LKB1 Suppresses p21-activated Kinase-1 (PAK1) by Phosphorylation of Thr109 in the p21-binding Domain*

The serine/threonine protein kinase LKB1 is a tumor suppressor gene mutated in Peutz-Jeghers syndrome patients. The mutations are found also in several types of sporadic cancer. Although LKB1 is implicated in suppression of cell growth and metastasis, the detailed mechanisms have not yet been elucidated. In this study, we investigated the effect of LKB1 on cell motility, whose acquisition occurs in early metastasis. The knockdown of LKB1 enhanced cell migration and PAK1 activity in human colon cancer HCT116 cells, whereas forced expression of LKB1 in Lkb1-null mouse embryonic fibroblasts suppressed PAK1 activity and PAK1-mediated cell migration simultaneously. Notably, LKB1 directly phosphorylated PAK1 at Thr109 in the p21-binding domain in vitro. The phosphomimetic T109E mutant showed significantly lower protein kinase activity than wild-type PAK1, suggesting that the phosphorylation at Thr109 by LKB1 was responsible for suppression of PAK1. Consistently, the nonphosphorylatable T109A mutant was resistant to suppression by LKB1. Furthermore, we found that PAK1 was activated in the hepatocellular carcinomas and the precancerous liver lesions of Lkb1(+/−) mice. Taken together, these results suggest that PAK1 is a direct downstream target of LKB1 and plays an essential role in LKB1-induced suppression of cell migration.

LKB1 is a serine/threonine kinase whose mutations have been found not only in Peutz-Jeghers syndrome patients (1)(2)(3) but also in various types of sporadic cancer (4 -6). These results suggest that LKB1 is a tumor suppressor gene. In addition, we and others (7)(8)(9) have previously shown that the heterozygous Lkb1 mutations in mice cause gastrointestinal hamartomas after 20 weeks of age and cause hepatocellular carcinomas (HCCs) 2 after 50 weeks (10). Notably, the Lkb1 gene in all HCC and hepatic precancerous lesions shows loss of heterozygosity (10,11). These phenotypes in Lkb1(ϩ/Ϫ) mice further indicate that LKB1 is a tumor suppressor gene.
In mammalian cells, LKB1 forms a complex with STE20related adaptor pseudokinase (STRAD) and scaffolding protein MO25, both of which are required for LKB1 enzymatic activity (12,13). It can phosphorylate and activate at least 14 kinases, including AMP-activated protein kinase (AMPK) and microtubule-associated protein/microtubule affinity-regulating kinases (MARKs) (5). Activation of AMPK by LKB1 leads to inactivation of mammalian target of rapamycin complex 1 via phosphorylation of the tuberous sclerosis complex 1/2, and this pathway has been implicated in tumor suppressor functions of LKB1. In addition to growth control, LKB1 also plays important roles in establishing cell polarity in mammalian cells (14). LKB1 regulates tight junction assembly and cell polarity through AMPK in mammalian cells (15,16), and we have shown that LKB1 suppresses tubulin polymerization by activating MARK microtubule-associated protein signaling (17). We have also reported that HCCs in Lkb1(ϩ/Ϫ) mice metastasize to the lungs (10). However, the mechanisms by which LKB1 suppresses cancer metastasis have not yet been explored. In this study, we have investigated the effect of LKB1 on cell motility whose acquisition occurs in early metastasis (18).
