H. Takemori, Laboratory of Cell Signaling and Metabolism, National Institute of Biomedical Innovation, 7-6-8, Asagi, Saito, Ibaraki, Osaka, 567-0085, Japan Fax: +81 72 641 9836 Tel: +81 72 641 9834 E-mail: email@example.com
Cyclic AMP responsive element (CRE)-binding protein (CREB) is known to activate transcription when its Ser133 is phosphorylated. Two independent investigations have suggested the presence of Ser133-independent activation. One study identified a kinase, salt-inducible kinase (SIK), which repressed CREB; the other isolated a novel CREB-specific coactivator, transducer of regulated CREB activity (TORC), which upregulated CREB activity. These two opposing signals are connected by the fact that SIK phosphorylates TORC and induces its nuclear export. Because LKB1 has been reported to be an upstream kinase of SIK, we used LKB1-defective HeLa cells to further elucidate TORC-dependent CREB activation. In the absence of LKB1, SIK was unable to phosphorylate TORC, which led to constitutive activation of CRE activity. Overexpression of LKB1 in HeLa cells improved the CRE-dependent transcription in a regulated manner. The inactivation of kinase cascades by 10 nm staurosporine in LKB1-positive HEK293 cells also induced unregulated, constitutively activated, CRE activity. Treatment with staurosporine completely inhibited SIK kinase activity without any significant effect on the phosphorylation level at the LKB1-phosphorylatable site in SIK or the activity of AMPK, another target of LKB1. Constitutive activation of CREB in LKB1-defective cells or in staurosporine-treated cells was not accompanied by CREB phosphorylation at Ser133. The results suggest that LKB1 and its downstream SIK play an important role in silencing CREB activity via the phosphorylation of TORC, and such silencing may be indispensable for the regulated activation of CREB.
Cyclic AMP-responsive element (CRE)-binding protein (CREB) is a transcription factor that plays an important role in numerous physiological events, such as cell proliferation, survival, tumorigenesis, glucose metabolism and memory, in a phosphorylation-dependent manner [1,2]. Upstream signals arriving at CREB are conveyed to transcriptional machineries via two distinct domains of CREB. The kinase-inducible domain (KID) is located in the N-terminal region, which contains an activating phosphoacceptor residue, Ser133. The other domain in the C-terminus is composed of a basic leucine zipper (bZIP) that is responsible for dimerization and binding to the CRE. Phosphorylation of Ser133 alters the affinity of KID to the KIX domain of CREB and p300, resulting in enhanced transcription of CRE-dependent genes . Development of a specific antibody that recognized phospho-Ser133 [4,5] enabled investigators to monitor the level of ‘activated CREB’, which has now significantly accelerated the studies of phosphorylation-dependent CREB activation.
Possible involvement of the bZIP domain of CREB in the regulation of CRE-dependent gene expression has been suggested by the results of two lines of research. One was initiated by an mRNA subtraction study to isolate a specific molecule induced in the adrenal gland under the stress of consuming a high-salt diet. The molecule isolated was a kinase, and, thus, we named it salt-inducible kinase (SIK) . SIK is a member of the AMP-activated protein kinase (AMPK) family . A gene database search found three isoforms of SIK, SIK1–3 [8,9]. In Y1 mouse adrenocortical tumor cells, levels of mRNA, protein and kinase activity of SIK1 were elevated within 30 min after the initiation of the cAMP–protein kinase A (PKA) cascade. Overexpression of SIK1 inhibited gene expression(s) induced by cAMP . Analyses of the promoter regions of such genes indicated that CREs in the promoters were the sites where SIK-mediated transcriptional repression occurred, and, thus, SIK1 was thought to repress CREB activity . Although SIK seemed not to phosphorylate CREB directly, it repressed CREB in a kinase-activity-dependent manner. Mapping the region where SIK1 exerted its repressive action suggested that SIK1 repressed CREB by acting on the bZIP domain . In reporter gene assays, overexpression of the kinase domain of SIK1 repressed CRE-dependent transcription completely, even when CREB was supposed to be fully activated by overexpression of PKA. We therefore thought that the phosphorylation of CREB at Ser133 was not sufficient for making ‘activated CREB’.
The second line of research began in an attempt to isolate novel factors that could modulate CREB activity by using high-throughput transformation assays [12,13]. Expression vectors containing full-length cDNAs were cotransformed with reporter vectors in HEK293 cells, and a new family of coactivators was identified. They were named transducer of regulated CREB activity (TORC) 1–3. The N-terminal region of TORCs formed a coiled-coil structure, which interacted with the bZIP domain of CREB . Once the TORCs had been overexpressed in HEK293 cells, CREB-dependent transcriptions were upregulated at, or beyond, the levels induced by cAMP. Activation of CREs by an overexpression of TORCs required CREB-, but not Ser133 phosphorylation, and, thus, TORCs were thought to be coactivators that did not require CREB phosphorylation. Cytochemical studies of TORCs, however, showed that the activating signals that could phosphorylate CREB, such as cAMP and Ca2+, also induced the nuclear import of TORCs [14,15], suggesting that combination of Ser133 phosphorylation and the binding of TORCs to the bZIP domain produces the fully ‘activated CREB’.
The above observations indicate that SIKs and TORCs share a common feature regarding the regulation of CREB activity, both acting on the bZIP domain of CREB in a phospho-Ser133-independent manner. Having examined this feature further, we found that SIK2 phosphorylated TORC2 at Ser171. The resulting phospho-TORC2 was exported from the nucleus to the cytoplasm, and this led to the apparent inactivation of the CREB activity . We also showed that PKA phosphorylated SIK1  and the phospho-SIK1 could not induce the nuclear export of TORC2 .
