Characterization of an Alternative Splice Variant of LKB1*

LKB1 is an upstream activating kinase for the AMP-activated protein kinase (AMPK) and at least 12 other AMPK-related kinases. LKB1 therefore acts as a master kinase regulating the activity of a wide range of downstream kinases, which themselves have diverse physiological roles. Here we identify a second form of LKB1 generated by alternative splicing of the LKB1 gene. The two LKB1 proteins have different C-terminal sequences generating a 50-kDa form (termed LKB1L) and a 48-kDa form (LKB1S). LKB1L is widely expressed in mouse tissues, whereas LKB1S has a restricted tissue distribution with predominant expression in the testis. LKB1S, like LKB1L, forms a complex with MO25 and STRAD, and phosphorylates and activates AMPK both in vitro and in intact cells. A phosphorylation site (serine 431 in mouse) and a farnesylation site (cysteine 433 in mouse) within LKB1L are not conserved in LKB1S raising the possibility that these sites might be involved in differential regulation and/or localization of the two forms of LKB1. However, we show that phosphorylation of serine 431 has no effect on LKB1L activity and that both LKB1L and LKB1S have similar patterns of subcellular localization. These results indicate that the physiological significance of the different forms of LKB1 is not related directly to differences in the C-terminal sequences but may be due to their differential patterns of tissue distribution.

LKB1 was originally identified as a tumor suppressor because inactivating mutations in the LKB1 gene lead to a cancer predisposition disease in humans, termed Peutz-Jeghers syndrome. LKB1 is a serine/threonine protein kinase that forms a complex with two other proteins, STRAD (STE20-related adaptor) (1) and MO25 (mouse protein 25) (2). STRAD shares extensive sequence identity with protein kinases but lacks sev-eral residues required for catalytic activity and therefore has been termed a pseudokinase. MO25 binds to the C-terminal 3 amino acids of STRAD (1,3), and binding of MO25 markedly increases the binding of STRAD to LKB1 (2). In addition to enhancing the catalytic activity of LKB1, it has been reported that binding of STRAD and MO25 causes LKB1 to relocalize from the nucleus to the cytoplasm (4). In addition, there has also been a report that claims that phosphorylation of serine 428 at the C terminus of human LKB1 (equivalent to Ser-431 in mouse) is necessary for LKB1 export from the nucleus (5). However, another study reported that phosphorylation at this site does not result in relocalization (6). The cytosolic localization of LKB1 appears to be important for its tumor suppressor functions, because mutants of LKB1 that are unable to enter the nucleus retain their full capability to suppress cell growth (7). LKB1 phosphorylates AMP-activated protein kinase (AMPK) 4 at a single site, Thr-172 within the T-loop segment of the catalytic ␣ subunit (8 -10). Phosphorylation of Thr-172 is essential for activation of AMPK (11). In addition to LKB1, Ca 2ϩ /calmodulin-dependent protein kinase kinase (12)(13)(14) and transforming growth factor-␤-activated protein kinase 1 (15) have been shown to phosphorylate and activate AMPK. Although a complete understanding of the physiological relevance of the different upstream kinases in the AMPK cascade is some way off, it appears that LKB1 has a predominant role in activating AMPK in some tissues. For example, deletion of LKB1 in liver led to an almost complete loss of AMPK activity (16), and in skeletal muscle a lack of LKB1 abolished activation of the ␣2-containing AMPK complexes (17). More recently, LKB1 has been shown to phosphorylate 12 AMPK-related kinases within their T-loop regions, and this phosphorylation is essential for their activation (18). LKB1 itself is phosphorylated on at least 8 serine/threonine residues. Four of these are autophosphorylation sites, whereas the remaining 4 sites are phosphorylated by other kinases (19). To date there is no evidence to suggest that phosphorylation at any of these sites has a direct effect on the activity of LKB1 in vitro (6,20,21). Moreover, several studies have reported that LKB1 activity is unaltered by stimuli that cause significant activation of AMPK (10,18,22), suggesting that LKB1 is constitutively active. However, despite the lack of evidence for the direct regulation of LKB1 catalytic activity, several studies have indicated that the function of LKB1 may be subject to indirect modes of regulation. Mutation of Thr-336 to a glutamic acid residue, but not to an alanine, prevented LKB1 from suppressing cell growth in G361 cells, suggesting that phosphorylation of this residue has an inhibitory effect on LKB1 activity (20). The physiological significance of this finding is unclear because phosphorylation of Thr-336 has only been reported to occur using manganese ATP as a substrate, rather than magnesium ATP, the physiologically relevant form of ATP (20). Ser-431 in mouse LKB1 (equivalent to Ser-428 in the human sequence) has been shown to be phosphorylated by a number of kinases, including cyclic AMP-dependent protein kinase (PKA), p90 ribosomal S6 protein kinase (21), and protein kinase C (5). Mutation of Ser-431 to either alanine or glutamic acid prevented LKB1 from suppressing cell growth (21), suggesting that this residue is required for some aspect of LKB1 function. Recently, a role for the PKA-dependent phosphorylation of LKB1 has been proposed based on studies examining LKB1 in neuronal polarization (23,24). Finally, it has been reported that phosphorylation of Ser-428 in human LKB1 is involved in relocalization of LKB1 from the nucleus to the cytoplasm and that this is required for the LKB1-mediated phosphorylation of AMPK (5).