The serine/threonine kinase PAK1 is known as a key regulator of cell motility, proliferation, differentiation, and survival (19). It can activate diverse signaling pathways, including LIM motif-containing protein kinase 1, mitogen-activated protein kinase (MAPK), and NF-B signaling (20). PAK1 interacts with the GTP-bound forms of Cdc42 or Rac, which cause PAK1 activation (21). In addition to activation by Cdc42/Rac, its activity is positively regulated also by other signaling molecules, including phosphoinositide 3-kinase (22), 3-phosphoinositidedependent kinase-1 (23), and AKT (24). PAK1 is overexpressed in various types of cancer, including breast (25,26), colon (27), and liver (28) cancers. A high level of PAK1 expression is correlated with more aggressive tumor behaviors such as metastatic phenotype and advanced tumor stages in hepatocellular carcinomas (28) as well as in colorectal cancer (27). Consistently, overexpression of constitutively active mutant PAK1 (T423E) helps develop malignant mammary tumors in a transgenic mouse model (29). These results indicate that PAK1 plays a key role in carcinogenesis and cancer metastasis. Here, we present evidence that LKB1 inhibits cell motility and PAK1-mediated signaling through direct phosphorylation of PAK1 at Thr 109 in the p21-binding domain.
Mice-Construction of Lkb1 knock-out mice has been described previously (10). We only used males due to the low incidence of nodular foci and HCCs in female Lkb1(ϩ/Ϫ) mice, as reported previously. All animal experiments were approved by the Animal Care and Use Committee of Kyoto University.
cDNA-The following cDNAs were isolated by standard PCR-based cloning techniques: human LKB1, NCBI accession no. NM_000455 and human STRAD, respectively, NCBI accession no. NM_001003787. HA and FLAG tags were attached to the N termini of LKB1 and STRAD, as described previously (17). LKB1 (kinase-dead, a catalytically inactive version of LKB1 with a D194A mutation (30)), was generated with QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol. Wild-type of PAK1, constitutively active mutant T423E, and a dominant-negative mutant H83LH86LK299R were kindly gifted by Dr. Jonathan Chernoff (Fox Chase Cancer Center). Appropriate PCR primers were used to generate the point mutants PAK1-T109E and PAK1-T109A with the QuikChange II site-directed mutagenesis kit. These BamHI/EcoRI fragments were subcloned into pCMV-Tag 3B (Stratagene), which has an N-terminal Myc tag and a cytomegalovirus promoter.
Recombinant Adenoviruses-Recombinant adenoviruses were constructed by Adeno-X expression system 1 (Clontech Laboratories, Mountain View, CA), according to the manufacturer's protocol. The titers of the recombinant adenoviruses were determined by the method previously reported by others (31).
Wound-healing Assay-Cells were infected with recombinant adenovirus; after 24 h of infection, wounds were incised by scratching the cell monolayer using 10-l pipette tips. Photographs were taken at 0 and 20 h after the wound was made. Cell migration was normalized so that 100% represents the migration distance of control cells.
Transfection-HEK293T cells were transfected with plasmid DNA of the indicated PAK1 and LKB1 expression vectors using Lipofectamine 2000 (Invitrogen). After 24 h of transfection, cell lysates were prepared for Western blotting or in vitro kinase assay. Lipofectamine 2000 RNAiMAX (Invitrogen) was used for small interfering RNA (siRNA) transfection according to manufacturer's protocol.
Statistical Analysis-Results of the experimental studies were reported as mean Ϯ S.D. Differences were analyzed by Student's t test. A value of p Ͻ 0.05 was regarded as significant.

