Salt-inducible Kinase 3 Signaling Is Important for the Gluconeogenic Programs in Mouse Hepatocytes*

Background: Salt-inducible kinases (SIKs) are capable of suppressing gluconeogenic gene expression in hepatocytes when they are overexpressed. Results: However, enhanced gluconeogenic programs are observed only in SIK3-defective hepatocytes. Conclusion: SIK3 is the major kinase that down-regulates gluconeogenesis. Significance: The present study proposes that SIK3 could be a new target of diabetic care. Salt-inducible kinases (SIKs), members of the 5′-AMP-activated protein kinase (AMPK) family, are proposed to be important suppressors of gluconeogenic programs in the liver via the phosphorylation-dependent inactivation of the CREB-specific coactivator CRTC2. Although a dramatic phenotype for glucose metabolism has been found in SIK3-KO mice, additional complex phenotypes, dysregulation of bile acids, cholesterol, and fat homeostasis can render it difficult to discuss the hepatic functions of SIK3. The aim of this study was to examine the cell autonomous actions of SIK3 in hepatocytes. To eliminate systemic effects, we prepared primary hepatocytes and screened the small compounds suppressing SIK3 signaling cascades. SIK3-KO primary hepatocytes produced glucose more quickly after treatment with the cAMP agonist forskolin than the WT hepatocytes, which was accompanied by enhanced gluconeogenic gene expression and CRTC2 dephosphorylation. Reporter-based screening identified pterosin B as a SIK3 signaling-specific inhibitor. Pterosin B suppressed SIK3 downstream cascades by up-regulating the phosphorylation levels in the SIK3 C-terminal regulatory domain. When pterosin B promoted glucose production by up-regulating gluconeogenic gene expression in mouse hepatoma AML-12 cells, it decreased the glycogen content and stimulated an association between the glycogen phosphorylase kinase gamma subunit (PHKG2) and SIK3. PHKG2 phosphorylated the peptides with sequences of the C-terminal domain of SIK3. Here we found that the levels of active AMPK were higher both in the SIK3-KO hepatocytes and in pterosin B-treated AML-12 cells than in their controls. These results suggest that SIK3, rather than SIK1, SIK2, or AMPKs, acts as the predominant suppressor in gluconeogenic gene expression in the hepatocytes.

Salt-inducible kinase (SIK) 2 is a member of the 5Ј-AMP-activated protein kinase (AMPK)-related kinase family, and the SIK subfamily is composed of three isoforms (1). We isolated SIK1 from the adrenals of rats fed with a high Na ϩ or K ϩ diet (2) and identified SIK2 by its sequence similarity (3). SIK3 was characterized by another group in a wide range analysis (4). Although multiple SIK substrates have been identified based on information from the SIK phosphorylation motifs, LX(R/K)(S/ T)XpSXXXL (5), only two types of substrates, CRTC (a coactivator for the cAMP response element (CRE)-binding protein (CREB)) and class IIa histone deacetylase (HDAC) (6), have been confirmed by independent research groups. Phosphorylation of these two substrates by SIKs results in a loss of their transcriptional regulatory activities by inducing nuclear export (5,7).
The other important aspect of SIK is as a feedback regulator in PKA signaling (8). When PKA activates CREB by phosphorylation, it also counters SIK actions by phosphorylating the C-terminal regulatory domain of SIK, which in turn promotes the dephosphorylation rate of CRTC, enhances nuclear accumulation of this coactivator, and increases CREB-dependent gene expression (5,7,9). Conversely, when PKA activity wanes, the reactivated SIK terminates CREB activity by inactivating CRTC (5,9). This is also the case for class IIa-HDAC (6). PKAdependent inactivation SIK facilitates the activation of class IIa-HDAC. Because class IIa-HDAC is the major suppressor of myocyte enhancer factor 2 (MEF2) (10), the MEF2-dependent transcription is suppressed by PKA and reactivated by SIK (6).
The first physiological relevance of SIKs in vivo was found in the mouse liver as a suppression of gluconeogeneic programs (11). CREB is one of the key transcription factors that up-regulate gluconeogenic gene expression (12) by binding to their gene promoters, such as phosphoenolpyruvate carboxykinase (Pepck), glucose-6-phosphatase catalytic subunit (G6pc), and peroxisome proliferator-activated receptor gamma coactivator 1␣ (Pgc1a; formal gene name is Ppargc1a). The knockdown of SIK1 protein in the mouse liver by RNAi-adenovirus vectors results in an increase in blood glucose levels, which is accompanied by the dephospho-form of CRTC2 and enhanced levels of gluconenogenic gene expression (11). However, the effects of SIK1 knock-out (KO) on glucose metabolism have not been elucidated (13). SIK2 also may play an important role in glucose metabolism (3, 14 -17).
Obvious abnormalities in glucose homeostasis have been found in SIK3-KO mice (18). SIK3-KO mice display a lean phenotype, resistance to a high fat diet, and excessive hypoglycemia. Although a response to insulin and lactate-induced gluconeogenesis in SIK3-KO mice appeared to be normal under fasting conditions, gluconeogenic gene expression was extremely higher in the SIK3-KO liver than in WT mice (18). In addition to the lean phenotype (small amount of energy storage), impaired bile acid, cholesterol, and retinoid homeostasis with skeletal abnormalities in SIK3-KO mice (19) can make it difficult to determine whether enhanced gluconeogenic programs in the SIK3-KO mice liver were the results of a cell autonomous action or systemic effects.