Here we describe the identification of a previously uncharacterized form of LKB1, generated by alternative splicing, that has a different C-terminal sequence from the previously studied form and lacks the equivalent residue corresponding to Ser-428/431. We have called the new form of LKB1, LKB1 S (for LKB1 short form) and the originally published form LKB1 L (corresponding to LKB1 long form). We show that the new form of LKB1 associates with STRAD and MO25, is catalytically active, and can activate AMPK in cells. LKB1 S is highly expressed in testis, with low levels of activity detected in other tissues. The predominant expression of LKB1 S in testis provides indirect evidence that it maybe involved in male fertility because male mice lacking LKB1 S are infertile. Furthermore, we show that phosphorylation of Ser-428/431 has no effect on LKB1 activity and is not involved in relocalization of the LKB1 complex between the nucleus and cytoplasm.

EXPERIMENTAL PROCEDURES
Materials-Forskolin, hydrogen peroxide, and oligonucleotide primers were purchased from Sigma-Aldrich. Mouse and human testis cDNA were obtained from Ambion. All of the microarray products were purchased from Agilent Technologies UK unless otherwise noted.
Tissue Harvesting and Preparation-Tissues from male mice (C57/Bl6), ϳ10 weeks old were harvested and immediately frozen in liquid nitrogen. Prior to analysis, the tissues were roughly chopped in two volumes of ice-cold buffer A (50 mM Tris, pH 7.5, 5 mM sodium pyrophosphate, 50 mM NaF, 1 mM EDTA, 0.25 M sucrose, 1 mM dithiothreitol, 4 g/ml trypsin inhibitor, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine) and then briefly homogenized with a rotor-stator homogenizer. Triton X-100 was added to the homogenates to 1% (v/v), and the homogenates were incubated on ice for 10 min before centrifugation at 16,000 ϫ g for 15 min to remove insoluble material. Protein content in the supernatant was determined using the Bradford assay (25).
Antibodies-We raised a rabbit polyclonal antibody against full-length mouse LKB1 L (expressed as a His tagged-fusion protein in Escherichia coli). A peptide (CMWKSQGAGLPGEE) corresponding to residues 390 -403 of mouse LKB1 S was used to raise anti-LKB1 S antibodies (termed anti-sLKB1) in rabbit. Antibodies were affinity-purified from the serum using the immunizing-peptide coupled to thiol-activated Sepharose. Affinity-purified sheep antibody raised against residues 24 -39 of human LKB1 was a kind gift from Prof. Dario Alessi (University of Dundee). Mouse monoclonal antibody recognizing both LKB1 L and LKB1 S (Ley37D/G6) and anti-STRAD␣ (STRAD N13) were from Santa Cruz Biotechnology. Mouse monoclonal antibody recognizing Ser(P)-428 in human LKB1 L , monoclonal anti-MO25, and phospho-CREB (Ser-133, 87G3) were from Cell Signaling Technology. FLAG (M2) and anti-CREB were from Sigma, and anti-glyceraldehyde-3-phosphate dehydrogenase was from AbCam. A pan-␤-specific AMPK antibody has been described elsewhere (26).
Western Blot Analysis-Primary antibodies were detected with the appropriate secondary antibody conjugated to either Alexa-Fluor 680 (Invitrogen) or IRDye800 (LI-COR) and scanned on the Li-COR Odyssey Infrared Imaging System. Quantification of results was performed using Odyssey software 2.0 (LI-COR Biotechnology).
LKB1 and AMPK Activity Assays-LKB1 activity was determined by a two-step assay that measures the ability of LKB1 to activate recombinant AMPK (10). For endogenous LKB1 S activity, mouse tissue extracts were incubated with protein A-Sepharose beads for 1 h at 4°C, and then LKB1 S was immunoprecipitated by incubation with rabbit anti-sLKB1antibody bound to protein A-Sepharose for 2 h at 4°C. LKB1 activity in transiently transfected cells was immunoprecipitated by incubation with anti-FLAG resin. In each case, the immune complexes were washed extensively with buffer A and then incubated with 0.2-1 g of recombinant AMPK (␣1␤1␥1) and 0.2 mM ATP, 5 mM MgCl 2 in buffer A (20 l of total volume) for 20 min at 37°C. AMPK activation of was quantified by removing an aliquot (5 l) of the reaction and measuring AMPK activity using the SAMS peptide assay (27). AMPK was immunoprecipitated from cell lysates (50 g of total protein) using an antipan-␤ AMPK antibody, and the activity was measured by phosphorylation of the SAMS peptide.
