data not shown). Among the candidates, p300 was shown to be phosphorylated by SIK2, a close paralogue of SIK1, in the liver, corroborating the usefulness of our approach (23).

We were particularly intrigued by the identification of PDE4D, a cAMP-specific phosphodiesterase among the candidates, since cAMP was known to be a negative regulator for SIKs through protein kinase A (PKA) in a number of tissues. Consensus sequence surrounding Ser¹³⁶ of PDE4D is conserved in other PDE4 isoforms (PDE4A–D) but not in other PDE isoforms (Fig. 3C, top). Furthermore, we found that the predicted SIK consensus sites in PDE4D are conserved in Coelomata including Drosophila melanogaster, which suggests that the circuitous regulation of cAMP and SIK could be conserved across eukaryotic organisms (Fig. 3C, bottom). Mass spectrometric analysis revealed that Ser¹³⁶ of PDE4D is indeed highly phosphorylated in the cell. In addition, an adjacent site (Ser¹⁴¹, also conserved in all PDE4 isoforms) of PDE4D was also shown to be significantly phosphorylated, although the nature of the candidate kinase for this site was not ostensible (Fig. 3D). Mutational analysis revealed that Ser to Ala mutation on Ser¹³⁶ in PDE4D completely ablated its SIK1-dependent phosphorylation, and Ser¹⁴¹ to Ala mutant PDE4D also affected SIK1-dependent phosphorylation levels to a lesser degree (Fig. 3E). Double point mutations on both Ser residues to Ala also completely abolished SIK1-dependent phosphorylation on PDE4D. Together with the mass spectrometry analysis, these data suggest that Ser¹⁴¹ either might be phosphorylated by SIK1 or could be required for the SIK1-mediated phosphorylation of PDE4D at the adjacent Ser¹³⁶. Indeed, these two sites were shown to be phosphorylated in a previous study, corroborating the significance of our finding (24).

Having seen the potential regulation of PDE4D by SIK1, we wanted to explore whether SIK1 could directly regulate cAMP signaling by controlling cAMP concentration in β-cells. Expression of SIK1 was highly induced after 30-min treatment of forskolin (>30-fold), picked at about 1 h, and stayed elevated up to 4 h in INS-1 β-cells (Fig. 4A, top). mRNA levels of SIK2 were reduced in response to forskolin treatment, while SIK3 mRNA did not fluctuate during the course of forskolin treatment. Similarly, SIK1 protein level was also induced after 1 h treatment of forskolin, peaked at 2 h, and remained higher at 4 h treatment of forskolin in INS-1 β-cell (Fig. 4A, bottom). The induction of SIK1 by forskolin treatment was also shown in primary islets (Fig. 4B). The induction of SIK1 expression by cAMP signaling was previously shown in hepatocytes and neuronal cells, suggesting that the similar mechanism is also conserved in β-cells (13,15). These data suggest that SIK1 could potentially be involved in the feedback inhibitory mechanism in reducing cAMP-dependent signaling at the later point after initial burst. In line with this idea, we observed the elevations of cAMP levels 1–2 h after the forskolin treatment, which were then reduced at 4 h forskolin treatment in the control cells (Fig. 4C, left). On the other hand, cAMP levels remained elevated during the course of experiment in SIK1-depleted INS-1 β-cells, showing indeed that SIK1 is involved in the feedback inhibition loop to reduce cAMP signaling at the later point. Depletion of SIK1 reduced phosphorylation of endogenous PDE4D, showing that SIK1 indeed functions as a specific kinase for PDE4D (Fig. 4C, right). Similar trends were also shown in SIK1-depleted primary islets, although the degree of elevation in cAMP was less than the result shown in INS-1 cells (Fig. 4D). In line with the increases in cAMP concentration upon SIK1 knockdown, insulin secretion was also significantly induced in SIK1-knockdown β-cells compared with the control cells (Fig. 4E and Supplementary Fig. 4A).