In addition to the issue of SIKs in gluconeogenesis, the role of AMPK in the regulation of CREB/CRTC-mediated gluconeogenic programs remains unclear (20 -22). A recent excessive study using mice that simultaneously lost the Ampk␣1/␣2 and Sik2 genes in the liver and a kinase inhibitor that inhibited all SIKs suggested that a loss of activity of all SIKs resulted in enhanced gluconeogenic programs, whereas the triple loss of AMPK␣1/␣2 and SIK2 left flawlessly managed gluconeogenic programs (23). Here, using cultured hepatocytes and a small compound, we have tried to discuss the important or indispensable role of SIK3 in the regulation of gluconeogenic programs in the liver.
The anti-AMPK, anti-phospho-AMPK, and anti-phospho antibodies were from Cell Signaling (Boston, MA), the anti-CREB antibody was from GenScript Corp. (Piscataway, NJ), the anti-GAPDH antibody was from WAKO (Tokyo, Japan), and anti-PHKG2 antibody was obtained from Santa Cruz Biotechnology, Inc. (Dallas, TX). Anti-pSIK3 (Thr(P)-163) was created against the keyhole limpet hemocyanin-conjugated peptide (CSNLFTPGQLLK(pT)W) in rabbits and purified with the same peptide. CRTC2 protein was separated on an 8% linear polyacrylamide gel, and other proteins were on 4 -20% gradient gels. AML-12 (alpha mouse liver-12) cells from American Type Culture Collection were cultured in DMEM-Ham's F-12 medium supplemented with 10% FBS and an insulin, transferrin, and selenium supplement (Thermo Fisher, Carlsbad, CA).
Pterosin B was extracted from Pteridium aquilinum (total 100 kg, wet) after soaking in 0.1% sodium bicarbonate at 70°C overnight. The ingredients in the P. aquilinum chloroform/ hexane (1:4) extract were separated by silica gel and charcoal column chromatography, and pterosin B was crystallized in chloroform by increasing the hexane content. Finally, we got 3 g of pterosin B whose purity was confirmed by nuclear magnetic resonance. Synthetic pterosin B (racemic) was obtained from Intelium Crop. (Tokyo, Japan).
Primary Hepatocytes-Hepatocytes were isolated from mice as described previously (23). Briefly, under isoflurane anesthesia, the mouse livers were perfused with Hanks' balanced salt solution, which contained 0.5 mM EGTA, followed by perfusion with liver digest medium (Thermo Fisher). Isolated hepatocytes were cultured in DMEM supplemented with 10% FBS, 100 nM insulin, and 1 M dexamethasone (1ϫ hepatocyte medium). Before the treatments, the hepatocytes were incubated with DMEM supplemented with 1% FBS, 10 nM insulin, and 0.1 M dexamethasone (0.1ϫ hepatocyte medium) for 12 h.
To prepare an adenovirus vector for SIK3 (WT, DA), the SIK3 cDNA fragments were amplified by PCR with the attB primers. The amplified products were then constructed into pDONR221 vectors by using BP clonase enzyme mix (Thermo Fisher). The resultant cDNAs were finally cloned into pAd/ DMV/V5-DEST Gateway vectors using LR clonase enzyme mix (Thermo Fisher).
To screen the SIK3 inhibitory compounds, we constructed the LexA reporter assay system. A DNA fragment containing 3ϫ LexA elements was prepared by annealing the oligonucleotides (5Ј-GATCTACTGTATATATATACAGTAGAGTA-CTGTATATATATACAGTACACTACTGTATATATAT-ACAGTA/5Ј-AATTTACTGTATATATATACAGTAGTGT-ACTGTATATATATACAGTACTCTACTGTATATATATA-CAGTA), and the fragment was ligated into the BglII/EcoRI site of the Renilla-Luc internal reporter vector (pRL-TK). To prepare the LEXA-CRTC2 expression vector, a DNA fragment for LEXA DNA-binding domain was amplified by PCR with primers (5Ј-GCAAAAAGCTAGATCATGAAAGCGTTAACG-GCCAGGCAA/5Ј-ACCATAATGAGAGTCCAGCCAGTCGC-CGTTGCGAA) and Escherichia coli genome DNA. An In-Fusion HD cloning kit (Clontech) was used to replace the DNA fragment for the GAL4 DNA-binding domain in the pM-CRTC2 vector with the LEXA fragments in pM-CRTC2. The HEK293 cells were placed into 96-well white-bottomed plates and transfected with the SIK3 expression vectors (WT or its empty vector; 5 ng), DNAbinding domain-linked expression vectors (pM-MEF2C and pMLexA-CRTC2; 5 ng), pTAL-GAL4 (20 ng), and pRL-LexA (20 ng) per well, using Lipofectamine 2000 (Thermo Fisher). To measure the reporter activity, we used the Dual-Luciferase reporter assay system (Promega). The cells were lysed with 10 l of passive lysis buffer, and all of the lysate was used for the assay.