Synthesis of cDNA Expression Constructs-cDNA encoding LKB1 S was amplified from human or mouse testis cDNA with LKB1 S -specific oligonucleotide primers using Phusion High-Fidelity DNA Polymerase (Finnzymes). The primers used were: human, forward primer (including FLAG tag (underlined) and EcoRI restriction site in italics), AAGCTTGAATTCATGGAT-TATAAAGATGATGACGATAAAGCCATGGAGGTGGT-GGACCCGCAG; and reverse primer (including XhoI restriction site in italics), TCTAGACTAGTCTCGAGCTACTGCG-GCGCCGGCCGAGAGG; mouse, forward primer (including FLAG tag (underlined) and EcoRI restriction site in italics), AAGCTTGAATTCATGGATTATAAAGATGATGACGAT-AAAGCCATGGACGTGGCGGACCCCGAG; and reverse primer (including XhoI restriction site in italics), TCTAGAC-TAGTCTCGAGCTACACTAAAGCCCCAAACCCC. The products were purified by agarose gel electrophoresis, digested with EcoRI and XhoI, and cloned into pCDNA3 mammalian expression vector (Invitrogen). The authenticity of the inserts were confirmed by DNA sequencing.
cDNA encoding human LKB1 L was amplified from an IMAGE clone with an N-terminal FLAG tag sequence engineered at the 5Ј end and cloned into pCDNA3 (a kind gift from Dr. Naveenan Navaratnam, Medical Research Council Clinical Sciences Centre). cDNAs encoding mouse LKB1 L and mouse LKB1 L harboring a mutation of aspartic acid 194 to alanine (LKB1 D194A ) were a kind gift from A. Ashworth. The cDNAs were amplified to include a FLAG tag at the N terminus and cloned into pCDNA3. The point mutations (S431A and S431E) were introduced into mouse LKB1 L using the QuikChange sitedirected mutagenesis kit (Stratagene) according to the manufacturer's protocol. The mutations were verified by DNA sequencing. cDNAs encoding human STRAD␣ and MO25␣ (a kind gift from Prof. Dario Alessi, University of Dundee) were cloned into pCDNA3.
Cell Transfection-Plasmid DNA was prepared using a Qiagen maxiprep kit according to the manufacturer's instructions. The cells were transfected using the calcium phosphate precipitation method (CalPhos TM mammalian transfection kit; Clontech), following the standard manufacturers protocol. After ϳ18 h the cells were washed in PBS and shocked with 10% (v/v) Me 2 SO for 2 min. The cells were treated 36 h post-transfection.
Immunofluorescence-CCL13 cells were plated on glass coverslips for transfection. 36 h post-transfection, the cells were fixed in 4% paraformaldehyde, permeabilized with PBS containing 0.1% (v/v) Triton X-100, and blocked with PBS containing 5% (v/v) donkey serum and 2% (w/v) bovine serum albumin before incubation with primary antibody. After thorough washing with PBS, the cells were incubated with the appropriate secondary antibody (conjugated with Alexafluour 488; Invitrogen). After further washing, the cells were mounted on coverslips with Vectashield and visualized on a Leica TCS SP1 confocal microscope. Image analysis was performed using Leica software.
Subcellular Fractionation-Mouse testis extract or HEK293 cell lysates were subjected to cell fractionation using a Qproteome Cell-Compartment kit (Qiagen) following the manufacturer's protocol.
Microarray Analysis-RNA was extracted from whole mouse testis using TRIzol reagent (Invitrogen) according to the manufacturer's instructions, followed by purification on an RNeasy column (Qiagen) according to the standard manufac-turer's protocol. RNA quality and concentration was assessed using an Agilent bioanalyzer. 1 g of RNA and 5 l of (1:2500 dilution) Agilent One-Color RNA Spike-In RNA were labeled with reagents supplied in the Agilent low RNA input linear amplification kit PLUS, one-color according to the manufacturer's instructions. The labeled cRNA was purified with the RNeasy mini kit (Qiagen) according to the manufacturer's protocol.
Hybridization and Scanning-The Agilent hybridization kit was used in conjunction with Agilent Mouse whole genome Oligo Arrays. 2 g of the labeled sample RNA were used for hybridization according to the Agilent one-color microarraybased gene expression analysis protocol. The hybridization was performed for 17 h at 65°C with 10 rpm rotation. The slides were then washed as described in the manufacturer's manual. The slides were then washed in acetonitrile for 1 min followed by 30 s in Agilent stabilization and drying solution. The slides were scanned with the Agilent G2565BA microarray scanner system.