Conversely, overexpression of WT, but not inactive mutant (T182A), SIK1 in INS-1 β-cells led to the decrease in cAMP concentration and reduced insulin secretion, corroborating the idea that SIK1 could directly affect cAMP concentration through the direct regulation of PDE4D (Fig. 5A and B and Supplementary Fig. 4B). In addition, restoration of SIK1 expression in SIK1^+/− β-cells by SIK1 adenovirus negated the increased GSIS, suggesting that the reduction in SIK1 levels was critical in enhancing the insulin secretion in β-cells (Supplementary Fig. 4C). Recently, SIK2 was shown to enhance insulin secretion in β-cells by phosphorylation and promotion of p35 proteins, leading to the increased calcium influx and the subsequent increase in insulin secretion (25). Indeed, we observed a slight increase in SIK2 mRNA levels in SIK1^+/− pancreas as well as SIK1^+/− islets compared with the WT, potentially suggesting that depletion of SIK1 may induce SIK2 expression, causing the enhanced GSIS in islets (Supplementary Fig. 5A). To verify this hypothesis, we tested the proven pan-SIK inhibitor (HG-9-91-01) in primary islets (26). Unlike the results from the previous report, the treatment of HG-9-91-01 led to the increased GSIS in WT islets. HG-9-91-01 treatment led to a slight increase in GSIS in SIK1^+/− islets, and it did not affect the GSIS in SIK1^−/− islets, suggesting that the inhibition of SIK1 activity was crucial in HG-9-91-01–dependent enhancement of GSIS in islets (Supplementary Fig. 5B). Indeed, the depletion of SIK1 did not affect SIK2 or p35 protein levels, showing that SIK1 did not regulate a SIK2-p35–dependent pathway in primary islets in C57/BL6 background (Supplementary Fig. 5C). To further exclude the possibility that the effect of SIK1 depletion on GSIS in β-cells could be due to the presence of SIK2, we infected adenovirus expressing short hairpin (sh)RNA for SIK2 in SIK1^+/− β-cells. Again, we observed an increase in GSIS from SIK1^+/− β-cells compared with the control. However, we did not observe any changes in GSIS from SIK1^+/− β-cells with up to 50% knockdown of SIK2, showing indeed that SIK2 did not have a significant effect on GSIS in SIK1-depleted β-cells at least under the C57/BL6 background (Supplementary Fig. 5D). SIK2 was as effective as SIK1 in inhibiting insulin secretion in INS-1 β-cells when ectopically expressed (Supplementary Fig. 5E). Finally, we prepared primary islets from either WT mice or SIK2 knockout mice under the C57/BL6 background and measured the GSIS. Unlike the result from the previous report, we observed a slight induction in insulin secretion from SIK2^−/− islets under both basal and high-glucose concentrations (Supplementary Fig. 5F). These data suggest that SIK2 might not be critically involved in insulin secretion from β-cells, at least under the C57/BL6 background.

We found that Ser¹³⁶ of PDE4D is a target of SIK1-mediated phosphorylation and the adjacent Ser¹⁴¹ could also be critical in this process. Since SIK1 overexpression decreased cAMP concentration while depletion of SIK1 increased it, we presumed that SIK1-mediated phosphorylation of PDE4 could enhance its activity toward the degradation of cAMP in β-cells. Thus, we generated adenovirus expressing Ser to Ala mutants of PDE4D in a hope of generating PDE4D proteins with reduced activity. Expression of WT PDE4D indeed reduced cAMP concentration as well as GSIS significantly. A single point mutation of PDE4D (S136A or S141A) did not affect its activity greatly. However, a mutant PDE4D with both Ser residues to Ala (S136A and S141A) displayed a drastically impaired ability in reducing both cAMP concentration and insulin secretion, showing that SIK1-dependent phosphorylation of PDE4D is critical in its ability to regulate cAMP concentration and subsequent insulin secretion in β-cells (Fig. 5C and D and Supplementary Fig. 6A). The phosphorylation status of CREB or CRTC2 were also distinct with PDE4D S136/141A mutant compared with WT, showing that SIK1-mediated control of PDE4D phosphorylation is critical in eliciting its activity toward a cAMP-dependent pathway (Supplementary Fig. 6B).

Finally, we wanted to verify whether SIK1-dependent regulation of intracellular cAMP concentration and insulin secretion from β-cells is mediated through PDE4. To this end, we utilized PDE4-specific inhibitor rolipram in our study. As shown before, adenovirus-mediated expression of SIK1 reduced cAMP levels and insulin secretion in INS-1 β-cells, and this effect was largely blunted by treatment of rolipram, showing indeed that SIK1-dependent regulation of PDE4 activity is critical in the regulation of insulin secretion in β-cells (Fig. 6A and B and Supplementary Fig. 6C). On the other hand, the PDE3B-specific inhibitor milrinone did not affect the SIK1-mediated inhibition of GSIS in INS-1 β-cells, showing that SIK1 did not regulate the insulin secretion in β-cells through a PDE3B-dependent manner (Supplementary Fig. 6D). Knockdown of PDE4D or treatment of PDE4D-specific inhibitor GEBR-7b also significantly blunted the SIK1-dependent repression of cAMP concentration and insulin secretion (Fig. 6C–F and Supplementary Fig. 6C, E, and F), showing that PDE4D could be a major isoform among PDE4 families in β-cells that is under the control of SIK1-dependent pathway. These data underscore the importance of SIK1-mediated control of cAMP signaling in the regulation of insulin secretion in β-cells through phosphorylation of PDE4D.