The importance of the phospho/dephospho regulation of TORC, by regulating CREB, was also shown as a physiological impact in hepatic gluconeogenesis . However, it remains to be clarified as to whether the phospho/dephospho regulation of TORC is one of the regulatory mechanisms for the CREB activity or the cascade indispensable for the CREB action. AMPK family kinases, including SIK, have flexible activation loops (A-loops) near their substrate-binding pockets. Phosphorylation in the A-loop induces a structural change in the catalytic site, which turns on the kinase activities. Recently, the LKB1 tumor suppressor kinase  was reported to be a major upstream activator of AMPK family kinases . LKB1 phosphorylates Thr residues in the A-loops of SIKs. By using LKB1-defective HeLa cells and a compound inhibiting TORC-kinases, including SIKs, we tried to elucidate the importance of the Ser133-independent activation of CREB. The results suggested that the phospho/dephospho regulation of TORC plays an indispensable role in the regulated activation of CREB.
All TORCs are substrates of SIK1
One of the downstream branches of the SIK-signaling cascade leads to the regulation of CRE-dependent transcription, and we recently succeeded in identifying TORC2, a CREB-specific coactivator, as an endogenous substrate of SIK2. Because mammals have three TORC isoforms, we decided to clarify which isoform could act as the endogenous substrate of SIK1. Figure 1A shows that the SIK-phosphorylation motif is highly conserved among the three isoforms. SIK1 was able to phosphorylate all the TORC peptides except for the S171A-TORC2 (Fig. 1B).
The levels of SIK1-dependent phosphorylation of TORC isoforms were also examined in cultured cells. By using COS-7 cells, glutathione-S-transferase (GST)-tagged full-length TORCs were coexpressed with contrasting SIK1 mutants; one was a kinase-defective mutant (K56M) and the other was a mutant constitutively phosphorylating TORC  (S577A mutant). As shown in Fig. 1C, the levels of phosphorylation at Ser171 of TORC2 and Ser163 of TORC3 were strikingly elevated in the presence of SIK1 (S577A) (see the lanes indicated by 577). However, the corresponding residue, Ser167 of TORC1, seemed to be phosphorylated even in cells expressing inactive SIK1 (the lanes indicated by 56), and its phosphorylation level was enhanced slightly in cells expressing SIK1 (S577A). In contrast to the phosphorylation at Ser167, binding of 14-3-3 to TORC1 was significantly enhanced by SIK1 (S577A), suggesting that SIK1 could phosphorylate TORC1, but some as yet unidentified kinases, other than SIK1, might phosphorylate at Ser167.
To evaluate the direct action of SIK1 on the transactivation activity of TORCs, assays were performed using Gal4-fused TORCs (Fig. 1D). SIK1 was able to completely inhibit the transactivation activities derived from all TORCs. Together, these results suggested that SIK1 could phosphorylate all TORCs and thereby repress their transactivation activities.
SIK1 is unable to induce the nuclear export of TORC in HeLa cells
The nucleo-cytoplasmic redistribution of TORC2 is important for both the stimuli-induced CRE activation and the SIK-mediated CRE repression. Interestingly, Bittinger et al. found that TORC2 and TORC3 never moved out of the nucleus in HeLa cells . Because HeLa cells lacked LKB1, which had been reported to phosphorylate SIKs and activate them , we thought that the impaired nuclear export of TORC2 in HeLa cells was a result of the loss of SIK activity . To test this, we compared the behaviors of TORCs in HeLa and COS-7 cells using a green fluorescent protein (GFP)-fusion technique. Because GFP–TORC1 was present in the cytoplasm of COS-7 cells (Fig. 2A), we were unable to see the SIK1-induced intracellular redistribution of GFP–TORC1 [compare SIK1(–) with SIK1(+)]. In contrast to TORC1, GFP–TORC2 clearly showed SIK1-dependent nuclear export. GFP–TORC3 also moved out of the nucleus in a slightly lower level. As expected, overexpression of SIK1 did not induce the intracellular redistribution of TORCs in HeLa cells (Fig. 2B).
Overexpression of LKB1 in HeLa cells restores the nucleo-cytoplasmic shuttling of TORC2
To elucidate the mechanism underlying the impaired nucleo-cytoplasmic shuttling of TORC in HeLa cells, we first tested whether overexpression of LKB1 in this cell line could restore the SIK1-induced nuclear export of TORC2 (Fig. 3A). As shown in the third panel of the ‘upper’ set, a small population of GFP–TORC2 moved to the cytoplasm in LKB1-overexpressing HeLa cells. Furthermore, expression of LKB1 and SIK1 in combination could completely induce the nuclear export of GFP–TORC2 (final panel). As expected, distribution of GFP–TORC2 was not influenced by the overexpression of LKB1 in LKB1-positive COS-7 cells (lower set).
The Thr182 of SIK1 is phosphorylated by LKB1, resulting in conversion from inactive SIK1 to the active form. The importance of phospho-Thr182 was also supported by the fact that substitution of the Thr with a negatively charged residue produced a constitutive active enzyme ; hence we prepared the T182E mutant. As shown in Fig. 3B, however, neither SIK1 (T182E) nor SIK1 (T182A) could enhance LKB1-supported nuclear export of GFP–TORC2 in HeLa cells.
Differential properties of the A-loops of the individual isoforms of SIK1–3
Because the SIK1 (T182E) mutant did not induce the nuclear export of TORC2, we assayed the kinase activity of this mutant. The T182E mutant, prepared as a GST-fusion protein using COS-7 cells, was much less active than wild-type SIK1 (Fig. 4A). As expected, neither SIK1 (T182A) nor a negative control mutant, K56M, showed kinase activities. The discrepancy between our T182E mutant and the mutant in previous reports [20,21] might be caused by the different sources, Escherichia coli or cultured cells. However, similar discrepancies have also arisen in studies of AMPK .