Data Extraction and Deposition into Gene Expression
Omnibus-For data extraction and quality control, the Agilent G2567AA feature extraction software (version 9.1) was used. The data files were deposited into the NCBI Gene Expression Omnibus to comply with Minimum Information About a Microarray Experiment requirements.
Gene Expression Analysis-The data were analyzed using Genespring GX 7.3 Expression Analysis software (Agilent Technologies). Text tab delimited raw files were imported into Genespring. Standard data transformation, chip, and gene normalizations were applied, in addition to standard QC procedures. Differences in testis gene expression between the two groups were identified using Volcano Plot analysis. The data were corrected for multiple testing corrections with adjusted p values using the method of Hotchberg and Benjamini with a user-defined false discovery rate of 5%.

RESULTS
We investigated the amount of LKB1 protein in mouse tissues following immunoprecipitation with a sheep polyclonal anti-LKB1 antibody. Western blot analysis of the resulting LKB1 immune complexes with an anti-LKB1 monoclonal antibody revealed a band migrating with a molecular mass of ϳ50 kDa in all of the tissues analyzed. In addition, a second cross-reacting band, migrating with a molecular mass of ϳ48 kDa, was detected in testis (Fig. 1A). The same pattern was observed using a different LKB1 antibody for immunoprecipitation (data not shown), suggesting that both bands are related to LKB1. We refer to the 50-kDa protein as LKB1 L (for long form) and to the 48-kDa protein as LKB1 S (for short form). Quantification of the blots showed that the 50-kDa protein is most highly expressed in brain and testis. Previous studies have shown that LKB1 requires binding of two other proteins, MO25 (2) and STRAD (1) for activity, and so we blotted the immune complexes with antibodies specific for the ␣ isoforms of these two proteins. As can be seen from Fig. 1, both MO25␣ and STRAD␣ co-immunoprecipitate with LKB1 in all of the tissues examined. In brain, a second slower migrating band was detected strongly by the STRAD␣ antibody. However, this band was not detected when a different LKB1 antibody was used for immuno-precipitation (data not shown), and so we remain uncertain as to the nature of this cross-reacting protein.
We ruled out protein degradation as a likely explanation for the appearance of the 48-kDa band, because it was only detected significantly in the testis, and similar patterns of detection were observed on multiple occasions in the presence or absence of pro-tease inhibitors in the buffers used for immunoprecipitation (data not shown). Next, we considered the possibility that alternative splicing of the LKB1 gene might give rise to LKB1 S . Upon re-examination of the mouse LKB1 gene, we identified a potential exon within the intron between exons VIII and IX of the previously determined gene structure (28). This additional exon, which we termed IXa (exon IX becoming IXb), is conserved within the human and mouse LKB1 genes ( Fig. 2A) and contains a termination codon. Alternative splicing of the LKB1 gene after exon VIII would lead to the different C-terminal amino acid sequences shown in Fig. 2B. Interestingly, the predicted amino acid sequence encoded by exon IXa lacks a phosphorylation site (Ser-428/431) and the farnesylation site (Cys-430/433) within the previously identified LKB1 sequence.
We raised an antibody (termed anti-sLKB1) against a predicted C-terminal peptide sequence (see "Experimental Procedures") of mouse LKB1 S and determined whether it could immunoprecipitate LKB1 from mouse tissues. Using this antibody, LKB1 activity, as determined by activation of recombinant AMPK, was readily detectable in immune complexes isolated from testis (Fig. 3A). Low, but detectable, activity was also observed in lung, heart, liver, and kidney, whereas no detectable activity was present in brain or muscle (Fig. 3A). Fig. 3B shows that the 48-kDa band is greatly depleted (greater than 85% of a control immunoprecipitation) from testis homogenates immunoprecipitated using the LKB1 S antibody. This confirms that the 48-kDa band seen in Western blots corresponds to LKB1 S . Immunoprecipitates of LKB1 isolated using an antibody raised to an N-terminal peptide sequence (which is present in both LKB1 L and LKB1 S ) were analyzed by Western blotting using antibodies to total LKB1 or LKB1 S (Fig.  3C). A doublet migrating at ϳ48 kDa, corresponding to the predicted molecular mass of the novel LKB1 protein, is recognized by both antibodies in testis but not in any other tissue tested. These results show that the 48-kDa protein highly expressed in testis is the alternative LKB1 splice variant and that this form of LKB1 is capable of phosphorylating and activating AMPK in vitro.
