Regulation of 5′-AMP-activated Protein Kinase Activity by the Noncatalytic β and γ Subunits

The mammalian 5′-AMP-activated protein kinase is a heterotrimer consisting of an α catalytic subunit and β and γ noncatalytic subunits, each of which is represented in a larger isoprotein family, related to the SNF1 kinase and its interacting proteins in yeast. In this study, we have used mammalian cell transfection to compare the activities of the two α subunit isoforms, α-1 and α-2, and to study the influence of the noncatalytic subunits on enzyme subunit association and activity. Expression of epitope-tagged protein subunits in COS7 cells indicates detectable but low level kinase activity for each of the two catalytic α subunits. Co-expression of α subunits with the β or γ subunits modestly increases kinase activity accompanied by the formation of α/β or α/γ heterodimers. Co-expression of all three subunits, however, is accompanied by a 50-110-fold increase in kinase activity with the formation of a heterotrimeric complex. In addition to binding of each noncatalytic subunit to the α subunit, the β and γ subunits bind to each other, likely resulting in a more stable heterotrimeric complex. The increase in kinase activity associated with expression of this heterotrimer is due both to an increase in enzyme-specific activity (units/enzyme mass) and to an apparent enhanced α subunit expression. Co-expression of a catalytically defective α subunit or the β/γ-binding COOH-terminal domain of the α subunit results in reduced heterotrimeric kinase activity. The synergistic positive regulatory roles for both the noncatalytic β and γ subunits of 5′-AMP-activated protein kinase contrasts with the Snf1p kinase, where only heterodimers of Snf1p and Snf4p seem to be required for maximum kinase activity.

Isolation of AMPK to homogeneity revealed that the catalytic subunit (␣) co-purifies with two other noncatalytic subunits (␤ and ␥) (12,13). Recent cloning data indicate that for each of these subunits, there exists at least one other mammalian protein isoform (14 -19). For example, we have recently reported the identification of two different catalytic subunits, ␣-1 and ␣-2, that are the products of unique genes with wide mammalian tissue expression (16). The ␣ subunit of the mammalian AMPK is related to the yeast SNF1 2 protein kinase family (14 -16, 18, 19). The Snf1p protein kinase is responsible for expression of the glucose-repressed genes during glucose starvation (20 -24). The noncatalytic ␤ and ␥ subunits of AMPK are related to proteins that interact with Snf1p. The AMPK-␥ subunits (␥1, ␥2, and ␥3) are homologous to the yeast protein, Snf4p (CAT3), 2 and the AMPK-␤ subunit is related to the yeast Sip1p/Sip2p/Gal83p family of proteins (14,17). All of these yeast proteins have been shown to associate with Snf1p, as judged by two-hybrid analysis, immunoprecipitation, and enzyme isolation (12,23,24). Genetic evidence, derived from yeast mutants, suggests that both Snf4p and the Sip1p/Sip2p/ Gal83p family of proteins positively regulate Snf1 protein kinase activity (20,21,23). However, direct demonstration by subunit recombination of these regulatory roles has not been reported.
Unsuccessful attempts to express an active ␣-2 isoform of the AMPK in bacteria, mammalian cells, and yeast have been reported (18,19), but these experiments did not take into account ␣ subunit heterogeneity and/or the possible necessary co-expression of the ␤ and ␥ subunits. In the present report, we have investigated the expression of AMPK subunits in COS7 cells and find that co-expression of the noncatalytic ␤ and ␥ subunits is required for optimal activity of the ␣ catalytic subunits.