Levels of Gluconenogenic mRNA Were Normalized by Tbp mRNA-HEK293 and AML-12 cells were transfected with the SIK3 expression vectors (pTarget hSIK3 WT, T411A, S493A, T411A/S493A (DA), or its empty vector; 50 ng), GAL4 DNAbinding domain-linked expression vectors (pM-MEF2C, pM-CRTC2, or its empty vector; 50 ng), pTAL-GAL4 (150 ng) (8), and the Renilla-Luc internal reporter (pRL-(Int-)TK, 50 ng) using Lipofectamine 2000 reagent (Thermo Fisher). To measure reporter activity, we used the Dual-Luciferase reporter assay system (Promega). The cells were lysed with 100 l of passive lysis buffer, and 10 l was used for the assay. The activities of firefly luciferase were normalized by those of Renilla luciferase. The reporter activity measurement was performed as described above. The cells were lysed with 100 l of passive lysis buffer, and 10 l was used for the assay. The activities of firefly luciferase were normalized by those of Renilla luciferase.
Glucose Production and cAMP Measurement-Primary hepatocytes were seeded on a 24-well plate (1.0 ϫ 10 6 ) and incubated in 1ϫ hepatocyte medium for 3 h and then in 0.1ϫ hepatocyte medium for 12 h. The AML-12 cells were plated in 24-well plate (1.0 ϫ 10 6 ) and incubated in DMEM/F-12 medium for 72 h (with medium change every 24 h) to increase the cellular lipid droplets. After the medium replacement with 0.1ϫ hepatocyte medium, the cells were further cultured for 12 h. The glucose production assay was performed as described (22). Briefly, the cells were washed with phosphate-buffered saline followed by adding 300 l of assay medium (DMEM without glucose or glutamate containing 1 mM sodium pyruvate, 10 mM lactate, and 100 nM dexamethasone) containing the compounds (Fsk, etc.) per well. After ϳ0 -8 h of incubation, the assay medium was harvested and centrifuged at 12,000 rpm for 10 min. Then 20 l of the supernatant was used for the assay. To measure the glucose levels, 180 l of reaction buffer (50 mM sodium phosphate, pH 7.0, 1 units/ml glucose oxidase, 0.1 units/ml horse radish peroxidase, 50 M 10-acetyl-3,7-dihydroxyphenoxazine) was added to each well of a 96-well black plate and incubated at room temperature for 30 min. The fluorescence at 590 nm was measured with excitation at 535 nm.
The AML-12 cells were placed into a 96-well plate at a density of 2.0 ϫ 10 4 . The pGloSensor-22F cAMP Plasmids (Promega) were transfected into the cells using Lipofectamine 2000. The cAMP levels were detected with chemiluminescence using GloSensor cAMP reagent (Promega).
Purification of SIK3 and in Vitro Kinase Activity-The pEBG-hSIK3 WT vector was transfected into the HEK293 cells. After 36 h of transfection, the cells were lysed with IP lysis buffer (50 mM Tris-HCl, pH 8.0, 5 mM EDTA, 5 mM EGTA, 2 mM DTT, 50 mM glycerol 3-phosphate, 50 mM NaF, 1 mM NaVO 4 , 0.5% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 14 g/ml aprotinin) and incubated with glutathione-Sepharose 4B (GE Healthcare) at 4°C for 15 min. The protein complexes were washed with IP lysis buffer and eluted with 10 mM of glutathione. Aliquots of purified SIK3 were subjected to an in vitro kinase assay and Western blot analyses.
Adenoviruses of hSIK3 WT and DA was transfected to AML-12 cells. After the stimulation with pterosin B and Fsk, the cells were lysed with 1 ml of IP lysis buffer and incubated with protein G-Sepharose (GE Healthcare) containing anti-SIK3 antibody at 4°C for 1 h. Purified SIK3 was eluted with 100 l of 3ϫ SDS and detected by Western blot analyses.
The purified SIK3 enzyme was incubated with coumarinlabeled CRTC2 peptides and compounds in 40 l of reaction buffer (5 mM Tris-HCl, pH 7.4, 1 mM ATP, 1 mM DTT, 5 mM MgCl 2 ). The reactions were performed at 25°C for 1 h and were stopped by the addition of 40 l of 3ϫ SDS sample buffer. The phosphorylated peptide was separated by electrophoresis on a 1.5% agarose gel in 50 mM Tris borate buffer (pH 8.5) and visualized by ultraviolet light (28).
The PHKG2 enzyme was also prepared by the same methods as for SIK3 and incubated with coumarin-labeled SIK3 T411 peptide (KKLSMRRHTVGVADP) or S493 peptide (KKPL-GRRASDGGANI) in 40 l of the reaction buffer (5 mM Tris-HCl, pH 6.8, 1 mM ATP, 1 mM DTT, 5 mM MgCl 2 ). The phosphorylated peptide was separated on the agarose gel in 50 mM Tris acetate buffer (pH 3.5).
Glycogen Measurement-AML-12 cells were lysed in 0.1 M sodium citrate buffer (pH 4.2) supplemented with 60 mM NaF, and the supernatants were recovered by centrifugation at 14,000 ϫ g for 5 min. The concentration of glycogen was measured using the EnzyChrom glycogen assay kit (BioAssay Systems, Hayward, CA). Reagents for periodic acid-Schiff stain were obtained from Mutokagaku Co. Ltd. (Tokyo, Japan).
Statistical Analyses-Data from each group were characterized by the means Ϯ S.D. Student's t test was used to assess all experimental data in Microsoft Excel.