Based on the sequence of mouse and human LKB1 S , we used specific oligonucleotide primers to amplify the cDNA from mouse or human testis. In both cases, single products of ϳ1.2 kb were obtained, and sequence analysis showed that they encoded the expected amino acid sequence of mouse or human LKB1 S (data not shown). The cDNA products were cloned into a mammalian expression vector with a FLAG tag engineered at the N termini. Fig. 4A shows overexpression of LKB1 S , LKB1 L , and a catalytically inactive mutant, LKB1 D194A in CCL13 cells, which do not express endogenous LKB1. To investigate the ability of these expressed pro-  teins to associate with MO25␣ and STRAD␣, cell lysates were immunoprecipitated using an anti-FLAG antibody to isolate the expressed LKB1. The resulting immune complexes were probed with antibodies specific for MO25␣ and STRAD␣. All three forms of LKB1 co-immunoprecipitate with endogenously expressed MO25␣ and STRAD␣ (Fig. 4B). The activity of endogenous AMPK immunoprecipitated from cells overexpressing LKB1 was measured using the SAMS peptide assay. Expression of either LKB1 L or LKB1 S results in activation of AMPK under basal conditions, compared with cells overexpressing LKB1 D194A . Overexpression of LKB1 D194A had no effect on AMPK activity compared with an untransfected control.
Previous reports have suggested that phosphorylation of Ser-428/431 may be responsible for regulation of LKB1 activity (6,21,23,24). Because the Ser-428/431 phosphorylation site is not present in either human or mouse LKB1 S , we wondered whether this site could be involved in differential regulation of the two splice forms. Therefore, to investigate this possibility we used the adenylate cyclase activator forskolin to activate the PKA pathway and increase phosphorylation of Ser-428/431 in LKB1 L . Forskolin treatment of CCL13 cells overexpressing LKB1 L resulted in increased phosphorylation of Ser-428, as well as CREB at Ser-133, a known target of PKA activation (29) (Fig. 5A). LKB1 activity from cells overexpressing LKB1 L , LKB1 S , or LKB1 D194A was determined. There was no change in the activity of either LKB1 L or LKB1 S following forskolin treatment, demonstrating that phosphorylation of Ser-428 in LKB1 L does not have a direct effect on activity (Fig. 5B). Because it has been reported previously that phosphorylation of Ser-428 increases the association of LKB1 with AMPK (5), we looked at the effect of forskolin treatment on endogenous AMPK activity. Forskolin treatment led to a modest increase in AMPK activity, and this was detected in cells expressing LKB1 L , LKB1 S , or LKB1 D194A (Fig. 5C). The finding that AMPK activity is increased with forskolin even in cells expressing catalytically inactive LKB1 indicates that this is not due to an effect on LKB1. Treatment with H 2 O 2 resulted in a marked increase in the activity of endogenous AMPK in cells overexpressing either LKB1 L or LKB1 S , demonstrating that AMPK is not fully activated by overexpression of LKB1.
It is possible that only a small proportion of the overexpressed LKB1 is phosphorylated by PKA in response to forskolin treatment so that the effect on LKB1 activity might be FIGURE 3. Tissue distribution of LKB1 S . A, mouse tissue homogenates (200 g of total protein) were immunoprecipitated using anti-LKB1 S antibody coupled to protein A-Sepharose. LKB1 activity in the immune complexes was measured by activation of recombinant AMPK. The activities are plotted as units/mg lysate, where 1 unit is the activity of LKB1 required to activate recombinant AMPK by 1 nmol/min/mg. B, LKB1 was immunoprecipitated from mouse testis homogenate (50 g of total protein) using anti-LKB1 S antibody, and the resulting immunoprecipitate (IP) and depleted supernatant (LKB1s SN) was blotted with anti-LKB1 mouse monoclonal antibody. The supernatant from a control immunoprecipitate using protein A in the absence of antibody is also shown (Control SN). C, immune complexes isolated by immunoprecipitation of mouse lysates (200 g of total protein) using sheep anti-LKB1 antibody were probed with anti-total LKB1 mouse monoclonal antibody and secondary antibody conjugated to Alexa-Fluor 680 and visualized in the green channel (top panel). The same blot was reprobed with anti-LKB1 S antibody and secondary antibody conjugated to IRDye800 and visualized on the Li-COR System in the red channel (bottom panel). The bands that were detected at both wavelengths (indicating recognition by both antibodies) are indicated by an asterisk. The migration of the 50-kDa molecular mass standard is indicated. B, lysates (250 g) were immunoprecipitated with anti-FLAG resin, and proteins in the immune complex were blotted with either anti-STRAD␣ or anti-MO25␣ antibodies. The migration of molecular mass standards are as indicated. In each case a representative blot of three similar blots from independent experiments is shown. C, endogenous AMPK was immunoprecipitated from 50 g of total protein using an anti-pan␤ antibody bound to protein A-Sepharose. AMPK activity in the immune complexes was determined using the SAMS peptide assay. The results are plotted as pmol/min/mg and are the means Ϯ S.E. of three independent experiments. underestimated. We therefore investigated the effect of mutating Ser-431 to either a negatively charged amino acid, glutamic acid (S431E) to mimic phosphorylation at this site, or an alanine (S431A), to prevent phosphorylation at this site. Fig. 6A shows expression of wild type and mutated LKB L protein in CCL13 cells. Mutation of Ser-431 to either alanine or glutamic acid had no effect on the activity of LKB1 as compared with wild type (Fig. 6B). In untransfected control cells, LKB1 was not detectable in this assay. Similarly, treatment with H 2 O 2 had no effect on LKB1 activity, as we have previously reported (10). Furthermore, mutation of Ser-431 had no effect on activation of endogenous AMPK under basal or H 2 O 2 -treated conditions (Fig. 6C).
