The Tumor Suppressor Kinase LKB1 Activates the Downstream Kinases SIK2 and SIK3 to Stimulate Nuclear Export of Class IIa Histone Deacetylases

Background: HDAC4, -5, -7, and -9 possess conserved motifs for phosphorylation-dependent 14-3-3 binding.

Results: SIK2 and SIK3 phosphorylate the deacetylases at the motifs to stimulate 14-3-3 binding.

Conclusion: The tumor suppressor kinase LKB1 activates SIK2 and SIK3 to promote trafficking of class IIa HDACs.

Significance: This study indicates that LKB1-dependent SIK activation is an important module upstream from class IIa HDACs.

Abstract

Histone deacetylases 4 (HDAC4), -5, -7, and -9 form class IIa within the HDAC superfamily and regulate diverse physiological and pathological cellular programs. With conserved motifs for phosphorylation-dependent 14-3-3 binding, these deacetylases serve as novel signal transducers that are able to modulate histone acetylation and gene expression in response to extracellular cues. Here, we report that in a PKA-sensitive manner the tumor suppressor kinase LKB1 acts through salt-inducible kinase 2 (SIK2) and SIK3 to promote nucleocytoplasmic trafficking of class IIa HDACs. Both SIK2 and SIK3 phosphorylate the deacetylases at the conserved motifs and stimulate 14-3-3 binding. SIK2 activates MEF2-dependent transcription and relieves repression of myogenesis by the deacetylases. Distinct from SIK2, SIK3 induces nuclear export of the deacetylases independent of kinase activity and 14-3-3 binding. These findings highlight the difference among members of the SIK family and indicate that LKB1-dependent SIK activation constitutes an important signaling module upstream from class IIa deacetylases for regulating cellular programs controlled by MEF2 and other transcription factors.

Introduction

Histone deacetylases have emerged as an important family of enzymes deacetylating not only histones but also many non-histone proteins. These enzymes are grouped into different classes based on sequence similarity to the yeast prototypes (1–4). With their C-terminal parts homologous to the deacetylase domain of yeast Hda1, class IIa HDACs⁴ (HDAC4, -5, -7, and -9) function as co-repressors of different transcription factors, most notably members of the myocyte enhancer factor 2 (MEF2) family (2, 4, 5). These HDACs are crucial regulators of a variety of MEF2-dependent physiological processes, including the development of skeletal muscle, bone, the vascular system, and regulatory T cells, as well as different pathological conditions such as cardiac hypertrophy, neurodegeneration, and leukemogenesis (5, 6). In addition, the HDAC4 mutation is directly linked to the brachydactyly mental retardation syndrome in patients with bone malformation and mental retardation, whereas murine Hdac7 has been identified as a new oncogene (7, 8). Therefore, class IIa deacetylases are important regulators of various physiological and pathological programs.

Each class IIa deacetylase possesses a unique N-terminal extension harboring a MEF2-binding site as well as three or four conserved motifs for serine phosphorylation and 14-3-3 binding (2, 4, 9). This binding promotes the cytoplasmic localization of class IIa HDACs through a combination of nuclear export sequence activation and nuclear localization signal inhibition (9–14), which then control the activity of MEF2. Thus, MEF2-dependent transcriptional repression is associated with dephosphorylation and nuclear localization of class IIa HDACs, and vice versa. A number of protein kinases have been identified to phosphorylate these conserved 14-3-3-binding motifs, including Ca²⁺/calmodulin-dependent protein kinases (CaMKs) (11, 15–19) and protein kinase D (9, 20–22). Stimuli that activate these kinases, e.g. intracellular [Ca²⁺] increase (23, 24) and VEGF treatment (25, 26), induce class IIa HDAC phosphorylation and nuclear export, leading to derepression of MEF2-dependent transcription.

Additional kinases have been reported for class IIa HDAC regulation, including AMP-activated protein kinase (AMPK) (27), microtubule affinity-regulating kinases (MARK2 and -3) (28, 29), and salt-inducible kinase 1 (SIK1) (30, 31). All four are activated by LKB1 (32, 33), so the interesting question is whether LKB1 itself regulates trafficking of class IIa HDACs. Mutations in the LKB1 gene play a causal role in Peutz-Jeghers syndrome (34, 35), and this kinase has emerged as a major tumor suppressor of lung cancer and other malignancies (36, 37). Downstream from LKB1, numerous studies have focused on AMPKs and established that through AMPK, LKB1 controls energy metabolism, mammalian target of rapamycin signaling, and protein translation (33, 34). Three recent reports reveal that mouse Lkb1 regulates the hematopoietic stem cell compartment in an AMPK-independent manner (38–40), reiterating the importance of other members of this kinase family. In mammals, a total of 14 kinases, including AMPKα1, AMPKα2, and 12 related ones, are downstream from and activated by LKB1 (32, 34), so we systematically investigated roles of these kinases in class IIa HDAC regulation, ultimately focusing on the SIK subfamily.

This subfamily is conserved from Caenorhabditis elegans to humans, and there are three members in mammals (41, 42). SIK1 was initially identified as a protein up-regulated in the adrenal glands of rats fed a high salt diet as well as in PC12 cells upon neuronal depolarization (43, 44). Chicken SIK1 was also cloned as a product induced by a winged helix transcription factor (45). Two SIK1 paralogs, SIK2 and SIK3 (also known as QSK), were found by database search based on sequence similarity (41). The three kinases share the catalytic domain located at the N-terminal part but show divergence in other regions. For example, SIK3 possesses a unique long C-terminal domain. SIK2 is highly expressed in adipose tissues (46), but SIK3 is ubiquitously expressed (41). Although SIK1 regulates cardiomyogenesis (47) and cancer metastasis (48), SIK2 is required for mitotic spindle formation (48) and insulin signaling (49). SIK2 phosphorylates CRTC2 and induces its 14-3-3 binding and nuclear export, inhibiting cAMP-response element-binding protein activity (50, 51). PKA phosphorylates SIK2 and reverses this effect (50, 51). Through SIK2 and CRTC2, LKB1 plays a key role in hepatic gluconeogenesis (52, 53). In addition to CRTC2, SIK2 phosphorylates p300 and regulates carbohydrate-responsive element-binding protein-dependent transcription in hepatic steatosis (54). Such a regulatory scheme remains to be established for SIK1 and SIK3, but the latter can promote CRTC2 to localize to the cytoplasm (52). Thus, compared with SIK1 and SIK2, much less is known about SIK3.

Here, we show that LKB1 activates SIK2 and SIK3 to phosphorylate class IIa HDACs and promote their cytoplasmic localization. Under the same experimental conditions, SIK1 is unable to do so. Different from SIK2, SIK3 also possesses unique properties, such as the ability to promote class IIa HDAC export independent of its kinase activity and to stimulate cytoplasmic localization of constitutively nuclear mutants of HDAC4 and HDAC7, highlighting the difference among the SIK family members. Moreover, PKA counteracts LKB1, SIK2, and SIK3 to inhibit the nuclear export of class IIa HDACs. These results thus identify the deacetylases as novel targets downstream from the LKB1-SIK2/3 signaling module and directly link this module to regulation of various MEF2-dependent cellular programs.

MATERIALS AND METHODS

Cell Culture

HEK293, HeLa, H1299, and C2C12 cells were maintained in Dulbecco's modified Eagle's medium (DMEM, Invitrogen), whereas A549 lung cancer cells were maintained in F-12/DMEM (1:1). All media were supplemented with 10% fetal bovine serum (FBS, Sigma), except for C2C12 myoblast differentiation, which entailed switching to DMEM containing 5% horse serum (Sigma). HEK293FT cells were maintained in DMEM containing 10% heat-inactivated FBS, plus 400 μg/ml G418, nonessential amino acids, pyruvate, and glutamate (FT medium). Lkb1^−/− mouse embryonic fibroblasts (MEFs) stably expressing GFP or FLAG-LKB1 were gifts from R. Jones (McGill University) (55). The fibroblasts were cultured in DMEM supplemented with 10% FBS.