There are three isoforms of SIK in the AMPK-related kinase family. Although the overall sequence of A-loops in SIKs is highly conserved (Fig. 4B), some variety is found in the N-terminal side of the LKB1-phosphorylatable Thr in SIK3. To see whether the SIK2 and SIK3 isoforms behave similarly to SIK1 with regard to LKB1-dependent phosphorylation of Thr, we prepared several mutants in which the corresponding Thr residues were substituted. Kinase assays of GST–SIK2s (Fig. 4C) produced results similar to those of SIK1s. However, GST–SIK3s (Fig. 4D) provided results quite different from the others. SIK3 (T163A) had a little peptide phosphorylation activity, and SIK3 (T163E) had activity as high as that of the wild-type enzyme.
SIK kinase activity is sufficient to induce the nuclear export of TORC2 in HeLa cells
The finding of the constitutive active SIK3 mutant, SIK3 (T163E), prompted us to investigate whether the kinase activity of SIK was sufficient to export TORC2 even under LKB1-defective conditions. Expression plasmids for GFP–TORC2 and SIK3s were cotransformed into HeLa cells. In LKB1-overexpressing HeLa cells, GFP–TORC2 was exported from the nucleus to the cytoplasm by either wild-type SIK3 or SIK3 (T163E) mutant (Fig. 5, upper). SIK3 (T163A) mutant was unable to enhance the nuclear export of GFP–TORC2. As expected, even in LKB1-nonexpressing HeLa cells (Fig. 5, lower), SIK3 (T163E) could induce the nuclear export of GFP–TORC2, although wild-type SIK3, induced little export. These results suggested that LKB1 could regulate the intracellular distribution of TORC2 through SIKs, and that SIK kinase activity might be sufficient to induce the nuclear export of TORC2.
Environments of CRE-dependent transcription in HeLa cells
Next, we examined the expression of an endogenous target of CREB, NR4A2 (Nurr1) gene (Fig. 6A). The level of NR4A2 mRNA was significantly induced by forskolin treatment in HEK293 cells. In HeLa cells, however, it had already been expressed moderately, and its level was not enhanced strongly by forskolin treatment. The level of 36B4 RNA, generally used as an internal standard, showed no change.
To find out why HeLa cells expressed NR4A2 mRNA constitutively, we compared the expression level and the status of components in the TORC–CREB system between HeLa and HEK293 cells. The mRNA levels of TORCs in HeLa cells did not differ substantially from those in HEK293 cells (Fig. 6B). However, protein levels of TORCs seemed to be much lower in HeLa cells than in HEK293 cells (Fig. 6C,D). TORC2 proteins in HEK293 cells are known to migrate as two bands on SDS/PAGE (Fig. 6C). A slowly moving form was the major form in nonstimulated HEK293 cells and was shown to be the phosphorylated form. However, the rapidly moving form was the major form in forskolin-treated cells and was shown to be the dephosphorylated form [14,15]. In HeLa cells, however, only the rapidly moving dephospho form was seen. Because our antibody raised against TORC1 was able to detect both TORC1 and TORC3 with equal efficiency, analyses of these two TORCs were carried in a single blot (Fig. 6D). Similarly to the case of TORC2, TORC1 and TORC3 also formed two bands in the gel. The cases of TORC1 and TORC3 were apparently the same as TORC2, but might be substantially different, because all of the bands responded to forskolin treatment in HEK293 cells and shifted to lower positions. However, in HeLa cells the bands remained at the same positions irrespective of treatment. It should be mentioned here that Ser133 phosphorylation of CREB might not be strong in forskolin-treated HeLa cells (Fig. 6E), suggesting that more than one step in the regulatory pathway of CREB might be impaired in HeLa cells.
LKB1 restores the accentual regulation of CRE-dependent transcription
To examine whether the constitutive expression of NR4A2 mRNA in HeLa cells suggested the impaired regulation of CREs, and if so, to test whether LKB1 could restore the forskolin-induced activation of CRE-dependent transcription, we tried to perform reporter assays. Because plasmid-based reporters could not provide a high enough level of reporter activities in HeLa cells (not shown), we prepared an adenovirus-mediated reporter system. As shown in Fig. 7A, weak enhancement of CRE activity by forskolin was observed in the control HeLa cells (LacZ). Coinfection with the LKB1-adenovirus (LKB1) repressed basal CRE activity to one-tenth of its original level within 24 h. At this time point, forskolin was unable to substantially induce CRE activity. At 48–72 h post infection, however, a large induction of the CRE activity by forskolin was observed. In addition, forskolin induced NR4A2 mRNA in LKB1 expressing HeLa cells (Fig. 7B).
To investigate whether the impaired regulation of CRE-dependent transcription in HeLa cells resulted from the dysfunction of the overall phosphorylation cascades in the SIK–TORC system or the particular combination of SIK- and TORC isoforms, the levels of proteins and the phosphorylation of individual isoforms were examined. As shown in Fig. 7C, the level of SIK1 protein was elevated slightly by forskolin in control cells (LacZ). When the LKB1-adenovirus was infected, the basal level of SIK1 decreased significantly, and the level was elevated prominently by forskolin. These results agreed with the fact that SIK1 gene expression depended on its own CREs . In vitro kinase assays using the SIK1 protein purified by immunoprecipitation indicated that overexpression of LKB1 restored SIK1 kinase activity in HeLa cells (see the panel indicated by 32P-ATP. As expected, Thr182 was not phosphorylated in the LKB1 nonexpressing HeLa cells. Because sensitivity of the anti-(phospho-Thr182 IgG) was less than that of the anti-(SIK1 IgG), we could not discuss whether Thr182 was phosphorylated in the LKB1-expressing cells when cells were not stimulated with forskolin (third lane from the left). After forskolin treatment, however, the level of SIK1 protein had risen sufficiently so that we were able to detect phospho-Thr182 (the final lane).