We then went on to look at cellular localization of LKB1 L and LKB1 S when overexpressed in CCL13 cells. When expressed alone, LKB1 S and LKB1 L are detected in both the nucleus and cytoplasm, as shown in Fig. 7A (left-hand panels). However, when co-expressed with MO25␣ and STRAD␣, there is a striking relocalization of LKB1 S and LKB1 L from the nucleus to the cytoplasm. Because overexpression of proteins may cause them to locate to compartments of the cell that are not representative of the endogenous protein, we looked at the localization of endogenous LKB1. We were unable to detect endogenous LKB1 by immunofluorescence in any cell types studied presumably because of limited sensitivity of the antibodies available. Therefore we used cell fraction- and are plotted as units/mg total lysate protein where 1 unit is the activity of LKB1 required to activate recombinant AMPK by 1 nmol/min/mg. C, endogenous AMPK was immunoprecipitated from 50 g of total protein using an anti-pan␤ antibody bound to protein A-Sepharose. AMPK activity in the immune complexes was determined using the SAMS peptide assay. The results are plotted as pmol/min/mg and are the means Ϯ S.E. of three independent experiments. FIGURE 6. Effect of S428 mutation on LKB1 activity. CCL13 cells were transfected with FLAG-tagged mouse wild type LKB1 L (WT), or LKB1 L harboring a mutation of serine 431 to either alanine (S431A) or glutamic acid (S431E), or left untransfected (UT). Prior to lysis, the cells were treated with or without 0.5 mM H 2 O 2 for 5 min. A, total cell lysates (50 g of protein) were blotted with anti-FLAG antibody to detect LKB1 expression. B, LKB1 activity in anti-FLAG immune complexes isolated from 100 g of total lysate protein was determined by activation of recombinant AMPK. The results are the means Ϯ S.E. of three independent experiments and are plotted as units/mg total lysate protein where 1 unit is the activity of LKB1 required to activate recombinant AMPK by 1 nmol/min/mg. C, endogenous AMPK was immunoprecipitated from 100 g of total protein using an anti-pan␤ antibody bound to protein A-Sepharose. AMPK activity in the immune complexes was determined using the SAMS peptide assay. The results are plotted as pmol/min/mg and are the means Ϯ S.E. of three independent experiments. ation of HEK293 cells, which express moderate amounts of endogenous LKB1 L , to determine its subcellular localization. Cells treated with and without forskolin were fractionated into membrane, cytosol, and nuclear enriched fractions and blotted for LKB1. Using this approach, we were only able to detect LKB1 in the cytosolic and membrane fractions, but not the nuclear fraction (Fig. 7B). Western blots were carried out on each cell fraction using antibodies to marker proteins known to be predominantly expressed in the cytosolic (glyceraldehyde-3-phosphate dehydrogenase), membrane (Na ϩ /K ϩ ATPase), or nuclear (CREB) fractions. To investigate whether endogenous LKB1 S and LKB1 L are located in different subcellular compartments we used mouse testis, because this is the only tissue in which both forms are expressed. Western blot analysis of the different subcellular fractions demonstrates that both LKB1 S and LKB1 L are predominantly localized in the cytosol, with a small amount of LKB1 S detected in the membrane fraction (Fig. 7C). However, there is no detectable LKB1 S or LKB1 L in the nuclear fraction.
To investigate the functional role of LKB1 S , a microarray analysis was carried out to investigate differential gene expression in the testis of LKB1 S knock-out mice compared with that of wild type. Table 1 lists all genes found to have significantly (p Ͻ 0.01) altered expression (Ͼ1.5-fold change relative to wild type).