Gluconeogenic Program Is Constitutively Up-regulated in SIK3-KO Mice
Hepatocytes-In cultured cell systems, SIK1-3 regulates energy metabolism by apparently sharing the same transcriptional modulators, CRTCs (27), and class IIa HDACs (29). However, in vivo, we observed that only SIK3-KO mice had dramatic phenotypes for energy metabolism (18). To confirm the indispensability of SIK3 in glucose metabolism in vivo, all SIKs KO mice with the same genetic background (C57BL/6J) were subjected to a glucose challenge after 4 h of starvation (Fig.  1A). Although the blood glucose levels in both the WT and SIK3-KO mice were significantly elevated in response to the glucose treatment, the maximum level was significantly lower in the SIK3-KO than WT mice. However, no significant difference in glucose clearance between the SIK1-KO or SIK2-KO mice and their WT mice was observed. Next, we examined the gluconeogenic potency in these mice by performing a lactate (a source of gluconeogenesis) challenge under a fed condition in which the gluconeogenic programs had not been run in the WT liver because of insulin actions. The blood glucose levels in the SIK3-KO mice were quickly increased to the levels of the WT mice after lactate treatment, whereas there was no significant response in the glucose level to lactate in WT, SIK1-KO, or SIK2-KO mice (Fig. 1B), suggesting a constitutively activated gluconeogenic ability in the SIK3-KO mice.
Although the gluconeogenic gene expression has been upregulated in SIK3-KO mouse liver (18), the systemic effects, e.g. hypoglycemia, made it difficult to discuss the cell autonomous actions in the SIK3-KO mouse liver. To eliminate the systemic effects, we prepared hepatocytes from SIK1-3 KO mice, and gluconeogenic programs were induced by the cAMP agonist Fsk. As shown in Fig. 2A, the SIK3-KO hepatocytes produced glucose more quickly than the WT hepatocytes (at 2 h post-Fsk treatment), despite a lower maximum glucose production (at 8 h). No significant difference in glucose production was observed between SIK1-KO or SIK2-KO and their WT hepatocytes. Gene expression analyses (Fig. 2B, genes related to gluconenogenesis: Pgc1a, Pepck, and G6pc) suggested that SIK3-KO hepatocytes possessed a high gluconeogenic potency even in unstimulated cells with almost fully enhanced gluconeogenic gene expression levels. Briefly, starvation (glucose/ glutamate-free) did not alter gluconeogenic gene expression levels in both WT and SIK3-KO mice hepatocytes. The stimulation of WT hepatocytes with Fsk up-regulated the levels of gluconenogenic gene expression up to (Pgc1a and Pepck) or beyond (G6pc) those increased levels in the SIK3-KO hepato-cytes. Again, we observed no significant difference in the gluconeogenic gene expression levels between in the SIK1-KO and SIK2-KO and WT mice hepatocytes.
Western blot analyses (Fig. 2C) detected a constitutively dephosphorylated CRTC2 in the SIK3-KO hepatocytes, but not in others. Curiously, Fsk-induced phosphorylation of CREB was not observed in the SIK3-KO hepatocytes. AMPK had already been phosphorylated in unstimulated SIK3-KO hepatocytes, suggesting an energy-depleted environment in SIK3-KO hepatocytes; however, the starved condition might not be one of the major factors for the enhanced gluconeogenic programs in SIK3-KO mice hepatocytes.
Pterosin B Inhibited SIK3 Downstream Cascades-To precisely characterize the SIK3 actions in defined culture systems, we chose the mouse hepatoma AML-12 cells. Despite a success-FIGURE 2. Enhanced gluconenogenic programs in SIK3-KO hepatocytes. A, hepatocytes were prepared from 8-week-old female mice (n ϭ 3 or 4) and cultured in a low serum (1%) medium for 12 h. Then cells were pretreated with a starvation (Starv.) medium (serum-and glucose/glutamate-free) for 1 h. Glucose production was induced by 10 mM lactate, 20 M Fsk, and 0.1 M dexamethasone. Means and S.D. are indicated. *, p Ͻ 0.05; **, p Ͻ 0.01. B, the hepatocytes were preincubated with a low serum (1%) medium for 12 h and incubated in a serum-free medium, a starvation medium alone, or supplemented with 20 M Fsk for 3 h. Total RNA was extracted for quantitative PCR. (n ϭ 3). C, the same treatment was used for Western blot analyses. ful knockdown of SIK1 and SIK2 protein in AML-12 cells, we could not knock down the SIK3 protein. Moreover, an in vitro screening for SIK3 inhibitors from our chemical library (ϳ2,500 compounds) resulted in only an identification of nonspecific inhibitors, such as the flavonoid fisetin (24). To modulate the SIK3-specific signal, we performed reporter-based screening using transcription regulators (class IIa-HDAC and CRTC2) that were suppressed by SIKs. Fig. 3A indicates the principles of the reporter system. Because SIK3 inhibited the suppressive action of class IIa-HDAC on the transcription factor MEF2, the firefly luciferase activity under the regulation of the GAL4 fusion MEF2 is up-regulated by SIK3 and down-reg-ulated by the SIK3 inhibitory compounds, such as the nonspecific kinase inhibitor staurosprine (STS) (27) (Fig. 3B). CRTC2 is also inhibited by SIK3; thus, the Renilla luciferase activity directly regulated by LEXA fusion CRTC2 was suppressed by SIK3, and STS up-regulated the LEXA-Renilla luciferase activity. The combination of these two reporters in the same cells produced ϳ25 times the S/N ratio, when SIK3 was inhibited by STS.