Plasmids and Antibodies

Constructs for GFP- and FLAG-tagged wild-type (WT) HDAC4, its triple mutant (TM) (S246A/S467/S632A), two single point mutants (V1066A and L1062A), as well as an expression plasmid for HA-14-3-3β have been described (14, 56). Both V1066A and L1062A were prepared on the HDAC4 mutant 1–1069, which lacks a short acidic stretch at the C-terminal end (14). An expression plasmid for mouse HDAC5 was provided by Khochbin and co-workers (57) and used to engineer GFP- and FLAG-tagged HDAC5 expression constructs by use of derivatives of pEGFP-C2 (BD Biosciences) and pcDNA3.1 (Invitrogen), respectively. The expression plasmid for GFP-HDAC7 was derived from a pEGFP-C2 derivative and an expression plasmid for mouse HDAC7 (GenBank^TM accession number AF207749). A cDNA clone for a shorter isoform of mouse HDAC7 was also obtained from H. Y. Kao (Case Western Reserve University) (see Refs. 58, 59). The truncation mutant 1–591 of HDAC7 was generated in an accidental subcloning experiment where a 0.84-kb HindIII/ApaI fragment from the coding sequence was inadvertently inserted after the HindIII site encoding Lys-590 and Leu-591, so this mutant contains the N-terminal 591 residues of murine HDAC7 and unrelated residues until an in-frame stop codon downstream. The construct for GFP-HDAC9 was engineered from a pEGFP-C2 derivative by linking an oligoduplex consisting of DAC103 (5′-AA TTC ATG CAC AGT AT-3′) and DAC104 (5′-GA TCA TAC TGT GCA TG-3′) to the coding sequence of human HDAC9. cDNA clones for SIK1, SIK2, SIK3, LKB1, CRTC2, and PKA were from Open Biosystems. The coding sequences of SIK1, SIK2, and SIK3 were subcloned into pcDNA3.1 derivatives with and without an HA tag. The sequence for PKA was subcloned into an HA tag-containing pcDNA3.1 derivative, and that for LKB1 was directly cloned into pcDNA3.1. Constructs for SIK1 and a kinase-dead mutant were also provided by Hiroshi Takemori (Osaka University) (46). The expression plasmid for mCherry-CRTC2 was constructed from a pEGFP-C2 derivative, in which the GFP coding sequence was replaced by an mCherry coding sequence. SIK2 and SIK3 mutants were generated by PCR-mediated site-directed mutagenesis using the Pfu polymerase system (Fermentas). All mutants were verified by automatic sequencing. The constitutively active CaMKIV mutant (60) and the 3×MEF2-luc reporter (61) have been described. The MCK-luc reporter construct was a gift from J. Nalbantoglu (Montreal Neurological Institute) (62). MCK-GFP was from J. McDermott (York University) (63). pLove-GFP lentivirus plasmid was from Addgene, and lentivirus plasmids bearing shRNAs targeting mouse LKB1 were from Open Biosystems as follows: shLKB1–2 (5′-CCGGG CCAAG TGAAG AAGGA AATTC TCGAG AATTT CCTTC TTCAC GTTGG CTTTT T-3′) and shLKB1–5 (5′-CCGGC ATCTA CACTC AGGAC TTCAC CTCGA GGTGA AGTCC TGAGT GTAGA TGTTT TT-3′). The LKB1 coding sequence was subcloned into pENTR11 (Invitrogen) and recombined with the pLenti6/V5-DEST lentivirus vector (Invitrogen) following the manufacturer's instructions to yield the pLenti6-LKB1 lentivirus construct.

An anti-phospho-Ser-246 (HDAC4) polyclonal rabbit antibody was prepared by immunization of rabbits with a peptide corresponding to residues 241–251 of human HDAC4 (LRKTApSEPNLKC, where pS is phospho-Ser). The Cys was added for conjugation to mcKLH (Pierce) for rabbit injection or to SulfoLink gel (Pierce) for affinity purification of the antibody from the rabbit antiserum. Specificity of the antiserum and the affinity-purified antibody toward phospho-Ser-246 versus nonphosphorylated peptides was confirmed by dot blotting (supplemental Fig. S1). An anti-HDAC4 polyclonal antibody was affinity-purified from the rabbit antisera previously developed (56). The affinity-purified antibody was highly selective for HDAC4 and did not cross-react with the other class IIa HDACs (supplemental Fig. S2).

Cell Transfection

For Western blotting (including immunoprecipitates), 2 × 10⁵ HEK293 or HeLa cells were plated in 6-cm plates, and transfections were performed with 10 μl of Superfect reagent (Qiagen) and 5 μg of total DNA. For fluorescence microscopy and reporter gene assays, 4 × 10⁴ HEK293, HeLa, or C2C12 cells were plated per well in 12-well plates. For HEK293 and HeLa cells, transfections were performed with 3 μl of Superfect (Qiagen) and 1.5 μg of total DNA or with 1.5 μl of Lipofectamine 2000 (Invitrogen) for HEK293 cells. For C2C12 cells, transfections were performed with 3 μl of Lipofectamine 2000 and 1.5 μg of total DNA. For HeLa cells, Superfect was used for fluorescence microscopy, but Lipofectamine 2000 was used for reporter gene assays. Further analyses were conducted 24–48 h post-transfection.

Co-immunoprecipitation

To determine 14-3-3 binding to HDAC4/5, expression plasmids for FLAG-tagged HDACs were transfected into HEK293 cells along with constructs for HA-SIK2 or HA-SIK3, and in some experiments an HA-14-3-3β construct was co-transfected. Transfections were performed in 6-cm plates using 10 μl of Superfect reagent and 5 μg of total DNA. About 48 h post-transfection, cells were washed twice with ice-cold phosphate-buffered saline (PBS) and lysed in 0.5 ml of buffer K (20 mm sodium phosphate, pH 7.0, 150 mm KCl, 30 mm sodium pyrophosphate, 0.1% Nonidet P-40, 5 mm EDTA, 10 mm NaF, 0.1 mm Na₃VO₄, 25 mm β-glycerophosphate, and protease inhibitors). For affinity purification of FLAG-tagged proteins, 200 μl of extracts were mixed with 10 μl of M2 agarose beads (Sigma) for rotation at 4 °C for 2 h. Following four washes with buffer K, bound proteins were eluted with a mixture containing 25 μl of buffer K and 2 μl of FLAG peptide (Sigma).

Immunoblotting

After addition of the 3× SDS sample buffer, soluble extracts and eluted immunoprecipitates were boiled for further separation by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked in PBS/Tween 20 (PBS-T) with 20% horse serum for 1 h at room temperature and then incubated overnight at 4 °C with anti-HA (Covance), anti-FLAG (Sigma), anti-14-3-3β (H-8; Santa Cruz Biotechnology), anti-phospho-HDAC4 (S246), anti-LKB1 (Ley 37D/G6, Santa Cruz Biotechnology), or anti-α-tubulin (Sigma) antibody. For detection of endogenous HDAC4, phospho-HDAC4 (Ser-246), LKB1, α-tubulin, and 5% skim milk in PBS-T was used as the blocking solution. Membranes were washed with PBS-T (six times for 8 min each) and incubated in the appropriate secondary antibody conjugated to HRP for 1 h at room temperature. After another set of washes (six times for 8 min each) with PBS-T, membranes were then incubated twice with PBS (5 min each) and once for 5 min with the Supersignal enhanced chemiluminescent solution (Pierce).

In Vitro Kinase Assays

HEK293T cells were transfected with the construct for FLAG-tagged HDAC5, SIK2, mutant K49M, SIK3, or mutant K37M. FLAG-tagged proteins were purified as above, except the final washing and the elution were performed with the kinase buffer (50 mm Tris, pH 7.5, 0.1 mm EGTA, 1 mm DTT, 10 mm NaF, 0.1 mm Na₃VO₄, and protease inhibitors) instead of buffer K. In vitro kinase (IVK) assays were performed in a total volume of 40 μl including 20 μl of IVK buffer (50 mm Tris, pH 7.5, 5 mm magnesium acetate, 5 mm MnCl₂, 0.1 mm EGTA, 1 mm DTT, 10 mm NaF, 0.1 mm Na₃VO₄, and protease inhibitors) along with 150 ng of FLAG-HDAC5, 10 ng of FLAG kinase (SIK2 WT, SIK2 K49M, SIK3 WT, or SIK3 K37M), and 100 pmol of ATP. Reactions were carried out at 30 °C for 1 h and stopped by the addition of 20 μl of an ice-cold 3× SDS sample buffer. Samples were boiled for separation by SDS-PAGE and transferred to nitrocellulose membranes, which were blocked in PBS, 0.15% Tween 20 (PBS-T) with 20% horse serum for 1 h at room temperature. For the kinase input blots, 100 ng of each FLAG-tagged kinase was resolved by SDS-PAGE. Membranes were incubated with anti-FLAG (Sigma, 1:1000) and anti-phospho-Ser-246 (HDAC4, 1:20,000) antibodies overnight at 4 °C for further processing as described above.