In the case of SIK2 (Fig. 7D), the protein level was not influenced by overexpression of LKB1. The levels of kinase activity and phospho-Thr175 were elevated in the LKB1-expressing HeLa cells. By contrast, the protein level of SIK3 increased in LKB1-expressing HeLa cells, and restoration of the kinase activity and phosphorylation at Thr163 also occurred (Fig. 7E).
Next, we examined the phosphorylation of TORCs in the same way (Fig. 7F). Similarly to the cases for HEK293 cells, overexpression of LKB1 restored phospho-/dephospho regulation of TORCs in HeLa cells. We describe briefly here the results of TORC1. A small part of the TORC1 population had already been phosphorylated in control HeLa cells (lanes indicated by LacZ), and LKB1 enhanced the phosphorylation of TORC1 (third lane from the left). Forskolin treatment stimulated the dephosphorylation of TORC1 a little (final lane). Results in Figs 7F and 1C suggest that multiple cascades might be operating differentially in the phosphorylation of TORC1, and these cascades may be classified into three categories, LKB independent, SIK independent and SIK dependent.
Finally, we assayed the level of CREB phosphorylation (Fig. 7G). Forskolin-induced phosphorylation of CREB at Ser133 was evident in LKB1-overexpressing HeLa cells. These results suggested that LKB1 could modulate the actions of all participants in the SIK–TORC–CREB cascade.
SIK activity restores the regulation of CRE-dependent transcription in HeLa cells
To obtain direct evidence of SIK-mediated phosphorylation of TORC in HeLa cells, the constitutive active SIK3 mutant was overexpressed in this cell line. As shown in Fig. 8A, the constitutive active SIK3 mutant, T163E, could phosphorylate TORC2 even in the absence of LKB1. Moreover, the T163E mutant could recover the forskolin-dependent induction of CRE activity without LKB1 (Fig. 8B).
Finally, it should be noted that the forskolin-induced CREB phosphorylation at Ser133 was restored by overexpression of SIK3 (T163E) (Fig. 8C). These observations suggested that restoration of SIK activity might be sufficient for repairing impaired CREB-dependent transcription in HeLa cells.
LKB1-mediated phosphorylation of SIK1 at Thr182 enhances phosphorylation at Ser577
We noticed a large discrepancy between the increase in total SIK1 activity and the decrease in the level of TORC phosphorylation in forskolin-treated LKB1-expressing HeLa cells (Fig. 7C,F). In this regard, we found a similar case when CREB was activated by PKA; PKA also phosphorylated SIK1 at Ser577, which diminished SIK1-mediated cytoplasmic retention of TORC2 . Interestingly, phosphorylation at Ser587 of SIK2 (corresponding to Ser577 of SIK1) was not obvious in LKB1 nontransformed HeLa cells (Fig. 7D, lower). To compare the specific level of SIK1 phosphorylation at Ser577 with that at Thr182 and the kinase activity, GST-fusion SIK1 was overexpressed in HeLa cells. As shown in Fig. 9A, Ser577 was not phosphorylated in control HeLa cells and was apparently less phosphorylated by forskolin treatment (indicated by LacZ). Overexpression of LKB1 induced phosphorylation at Ser577, and its level was significantly elevated after forskolin treatment (indicated by LKB1), suggesting that phospho-Ser577 might be the result of an autophosphorylation of SIK1. Other indicators, such as phospho-Thr182 and kinase activities, depended on LKB1, but not on the forskolin treatment.
When Ser577 is phosphorylated, the phospho-SIK1 moves to the cytoplasm. Using this property, we examined the LKB1-initiated autophosphorylation of SIK1 at Ser577. As shown in Fig. 9B, in control HeLa cells (–), GFP–SIK1 was localized only in the nucleus. When GFP–SIK1 was coexpressed with LKB1 or PKA, part of the SIK1 population moved to the cytoplasm. LKB1-induced nuclear export of SIK1 was abolished by the T182A substitution (Fig. 9C). Substitution at Ser577 completely inhibited both LKB1- and PKA-induced nuclear export of SIK1.
Finally, we tested the level of TORC phosphorylation using wild-type and S577A-SIK1 (Fig. 9D). In COS-7 cells, the Ser577 mutant SIK1 phosphorylated GST-fusion TORC2 more efficiently than the wild-type. These observations suggested that the PKA-phosphorylatable Ser577 also acted as the autophosphorylation site, and that phospho-Ser577 might be a critical modulator of the TORC phosphorylation activity of SIK1. In this context, LKB1 might also play important roles in the attenuation step of the phosphorylation of TORC. The SIK3 T163E-mutant having the additional mutation at Ser493, equivalent to S577A of SIK1, also suggested the importance of the Ser phosphorylation (Fig. 8B).
Inhibition of kinase cascades activates CRE-dependent transcription constitutively
The constitutive activation of CRE-dependent transcription in HeLa cells (Fig. 6A,B) might be due to inactivation of the phosphorylation cascades from LKB1 to TORCs, suggesting, paradoxically, that inhibition of the kinase cascades could mimic impaired CREB regulation even in LKB1-positive cells. To test this possibility, we performed CRE-reporter assays in HEK293 cells in the presence of various kinase inhibitors. As shown in Fig. 10A, no specific kinase inhibitor could activate the CRE. However, staurosporine (STS; 10 nm), a nonspecific kinase inhibitor , induced CRE reporter activity to levels as high as forskolin. Moreover, staurosporine upregulated transcription of the NR4A2 gene to a level higher than that elevated by forskolin treatment (Fig. 10B).