Knockdown of Endogenous LKB1 Enhances Cell
Migration and PAK1 Activity in HCT116 Cells-We have reported previously that HCCs in Lkb1(ϩ/Ϫ) mice metastasize to the lungs (10). Consistently, LKB1 also appears to suppress lung cancer metastasis (33). To investigate the mechanism(s) by which HCT116 cells were transfected with two different specific siRNAs against human LKB1. Control cells were transfected with a scrambled (Scr.) siRNA pool. After 24 h of plating, scratches were made with 10-l pipette tips. Photographs were taken at 0 and 20 h after the wound was made. B, cell migration was normalized so that 100% represented migration distance of control cells. Error bars indicate S.D. The asterisk indicates significant increases compared with control cells (p Ͻ 0.001). C, cellular levels of phospho-PAK1 (Ser 144 ), total PAK1 and in vitro PAK1 activity in LKB1 knockdown HCT116 cells. Lysates were prepared from HCT116 cells transfected with the indicated siRNAs. The levels of phospho-PAK1 (Ser 144 ) and PAK1 were determined by Western blotting with the indicated antibodies. Lysates were prepared from HCT116 cells transfected with the indicated siRNAs and immunoprecipitated using an anti-PAK1 antibody. The precipitates were used for in vitro kinase assays using GST-VASP-(158 -277) as a substrate. The phosphorylation of GST-VASP was visualized using BAS-5000 Bio-imaging Analyzer. D, induction of PAK1-K299R suppressed the increase in migration in LKB1 knockdown cells. The cells transfected with the indicated siRNAs were then infected with lentivirus lenti-PAK1-K299R or lenti-vector. At 24 h post-infection, cultures were scratches were made with 10-l pipette tips. Photographs were taken at 0 and 20 h thereafter. Cell migration was normalized so that 100% represented the migration distance of control cells. Error bars indicate S.D. The asterisk represents significant increases compared with control cells (p Ͻ 0.001).
LKB1 suppressed metastasis of cancer cells, we determined the effect of LKB1 knockdown on cell migration using a woundhealing assay. As test cells, we used a human colon cancer cell line HCT116 that expressed a significant level of LKB1 endogenously. HCT116 cells transfected with LKB1 siRNA (1 and 2) migrated 1.5ϫ faster than those with scrambled siRNA pool (Fig. 1, A and B, Scr.). Because PAK1 is an important regulator of cell migration (34), we investigated a possible role of PAK1 in cell migration enhanced by LKB1 knockdown. To determine the PAK1 activity in LKB1 knockdown cells, we then analyzed the level of phospho-PAK1 (Ser 144 ), its active form (35). Knockdown of LKB1 increased the cellular level of the phosphorylated PAK1, without affecting that of total PAK1 (Fig. 1C). These results suggest that knockdown of LKB1 increases the PAK1 activity. To confirm this interpretation, we determined the effect of LKB1 knockdown on in vitro PAK1 activity. Because histone H4, a widely used substrate for PAK1, was phosphorylated by LKB1 (data not shown), we screened in silico candidate substrate sequences for PAK1 using the consensus phosphorylation motif ((K/R)(R/X)(X)(pS/pT)) (34) as a query. As a candidate, we identified VASP-(158 -277), which is a major substrate for both PKA and PKG (32,36). We first verified that recombinant PAK1 efficiently phosphorylated GST-VASP-(158 -277) (data not shown). We found that PAK1 activity was significantly higher in LKB1 knockdown cells than in the control cells (Fig. 1C). To further investigate whether enhanced PAK1 activity contributed to the cell migration increase in LKB1 knockdown cells, we tested the effect of a dominantnegative mutant PAK1-K299R in LKB1 knockdown cells. Notably, expression of PAK1-K299R suppressed the increase in migration of these cells (Fig. 1D). These results suggest that LKB1 inhibits PAK1 activity, which can suppress cell migration.