Reporter Gene Assays

MEF2-dependent luciferase assays were performed with the 3×MEF2-luc reporter as described previously (61). A CMV-driven β-galactosidase reporter construct was co-transfected as the internal control.

Myogenesis

Myogenesis was performed as described previously (61, 63). Plasmids encoding the MCK-GFP reporter and the mCherry vector were co-transfected into C2C12 cells in the presence or absence of the constitutively active mutant of SIK2 (S587A, HA-tagged) or the kinase-dead mutant (HA-K49M). After a 48-h recovery in the growth medium (GM) containing DMEM and 10% FBS, the myoblasts were incubated in DMEM containing 5% horse serum (differentiation medium, DM) for 48 h. After 48 h in DM, myotubes were detected by live green fluorescence microscopy. Transfection efficiency was monitored by visualization of mCherry expression via live red fluorescence microscopy. To quantify myogenesis, transfection was also performed with an MCK-luciferase (MCK-luc) reporter plasmid with a β-galactosidase expression plasmid for monitoring the transfection efficiency. Forty eight hours after switching from GM to DM, cells were lysed in the luciferase lysis buffer (0.2 m potassium phosphate (17 mm KH₂PO₄ + 183 mm K₂HPO₄), 1% Triton X-100, 2 mm DTT, 10% glycerol), and the luciferase activities were assessed on a luminometer. The values were divided by β-galactosidase values, and a value of 1.0 was arbitrarily assigned to the reporter-only condition to facilitate normalization among three independent experiments.

Lentivirus Production and Infection

Lentivirus plasmids were transfected into HEK293FT cells using the reverse transfection method. Briefly, 10 μg of lentivirus plasmid, along with the helper vectors psPAX2 (6.5 μg) and pMD2.G (3.5 μg), were incubated with 45 μl of Lipofectamine 2000 according to the manufacturer's instructions and then added to 10-cm plates. HEK293FT cells (8 × 10⁶) were seeded onto the DNA/Lipofectamine mixture. Throughout the transfection and virus production phases, HEK293FT cells were grown in FT medium in the absence of antibiotics. Virus-containing media were collected from the plates on the 2nd and 3rd days after transfection, filtered, pooled, and frozen in aliquots at −80 °C. For infection, H1299 or A549 cells were seeded in 6-well plates (1 × 10⁶) in the FT cell medium containing 6 μg/μl Polybrene but no antibiotics. Cells were split and re-plated in 10-cm dishes 48 h after infection, and after an additional 48 h, cells were harvested for immunoblotting if necessary.

Immunofluorescence Microscopy

Lkb1^−/− MEFs stably infected with either empty vector or FLAG-LKB1 were seeded on glass coverslips. Once the cells had reached ∼60% confluency, coverslips were washed twice with PBS, fixed with 5% paraformaldehyde for 20 min, and then washed three times with PBS. Cells were then permeabilized with 0.2% Triton X-100 for 10 min, followed by washing three times with 100 mm glycine (5 min each). After blocking with IF buffer (PBS with 0.2% Triton X-100 and 0.05% Tween 20) containing 2% BSA for 45 min, cells were incubated overnight at 4 °C with anti-HDAC4 antibody (1:500). After primary antibody incubation, cells were rinsed three times with IF buffer (5 min) and then incubated in AlexaFluor 568 goat anti-rabbit IgG secondary antibody (Invitrogen, Molecular Probes) (1:1000) at room temperature for 45 min, followed by washing three times with IF buffer (5 min). Cells were then counterstained with DAPI for 5 min at room temperature, washed twice with Nanopure H₂O (5 min each), and mounted for microscopic analysis.

Statistical Analysis

Data were presented as means ± S.E. or S.D. as specified. For experiments with more than two conditions, one-way analysis of variance was performed with Bonferroni post hoc tests. For experiments with only two conditions, unpaired two-tailed Student's t tests were performed with Prism 5 (Graphpad). p < 0.05 was considered to be statistically significant.

RESULTS

SIK2 and SIK3 Induce Nuclear Export of Class IIa HDACs

Four known class IIa HDAC kinases belong to the AMPK-related superfamily (27–30, 42), and 13 different kinases are downstream from and activated by the LKB1 master kinase (32, 34, 35). So, two intriguing issues are as follows: (i) whether other members of the AMPK family play a role in class IIa HDAC regulation and (ii) whether LKB1 activates the AMPK family of kinases to regulate class IIa HDAC trafficking. To address these issues, we systematically investigated the effect of the AMPK-related kinases on class IIa HDAC trafficking. For this, we tested their ability to increase cytoplasmic localization of class IIa HDACs upon ectopic expression in HEK293 cells. Specifically, we transfected HEK293 cells with expression plasmids for GFP-tagged HDAC4, -5, -7, or -9 together with empty vector or expression plasmids for several AMPK-related kinases. Western blotting was performed with an anti-GFP antibody to verify that GFP-HDACs were correctly expressed (supplemental Fig. S3). Except for some effects of the trimeric AMPK complex (supplemental Fig. S4), expression of MARK1, MARK2, MARK3/C-TAK1, NUAK1/ARK5, NUAK2/SNARK, SNRK, or NIM1 (all related to AMPK and, except NIM1, activated by the LKB1 master kinase (32, 34)) in HEK293 cells had no significant effects on the subcellular localization of HDAC5 (data not shown). Although it was reported to induce nuclear export of HDAC5 in C2C12 cells (30, 31), we failed to detect any major effects of SIK1 expression on class IIa HDAC subcellular localization in HEK293 cells (Fig. 1A, panels v–viii), except for a relatively small change in the proportion of cells with GFP-HDAC9 displaying a pancellular localization pattern versus a nuclear pattern (Fig. 2). In the same experimental settings, expression of SIK1 caused a major increase in cytoplasmic accumulation of CRTC2 (supplemental Fig. S5A), indicating that the kinase expressed was functional. Thus, this kinase did not function in our experimental system.

FIGURE 1.

Effects of SIKs on subcellular localization of class IIa HDACs. A, HEK293 cells were transiently transfected with expression plasmids for GFP-HDAC4, -5, -7, and -9, in combination with pcDNA3.1 or an expression plasmid encoding SIK1, SIK2, or SIK3 as indicated. The next day, fluorophore-tagged proteins were visualized via live green fluorescence microscopy. Cytoplasmic aggregation was also observed with GFP-HDAC4 in the absence of SIK3, but the biological significance remains unclear. B, HEK293 cells were transiently transfected with the expression plasmid for GFP-HDAC5 or GFP-HDAC9 together with either the kinase-dead mutants of SIK2 (K49M) or SIK3 (K37M) as indicated. GFP-tagged proteins were visualized as in A.

FIGURE 2.

Quantification of the subcellular localization of class IIa HDACs. Transient transfection and live green fluorescence microscopy were performed as depicted in Fig. 1A. At least 100 cells were counted for each condition over three independent experiments. Data are presented as mean ± S.E. (error bars).

In contrast to SIK1, ectopic expression of SIK2 or SIK3 caused dramatic relocalization of GFP-HDAC5 (Fig. 1A, panels x versus ii and panels xiv versus ii) and GFP-HDAC9 (Fig. 1A, panels xii versus iv and panels xvi versus iv) from the nucleus to the cytoplasm. Quantification validated this finding (Fig. 2, B and D). In the cases of HDAC4 (Fig. 1A, panel i) and HDAC7 (Fig. 1A, panel iii), which displayed a cytoplasmic and/or pancellular localization pattern in HEK293 cells, the effect was necessarily minor compared with that observed with HDAC5 and HDAC9. Nevertheless, expression of SIK2 and SIK3 resulted in a more completely cytoplasmic localization pattern compared with empty vector control (Figs. 1A and 2, A and C). The difference that we observed between SIK1 and SIK2/3 was intriguing. These three kinases, however, behaved similarly in regulating CRTC2 in both HEK293 and HeLa cells (supplemental Fig. S5), indicating that the functional difference among the kinases is somehow specific to the HDACs.