Staurosporine has been classified as a PKC inhibitor (Fig. S1A). However, another PKC-specific inhibitor, bisindolylmaleimide I (Bis) did not induce any CRE reporter activity in our assay system (Fig. 10A), suggesting that PKC might not be the kinase responsible for staurosporine-induced CRE activity.
To investigate the phosphorylation status of TORC and CREB in staurosporine-treated cells, GST-tagged TORC2 and endogenous CREB were examined in COS-7 cells (Fig. 10C). Forskolin induced both the dephosphorylation at Ser171 and the decrease in the level of bound 14-3-3. It enhanced the phosphorylation of CREB at Ser133, of course. As expected, staurosporine completely inhibited the phosphorylation of TORC2 and did not enhance the phosphorylation of CREB.
Because staurosporine significantly blocked SIK1-mediated CRE repression (not shown), the efficacy of staurosporine on SIK1 was estimated by measuring its IC50 as regards the kinase activity. The in vitro IC50 was ∼ 0.15 nm (Fig. S1B). To evaluate SIK1 inhibition in vivo, the difference between forskolin-induced CRE-reporter activity and its activity in the presence of SIK1 S577A was used. An in vivo IC50 of staurosporine against the exogenously expressed SIK1 S577A mutant was ∼ 5.0 nm (Fig. S1C). These results suggested that the kinase activity of endogenous SIK might be inhibited by staurosporine at a dose as low as that against PKC.
To examine whether staurosporine-induced dephosphorylation of TORC2 was accompanied by its nuclear accumulation, GFP–TORC2 was expressed in HeLa cells in the presence or absence of the LKB1–SIK cascades, and the cells were treated with staurosporine (Fig. 10D). Staurosporine inhibited the nuclear export of GFP–TORC2 in all the cases tested. These results, however, might indicate two possibilities, namely that staurosporine either inhibited SIKs directly or blocked the upstream cascades of SIKs, including LKB1.
To clarify this point, COS-7 cells that had been expressing GST-tagged SIK1 were treated with staurosporine and GST–SIK1 protein was purified (Fig. 10E). SIK1 enzyme purified from the staurosporine-treated cells was phosphorylated at Thr182 but did not show any kinase activities. Because Thr172 of AMPKα1 (corresponding to Thr182 of SIK1) is also phosphorylated by LKB1, we performed the same experiment using GST–AMPKα1 (lower left). Neither the phosphorylation level at Thr172 nor the kinase activity of AMPKα1 was affected by staurosporine treatment. Forskolin treatment did not alter the levels of Thr phosphorylation or the kinase activity of SIK1 and AMPKα1. Similar experiments using other SIKs were also performed (right). Staurosporine inhibited the kinase activity of SIK2 and SIK3 in cultured cells. These results suggested that staurosporine inhibited the kinase activities of SIKs directly without blocking the upstream cascades.
Inhibition of SIK may be a direct cause of staurosporine-induced CREB activation
If the major cause of the staurosporine-induced dephosphorylation of TORC is the direct inhibition of SIKs, CRE activity may be re-repressed by introducing a staurosporine-resistant mutation into SIK. A model of staurosporine accommodated in the ATP-binding pockets of SIKs suggested possible interactions between three oxygen atoms, each derived from three glutamic acid residues, and the nitrogen atoms of staurosporine. Two of the three oxygens were constituents of peptide bonds, whereas the third was in a gamma-carboxyl group. We focused our attention on the glutamic acid residue whose gamma-carboxyl group interacted with staurosporine, and decided to replace it with an aspartic acid residue (Fig. 11A shows the case of SIK2). Thus mutants SIK1 (E110D), SIK2 (E103D) and SIK3 (E91D) were prepared. The kinase activity of SIK1 (E110) and SIK3 (E91) was lower than those of their parents (not shown), and therefore these mutants did not seem suitable for the analysis of staurosporine action. Mutant SIK2 (E103D), however, showed kinase activity comparable with the parent and the activity was not influenced by the addition of staurosporine in vitro (Fig. 11B). SIK2 (E103D) in the reporter assay systems repressed forskolin-induced CRE activity as strongly as the wild-type (Fig. 11C, left). When staurosporine was added to this assay system, the SIK2 (E103D)-dependent repression was less influenced than the wild-type (right). These results suggested that staurosporine may inhibit SIK directly, which might result in the dephosphorylation of TORC followed by the constitutive activation of CRE-dependent transcription.
AMPK might not regulate the CREB activity directly
Recently, Shaw et al.  showed that LKB1 downregulated the expression of gluconeogenic genes in the liver, possibly via activation of AMPK. Koo et al.  also showed that AMPK possibly phosphorylated TORC2 and blocked its nuclear accumulation in the liver. However, the degree to which an AMPK agonist enhanced the phosphorylation of TORC2 was lower than that by overexpressed SIKs. Moreover, it had not been shown whether CRE was the element responsible for the AMPK agonists-induced downregulation of gluconeogenic gene expression. Finally, a low dose of staurosporine (10 nm) could induce CRE-dependent transcription without significant effects on AMPK in living cells (Fig. 10E; also see in vitro IC50 in Fig. S1B). To examine the potential of AMPK-related kinases for CREB-regulatory activity, we performed reporter assays using a CRE reporter and a Gal4–TORC2 reporter. As shown in Fig. 12A, overexpression of the kinase domains (amino acids 1–312) of AMPKα1 and -α2 did not repress the reporter activities derived from the CRE reporter plasmid. This was also the case when Gal4–TORC2 was used in the reporter system (Fig. 12B). All enzymes were expressed as active kinases in COS-7 cells and could phosphorylate the TORC2 peptide in vitro (Fig. 12C).