DISCUSSION
Here we describe a previously uncharacterized splice variant of LKB1 (LKB1 S ), which has a different C-terminal amino acid sequence compared with that of the previously reported form of LKB1 (LKB1 L ) (30). The two proteins are products of the same gene, generated by alternative splicing of the final coding exon. LKB1 S is the product of exons I-VIII and exon IXa, whereas LKB1 L is product of exons I-VIII and exon IXb. In mouse, LKB1 S is expressed predominantly in the testis. In a previous study, two pools of LKB1 activity were purified from rat liver and were shown to contain different molecular mass forms of LKB1, as judged by Western blotting (8). It is possible that these different forms correspond to the rat equivalent of LKB1 L and LKB1 S . Sakamoto et al. (17) described a faster migrating protein in testis detected in Western blots probed with an LKB1 antibody, which was absent from mice homozygous for a floxed allele of LKB1. Importantly, the method by which the floxed allele was generated in that study would result in the absence of LKB1 S expression. This is because a cDNA cassette encoding exons V-VIII and IXb of LKB1 L was used to replace part of the LKB1 gene, eliminating any possibility of alternate splicing. We used this animal model to investigate whether there is an absence of LKB1 S in the testis of these mice. No activity was detected in immune complexes isolated from testis of these animals using anti-sLKB1 antibodies for immunoprecipitation (data not shown). The simplest interpretation of these results is that the lower molecular mass band seen by Sakamoto et al. (17) is LKB1 S . Our finding that LKB1 S is predominantly expressed in the testis leads us to speculate that the male sterility of the LKB1 floxed mice described in the earlier study is caused by a lack of LKB1 S . In preliminary studies we have found that the absence of LKB1 S leads to a defect in spermatogenesis. 5 It will be interesting to investigate more fully the role of LKB1 S in spermatogenesis and in particular its relationship with SNRK (Snf1-related kinase), a testis specific AMPK-related kinase.
In mouse tissues we show that LKB1 forms a complex with MO25␣ and STRAD␣. Two isoforms (␣ and ␤) of both MO25 and STRAD have been identified (1,2). However, we have been unable to examine the association of LKB1 with the ␤ isoforms of MO25 and STRAD because of the lack of available antibodies. Further studies are required to determine whether LKB1 L and LKB1 S interact with both the ␣ and ␤ isoforms of MO25 and STRAD or whether there is some selectivity in formation of the complex in vivo. Using a recombinant expression system, LKB1 L was shown to form a complex with both the ␣ and ␤ isoforms of MO25 and 5 F. C. Denison and A. Woods, unpublished results. The cells were fixed in 4% paraformaldehyde and stained with a mouse monoclonal anti-LKB1 antibody (Ley37D/G6) that recognizes both LKB1 L and LKB1 S . Primary antibodies were detected with fluorescently linked secondary antibodies and visualized using a Leica TCS SP1 confocal microscope. B, HEK293 cells were treated with or without 20 M forskolin for 30 min and fractionated into cytosolic (C), membrane (M), and nuclear (N) enriched fractions. An equal volume of each fraction was blotted with an anti-LKB1 antibody (Ley37D/G6). The same fractions were used to determine the expression of marker proteins for each of the fractions (Cytosolic, glyceraldehydes-3phosphate dehydrogenase (GAPDH); Membrane, Na ϩ /K ϩ ATPase; Nuclear, CREB). C, cytosolic, membrane and nuclear fractions obtained from mouse testis were blotted with an anti-LKB1 antibody (Ley37D/G6). In each case, the blots shown are representative of blots obtained from three independent experiments. The migration of molecular mass markers is indicated.

TABLE 1 Differential gene expression in the testis of wild type versus LKB1 S knock-out mice
Microarray analysis was carried out to determine changes in gene expression in the testis of LKB1 S knock-out mice compared with that of wild type. Those genes with a fold change (FC) of greater than 1.5 compared with wild type and a p Ͻ 0.01 are listed. The accession numbers are shown in brackets. Negative values refer to decreased expression in the LKB1 S knock-out relative to wild type. STRAD (2). In our current study, we show that LKB1 S expressed in CCL13 cells (lacking endogenous LKB1) forms an active complex with endogenous MO25␣ and STRAD␣. Overexpression of either LKB1 L or LKB1 S in cells that do not express endogenous LKB1 showed that both forms were able to activate endogenous AMPK to a similar extent, showing that both forms are functionally equivalent, at least under these conditions. At present, there is some controversy regarding the regulation of LKB1 by post-translational modification. There have been several reports in the literature specifying a role for phosphorylation of Ser-428/431 by upstream kinases. Two recent studies have suggested an important role for phosphorylation of Ser-428/431 in axon differentiation (24) and neuronal polarization via activation of SAD-A and SAD-B (also known as brain-specific kinase 1 and 2) by LKB1 (23). However, we recently reported that phosphorylation of LKB1 by PKA had no effect on brain-specific kinase 1 or 2 (31). Another study reported that phosphorylation of LKB1 at Ser-431 did not alter its ability to phosphorylate p53, but S431A mutation abolished the ability of LKB1 to arrest cell growth (21). In our study, activation of the PKA pathway by forskolin increased Ser-431 phosphorylation of LKB1, but this had no effect on LKB1 activity. We did observe a small increase in AMPK activity upon forskolin treatment, but this appears to be independent of LKB1, because the effect is observed following expression of a catalytically inactive form of LKB1. These results show that phosphorylation of LKB1 on Ser-428/431 does not alter its activity and therefore does not support the notion that this site is involved in differential regulation of LKB1 L relative to LKB1 S .