At the sequence level, the three kinases display high similarity within the N-terminal kinase domain, whereas the C-terminal part is highly divergent (supplemental Fig. S6). Interestingly, SIK3 is quite conserved from fish to human even in the C-terminal domain and is more homologous than the corresponding regions of SIK1 and SIK2 (supplemental Fig. S7). Notably, fly SIK3 is also quite divergent from the mammalian counterpart (supplemental Fig. S7). Therefore, even though they are members of the same family, these kinases possess sequence characteristics that may underlie their functional differences. Because their effects on class IIa HDAC localization were so dramatic, we chose to focus on SIK2 and SIK3 for subsequent experiments.

To investigate whether the cytoplasmic accumulation of class IIa HDACs in response to ectopic expression of SIK2/3 was due to stimulation of nuclear export or inhibition of nuclear import, we treated cells with leptomycin B, an inhibitor of the CRM1 nuclear export receptor (64). Because it inhibits CRM1-dependent nuclear export of class IIa HDACs without having any effects on nuclear import (17, 56), leptomycin B represents a useful tool to separate these two opposing processes. As shown in supplemental Fig. S8, SIK2/3-mediated accumulation of GFP-HDAC5 in the cytoplasm was abolished by leptomycin B treatment, indicating that the kinases stimulate nuclear export of HDAC5.

To elucidate the molecular mechanisms underlying the observed nuclear export, we analyzed a SIK2 mutant lacking the C-terminal regulatory domain (supplemental Fig. S9A). Although less dramatically than full-length SIK2 (Figs. 1 and 2), this mutant was able to promote export of SIK2 (supplemental Fig. S9, B and C), suggesting that the kinase domain is sufficient. To determine whether the kinase activity is necessary, we engineered kinase-dead mutants of SIK2 and SIK3 by substituting a key lysine residue in the catalytic center with methionine (K49M for SIK2 and K37M for SIK3). These mutations have been shown to render SIK2 and SIK3 catalytically inactive (46, 52). As described below (Fig. 3), both mutants behaved as expected and did not phosphorylate HDAC5. Mutant K49M was unable to induce nuclear export of HDAC5 or HDAC9, indicating that the kinase activity of SIK2 is necessary for the effect (Fig. 1B). Surprisingly, mutant K37M retained the ability to cause nuclear export of the HDACs, indicating that the kinase activity of SIK3 is dispensable for promoting the nuclear export (Fig. 1B). These results reiterate differences among the three kinases.

FIGURE 3.

SIK2 and SIK3 phosphorylate HDAC4/5 and stimulate 14-3-3 binding. A, IVK assay was performed with FLAG-tagged proteins affinity-purified from HEK293T cells. Reactions were performed with the indicated combinations of purified proteins (150 ng of FLAG-HDAC5 and 10 ng of FLAG-tagged kinase) and 100 pmol of ATP in IVK buffer for 1 h at 30 °C. Reactions were stopped by the addition of an SDS sample buffer and then boiled for 5 min prior to SDS-PAGE. Western blotting was performed using antibodies directed against the proteins indicated to the right of the blots. For the input blot, 100 ng of purified FLAG-SIK2 or FLAG-SIK3 WT or kinase-dead was subjected to SDS-PAGE and Western blotting using the anti-FLAG antibody. B, HEK293 cells were transiently transfected with expression plasmids encoding the FLAG-tagged wild-type (WT) (lanes 3–5) or triple mutant (TM) of HDAC4 (lanes 6 and 7). In this mutant, three phosphorylatable serine residues that serve as 14-3-3-binding sites were mutated to alanine (S246A/S467A/S632A). endog., endogenous. In some cases, an expression plasmid for HA-tagged WT (lanes 2, 4, and 7) or K49M of SIK2 (lane 5) was co-transfected with the FLAG-HDAC4 construct. Forty eight hours post-transfection, cells were harvested in buffer K, and FLAG-tagged proteins were immunoprecipitated on M2-agarose beads and eluted with FLAG peptide. Soluble extracts (Input) and immunoprecipitates (IP) were separated by SDS-PAGE, and immunoblotting was performed with antibodies against the proteins indicated to the right side of the blots. C, same as in B except that FLAG-HDAC5 was expressed alone (lane 1) or together with either HA-SIK2 (lane 2) or HA-SIK3 (lane 3) as indicated. D, same as in B except that HA-14-3-3β was expressed alone (lane 1) or together with FLAG-HDAC5 (lane 2), or FLAG-HDAC5 and HA-SIK2 (lane 3).

SIK2 and SIK3 Stimulate Class IIa HDAC Phosphorylation and 14-3-3 Binding

The observation that the kinase activity of SIK2 is crucial for its ability to induce nuclear export of HDAC5 and HDAC9 prompted us to examine whether SIK2 and SIK3 can phosphorylate class IIa HDACs. For this, we performed in vitro kinase assays with purified FLAG-tagged wild-type and mutants of SIK2 and SIK3. Phosphorylation was detected by Western blotting with the antibody that recognized phospho-Ser-246 of HDAC4 and the corresponding phosphorylated residue (Ser-259) of HDAC5 (supplemental Fig. S1) (4). When wild-type SIK2 or SIK3 was incubated with purified FLAG-HDAC5 in the presence of ATP, Ser-259 phosphorylation of HDAC5 increased (Fig. 3A). In contrast, no phosphorylation was observed when mutant K49M or K37M was incubated with FLAG-HDAC5 (Fig. 3A). Consistent with this, unlike wild-type SIK2, the K49M mutant failed to stimulate phosphorylation of HDAC4 at Ser-246 in HEK293 cells (Fig. 3B, top two panels). These results underscore further the importance of the SIK2 kinase activity.

Nuclear export of class IIa HDACs is controlled by phosphorylation of conserved serine residues and association with 14-3-3 proteins, which use these conserved phosphoserine residues as docking sites (10, 11, 17, 56, 65, 66). We thus performed immunoprecipitation to test whether expression of the two kinases increases class IIa HDAC phosphorylation and 14-3-3 binding. For this, we co-expressed HA-tagged SIK2 or SIK3 along with FLAG-HDAC4 or FLAG-HDAC5 and performed Western blotting with the anti-phospho-Ser-246 antibody (supplemental Fig. S1). The phosphorylation increased when SIK2 or SIK3 was co-expressed along with HDAC4 or HDAC5 (Fig. 3, B, 4th lane, and C, 2nd and 3rd lanes). This phosphorylation corresponded to increased 14-3-3 binding. Similar results were obtained either with endogenous 14-3-3 proteins (Fig. 3B, 4th lane) or HA-tagged 14-3-3β (Fig. 3D, 3rd lane). SIK3 expression also stimulated 14-3-3 binding (data not shown). Moreover, no increase in 14-3-3 binding was observed when mutant K49M was expressed (Fig. 3B, 5th lane). These results demonstrate that SIK2 and SIK3 are two class IIa HDAC kinases able to induce phosphorylation in vitro (Fig. 3A) and in vivo (Fig. 3, B–D).

SIK3 Stimulates Nuclear Export of Class IIa HDACs through a Unique Mechanism

HDAC4 possesses a nuclear export sequence (NES) at its C-terminal end (Fig. 4A) (12, 14). To elucidate the potential role of this NES in the nuclear export of HDAC4 induced by SIK2 and SIK3, we analyzed how these two kinases may promote nuclear export of three HDAC4 mutants as follows: 1–1040 lacks the NES entirely, whereas V1066A and L1062A only contain point mutations within the NES (Fig. 4A). As reported (14), these three mutants were predominantly nuclear (Fig. 4B). Interestingly, although SIK2 failed to alter the nuclear localization of these mutants, SIK3 induced their nuclear export (Fig. 4B). Like SIK3, the well known class IIa kinase CaMKIV also stimulated the export of these three HDAC4 mutants (Fig. 4B). We also analyzed the HDAC4 truncation mutant 1–669, which lacks the C-terminal part, including the NES and the deacetylase domain (Fig. 4A). Interestingly, SIK3 promoted export of this mutant 1–669, and neither SIK2 nor CAMKIV was able to do so (Fig. 4B). Western blotting revealed that this truncation mutant and 1–1040 were properly expressed in the absence or presence of SIK3 (supplemental Fig. S10A). Therefore, SIK3 appeared to possess a unique ability to promote nuclear export of HDAC4.