Because the signal transduction of AMPKα is accelerated by the interaction with β and γ subunits, the full-length AMPKα subunit was expressed in the CRE reporter assay in the presence or absence of β and γ subunits. As shown in Fig. 12D, no combination of AMPKαs/β1/γ1 had a significant effect on CRE activity. Exogenously expressed AMPKα may be active, as it was highly phosphorylated at Thr172 (Fig. 12E) . Moreover, although endogenous AMPKα may have a molecular mass of ∼ 65 kDa, anti-(phospho-Thr172 IgG) did not react with bands around 65 kDa, suggesting that the level of the exogenously expressed AMPK might be much higher than that of endogenous enzyme in our system. These results suggest that the potential of AMPK as the regulator of CRE-dependent transcription might be lower than those of SIKs in living cells.
Finally, we would like to briefly mention MARK4. Because the kinase domains of SIKs are highly conserved with those of four other kinases that belong to the MARK subfamily, we tested the possibility of MARKs as CREB regulators by using one MARK, MARK4. The kinase domain of MARK4 also showed CREB-repression activity. In addition, the kinase activity of the GST-fusion MARK4 was inhibited by 10 nm staurosporine in living COS-7 cells (not shown).
LKB1 is the kinase responsible for causing Peutz–Jeghers syndrome  and has been reported to activate AMPK-related kinases by phosphorylating Thr residues in the A-loops . Because patients with Peutz–Jeghers syndrome are at a high risk of developing cancers of epithelial tissue origin, LKB1 has been suggested as a tumor suppressor kinase .
SIK1, a member of the AMPK family, was shown to be a downstream kinase of LKB1, with its Thr182 being phosphorylated by LKB1 . Meanwhile, SIK, when overexpressed in cells, was shown to inhibit CREB activity by phosphorylating, and repressing, CREB-specific coactivators, TORCs. TORC1 has been independently identified as a candidate for inducers of salivary gland tumors and it has been called mucoepidermoid carcinoma translocated 1 (MECT1) . The genomic rearrangement of t(11;19), which is often associated with mucoepidermoid carcinoma, produces a fusion protein that has the N-terminal CREB binding region (1–42 amino acids) of TORC1/MECT1 and the transcriptional activation domain of other transcription factor, Mastermind-like gene family 2 (MAML2). The resultant chimeric protein, MECT1–MAML2, binds to CREB, activates CRE-mediated transcription  and induces foci formation in RK3E cells . Because the MECT1–MAML2 fusion protein lacks Ser167 of TORC1/MECT1, the SIK phosphorylation site (Fig. 1A), we speculate that loss of the SIK-mediated repression of the MECT1–MAML2 activity might cause the uncontrolled induction of CRE-dependent transcriptions, which might result in tumor formation. This is also supported indirectly by a recent report that CREB has carcinogenic potential, when its cellular level is kept high .
The question regarding whether the constitutive activation of CREB is one of the causes of Peutz–Jeghers syndrome remains. Loss of one allele of the LKB1 gene is sufficient to induce Peutz–Jeghers syndrome [30–32], which enhances the frequency of polyp formation. However, the polyps in Peutz–Jeghers syndrome show resistance to transformation, and the progression of carcinogenesis requires complete loss of the LKB1 gene . However, neither knockdown of the LKB1 protein, using an siRNA technique, nor overexpression of LKB1 affected CREB activity in HEK293 cells (not shown). This was also supported by the result that overexpression of LKB1 in COS cells did not alter the intracellular distribution of GFP–TORC2 (Fig. 3A). Therefore, we suppose that the constitutive activation of CREB seen in HeLa cells may be due to complete loss of the LKB1 protein, and, thus, the impaired CREB regulation is unable to account for the pathogenesis of Peutz–Jeghers syndrome caused by the haploinsufficiency of the LKB1 allele.
Recent studies [20,34] have suggested similar properties for the A-loops of SIK1 and SIK3 (also called by Qsk). Levels of kinase activities of SIK1 (T182E) and SIK3 (T163E) were indistinguishable from those of wild-type enzymes phosphorylated by LKB1 [20,21]. A phospho-amino acid binding protein, 14-3-3, was shown to associate with phospho-Thr in the A-loop of SIK1 and SIK3, which resulted in enhancing their kinase activities . Association of 14-3-3 with the A-loop of SIK2 (QIK) was reported as uncertain. At least under our assay conditions, however, SIK3 (T163E), but neither SIK1 (T182E) nor SIK2 (T175E), showed kinase activity as high as its wild-type enzyme (Fig. 4). It remains to be examined whether 14-3-3 could bind to the glutamic acid of SIK3 (T163E), but not to those of SIK1 (T182E) or SIK2 (T175E), although to our knowledge no report has shown glutamic acids acting as acceptors for 14-3-3. In addition, Al-Hakim et al. suggested that the binding of 14-3-3 to the A-loop of SIK1 might induce the cytoplasmic localization of SIK1. However, as we show here, phosphorylation at Ser577, which occurred either by the action of PKA or by autophosphorylation, might cause the cytoplasmic localization of SIK1 (Fig. 9B,C). In this context, phosphorylation at Thr182 in the A-loop might be a prerequisite step for the cytoplasmic localization of SIK1.