Expression of LKB1 in Lkb1-null MEFs Inhibits PAK1 Activity and Cell Migration-To investigate the role of LKB1 in suppression of PAK1-mediated signaling, we further evaluated the effect of forced LKB1 expression on PAK1 activation in mouse embryonic fibroblasts that lack wild-type LKB1 (MEF3-2). To this end, we used the recombinant adenovirus encoding LKB1/STRAD (Adv-LKB1) that contained a transcription termination cassette (STOP sequence) flanked by the loxP sequences and expressed LKB1/STRAD by co-infection of MEF3-2 with Adv-Cre, which removed the STOP sequence from Adv-LKB1. The floxed inducible system was employed because LKB1 had some cytotoxic effects on host HEK293 cells and hampered production of the LKB1-expressing recombinant adenovirus. Such cytotoxicity was not observed with MEF3-2 cells up to 48 h post-induction. (Fig. 2A) (17). The cells infected with Adv-LKB1 alone showed similar activity of PAK1 to those infected with Adv-Cre alone (control cells) (Fig. 2B). Notably, forced expression of LKB1/ STRAD by co-infection of Adv-LKB1 and Adv-Cre decreased the cellular level of the phosphorylated PAK1 but not that of total PAK1 and inhibited the PAK1 activity by ϳ60% as compared with that in the control cells infected with Adv-Cre or Adv-LKB1 alone (Fig. 2B). Furthermore, forced expression of LKB1/ STRAD reduced cell migration in MEF3-2 cells (to 35% of Adv-Cre alone) (Fig. 2C). To directly test whether LKB1 reduces cell migration through inhibition of PAK1, we infected MEF3-2 cells with various multiplicity of infection of the adenovirus encoding the constitutively active mutant of PAK1 (Adv-PAK1-T423E) together with Adv-Cre/Adv-LKB1. Co-infection with Adv-PAK1-T423E has recovered the LKB1-induced reduction of cell migration in a dose-dependent manner (Fig. 2C). These results suggest that LKB1 can contribute to the suppression of cell migration through inhibition of PAK1.
LKB1 Activity Is Necessary for Suppression of PAK1-PAK1 can be activated by recruitment of the active forms of Cdc42/ Rac and phosphorylation by 3-phosphoinositide-dependent kinase-1 (19). Thus, we investigated whether the protein kinase activity of LKB1 was necessary to suppress the PAK1 activation. HEK293T cells were transiently transfected with PAK1 and/or LKB1/STRAD expression vector to determine PAK1 activity. The wild-type LKB1 (wt) inhibited the activity significantly, whereas the kinase-dead mutant of LKB1 (kinase-dead) did not affect activity (Fig. 3A). These results indicated that LKB1 activity was necessary for PAK1 inhibition. Accordingly, we next hypothesized that LKB1-mediated phosphorylation of PAK1 reduced its kinase activity. To test the possibility, we preincu-bated recombinant PAK1 with various amounts of LKB1 in the presence of ATP and then added the substrate of PAK1 to determine its kinase activity. LKB1 directly inhibited the PAK1 activity in vitro in a dose-dependent manner (Fig. 3B). These results indicate that LKB1-mediated PAK1 phosphorylation decreases its kinase activity.

LKB1 Inhibits PAK1-mediated Signaling
strate for LKB1 and that Thr 109 is a likely target of phosphorylation by LKB1, although it is possible that LKB1 phosphorylates other residues of PAK1 in addition to Thr 109 . To determine whether LKB1 phosphorylates PAK1 at Thr 109 in vivo, we performed a series of mass spectrometric analyses of PAK1 purified from HEK293T cells transiently co-transfected with plasmids encoding PAK1 and LKB1/STRAD (supplemental Fig. S1). The protonated molecular ions of peptide LLQTSNITK-(106 -114) obtained from tryptic digestion of wt-PAK1 were detected in its nonphosphorylated form ([M ϩ H] ϩ at m/z 1017.55) and in its monophosphorylated form ([M ϩ H] ϩ at m/z 1097.54). To further confirm the specific phosphorylation site of the peptide, we used HEK293T cells co-transfected with plasmids encoding PAK1-T109A and LKB1/STRAD for another set of mass spectrometric analyses. The molecular ion of LLQASNITK appeared at m/z 987.59 corresponding to the substitution of threonine at residue 109 with alanine. Notably, the corresponding monophosphopeptide was not detected in PAK1-T109A, excluding phosphorylation at the other Thr or Ser residue of the peptide. These results are consistent with our interpretation that LKB1 phosphorylates PAK1 at Thr 109 in vivo.