FIGURE 4.

Unique properties of SIK3-mediated nuclear export of HDAC4 and HDAC7. A, schematic diagram for each HDAC4 mutant compared with the wild type. NLS, nuclear localization signal; DAC, deacetylase domain. B, HEK293 cells were transfected with expression plasmids for the indicated GFP-HDAC4 mutants along with pcDNA3.1 or the expression construct for SIK2, SIK3, or CaMKIV. Proteins were visualized using green fluorescence microscopy (right panels). Note that the TM, 1–1040, and 1–669 form nuclear dots, with the former two displaying cytoplasmic dots in the presence of SIK3; the significance of the nuclear and cytoplasmic dots remains unclear. C, top panel, schematic diagram of HDAC4 protein with phosphorylatable serines (Ser-246, Ser-467, and Ser-632) and hypothetical phosphorylatable serine (1036) are highlighted. Bottom panel, sequence alignment of three known phosphorylation sites (Ser-246, Ser-467, and Ser-632) and fourth hypothetical phosphorylation site (Ser-1036), are highlighted in a red box, plus the surrounding sequence. Identical amino acids are highlighted in gray boxes. D, HEK293 cells were transfected with the plasmid expressing GFP-tagged S246A/S467A/S632A/S1036A (quadruple mutant, QM) together with pcDNA3.1 or a SIK3 expression plasmid. Fluorescence microscopy was performed as in A. E, schematic diagram for HDAC7 and its truncation mutant 1–591. The domains are labeled as in A. The serine residue (S) corresponding to the fourth 14-3-3-binding site is highlighted in green. F, same as in B except that the construct for GFP(1–591) was transfected with the indicated kinase expression plasmids.

In addition to the NES, HDAC4 possesses three known motifs to mediate 14-3-3 binding (Fig. 4A), and these motifs are crucial for the nuclear export of HDAC4. Related to this, previous studies have established that due to deficient 14-3-3 binding, the HDAC4 TM (S246A/S467A/S632A, Fig. 4A) is resistant to nuclear export, as shown for the equivalent mutants of HDAC5/7/9 (11, 15, 18, 22). As expected, CaMKIV did not stimulate nuclear export of this mutant (Fig. 4B). We next tested whether it is also the case for SIK2 and SIK3. Western blotting revealed that this HDAC4 mutant was properly expressed in the absence or presence of SIK2 or SIK3 (supplemental Fig. S10A). Consistent with the lack of 14-3-3 binding to the mutant (Fig. 3B), SIK2 was unable to induce the nuclear export (Fig. 4B). Surprisingly, unlike SIK2 and CaMKIV, SIK3 induced the nuclear export (Fig. 4B). Together, these results underscore the intrinsic difference that these kinases have in using sequence elements to promote nuclear export of HDAC4.

While addressing the unique property of SIK3, we noticed a fourth potential 14-3-3-binding site at Ser-1036 of HDAC4 (Fig. 4C), so we constructed the quadruple mutant (S246A/S467A/S632A/S1036A) to investigate whether this site affects the ability of SIK3 to induce nuclear export of HDAC4. However, SIK3 induced nuclear export of this mutant as well (Fig. 4D), indicating that Ser-1036 of HDAC4 is not responsible for nuclear export induced by SIK3.

We then asked whether the unique property of SIK3 is specific to HDAC4 or whether other class IIa HDACs can be similarly regulated. An HDAC7 mutant, 1–591, displayed the same shuttling dynamics as the HDAC4 NES mutants, as both SIK3 and CaMKIV, but not SIK2, could induce nuclear export of this mutant (Fig. 4F). Compared with the HDAC4 mutant 1–669, which was unresponsive to CaMKIV (Fig. 4B), this HDAC7 mutant possesses an extra 14-3-3-binding site at the nuclear localization signal (Fig. 4E), suggesting the importance of this site and the intrinsic difference between HDAC4 and HDAC7 in response to CaMKIV signaling. Together, the above results indicate that SIK3 has properties distinct from SIK2 and other class IIa HDAC kinases such as CaMKIV.

To gain further mechanistic insights, we expressed the three SIK kinases as mCherry fusion proteins and compared their subcellular localization. Although SIK1 formed predominantly nuclear dots, both SIK2 and SIK3 were exclusively cytoplasmic (Fig. 5A). This is consistent with the subcellular localization reported for the kinases in other cells (41, 46, 67). Western blotting indicated that the kinases were properly expressed (supplemental Fig. S10B, bottom panel). As expected, only the SIK3 fusion protein promoted export of the TM mutant of HDAC4 (Fig. 5B). Although this may explain the inability of SIK1 in promoting class IIa export in our experimental system, it does not account for the difference between SIK2 and SIK3. At the sequence level, SIK3 is quite different from SIK2 at the C-terminal domain (supplemental Figs. S6 and S7), so we engineered two truncation mutants corresponding to the N- and C-terminal domains of SIK3 (Fig. 5C). Compared with the mutant SIK3N, the C-terminal mutant SIK3N dramatically promoted nuclear export of the TM mutant of HDAC4 and, more importantly, co-localized very nicely with it in cytoplasmic vesicles (Fig. 5, B and C), indicating a novel mechanism through which the C-terminal domain of SIK3 stimulates nuclear export of HDAC4 through co-trafficking.

FIGURE 5.

Effect of SIKs on subcellular localization of the HDAC4 triple mutant. A, expression plasmids for mCherry or its kinase fusion proteins were separately transfected into HEK293 cells. Representative live red fluorescence images were taken 18 h post-transfection. Note the distinct subcellular localization that the three kinases display. B, expression plasmid for GFP-TM was transiently transfected into HEK293 cells with an expression vector for mCherry or its kinase fusion proteins as indicated. Representative live green and red fluorescence microscopic images were taken 18 h post-transfection. Note the striking co-localization of SIK3C with the HDAC4 mutant in cytoplasmic vesicles. C, schematic representation of human SIK3 and two truncation mutants. The kinase domain is boxed in green, and the C-terminal regulatory region is denoted as a solid line. Within the C-terminal domain, four conserved PKA phosphorylation sites (supplemental Fig. S7) are shown in two small boxes marked with the letter S or T. The plus signs denote approximate impact on the subcellular localization of the HDAC4 mutant.

Differential Effects of SIK2 and SIK3 on MEF2-dependent Transcription

Many of the biological effects that class IIa HDACs have can be attributed to their inhibition of transcription dependent on the MEF2 family of transcription factors (5, 68). Thus, we used MEF2-dependent transcription as a read-out of the class IIa HDAC activity regulated by SIK2 and SIK3. For this, we first performed reporter gene assays with the 3×MEF2-luc construct, in which luciferase expression was driven by three MEF2-binding sequence elements. When HDAC4, HDAC5, HDAC7, or HDAC9 (Fig. 6, A–D) was co-expressed with MEF2D, there was a major suppression of MEF2 transcriptional activity. However, upon the addition of SIK2, this suppression was abrogated. This result was expected because SIK2 caused cytoplasmic localization of these HDACs, rendering them unable to repress the MEF2D transcriptional activity. Moreover, both the triple mutants of HDAC4 and HDAC7 were resistant to SIK2-mediated derepression of MEF2 activity (Fig. 6, A and C), further supporting the role of 14-3-3 binding and nuclear export in SIK2-mediated regulation of class IIa HDACs.

FIGURE 6.