Identified as members of AMPKRK, the SIK family of kinases was shown to be critical regulators of transcription. Both SIK1 and SIK2 were shown to inhibit CREB-dependent transcription by conferring the inhibitory phosphorylation on CRTCs in the liver, resulting in inhibition of hepatic gluconeogenesis (13,27). In addition, both kinases were shown to control lipid metabolism in the liver; SIK1 inhibits lipid metabolism by phosphorylation and inhibition of SREBP-1c, while SIK2 represses lipogenesis by indirect control of ChREBP through inhibition of p300 (14,23). SIK1 is also shown to control myogenesis by inhibitory phosphorylation of histone deacetylase 5 in the skeletal muscle, while SIK2 might be involved in the adipogenesis via controlling activity of CRTC2 or insulin receptor substrate 1 in the adipocytes (12,16,28). SIK3 is ubiquitously expressed in the mammalian tissues and may participate in the regulation of systemic glucose or lipid metabolism based on the studies utilizing knockout mice for SIK3 (29). These data accentuate the specific roles of SIK family of kinases in the regulation of metabolic processes in various tissues.

We attempted to isolate potential phospho-targets of SIKs based on the in silico approach using the consensus SIK motif using the CRTC2 sequences. Interestingly, we identified ULK1 and BRAF as potential candidates for SIK1-dependent phospho-substrates, which were confirmed by in vitro kinase assay. Recent reports showed that both of these proteins are targets of AMPK, an important regulator of metabolic pathways as well as cellular proliferation (30,31). It will be of interest to investigate whether SIKs and perhaps other members of AMPKRK could also be involved in the regulation of these proteins in the future. On a similar note, we could speculate that other identified targets of SIK1 in the current study might be also regulated by AMPK in certain conditions. Delineation of specific roles of AMPK and its related kinases in the regulation of various cellular signaling pathways warrants further research.

By using an SIK consensus motif, followed by in vitro kinase assay, we identified PDE4D as a substrate for SIK family of kinases. By mass spectrometry, we discovered that the evolutionarily conserved Ser¹³⁶ is a phospho-target of SIK1 and the adjacent Ser¹⁴¹ might also be critical in SIK1-meditated phosphorylation. Further investigation is necessary to clarify the role of Ser¹⁴¹ in the SIK1-dependent phosphorylation of the adjacent Ser¹³⁶. The fact that Ser¹⁴¹ was also conserved among various species also suggests the importance of this amino acid in the proper regulation of PDE4D by SIK1. Ser¹⁴¹ might be targeted by members of map kinases by a sequence prediction program, although the exact identity of a potential kinase should be elucidated by a further investigation.

While we were preparing this manuscript, a report by Sakamaki et al. (25) suggested that SIK2 enhanced insulin secretion from the β-cells by promotion of p35 degradation. To explore the possibility that SIK1 could also control the insulin secretion via a p35-dependent pathway, we measured both SIK2 and p35 protein levels in SIK1-depleted primary islets. As shown in Supplementary Fig. 5C, we did not observe any changes in either SIK2 or p35 protein levels, showing that SIK1 deficiency did not affect a SIK2-p35–dependent pathway. The notion was further confirmed by the fact that knockdown of SIK2 did not affect the GSIS in SIK1^+/− β-cells (Supplementary Fig. 5D). While we observed an increase in the GSIS in SIK2^−/− islets compared with the control, Sakamaki et al. (25) showed a reduction in the GSIS in SIK2^−/− islets (Supplementary Fig. 5F). In addition, we were able to observe an increase in the GSIS by the treatment of pan-SIK inhibitor HG-9-91-01, in primary islets, while Sakamaki et al. reported the opposite result (Supplementary Fig. 5B) (25). Several differences between their system and our system might account for the discrepancy. First, Sakamaki et al. (25) used a pancreas-specific–temporal Cre system to knockout SIK2 (1–2 weeks post–tamoxifen injection), while we used systemic, chronic SIK2 KO models. Although we would expect to observe a similar effect on GSIS upon SIK2 depletion in the isolated islets, we cannot completely exclude the possibility that chronic SIK2 depletion might affect the β-cell properties. We did not observe a significant difference in the insulin content between WT islets and SIK2^−/− islets. Second, we used C57/BL6 background for our SIK1 or SIK2 KO mice (SIK1 mice were generated by using C57/BL6 albino embryonic stem cells and were backcrossed for 10 generations once we obtained the SIK1^+/− mice; we purchased SIK2^+/− mice under the C57/BL6/129SV background and backcrossed them for 10 generations), while Sakamaki et al. used mice under the FVB background (16,25). Finally, we directly obtained HG-9-91-01 from the original investigator who developed and tested the reagent (Dr. Nathanael Gray), unlike Sakamaki et al., who obtained the chemical from the indirect sources. Further study is necessary to determine whether the genetic background or a slight difference in chemicals indeed affected the discrepancy. As least in our system, we could safely conclude that the effect of SIK1 depletion on GSIS is not through the SIK2-p35–dependent pathway.