Phosphorylation of PAK1 at Thr 109 by LKB1 Reduces Its Activity-We have demonstrated that the kinase activity of LKB1 is necessary to inhibit that of PAK1 (Fig. 3A). Therefore, we next investigated whether phosphorylation of Thr 109 by LKB1 regulated PAK1 activity. Namely, we constructed a PAK1-T109E mutant whose amino acid residue 109 was converted to glutamic acid from threonine. This substitution mimics the phosphorylation of threonine with the acidic moiety of glutamic acid. As an inactive (negative) control, we also constructed the alanine substitution PAK1-T109A (Fig. 5A). We transiently transfected HEK293T cells with plasmids encoding these constructs of PAK1: wt, constitutivelyactive (T423E), dominant-negative (K299R), T109E, and T109A, respectively (Fig. 5A). We then determined the in vitro PAK1 activities in the cell lysates using GST-VASP-(158 -277) as substrate. The PAK1 mutant T423E showed the highest activity (2.1ϫ), whereas mutant K299R with defective kinase (kinase-dead) had only 2% of the wt-PAK1 activity. Notably, T109E mutant showed significantly reduced PAK1 activity (36% of wt), whereas T109A mutant had a similar activity to the wt (98%) (Fig. 5B). These results suggest that phosphorylation of PAK1 at Thr 109 decreases its kinase activity. To confirm the role of phosphorylation at Thr 109 in suppression of PAK1 activity by LKB1, we next tested nonphosphorylatable PAK1-T109A in HEK293T cells. Co-expression of LKB1 failed to suppress the activity of PAK1-T109A (Fig. 5C), in sharp contrast to the effect on wt-PAK1 (Fig. 3A). These results strongly suggest that LKB1 suppresses the PAK1 activity by phosphorylating it at Thr 109 .
Activation of PAK1 in Lkb1(ϩ/Ϫ) Mouse HCCs and Lkb1(Ϫ/Ϫ) MEFs-We have previously demonstrated that Lkb1(ϩ/Ϫ) mice develop HCC after 50 weeks of age and that HCCs and nodular foci of Lkb1(ϩ/Ϫ) liver show loss of the Lkb1 heterozygosity (10,11). To test the involvement of LKB1 in PAK1-mediated signaling in vivo, we next determined the level of phosphorylated form of PAK1 in HCCs of Lkb1(ϩ/Ϫ) mice, using two different types of phospho-specific PAK1 antibodies that recognized the active form of PAK1. Notably, we found that the level of phosphorylated PAK1 was increased in HCCs (H1 and H2) and in precancerous lesions (nodular foci, F1 and F2), compared with the normal liver of Lkb1(ϩ/Ϫ) or Lkb1(ϩ/ϩ) mice, while the levels of total PAK1 in nodular foci and HCCs were similar to those in normal liver (Fig. 6A). These results are therefore consistent with the interpretation that loss of LKB1 induces PAK1 activation. We further investigated the status of PAK1 activation in Lkb1(Ϫ/Ϫ) MEFs (MEF3-2) and wt MEFs. Both the level of phosphorylated PAK1 and in vitro PAK1 activity were significantly higher in Lkb1(Ϫ/Ϫ) MEFs than those in wt MEFs (Fig. 6B). These data together suggest that loss of LKB1 contributes to the activation of PAK1.