To summarize, although CREB is believed to be active when it is phosphorylated by kinases cascades, this study gives a new insight that CREB has a constitutive active potency, which is also repressed by kinase cascades involving LKB1 and SIKs. In this study, however, to restore the regulated activation of CREB, we used overexpression systems, which might not reflect physiological levels. To elucidate the contribution of individual TORC isoforms and TORC kinases, including SIKs, to CRE-dependent transcription in living cells, it would be helpful to develop strategies using the combination of RNAi. But when the total number of the TORC kinases, their cross-talks and feedback regulations are considered, these strategies would require a quite formidable effort. At present, therefore, our strategy to use staurosporine or LKB1-defective cells is the only practical method to extract the constitutive active potency of CREB without classical agonists. Further elucidation of individual isoforms of SIK and TORC will be needed.
Cells, chemicals and antibodies
HeLa S9 cells were purchased from the Health Sciences Research Resource Bank (Osaka, Japan) and used as HeLa cells in this study. Because the round shape of HeLa S9 cells was not suitable for cytochemical analyses, flat cells were re-screened and used for the cytochemistry. 293A cells were from Invitrogen (Carlsbad, CA) and used as HEK293 cells in this study. COS-7 cells were in a collection of our laboratory. We used HEK293 cells and COS-7 cells as LKB1-positive cells. Both originated from kidney fibroblasts and showed a good response to forskolin; therefore, we used these cells for reporter assays. Because HEK293 cells are human cells, we used this cell line for mRNA analyses using real-time PCR. Because COS-7 cells have an SV40-ori which amplifies plasmids episomally, we used this cell line for the overexpression of SIKs or TORC2. In addition, HEK293 cells barely attached onto cover slips. Therefore, we used COS-7 cells for cytochemical analyses.
4′,6-Diamidino-2-phenylindole (DAPI) dilactate was purchased from Molecular Probes (Eugene, OR), staurosporine, H89, KN93, bisindolylmaleimide, PD98059, Wortmannin, SB203580, KT5823 and Hypericin were from Calbiochem (San Diego, CA) and forskolin was from Sigma-Aldrich (St. Louis, MO). Anti-CREB IgGs were from Cell Signaling Technology (Beverly, MA). An anti-(LKB1 IgG) was from Santa Cruz (Santa Cruz, CA). An anti-hemagglutinin (HA)-tagged IgG was from Roche (Indianapolis, IN). An anti-(phospho-AMPKα) serum (pT172) was purchased from Cell Signaling (Danvers, MA).
An anti-(SIK1 IgG) was purified using the SIK1 peptide (499–776). An anti-SIK3 serum was raised against a human SIK3 peptide (1106–1263). A TORC2 antiserum was raised against the mouse TORC2 peptide (563–699), and the specific IgG was purified using the same peptide. TORC1- and TORC3-antisera were raised against human TORC1 peptide (551–650) and TORC3 peptide (480–619), respectively. Because these antisera cross-reacted with both TORC1 and TORC3, the IgG was purified from the anti-TORC1 serum using the TORC3 peptide. The resultant anti-(TORC1/3 IgG) could recognize both TORC1 and TORC3 with equal efficiency. The above peptides were prepared as GST-fusion peptides in E. coli using a pGEX-6P3 expression vector and cleaved from GST by PreScission Protease (Amersham Biosciences). A phospho-Thr182 SIK1 antiserum was raised against a keyhole limpet hemocyanin-conjugated peptide of KSGELLApTWCGSPPY (human SIK2 A-loop; pT means phospho-Thr). The specific IgG was purified with an affinity column immobilized with a peptide of KPGEPLSpTWCGSPPY (SIK1 A-loop). The phospho-Thr182 SIK2 cross-reacted with all phospho-Thr residues of SIK1-3. Supplementary material (Fig. S1D) shows the specificity of the antiphospho-T182 IgG. The anti-(phospho-SIK1) (pSer577) serum was as described in Takemori et al., and the anti-(phospho-TORC2) (pSer171) serum was as in Screaton et al. .
The cDNAs for rat AMPKs have been described previously . Human SIK3 and MARK4 cDNAs (KIAA0999 and KIAA1860, respectively) were gifts from Kazusa DNA Institute (Chiba, Japan). The KIAA0999 clone had an extra 5′-UTR that might come from a 5′-upstream gene [human ankyrin repeat and MYBD domain containing 1 mRNA (Accession no. BC033495)]. Because this 5′-UTR was not found in a mouse SIK3 mRNA (Accession no. NP_081774), we supposed that the initial methionine codon (ATG) might be at 438–440 of the KIAA0999 clone (Accession no. BAA76843). As the result, SIK3 is likely to contain 1263 amino acids.
The mammalian expression vector, pCMVsport6, containing full-length hTORC1 (IMAGE: 4938995), mTORC2 (IMAGE: 5345301) and hTORC3 (IMAGE: 6470060) cDNAs, were purchased from Invitrogen. Site-directed mutagenesis was carried out using a kit, Quick change mutagenesis (Stratagene) to create restriction sites. The resultant cDNAs were digested with BamHI, BglII, or SpeI and NotI (in the multicloning site), and the TORCs cDNA fragments were ligated into the BamHI–NotI or the SalI–NotI site of a GFP expression vector pEGFP-C for cytochemical studies, a pENTR-1A plasmid (Invitrogen) for adenovirus preparation and a pM vector (Gal4-fusion) for reporter assays.
Mouse LKB1 cDNA fragment was amplified from total RNA of mouse adrenal tumor Y1 cells by RT-PCR using primers based on the mouse LKB1 cDNA sequence (Accession no. AF145287). The sequences of the primers were 5′-TTTACTAGTATGGACGTGGCGGACCCCGAG and 5′-TTTGCGGCCGCTCACTGCTGCTTGCAGGCCGAG.