In this report, we found that haploinsufficiency of SIK1 in mice led to the lower blood glucose levels with improved glucose tolerance when challenged with high-fat diet. Insulin secretion from SIK1^+/− islets was enhanced compared with the WT control, due in part to the elevation in cAMP concentration. Conversely, overexpression of SIK1 reduced cAMP concentration and insulin secretion from the β-cells, and this effect was largely blunted by either the treatment of PDE4 inhibitor or PDE4D knockdown. Expression of WT PDE4D, but not S136/141A mutant PDE4D, reduced cAMP concentration and insulin secretion, showing indeed that SIK1-dependent phosphorylation of PDE4D is critical in the regulation of cAMP-dependent signaling pathway in β-cells. It is noteworthy that cAMP signaling and PKA suppress SIK1 activity by inhibitory phosphorylation at its Ser⁵⁷⁷ residue (32). We postulated that SIK1 serves as a controller of cAMP-dependent insulin secretion of β-cells by enhancing PDE4D activity, thus reducing insulin secretion from β-cells under basal conditions. Upon feeding, increased glucagon-like peptide 1–dependent signaling activates PKA in β-cells, thus augmenting cAMP signaling and resultant insulin secretion. Increases in cAMP signaling in turn activate a CREB/CRTC2-dependent transcriptional program, leading to the induction of SIK1 expression. SIK1 can then phosphorylate and activate PDE4D in β-cells, thus promoting the degradation of cAMP and the termination of the signal as a part of a feedback inhibitory loop (Fig. 7). Lack of SIK1 led to the increased cAMP concentration and subsequent increase in insulin secretion from β-cells both in vitro and in vivo. In line with this idea, we observed a decreased expression of SIK1 in β-cells from high-fat diet or genetic mouse models of obesity and insulin resistance, which can potentially be linked to the compensatory increases in insulin secretion and hyperplasia of β-cells in those mice (Supplementary Fig. 6G). Taken together, we propose that SIK1 is an endogenous regulator of cAMP signaling pathway in the pancreatic β-cells by controlling PDE4D activity and subsequent insulin secretion in this tissue.

Acknowledgments. The authors thank Dr. Nathanael S. Gray (Dana-Farber Cancer Institute) for providing HG-9-91-01 and critical comments and Dr. Hye-Sook Han (Korea University) for critical review of this manuscript.

Funding. This work was supported by the National Research Foundation of Korea (grant nos. NRF-2010-0015098 and NRF-2012M3A9B6055345), funded by the Ministry of Science, ICT & Future Planning, Republic of Korea; a grant of the Korean Health Technology R&D Project (grant A111345 and HI13C1886), Ministry of Health and Welfare, Republic of Korea; and a grant from Korea University. D.-K.S. was supported by the National Research Foundation of Korea (NRF-2014R1A5A2010008), funded by the Ministry of Science, ICT & Future Planning, Republic of Korea. All authors declare no competing financial interests.

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

Author Contributions. M.-J.K., S.-T.K., and S.-H.K. designed research. M.-J.K., S.-K.P., J.-H.L., C.-Y.J., D.J.S., J.-H.P., Y.-S.Y., and J.P. performed research. M.-J.K., S.-K.P., D.J.S., Y.-S.Y., K.-G.P., D.-K.S., H.C., S.-T.K., and S.-H.K. analyzed data. M.-J.K., H.C., S.-T.K., and S.-H.K. wrote the manuscript. S.-T.K. and S.-H.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.