DISCUSSION
In this study, we have provided evidence that LKB1 inhibits PAK1 activation by direct phosphorylation of PAK1 at Thr 109 . We first showed that knockdown of LKB1 increased cell migration and PAK1 activity in HCT116 cells (Fig. 1), whereas introduction of LKB1 in Lkb1-null MEFs reduced cell migration and PAK1 activity simultaneously (Fig. 2). Notably, a constitutively active mutant of PAK1 (PAK1-T423E) recovered the LKB1mediated reduction of cell migration in a dose-dependent manner (Fig. 2C). Because HCCs in Lkb1(ϩ/Ϫ) mice can metastasize to the lung (10), it is conceivable that inhibition of PAK1 by LKB1 contributes to suppression of cancer metastasis. We also demonstrated that LKB1 directly phosphorylated PAK1 at Thr 109 both in vitro (Fig. 4) and in vivo (supplemental Fig. S1). Importantly, the phosphomimetic PAK1-T109E mutation reduced its activity (Fig. 5B), whereas the activity of a nonphosphorylatable mutant PAK1-T109A was not suppressed by LKB1 (Fig. 5C). These results indicate that phosphorylation of Thr 109 is both necessary and sufficient for inhibition of PAK1 activity by LKB1, although it is possible that LKB1 phosphory-lates other residues of PAK1 in addition to Thr 109 (Fig. 4C). Consistently, we found that active forms of PAK1 were increased in the nodular foci and HCCs in Lkb1(ϩ/Ϫ) mice, compared with the normal liver of Lkb1(ϩ/Ϫ) and Lkb1(ϩ/ϩ) mice (Fig. 6A).
Our present results indicate that PAK1 is a novel LKB1 substrate. So far, 14 kinases have been identified as LKB1 substrates, including AMPK␣1, AMPK␣2, MARK1, and MARK2. LKB1 phosphorylates these kinases at the LXT motif in the activation loop (5,37), and such phosphorylations can induce kinase activation. In contrast, LKB1-induced phosphorylation of PAK1 at Thr 109 in the LXT motif of PBD causes a decrease in the kinase activity (Fig. 5B). Because this motif is well conserved among group I PAKs (PAK1, PAK2, and PAK3), it is likely that regulation of the PAK1 activity by LKB1 applies to PAK2 and PAK3 as well. According to the structural analysis of PAK1, Thr 109 is located at the helix loop I␣ in an inhibitory switch (38) and exposed to the outer space. In addition, the L107F mutation of PAK1 prevents the interaction between the autoinhibitory region and the C-terminal catalytic region, making the constitutively active kinase (39). Because this mutant lacks the LXT motif, the constitutive activation is likely due to the loss of negative regulation by LKB1. We therefore speculate that the phosphorylation of this Thr residue enhances the interaction between the autoinhibitory (N-terminal regulatory) and kinase domains, leading to inhibition of the PAK1 activity. Further experiments are required to elucidate the precise role of Thr 109 phosphorylation in PAK1 conformation.
It has been reported recently that LKB1 maintains cancer cell polarity by recruiting Cdc42 to the leading edges and that inhibition of LKB1 reduces PAK1 autophosphorylation (40). In contrast, we have found that LKB1 suppresses PAK1 activity by direct phosphorylation. The reason for the discrepancy is not clear but may be related to the methods used to overexpress LKB1 (transient transfection of LKB1 expression vector alone versus co-infection of LKB1 with STRAD using adenovirus vector) or the cell context, genetic constitution such as p53 mutation status (non-small cell lung cancer cells (p53 mutant) versus HEK293T or Lkb1-null MEFs (wt-p53)). Notably, LKB1 plays key roles in growth arrest or apoptosis through a p53-dependent mechanism (41,42). Thus, it is conceivable that LKB1-mediated PAK1 inhibition may be dependent on the p53 status.
Numerous PAK1 substrates and/or interacting proteins have been identified that activate diverse signaling processes involved in cell motility, morphology, growth, survival, and epithelial-mesenchymal transition (19). For example, PAK1 stabilizes microtubules through stathmin/Op18 phosphorylation (43,44). Therefore, LKB1 also may contribute to microtubule destabilization by regulating PAK1, whereas we have reported previously that LKB1 inhibits tubulin polymerization through activation of MARK2 (17). Further investigation is needed to address the role of LKB1 in various downstream processes of PAK1 other than cell motility. In conclusion, our present results suggest that PAK1 is a novel downstream target of LKB1, and that LKB1-mediated PAK1 phosphorylation may play important roles in tumor suppression in the liver.