SIK2 but not SIK3 derepresses MEF2 transcriptional activity. A–D, HEK293 cells were transfected with the luciferase construct 3×MEF2-Luc along with HA-SIK2 or HA-SIK3 plasmids as denoted by lanes 2 or 3, respectively. A CMV-driven β-galactosidase reporter construct was co-transfected as the internal control. Where indicated, an MEF2D expression plasmid was co-transfected along with expression plasmids for wild-type or triple mutants (TM) of HDAC4 (A), HDAC5 (B), the wild-type or triple-mutant of HDAC7 (C), and HDAC9 (D). Western blotting shows equal expression of HA-SIK2 and HA-SIK3 (A, inset). The luciferase activity was measured with a luminometer and divided by the β-galactosidase activity to normalize the transfection efficiency. The value for the 1st column in all panels (reporter genes with no effector plasmids transfected) was arbitrarily assigned as 1.0. Data are presented as means ± S.E. (error bars) from three independent experiments. E, GFP-MEF2D expression plasmid was transfected into HEK293 cells along with pcDNA3.1 or an expression plasmid for SIK2 or SIK3 as indicated. Subcellular localization of GFP-MEF2D was determined by live green fluorescence microscopy. F and G, expression plasmid for GFP-HDAC4 (F) or GFP-HDAC5 (G) was transfected into HEK293 cells with a MEF2D expression plasmid in the presence of pcDNA3.1 or the expression plasmid for SIK2 or SIK3 as indicated. Subcellular localization of GFP-MEF2D was monitored by live green fluorescence microscopy.

In the case of SIK3, which also induced phosphorylation and nuclear export of class IIa HDACs (Figs. 1–3), there was no derepression of MEF2 activity (Fig. 6, A–D). The lack of effect with SIK3 was not due to nuclear exclusion of MEF2, as we saw no change in MEF2D localization following expression of SIK2 or SIK3 (Fig. 6E), nor was it due to MEF2-mediated trapping of class IIa HDACs in the nucleus. Expression of MEF2D did increase the nuclear localization of HDAC4 (Fig. 6F), as reported previously for MEF2C (14). Both SIK2 and SIK3 overcame this change of subcellular localization to induce nuclear export of HDAC4 (Fig. 6F) and HDAC5 (Fig. 6G). Thus, as expected for a class IIa HDAC kinase, SIK2 stimulated MEF2 transcriptional activity in the presence of these HDACs, whereas SIK3 unexpectedly lacked this ability, suggesting that nuclear export and derepression of transcription may not be strictly correlated. It will be interesting to investigate whether SIK3-mediated export plays a role in regulating other nuclear functions of class IIa HDACs or whether these HDACs have functions in the cytoplasm.

SIK2 Rescues Class IIa HDAC-mediated Repression of Myogenesis

Because of the well known role of MEF2 in skeletal muscle physiology (5, 68), we tested whether SIK2 also causes class IIa HDAC export and MEF2 derepression in C2C12 cells, a widely used mouse myoblast cell line (69). In contrast to its effect in HEK293 cells, wild-type SIK2 did not cause HDAC5 nuclear export or derepression of MEF2-dependent transcription in C2C12 cells (Fig. 7, A and B). We postulated that this is due to strong cAMP/PKA signaling in these cells (70), so we used the S587A mutant of SIK2, which is resistant to PKA-mediated phosphorylation (51). Indeed, this mutant was competent in inducing HDAC5 cytoplasmic localization in addition to derepressing MEF2 activity in C2C12 cells (Fig. 7, A and B, and supplemental Fig. S11A).

FIGURE 7.

SIK2 reverses class IIa HDAC-mediated inhibition of myogenesis. A, C2C12 cells were transiently transfected with a GFP-HDAC5 expression plasmid together with pcDNA3.1 or the expression plasmid encoding the wild-type or the S587A mutant of SIK2. Live green fluorescence microscopy was performed to monitor the subcellular localization of GFP-HDAC5. B, C2C12 cells were transfected with the 3×MEF2-luc reporter construct (containing three MEF2-binding sites upstream from the luciferase gene) together with the indicated plasmids as well as a CMV-driven β-galactosidase reporter construct as the internal control. The reporter activities were measured and processed as in Fig. 6, A–D. C, MCK promoter construct containing the GFP gene and an mCherry expression vector were co-transfected into C2C12 cells. Two days post-transfection, cells were shifted from GM (10% FBS) to DM (5% horse serum). Live green fluorescence microscopy was performed to monitor MCK promoter activity and myotube formation following 2 days in DM. D, C2C12 cells were transfected with a plasmid bearing the luciferase gene driven by the muscle creatine kinase promoter (MCK-luc) along with the indicated plasmids and a β-galactosidase expression vector. Two days post-transfection, cells were shifted from the growth medium to the differentiation medium, and luciferase assays were performed after 2 days in DM. *** denotes p < 0.001 versus the 1st bar.

To assess whether the kinase activity of SIK2 is relevant in a more physiological setting, we performed myogenesis assays in C2C12 cells, which readily form myotubes when switched from regular medium containing 10% FBS (growth medium, GM) to differentiation medium (DM) containing 5% horse serum. To visualize live myogenesis, we transfected C2C12 cells with a GFP construct driven by the MCK promoter (MCK-GFP). The MCK promoter contains a MEF2-binding site and is heavily dependent on MEF2 activity for activation in myogenic cells (71). This reporter gene provides a convenient read-out of MEF2 activity in the context of a native promoter fragment and allows the live visualization of myotubes due to green fluorescence of the expressed GFP protein (63). Consistent with its dependence on MEF2 activity, myotube formation was strongly repressed by co-transfection of HDAC4, whereas in the absence of HDAC4, MCK-GFP was highly expressed after 48 h in DM (Fig. 7C). This repression was rescued by co-expression of constitutively active SIK2 (S587A) but not by kinase-dead SIK2 (K49M) (Fig. 7C). To quantify this result, we repeated the same assay using a different reporter gene construct. In this assay, we used the MCK-Luc reporter construct, in which an MCK promoter fragment controls luciferase expression. Consistent with the results for the MCK-GFP experiment, the MCK-Luc quantification showed that expression of the S587A mutant but not the K49M mutant rescued HDAC4-mediated suppression of the MCK promoter activity during C2C12 differentiation (Fig. 7D). Thus, SIK2 rescued HDAC4-mediated repression of myogenesis, and this effect was associated with SIK2-induced phosphorylation-dependent nuclear export.

LKB1 Acts Upstream from SIK2 and SIK3

SIK2 and SIK3 are members of the AMPK-related kinase family, whose members are targets of the upstream LKB1 master kinase (34, 35). LKB1 phosphorylates AMPK-related kinases on a threonine residue in the activation loop, an event crucial for their kinase activities (32). Thus, we investigated whether LKB1 is necessary for SIK2/3 to act as class IIa HDAC kinases. To address this, we employed LKB1-deficient HeLa cells (72). In these cells, neither SIK2 nor SIK3 caused nuclear export of GFP-HDAC5 (Fig. 8A). This result is in stark contrast to what we showed with HEK293 cells, i.e. robust cytoplasmic accumulation of HDAC5 when co-transfected with SIK2 or SIK3 (Figs. 1 and 2). Compared with HEK293 cells, HeLa cells possess other genetic differences in addition to the lack of LKB1 expression, so we performed add-back experiments to assess the specificity of the observed effect. Although SIK2/3 did not cause nuclear export of HDAC5 in HeLa cells, this ability was restored upon ectopic expression of LKB1 (Fig. 8A and supplemental Fig. S11B). Similar to our observations with HEK293 cells, SIK1 failed to induce HDAC5 nuclear export even when LKB1 was co-expressed (Fig. 8A). In contrast, co-expression of SIK1 and LKB1 caused cytoplasmic localization of CRTC2 (supplemental Fig. S5B). Consistent with these results on subcellular localization, co-expression of LKB1 along with SIK2 was necessary for SIK2 to derepress MEF2 activity in HeLa cells (Fig. 8B). In addition, co-expression of LKB1 was needed for SIK2 to induce Ser-246 phosphorylation of endogenous HDAC4 in these cells (Fig. 8C). To a lesser extent, the same result was seen for phosphorylation of the corresponding residue (Ser-155) of endogenous HDAC7 (Fig. 8C). These results point to the importance of the upstream kinase LKB1 for the class IIa HDAC kinase activity of SIK2 and SIK3.