Using these reporters and HEK293 cells, we identified pterosin B, an ingredient in P. aquilinum, as a candidate for SIK3 signaling inhibition (Fig. 3C). The action of pterosin B on MEF2 and CRTC2 were also confirmed in separately assayed firefly luciferase reporter systems in the AML-12 cells (Fig. 3D). The structural-activity relationship assay suggested that only pterosin S failed to inhibit SIK3 signaling (Fig. 3E). Synthetic pterosin B (racemic) also inhibited SIK3 signaling, indicating that pterosin B, not non-pterosin contaminants, inhibited SIK3 signaling.
To visually evaluate the inhibitory actions of pterosin B on SIK3 signaling, GFP fusion HDAC5 (a class IIa-HDAC) and CRTC2 were expressed together with SIK3. Both HDAC5 (Fig.  4A) and CRTC2 (Fig. 4B) were localized in the cytoplasm in the SIK3-overexpressing cells, and pterosin B inhibited this cytoplasmic localization, suggesting that pterosin B was able to inhibit downstream cascades of SIK3 signaling.
Pterosin B Does Not Affect SIK3-KO Hepatocytes-To examine whether pterosin B inhibits SIK3 signaling and mimics the glucose metabolism observed in SIK3-KO hepatocytes (Fig. 2), both WT and SIK3-KO hepatocytes were treated with pterosin B, and glucose production was monitored (Fig. 5A). Pterosin B quickly increased the medium glucose levels in WT hepatocytes, which was accompanied by enhanced expression of gluconeogenic genes (Fig. 5B). However, no further enhancement of glucose production or gluconeogenic gene expression by pterosin B was observed in the SIK3-KO hepatocytes, suggesting that pterosin B-induced gluconeogenesis may be mediated by SIK3. Prolonged treatment with pterosin B was found to decrease glucose levels in the WT hepatocyte medium, and G6pc gene expression in SIK3-KO hepatocytes was down-regulated by pterosin B.
Pterosin B Represses SIK3 Signaling via the C-terminal Regulatory Region-To test the specificity of pterosin B toward SIKs, SIK1-3 was examined in the same reporter assay in the HEK293 cells (Fig. 6A). The up-regulation of MEF2 and downregulation of CRTC2 activities by SIK1 or SIK2 were not affected by pterosin B, indicating that pterosin B specifically inhibited SIK3 signaling. On the other hand, pterosin B did not inhibit SIK3 kinase activity up to 1 mM, whereas the strong pan-SIK inhibitor HG9-91-01 (30) completely inhibited SIK3 kinase activity even at 1 M (Fig. 6B).
Because SIK3 was activated by phosphorylation at Thr-163 by the upstream kinase LKB1 (4), we examined the phosphor-   JULY 17, 2015 • VOLUME 290 • NUMBER 29 ylation status at this site on the overexpressed SIK3 in HEK293 cells (Fig. 6C). Western blot analyses with anti-Thr(P)-163 revealed that pterosin B did not alter the phosphorylation level at this site. The alternative candidate sites for inhibition of the SIK3 actions were in the C-terminal regulatory domains of Thr-411 and Ser-493 (Fig. 6D) (31). The phosphorylation of these sites suppressed the CRTC inhibitory activity of SIK3. To confirm the importance of the C-terminal domain, we performed reporter assays with various SIK3 mutants. The truncation of the C-terminal domain (kinase domain only) converted SIK3 into a pterosin B-resistant mutant (Fig. 6D). The actions of double mutant SIK3 (DA: T411A and S493A) were also not affected by pterosin B (Fig. 6E), despite a small difference between the MEF2 and CRTC2 systems (Thr-411 may be more critical for CRTC2 regulation than Ser-493).

SIK3 Is the Major CRTC Kinase in Hepatocytes
Indeed, the phosphorylation levels at Thr-411 and Ser-493 were up-regulated in mouse hepatoma AML-12 cells treated with pterosin B, which was also found in Fsk-treated cells (Fig.  7A). Ser-493 may be phosphorylated by PKA (1), and overexpression of PKA in HEK293 cells mimicked the pterosin actions on MEF2 and CRTC2 (Fig. 7B). However, the PKA inhibitor H89 did not block pterosin B-dependent suppression of SIK3 signaling (Fig. 7C). Although Thr-411 is probably phosphorylated by PKA or CaMKs, the overexpression of CaMK I/II or the treatment with the CaMK inhibitor KN62 disproved the involvement of these kinases in the pterosin B-dependent suppression of SIK3 signaling (Fig. 7, B and C). In addition, the cAMP-responsible luciferase system showed that pterosin B reduced the intracellular cAMP levels induced by Fsk (Fig. 7D). These results suggested that unknown kinases, rather than PKA or CaMK I/II, might phosphorylate SIK3 in Thr-411 and Ser-491, which inhibited SIK3 signaling.