To prepare adenovirus vectors, cDNA fragments and reporter DNA(s) in the pENTR-1A vector were transferred onto pAd/CMV/V5-DEST or pAd/PL/V5-DEST adenovirus-DNA using the Gateway system (Invitrogen). pAd-lacZ was used as a standard to determine the titers of individual viruses. When HeLa cells were infected with adenovirus at > 50 moi (multiplicity of infection for HEK293 cells), some of cells died within 72 h. Therefore, the total moi of the adenovirus was adjusted to 30 plus 6 for reporters (three for the CRE-fLuc adenovirus and three for the internal standard TK-rLuc adenovirus).
Immunoprecipitation, in vitro kinase assay, reporter assay, cytochemical analysis and quantitative PCR analysis (real-time PCR)
Immunoprecipitation and in vitro kinase assay was performed as described previously . The methods for reporter assays have been described previously . To introduce plasmids into the cells, we used LipofectAMINE 2000 (Invitrogen)/Escort V (Sigma-Aldrich) mixture (1:1) was used in this study. Luciferase activities were measured by using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI). For the CRE-reporter assay, HEK293 and HeLa cells (2 × 104 per well) were transformed with SIKs expression plasmids (pIRES, pEBG, pTarget or pIRES: 0.1–0.2 µg), the CRE-luciferase reporter [pTAL-CRE or its empty reporter (pTAL): 0.2 µg], the PKA expression plasmid (pIRES-PKA: 0.1 µg) and 0.03 µg of an internal standard Renilla luciferase vector [pRL-SV40 for standard assays or phRL-TK(Int–) for assays with TORC expression plasmids]. To measure the transcriptional activities of TORCs, HEK293 cells were transformed with Gal4 DNA binding domain-linked TORC expression vectors (pM-TORCs: 0.05 µg) and the 5 × GAL4-luciferase reporter plasmid (pTAL-5 × GAL4: 0.2 µg).
For fluorescent cytochemical analyses, cells were cultured on poly L-lysine coated coverslips (18-mm) (Matsunami Co. LTD, Tokyo, Japan) using a 12-well dish. COS-7 and HeLa cells (< 1 × 104) were transformed with 0.2–0.5 µg of expression vectors for GFP-TORCs in the presence or absence of pEBG-SIKs or -LKB1 (∼0.3 µg). After 16 h, the cells were fixed with 1 mL of 4% paraformaldehyde dissolved in NaCl/Pi for 15 min, stained with DAPI (1 ng mL−1 in 0.01% Triton X-100/NaCl/Pi) for 5 min, and then washed with NaCl/Pi four times. The cells on the coverslip were embedded onto a slide glass using 50% glycerol. On average, about 80% of cells showed detectable GFP-TORCs signals in independent duplicate experiments.
Total RNA was extracted using the RNeasy kit (Qiagen, Valencia, CA). cDNAs were prepared by reverse transcription (RT) from one microgram total RNA using Superscript III and random primers (Invitrogen). One-hundredth of the RT products and standard plasmids were subjected to real-time PCR analyses using the IQ SYBR Green Supermix (Bio-Rad). The PCR program included 10 min of denaturation at 95 °C and then 40 cycles at 95 °C for 15 s, 58 °C for 30 s and 72 °C for 30 s.
The specific primers were used (TORC1-F: 5′-CACCTGGCTCCTCTCCACA; and TORC1-R: 5′-AGCTGCTGCTCCAGAGACA, TORC2-F: 5′-TGACCTCACCAACCTGCACT; and TORC2-R: 5′-GTGAGTCATGGTGTGGGT?hangover 0>CA, TORC3-F: 5′-CAACATCCCAGCTGCTATGA; and TORC3-R: 5′-GATGCGTTGGGAACAGATGT, NR4A2-F: 5′-ACCAGAACTACGTGGCCACTA; and NR4A2-R: 5′-TCGAAGCGCATCTGGCAACTA, 36B4-F: 5′-TGTGTGTCTGCAGATCGGGT; and 36B4-R: 5′-TGGATCAGCCAGGAAGGCCT).
Modeling of a three-dimensional structure of SIK1 with staurosporine
The three-dimensional structure of the SIK2 kinase domain was modeled from the primary sequence of the rat SIK2 kinase domain using SWISS MODEL (http://swissmodel.expasy.org/SWISS-MODEL.html)  which predicts and constructs the three-dimensional structure according to the sequence homologies. Using insight ii software (Accerlys, San Diego, CA), the model of SIK2 was superimposed on Chk1 from Chk1–staurosporine complex, the structure of which had already been determined (PDBID 1NVR) , and the staurosporine was placed at the corresponding position of the SIK1 to the staurosporine-binding site on the Chk1. After energy minimization of the complex model, we obtained a final model of the SIK2–staurosporine complex. We next calculated the electrostatic potential on the molecular surface  of SIK2 using the program SCB developed by Nakamura & Nishida . The illustration was prepared using molfeat v2.2 (FiatLux Co. Ltd, Tokyo, Japan).
We are grateful to Dr Marc R. Montminy (Salk Institute, USA), Dr K. Morohashi (National Institute for Basic Biology, Japan) and Dr T. Sugawara (Hokkaido University, Japan) for providing us the EXV-1, CYP11A and StAR reporter plasmids, respectively. These plasmids were used in the original version of this manuscript. We thank Dr D. Carling (MRC Clinical Sciences Centre, UK) for critical evaluation of our data. The cDNAs for SIK3 and MARK4 were gifts from the Kazusa DNA Institute, Japan. This study was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology and Ministry of Health, Labor and Welfare Japan, and a grant from the Technology Research Grant Program in ‘03’ from the New Energy and Industrial Technology Development Organization (NEDO) of Japan.