There have been a number of studies investigating the subcellular localization of LKB1. LKB1 L overexpressed in COS cells was localized mainly to the nucleus, but co-expression with STRAD caused some relocalization of LKB1 L to the cytoplasm (1), and co-expression with STRAD and MO25 resulted in predominantly cytoplasmic localization of LKB1 L (2). In our present study, LKB1 L or LKB1 S expressed alone were distributed in both the cytoplasm and the nucleus. Co-expression with MO25␣ and STRAD␣ resulted in a striking redistribution of both LKB1 L and LKB1 S to a largely cytoplasmic localization (Fig. 7). Our interpretation of these results is that when LKB1 is expressed alone a proportion of it forms a complex with endogenous MO25 and STRAD and is localized in the cytoplasm, whereas some remains unassociated and localizes to the nucleus. In these heterologous expression systems, the amount of LKB1 that is unassociated and localized in the nucleus will depend on the relative levels of expressed LKB1 with endogenous MO25 and STRAD. When all three proteins are co-expressed, the amount of unassociated LKB1 is relatively low, and therefore LKB1 localizes predominantly to the cytoplasm. A recent study reported that STRAD␣ prevents shuttling of LKB1 from the cytoplasm to the nucleus by competing with importin ␣ for binding of LKB1 (32), which could explain the fact that if STRAD␣ is limiting, LKB1 localizes to the nucleus.
We found that endogenous LKB1 L in HEK 293 cells was undetectable in the nucleus but was readily detected in the cytosolic and membrane fractions. Interestingly, treatment with forskolin appeared to cause a decrease in the amount of LKB1 L in the membrane fraction. Two previous studies, how-ever, reported that phosphorylation of Ser-431 did not alter the amount of LKB1 associated with the membrane fraction (6,21). Given these conflicting results it will be important to investigate further whether phosphorylation of Ser-428/431 could play some role in membrane localization of LKB1 L . Other studies have implicated a role for phosphorylation of Ser-428/431 in regulating the nuclear export of LKB1 L . Phosphorylation of Ser-428 by PKChas been reported to lead to the nuclear export of LKB1 L in endothelial cells (5,33,34). However, our data and that of others (1,2) shows that LKB1 in complex with MO25 and STRAD is not detected in the nucleus. Moreover, in HEK293 cells we have been unable to detect endogenous LKB1 L in the nuclear fraction irrespective of the conditions used to treat the cells. This finding suggests that LKB1 L is excluded from the nucleus independently of the level of Ser-428/431 phosphorylation. We cannot rule out the possibility that LKB1 might be present in the nucleus in certain cell types. However, using mouse testis, the one tissue in which both LKB1 L and LKB1 S are highly expressed, we found no evidence to suggest that either form is present in the nucleus in vivo.
One of the obvious differences between LKB1 L and LKB1 S is the presence of a farnesylation site (Cys-430/433) in LKB1 L that is not conserved in LKB1 S . Farnesylation of Cys-430/433 has been shown previously to be involved in the membrane localization of LKB1 L (6). To examine the subcellular localization of LKB1 S , we used mouse testis, because this is the only tissue with readily detectable expression of LKB1 L and LKB1 S . Both forms were detected primarily in the cytosolic fraction, although we did detect some LKB1 S in the membrane fraction. It is possible, therefore, that LKB1 S may be able to associate with the membrane fraction even though it lacks a farnesylation motif. Whatever the role of the farnesylation modification in LKB1 L , it is clearly not required for a functional LKB1 complex, because LKB1 S in complex with MO25, and STRAD activates AMPK with a similar efficiency to the LKB1 L complex.
In this study we have identified a new form of LKB1 generated by alternative splicing. The finding that this form is predominantly expressed in testis suggests that it may play a specific role in this tissue. As a first step to explore the physiological role of LKB1 S , we conducted a microarray analysis of gene expression in testis. Our results show that deletion of LKB1 S results in altered expression of many genes ( Table 1), demonstrating that LKB1 S has a specific and nonredundant role in testis. This is consistent with the observation that the only detectable phenotype in mice lacking LKB1 S is male infertility, which may be caused by a defect in spermatogenesis (17,35). 6 In future studies it will be interesting to explore the consequences of this altered gene expression in fertility, which may lead to new insights into causes of male sterility.