FIGURE 8.

LKB1 is required for SIK2/3-mediated class IIa HDAC nuclear export. A, HeLa cells were transiently transfected with the GFP-HDAC5 expression plasmid together with the indicated constructs. Th subcellular localization was monitored by live green fluorescence microscopy. B, HeLa cells were transfected with the 3×MEF2-luciferase construct together with the indicated expression plasmids. Reporter activities were measured and processed as in Fig. 6, A–D. C, HeLa cells were transfected with the HA-SIK2 construct (2nd and 3rd lanes) together with (3rd lane) or without (2nd lane) the LKB1 expression plasmid. Cells were harvested in buffer K and subjected to SDS-PAGE followed by Western blotting with antibodies against the proteins indicated on the right side of the blots. The position of the band representing either HDAC4 or HDAC7 is indicated to the right of the anti-phospho-Ser-246 blot (top).

PKA Reverses SIK2/3-dependent Nuclear Export of HDAC5

Previous reports with CRTC2 (51, 52) and our experiments in C2C12 cells suggest that SIK-induced HDAC5 cytoplasmic accumulation might be sensitive to PKA signaling, so we next sought to test this hypothesis. Like its reported effect on subcellular localization of CRTC2 (32, 52), expression of PKA prevented cytoplasmic localization of HDAC5 induced by SIK2/LKB1 and SIK3/LKB1 in HeLa cells (Fig. 8A). We also observed this effect in HEK293 cells without co-expression of LKB1 (data not shown). This result demonstrates that PKA exerts a dominant effect over SIK2/3. Moreover, it suggests that multiple upstream signals influence SIK2/3 activity and ultimately affect the downstream targets of SIK2/3, such as class IIa HDACs.

LKB1 Promotes Phosphorylation and Export of Endogenous Class IIa HDACs

Because of the importance of LKB1 in SIK2/3-induced nuclear export of class IIa HDACs, we next tested whether endogenous LKB1 expression is required for proper subcellular localization of class IIa HDACs. To address this issue, we employed complementary approaches as follows: knockdown or add-back of LKB1 and genetic deletion of the Lkb1 gene. We examined whether LKB1 expression regulates the HDAC4 Ser-246 phosphorylation status. For this, we used two lung cancer cell lines, one of which lacks ectopic expression of LKB1 (A549 cells) and the other of which expresses LKB1 (H1299 cells). Expression LKB1 in A549 cells induced Ser-246 phosphorylation of HDAC4 and Ser-155 phosphorylation of HDAC7 (Fig. 9A). Western blotting confirmed that LKB1 expression was only detectable in cells infected with an LKB1-expressing lentivirus (Fig. 9A). The opposite effect was observed when LKB1 was knocked down in cells that normally express it. Specifically, LKB1 was knocked down in H1299 cells using lentiviruses bearing shRNAs targeting two different sequences of human LKB1. LKB1 expression was virtually undetectable after infection with lentivirus for either shLKB1–2 or shLKB1–5, whereas a scrambled shRNA had minimal effects (Fig. 9B). Compared with the scrambled shRNA, both shRNAs led to a decrease in HDAC4 Ser-246 phosphorylation and, to a lesser extent, in HDAC7 Ser-155 phosphorylation (Fig. 9B).

FIGURE 9.

LKB1 regulates phosphorylation and export of endogenous HDAC4. A, A549 cells (LKB1-negative) were left untreated (NT, lane 1) or infected with lentiviruses containing the coding sequence for GFP (lane 2) or LKB1 (lane 3). Whole cell extracts were subject to SDS-PAGE followed by immunoblotting with antibodies against the indicated proteins. The position of the band representing either HDAC4 or HDAC7 is indicated to the right of the phospho-Ser-246 blot (top blot). The size of the indicated proteins in kDa is shown to the left of the blots. B, H1299 cells (LKB1-positive) were infected with lentivirus for a scrambled shRNA (SCR, lane 1) or shRNAs targeting LKB1 (LKB1–2, lane 2; LKB1–5, lane 3). Western blotting was performed as in A. C, soluble extracts were prepared as in A from Lkb1^−/− MEFs stably integrated with an empty vector (Vector Laboratories) or the same vector expressing FLAG-tagged LKB1 (LKB1) and Western blotting was performed as in A. D, immunofluorescence microscopy of Lkb1^−/− MEFs stably integrated with an empty vector (Vector Laboratories) or the same vector expressing FLAG-tagged LKB1 (LKB1). Cells were subject to indirect fluorescence microscopy with the affinity-purified anti-HDAC4 antibody (left panels) and counter-stained with DAPI to visualize the nuclei (middle panels). E, quantification of the subcellular localization of endogenous HDAC4. The percentage of cells with nuclear exclusion of HDAC4 was determined in three independent experiments. Data are presented as means ± S.E. (error bars). ** denotes p < 0.01 versus the vector group.

We also performed studies using Lkb1^−/− MEFs to complement the above two approaches. Ser-246 phosphorylation of HDAC4 was higher in Lkb1^−/− MEFs with stable ectopic expression of LKB1 than that in the same MEFs stably infected with the corresponding empty vector (Fig. 9C). This was consistent with the dependence of HDAC4 Ser-246 phosphorylation on LKB1 expression in the lentivirus experiments (Fig. 9, A and B). Albeit to a lesser extent, a similar effect was observed with Ser-155 phosphorylation of HDAC7 (Fig. 9C). Finally, we performed indirect immunofluorescence microscopy to examine the subcellular localization of endogenous HDAC4 in Lkb1^−/− MEFs. The anti-HDAC4 antibody was highly specific to HDAC4 and did not react with any other class IIa HDAC (supplemental Fig. S2). HDAC4 localization in Lkb1^−/− MEFs was predominantly pancellular, although it was more cytoplasmic in Lkb1^−/− MEFs with restored LKB1 expression (Fig. 6D). We quantified this effect by observing the number of cells displaying nuclear exclusion (a marked absence of staining in the nucleus) of HDAC4. A graph summarizing the data from three independent experiments shows an ∼2-fold increase in HDAC4 nuclear exclusion in Lkb1^−/− cells re-expressing LKB1 versus cells lacking LKB1 expression (Fig. 9E, 55% versus 27%, p = 0.0017). Together, these results demonstrate that LKB1 plays an important role in the phosphorylation and subcellular localization of endogenous HDAC4.

DISCUSSION

Because of their unique domain organization, HDAC4 and other class IIa members of the deacetylase family have emerged as novel signal transducers able to process different signaling cues to regulate histone deacetylation and gene expression in the nucleus (2–5). Two families of kinases, including CaMK (11, 15–19) and protein kinase D (20–22), have been well established to phosphorylate and stimulate cytoplasmic accumulation of these deacetylases. Here, we have identified SIK2 and SIK3 as two new class IIa HDAC kinases (Fig. 10). Both phosphorylated HDAC4 and HDAC5 in vitro and in vivo (Fig. 3). The phosphorylation promoted 14-3-3 binding to both HDACs (Fig. 3 and data not shown), induced their nuclear export (Figs. 1 and 2), and stimulated MEF2 transcriptional activity (Fig. 6). In a more biological setting, the constitutively active SIK2 mutant S587A reversed HDAC4-mediated inhibition of myogenesis (Fig. 7), a cell differentiation program in which MEF2 plays a critical role (5). Interestingly, SIK3 possessed unique characteristics not shared by SIK2, including the following: (i) the ability to induce nuclear export of HDAC5 and -9 independent of kinase activity (Fig. 1B), and (ii) the ability to induce nuclear export of HDAC4 and -7 mutants lacking three 14-3-3-binding sites and/or the nuclear export signal sequence (Fig. 4).

FIGURE 10.