Pterosin B-resistant SIK3 Decreases Glucose Production and CRTC2 Dephosphorylation in AML-12 Cells-To specifically examine the signaling in pterosin B-induced gluconeogenesis and the relevance of the importance of the C-terminal domain of SIK3, we examined glucose production in AML-12 cells after pterosin B treatment. Pterosin B increased glucose production as quickly as Fsk, whereas HG9-91-01 increased it gradually (Fig. 8A). At 8 h post-stimulation, however, the glucose level in pterosin B-treated cells was returned to the basal level. Because the higher potential of pterosin B on glucose production than HG9-91-01 suggested the presence of SIK-independent pathways in pterosin B actions, we examined pterosin B-induced glucose production in the AML-12 cells that had been transfected with SIK3-adenoviruses. Overexpression of SIK3 lowered the pterosin B-induced glucose production, which was more evident in cells that expressed pterosin B-resistant SIK3 DA mutant than in pterosin B-sensitive SIK3 (WT) (Fig. 8B). Next, we examined the levels of gluconeogenic gene expression in the pterosin B-treated AML-12 cells. Cells were pretreated with serum-free medium for 1 h and then incubated with pterosin B or other compounds in the absence of glucose for 3 h. Pterosin B up-regulated Pgc1a, Pepck, and G6pc mRNA levels with different efficacies, which was also represented by Fsk and HG9-91-01 with different target specificities (Fig. 8C). Again, we confirmed whether SIK3 overexpression was able to suppress pterosin B-induced gluconeogenic gene expressions. All expressions were suppressed by SIK3 overexpression, and SIK3 DA was more efficient than SIK WT (Fig. 8D).
We then examined the protein status involved in gluconeogenesis. Pterosin B quickly induced the dephosphorylation of CRTC2 in AML-12 cells (Fig. 8E), which was also observed in Fsk-and HG9-91-01-treated cells. CREB phosphorylation at Ser-133 was up-regulated with starvation time in the control (DMSO) cells. However, the treatment with pterosin B or HG9-91-01 inhibited the CREB phosphorylation. An increase in AMPK phosphorylation levels at Thr-172 was further up-regulated in pterosin B-treated cells and down-regulated in Fsktreated cells. HG9-91-01 apparently showed no effect on the AMPK phosphorylation levels. Similarly to other assays, SIK3 overexpressionreturnedthelevelsofpterosinB-induceddephosphorylation of CRTC2 and phosphorylation of CREB (Fig. 8F), which was, again, more evidently with SIK3 DA than SIK3 WT. These results suggested that dephosphorylation of CRTC2 by pterosin B might induce gluconeogenic programs, and the inactivation of SIK3 caused by enhanced phosphorylation of its C-terminal domain might be a mechanism of these programs.
Involvement of PHKG2 in Pterosin B-mediated Inactivation of SIK3-Finally, we tried to identify the molecules that mediate pterosin B action on SIK3, which could be a kinase or a phosphatase. Some candidates were overexpressed in the MEF2 and CRTC2 reporter systems in HEK293 cells, and SIK3-dependent activation and repression were evaluated in the presence or absence of pterosin B (Fig. 9A). We previously reported that the phosphatases PP1, PP2A, and calcineurin modulate CRTC2 activity (32). However, these phosphatases modulate reporter activities of either MEF2 or CRTC2. Pyruvate dehydrogenase kinase and pyruvate dehydrogenase phosphatase are key regulators in mitochondria and regulate acetyl-CoA production by modulating pyruvate dehydrogenase activity (33). Pyruvate dehydrogenase activity is inextricably linked with pyruvate carboxylase, producing oxaloacetic acid, which is the initial metabolite of gluconeogenesis. However, pyruvate dehydrogenase kinase(s) and pyruvate dehydrogenase phosphatase(s) did not modulate SIK3-dependent MEF2 or CRTC2 activities.
When gluconeogenesis is activated in hepatocytes, glycogenolysis could also be activated. The key molecules regulating glycogenolysis are phosphorylases and are activated by phosphorylase kinase catalytic gamma 2 (PHKG2). Indeed, PHKG2, but not the muscle type PHKG1, inhibited the SIK3-dependent regulation of both MEF2 and CRTC2, which is accelerated by pterosin B (when SIK3 is inhibited, the fold change approaches 1). In addition, 3 h of treatment with pterosin B decreased the glycogen content in AML-12 cells in a dose-dependent manner (Fig. 9B), suggesting that pterosin B stimulates glycogenolysis.
An in vitro kinase assay of PHKG2 and SIK3 peptide corresponding to the regions of Thr-411 and Ser-493 indicates that PHKG2 can phosphorylate SIK3 (Fig. 9C). GST pulldown in AML-12 cells suggests that PHKG2 binds to SIK3 in response to pterosin B (Fig. 9D) in a Thr-411 or Ser-493 phosphorylation-independent manner (Fig. 9E). The association with PHKG2 was only observed when SIK3 was used as bait (Fig. 9F).
On the other hand, the overexpression of PHKG2 in AML-12 cells had less of an effect on the SIK3 cascades downstream, probably because of the limited amount of SIK3 protein. To show the indispensable role of PHKG2 in pterosin B-mediated SIK3 inactivation, knockdown experiments were performed in AML-12 cells. Two different miRNA vectors for mouse PHKG2 decreased the protein levels and lowered the phosphorylation levels of SIK3 Thr(P)-411 and Ser(P)-493 (Fig. 9G), in contrast to CRTC2 phosphorylation levels. Pterosin B-induced gluconeogenic gene expression was also suppressed by PHKG2 knockdown (Fig. 9H), suggesting that PHKG2 is a new regulator of SIK3 as elicited by pterosin B.