LKB1-SIK2/3-class IIa HDAC pathway as a novel signaling module. A, schematic showing effects of the PKA-sensitive LKB1-SIK2/3 pathway on nucleocytoplasmic shuttling of class IIa HDACs. Although not illustrated here, HDAC phosphorylation and 14-3-3 binding can also occur in the nucleus. Although HDAC4, -5, and -9 each contain three 14-3-3-binding sites, HDAC7 possesses four; for simplicity, only three are shown here. The dotted line denote links that await formal demonstration or further investigation. B, schematic illustrating major targets downstream from LKB1 and the predominant physiological processes that they regulate. The 14 LKB1-activated kinases belong to several groups (32, 34) and lead to formation of different signaling branches downstream from LKB1. The AMPK branch has been most widely studied and plays a major role in metabolic control (33, 34, 55), but recent studies indicate that LKB1 regulates the hematopoietic stem cell compartment in an AMPK-independent manner (38–40), suggesting the importance of other groups. Downstream from the SIK group are three signaling axes, with the first two controlled by the CRTC2 and p300 transcriptional co-activators (52–54) and the third one being the SIK2/3-class IIa HDAC pathway characterized herein and in another study (78). Although only MEF2 is illustrated here, other transcription factors may also interact with class IIa HDACs and are subject to regulation by the LKB1-SIK2/3 axis. In addition, it remains to be investigated if this pathway regulates potential roles of class IIa HDACs in the cytoplasm. There is also evidence that LKB1 modulates cell polarity through amphid-defective (SAD) and microtubule-associated protein/microtubule affinity-regulation kinases (MARK) (34, 79, 80) and controls cell senescence (81) and adhesion (82) via NUAK1. CREB, cAMP-response element-binding protein.

In support of the functional differences among the three members of the SIK family, they only showed sequence conservation in their kinase domain, whereas their C-terminal domains are quite divergent (supplemental Fig. S6). Compared with SIK1 and SIK2, SIK3 is much larger due to its extended C-terminal domain (1311 residues versus 783 or 926 residues, supplemental Figs. S6 and S10B). In agreement with their differences at both the sequence and functional levels, these three kinases also displayed different subcellular localization (Fig. 5). Moreover, they form distinct multiprotein complexes with different sets of binding partners (67). Furthermore, genetic deletion of mouse Sik3 function cannot be compensated by Sik1 or Sik2 (73). All of these findings reiterate the differences among the three kinases.

The unique C-terminal extension of SIK3 (supplemental Figs. S6 and S7) (41) appeared to be responsible for some of the unique characteristics (Fig. 5). SIK3 is likely to control nuclear export of class IIa HDACs in kinase activity-dependent and -independent manners through the N-terminal kinase core and C-terminal regulatory domain, respectively. The relative contribution of these two mechanisms may be cell context-dependent, which may explain the puzzle that LKB1 is not required for SIK3 activation in HEK293 (Fig. 1B) but is required in HeLa cells (Fig. 8A). How LKB1 acts through both mechanisms is an interesting issue awaiting further investigation. Also puzzling is the inability of SIK3 to regulate MEF2-dependent transcription (Fig. 6). This may suggest that nuclear export of class IIa HDACs and derepression of MEF2-dependent gene expression are not strictly correlated. Alternatively, targets other than MEF2 are downstream from SIK3 and class IIa HDACs (Fig. 10B). Related to this, RT-PCR analysis revealed that LKB1 knockdown in H1299 or knock-out in MEFs did not affect expression of well established MEF2 targets (data not shown). It is also possible that class IIa HDACs may play a role in the cytoplasm and that SIK3 may regulate this cytoplasmic function (Fig. 10B).

Using gain-of-function and loss-of-function approaches, we demonstrated that LKB1 is a crucial activator upstream from SIK2 and SIK3 (Figs. 8 and 9). In HeLa cells, which do not express LKB1 (72), SIK2 and SIK3 failed to alter nuclear localization of HDAC5 unless LKB1 was co-expressed (Fig. 8A). This is consistent with reports that SIK2 and SIK3 rely on LKB1 for the kinase activity (32) and the ability to cause CRTC2 nuclear export (52). Moreover, phosphorylation of endogenous HDAC4 and HDAC7 increased upon ectopic LKB1 expression in A549 cells (Fig. 9A), and the phosphorylation decreased after LKB1 knockdown in H1299 cells (Fig. 9B). We also showed that in MEFs Lkb1 is required for phosphorylation of endogenous mouse Hdac4 and Hdac7 (Fig. 9C), as well as for cytoplasmic localization of endogenous mouse Hdac4 (Fig. 9, D and E). Together, these findings suggest a new model where, through SIK2 and SIK3, LKB1 regulates phosphorylation-dependent nuclear export of class IIa HDACs (Fig. 10A). It should be noted, however, that this does not mean LKB1 is the sole regulator upstream from SIK2/3. Of relevance, knockdown of LKB1 in HEK293 cells only had modest effects on the cytoplasmic localization of HDAC5 stimulated by SIK2 (data not shown), suggesting LKB1-independent activation of SIK2 in these cells.

The importance of LKB1 is consistent with reports that other LKB1-dependent kinases, including AMPK (27, 74), MARK2/3 (28, 29), and SIK1 (30, 31), serve as class IIa HDAC kinases. Although we could not detect major effects in our assay systems, these kinases may play important roles in other cellular contexts. Related to this, in our assay systems the effects of SIK2 varied from cell line to cell line. For example, SIK2 and SIK3 were active and did not require exogenous LKB1 in HEK293 cells (Figs. 1 and 2), whereas ectopic expression of this kinase was required to activate SIK2 and SIK3 in HeLa cells (Fig. 8). In C2C12 myoblasts, the effect of wild-type SIK2 was small, and LKB1 expression did not help (data not shown), but the S587A mutation activated SIK2 (Fig. 7). Therefore, the cellular state controls the activity of the LKB1-SIK signaling module. Whether this is true for other LKB1-activated kinases remains an interesting question awaiting further investigation. One challenge in the future is to map out the cellular contexts where SIKs and other kinases, such as CaMK and protein kinase D (2, 4), contribute to spatiotemporal regulation of class IIa HDACs. Such efforts will provide links to cellular signaling networks and shed light on how to target the regulation for therapeutic intervention.

As a negative regulator of SIKs, PKA prevented nuclear export of HDAC5 induced by expression of LKB1/SIK2/3 in HeLa cells (Fig. 8A), suggesting that the kinase activity of SIK2/3 is influenced by the competing actions of LKB1 and PKA (Fig. 10A). Consistent with this, mutant S587A (51, 52) but not wild-type SIK2 promoted HDAC5 export and stimulated MEF2-dependent transcription and myogenesis (Fig. 7). It should be noted that SIK2 possesses another potential PKA site (supplemental Fig. S6), and SIK3 possesses four such sites (supplemental Fig. S7). Thus, PKA is likely to be important upstream from SIK regulation, and further investigation is needed to fully establish this. PKA also phosphorylates LKB1 (34) and MEF2D (63) and controls the phosphorylation of class IIa HDACs (75–77). Thus, cAMP/PKA signaling regulates MEF2-dependent transcription through multiple mechanisms (Fig. 10A).

Along with two recent reports that focus on the fly LKB1-SIK3 and mammalian LKB1-AMPK axes upstream from class IIa HDACs (74, 78), this study places the SIK2/3-class IIa HDAC signaling module as a novel branch downstream from LKB1 (Fig. 10B), in parallel to SIK-mediated regulation of gluconeogenesis and lipogenesis through the transcriptional co-activators CRTC2 and p300, respectively (52–54). LKB1 also signals through AMPKs to control metabolism (33, 34, 74), activates synapses of the amphid-defective (SAD) and microtubule-associated protein/microtubule affinity-regulation kinase (MARK) to modulate cell polarity (34, 79, 80), and regulates senescence and cell adhesion via novel (nua) kinase family 1 (Fig. 10B) (81, 82). Mutations in the LKB1 gene play a causal role in Peutz-Jeghers syndrome (34, 35). In addition, LKB1 controls hematopoietic stem cells (38–40) and functions as a tumor suppressor (36, 37). Related to the latter, mouse Hdac7 was recently identified as a new oncogene (8). Furthermore, HDAC4 deletion is linked to the brachydactyly mental retardation syndrome (7). Class IIa HDACs are transcriptional co-repressors with critical roles in diverse developmental and pathological programs (5, 6), so the LKB1-SIK2/3-class IIa HDAC signaling axis identified here (Fig. 10) provides novel mechanistic insights into related clinical conditions and suggests new molecular candidates that may be mutated in such conditions.