Discussion
The first clue about the involvement of SIK in the regulation of gluconeogenesis in the liver was based on the fact that SIK1 suppressed the CREB activity via inactivation of CRTC2 (11). During the proof of this suppression, the AMPK activator 5aminoimidazole-4-carboxamide-1-s-D-ribofuranoside (AICAR) also represented the similar CRTC2 inactivation activity (11), which might deceive us into discussing the role of AMPK in gluconeogenic programs.
AMPK generally acts as energy-saving factors, whereas gluconeogenesis consumes energy (21,34,35). In this context, AMPK-mediated suppression of gluconeogenic programs is apparently reasonable. However, overexpression of AMPK in HEK293 cells fails to suppress the CRTC2 transactivating activity despite its ability as the CRTC2 kinase in vitro (27). If a kinase directly phosphorylated their putative substrate, the kinase would never choose a cell type, hepatocytes, or model cells (HEK293), and regulate the cellular downstream cascades of the substrate in the same manner, which may deny that AMPK is the CRTC2 kinase in hepatocytes. However, it does not deny the contribution of AMPK to gluconeogenesis in hepatocytes (36), although we have to mention here that other AMPK family kinases, MARKs (37,38), may be capable of phosphorylating CRTC in cells (32,39).
Other aspects about the upstream regulator of AMPK and SIK confused the discussion. LKB1 commonly phosphorylates AMPK and SIK and activates them (4). The loss of LKB1 in the liver results in hyperglycemia caused by enhanced gluconeogenesis (20,40). The AMPK-CRTC2 cascade was first proposed to be the responsible cascade for gluconeogenic programs caused by LKB1 deficiency (20). However, some reports have provided negative evidence about this issue (22,23). The liverspecific double knock-out mice with defective genes for Ampk␣1/␣2 have been found to have less dysregulation in glucose metabolism in vivo (22). The basal level of CRTC2 phosphorylation in the AMPK␣1/␣2-deficient hepatocytes is as high as that in WT hepatocytes. A recent extensive study using triple knock-out mice (AMPK␣1/␣2 and SIK2) showed an intact gluconeogenic regulation in the liver of this mouse (23). In contrast, the pan-SIK inhibitor HG9-91-01 that failed to inhibit AMPK induced the gluconeogenic programs followed by glucose production. Because the HG9-91-01-induced gluconeogenic programs are suppressed by overexpression of any drug-resistant isoforms of SIK1-3 (23), the presence of at least one SIK isoform may be sufficient to manage the gluconeogenic programs in hepatocytes.
Does AMPK really phosphorylate CRTC2 in hepatocytes? The phosphorylation level of CRTC2 in the cAMP-treated hepatocytes is increased by AMPK-activating compounds, metformin (41), and AICAR (42), which is not observed in AMPK␣1/␣2-deficient hepatocytes, suggesting unknown mechanisms by which AMPK can regulate CRTC phosphorylation (22,23). It could be speculated that AMPK may inhibit the dephosphorylation steps of CRTC rather than the phosphorylation steps.
The next question that arose was whether all SIK isoforms equally suppressed gluconeogenic programs. SIK3-KO mice liver expressed gluconeogenic genes at a high level (18) and constitutively activated (the levels were comparable to those stimulated by Fsk) gluconeogenic programs in the SIK3-KO hepatocytes have been found in the present study, which may mimic LKB1-KO hepatocytes (20), but not AMPK␣1/␣2and SIK2-triple KO hepatocytes (23). These results suggest that SIK3 may be the major regulator of gluconeogenic programs under the control of LKB1 in hepatocytes.
A number of reports succeeded to performing SIK3 knockdowns, and some showed impaired cholesterol transport in SIK3 knockdown cells (43), which explained the phenotype of cholesterol accumulation in the SIK3-KO liver (18). However, some phenotypes caused by SIK3 knockdown were observed in HeLa cells in which SIK3 could not possess kinase activity because of LKB1 deficiency (27). Although it was difficult to conclude whether the results in knockdown experiments were from on or off target effects, it was also true that we failed to knock down SIK3 even in HEK293 cells. Thus, we tried to identify SIK3-specific inhibitors.
Pterosin B suppressed SIK3 downstream cascades, including CRTC2 phosphorylation, despite not being capable of inhibiting SIK3 kinase activity. Pterosin B may modulate an unknown kinase or phosphatase signals and increase the self-inhibitory function of the C-terminal regulatory domain of SIK3. This mechanism was specific to SIK3, not SIK1 or SIK2; thus, we thought that the loss of SIK3 or its signaling resulted in an enhancement of gluconeogenic programs in hepatocytes via CRTC2 dephosphorylation. However, it was also true that pterosin B and Fsk induced higher levels of glucose production than that by the pan-SIK inhibitor HG9-91-01 (23). Because glucagon and Fsk increased cellular cAMP levels and activated PKA followed by CREB phosphorylation, the synergistic actions between dephosho-CRTC2 and phospho-CREB might result in higher potency of the gluconeogenic programs than the solo action of dephosho-CRTC2 produced by HG9-91-01 (23). On the other hand, pterosin B suppressed CREB phosphorylation levels in common with HG9-91-01. Pterosin B may activate an unidentified cascade up-regulating gluconenogenic programs in which SIK3 was fatefully inactivated as by PKA signaling.
Re-evaluation of metabolic regulators whose activities were modulated along with gluconeogenesis suggests that PHKG2 is a candidate for SIK3-inactivating kinase. PHKG2 belongs to the CaMK family (44) and shares phosphorylation motifs with CaMKs and PKA (45). A major role of PHKG2 is the initiation of glycogen breakdown in response to glucagon-cAMP-PKA or Ca 2ϩ signaling by phosphorylating phosphorylase (46). Muta-tions in the PHKG2 gene cause type IXc glycogen storage disease (liver glycogenosis), which is characterized by hypoglycemia, lactic acidosis, and cirrhosis (47). Although glucagon response was found in type IX glycogen storage disease patients (48), some patients with PHKG2 gene mutations were reported to not respond or to only weakly respond to glucagon (49). In addition, type Ia glycogen storage disease, G6Pc deficiency (50), and PEPCK deficiency (liver isoform PCK1) (51) commonly cause lactic acidosis, suggesting the dysregulation of gluconeogenesis in PHKG2-mutated patients. A decrease in the glycogen content in pterosin B-treated AML-12 cells, capability of SIK3 phosphorylation by PHKG2 in vitro, specific association of PHKG2 with SIK3, and weakened action of pterosin B in PHKG2 knocked down AML-12 suggest that pterosin B inactivates SIK3 via the activation of PHKG2.
Pterosin B may activate both glycogenolysis and gluconeogenesis (Fig. 9I) and is therefore capable of quickly producing glucose (ϳ4 h post-treatment). Once the glycogen is depleted (SIK3-KO hepatocytes are also under an energy deficiency state), pterosin B may be unable to promote glucose production, which decreases the rate of glucose production later on (8 h). In addition, unexplained weaker expression of G6pc by pterosin B (compared with by Fsk or HG9-91-01) may result in insufficient glycogenolysis and gluconeogenesis, which is also observed in SIK3-KO hepatocytes, suggesting that secondary effects, such as modulation of AMPK signaling via glycogen metabolism (52), may affect G6pc expression, probably via a CRTC2-independent mechanism (e.g. through forkhead transcription factor FKHR) (53).
Pterosin A, a derivative of pterosin B, was reported to lower, not increase, blood glucose levels in diabetic animals (stroptozotocine-treated, high fat-fed, and db/db mice) (54). Pterosin A simultaneously activated conflicting signals in energy metabolism, AKT, and AMPK in the skeletal muscle and livers. In the liver, pterosin A increased glycogen content and decreased cAMP/dexamethasone-induced Pepck gene expression, which was a mechanism of pterosin A-mediated blood lowering effects. However, pterosin B lowered the glycogen content and initiated gluconeogenic programs, followed by glucose production. A number of assay conditions were different between ours and the pterosin A experiments (e.g. the absence or presence of serum and pterosin alone or with gluconenogenic inducers); we noted here that the most critical difference might be from acute or chronic effects. It was true that pterosin B resulted in a lowered glucose production at later phase (after 8 h), and pterosin A experiments for mRNA analyses were performed after 24 h, suggesting that pterosin(s) may act as bilateral FIGURE 9. PHKG2 inactivates SIK3 in response to pterosin B. A, a reporter assay was performed in HEK293 cells. Some kinases and phosphatases were overexpressed together with GAL4-MEF2 or GAL4-CRTC2 reporters (see Fig. 2B). The bars indicate fold activation (MEF2) or repression (CRTC2) by SIK3 overexpression (n ϭ 3). In SIK3 without transcriptional regulatory activity, the fold change (activation or repression) approaches 1 (no change). Pterosin B was added at 300 M. B, AML-12 cells that had been cultured for 72 h with daily medium change were incubated with various concentrations of pterosin B (0 -300 M) for 3 h. The glycogen concentrations were then measured (left panel). Intracellular accumulation of glycogen was observed after periodic acid-Schiff staining (right panel). C, the GST fusion PHKG2 enzyme was overexpressed in COS-7 cells and purified using a glutathione column. Fluoro-peptides corresponding to SIK3 Thr-411 and Ser-493 were incubated with GST-PHKG2 in the presence of ATP. Phosphorylated peptides were separated by electrophoresis on agarose gel. D, GST-SIK3 were overexpressed in AML-12 cells in the presence of Halo-tagged PHKG1/2 and pulldown by glutathione-Sepharose after 3 h of pretreatment with pterosin B (300 M). PHKG1/2 and phospho-SIK3 (pS) were detected by anti-Halo tag antibody and anti-pS493, respectively. E, GST-SIK3 WT and T411A/S493A mutant (DA) were used. F, GST SIK1-3 were used. G, PHKG2 protein was knocked down in AML-12 cells by transformation with miRNA plasmid vectors. H, the same sequences of miRNA were transferred into an adenovirus vector and knocked down PHKG2 protein in AML-12 cells to monitor pterosin B-induced gluconeogenic gene expression (300 M, 2 h). n ϭ 3. **, p Ͻ 0.01 (compared with the control group, LacZ-pterosin B). I, hypothetical model of SIK3 signaling. effectors in glucose metabolism. A reduction in glucose production (content in the culture medium) at the late phase was also observed in SIK3-KO mouse hepatocytes, and hypoglycemia was also the representative phenotype of these mice. These results suggested that SIK3-KO hepatocytes and pterosin B-treated AML-12 cells may share similar mechanisms that may be linked to glycogen breakdown. Further studies on SIK3-mediated regulation of gluconeogenesis, glycogenolysis, and an uncharacterized crosstalk between AMPK and SIK3 signaling are needed.