EXPERIMENTAL PROCEDURES
Plasmid Construction-For mammalian cell expression, AMPK subunit cDNAs were cloned into either pEBG vector (␣-1 and ␣-2 subunits) or pMT2 vector (␤ and ␥-1 subunits) (25). In the former vector, the cloned insert is preceded by a glutathione S-transferase (GST) se-quence, which enables isolation of expressed protein on glutathioneagarose (25). The pMT2 vectors employed also had 5Ј-end epitope tags (peptide sequences derived from either hemagglutinin (HA) or c-myc protein), which enabled detection of expressed proteins by immunologic techniques employing epitope-specific antibodies (25). In the current studies, rat cDNAs for the ␣-1, ␣-2, and ␤ subunits were employed; the ␥ subunit sequence was cloned from a human cDNA library, as previously reported (15)(16)(17). The latter represents the ␥-1 subunit of the ␥ isoform family (17); it is referred to throughout this report simply as ␥.
In addition to these native subunit sequences, two other pEBG expression plasmids were employed in efforts to generate a kinase inhibitor. The first was constructed after site-specific mutagenesis of the ␣-1 subunit at its catalytic Lys 45 , as below. The second of these plasmids was constructed using the COOH-terminal half of the ␣-1 subunit (amino acids 262-548), which omits the catalytic kinase subdomain (16).
Site-directed Mutagenesis-Site-directed mutagenesis of the ␣-1 subunit catalytic lysine (Lys 45 ) to arginine was performed using the transformer site-directed mutagenesis kit and the trans oligonucleotide AflIII/BglII (Clontech). NotI linkers were added to the 5Ј and 3Ј termini of ␣-1 cDNA in pBluescript. The mutant oligonucleotide (5Ј-TTGCTGT-GAGGATCCTCAACCGGC-3Ј) (Midland Certified Reagent Company) was phosphorylated with polynucleotide kinase, and the mutagenesis was performed as per the manufacturer's protocol. Following bacterial transformation and colony isolation, inserts were sequenced on an ABI DNA sequencer for verification of mutation. The proper clone was then cut from pBluescript with NotI, and the mutated ␣ 1 cDNA (␣-1 K-ϾR) was then ligated into the NotI site of pEBG for cellular expression.
Cell Culture, Transfection, and Lysis-COS7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin/streptomycin, and glutamine. Cells were plated at a density of 3.5 ϫ 10 5 cells/well in six-well tissue culture plates the day before transfection. On the day of transfection, cells were washed once with serum-free, antibiotic-free medium and then transiently transfected using the lipofectamine reagent (Life Technologies, Inc./ BRL). Each well was transfected with a total of 2 g of DNA in 10 l of lipofectamine in 1.0 ml of the same medium by incubation for 5 h at 37°C in a 5% CO 2 incubator. In all the transfections reported herein, the total amount of DNA and the total amount of each vector per well were kept constant by inclusion, where appropriate, of expression plasmid lacking insert cDNA. After 5 h, an equal volume of medium containing 20% fetal bovine serum and antibiotics was added, and the cells were incubated overnight. The following day the medium was changed to the standard growth medium, and cells were then incubated for an additional 24 h.
For harvesting of expressed cell proteins, cells were washed two times in ice-cold phosphate-buffered saline and then lysed on ice for 15 min in 250 l of buffer containing Tris-Cl (20 mM, pH 7.4), NaCl (50 mM), NaF (50 mM), Na 4 P 2 O 7 (5 mM), dithiothreitol (2 mM), sucrose (0.25 M), Triton X-100 (1%), and several protease inhibitors, as in (9). After 15 min the cells were scraped from the wells, and the lysate from two identical wells were combined followed by vigorous vortexing for 15 s. Lysates were then centrifuged at 14,000 rpm at 4°C to remove cellular debris. This supernatant was then divided into aliquots for adsorption to glutathione-agarose beads or for preparation of denatured samples for SDS-polyacrylamide gel electrophoresis analysis.
AMPK Assay-200 l of cell lysate was added to 200 l of lysis buffer and 40 l of a 1:1 suspension of glutathione-agarose beads (Sigma). The tubes were incubated on a rotator at 4°C for 3 h. After 3 h the beads were harvested by brief centrifugation and washed successively (two times with each buffer; 250 l/wash; 4°C) with lysis buffer, high salt wash buffer (Tris-Cl (0.1 M, pH 7.4), LiCl (0.5 M), Triton X-100 (0.1%), and dithiothreitol (1 mM)) and 4 ϫ AMPK assay buffer (NaHEPES (0.24 M, pH 7.4), NaCl (0.48 M), dithiothreitol (4 mM)). The beads were then resuspended in 20 l of 4 ϫ AMPK assay buffer, and AMPK activity was assessed by measurement of incorporation of 32 P into the SAMS substrate peptide, as in Ref. 9. Under these conditions, routine assays were linear for up to 10 min; in preliminary experiments (not shown), preelution of the adsorbed proteins with free glutathione gave total activity similar to immobilized enzyme.
Immunoblotting of Lysate and Glutathione-Agarose Adsorbed Proteins-Cell lysates or glutathione-agarose adsorbed proteins were also analyzed by immunoblotting after SDS-polyacrylamide gel electrophoresis. The glutathione-agarose adsorbed proteins were eluted by incubation of beads at 95°C in a 2% SDS buffer prior to preparation of the gel sample. Immunoblotting was performed with a panel of antibodies including anti-GST (Sigma), anti-HA, anti-c-myc, and/or with subunit-specific anti-peptide antibodies. Antibody binding was detected with the appropriate secondary antibody conjugated with horseradish peroxidase and with enhanced chemiluminescence (Amersham Corp.). Exposed films (Kodak XAR5) were analyzed on a Molecular Dynamics Scanning Densitometer and quantitated using IPLab Gel software.
Materials-COS7 cells were obtained from American Type Culture Collection. Media, fetal bovine serum, and lipofectamine were purchased from Life Technologies, Inc. [␥-32 P]ATP was purchased from ICN. Polyvinylidene difluoride membrane for immunoblotting was obtained from Millipore. Glutathione-agarose, secondary antibodies, and most chemicals were purchased from Sigma. The pEBG and pMT2 vectors and antibodies against the HA and c-myc epitopes were kindly supplied by Drs. John Kyriakis and Joseph Avruch (Massachusetts General Hospital).

RESULTS
Expression of the ␣-1 and ␣-2 Catalytic Subunits-The 5Ј-AMP-activated protein kinase consists of one catalytic subunit and two noncatalytic subunits (12). Two isoforms of the catalytic subunit of AMPK have recently been identified and have been designated ␣-1 and ␣-2 (15,16). Previous attempts by others to express an active ␣-2 isoform in mammalian cells, bacteria, and yeast were unsuccessful (18,19). In our initial experiments, we therefore examined the expression of these two isoforms as GST fusion proteins by transient transfection of COS7 cells and assessed kinase activity after the adsorption of expressed protein to glutathione-agarose. As shown in Fig.  1A, a 90-kDa fusion protein for each isoform is readily detected in the glutathione-agarose adsorbates using an anti-GST antibody. This molecular mass is that predicted from ␣ sequence (63 kDa) and the GST fusion sequence (27.7 kDa). SAMS peptide phosphotransferase activity of each of the expressed catalytic subunits is readily detectable in the glutathione-agarose adsorbates (Fig. 1B). Although the ␣-2 expressed activity appears to be lower in these representative experiments, correction for the mass of the expressed subunit indicates nearly equal specific activities (units kinase activity corrected for fusion protein expressed) for each under these assay conditions (data not shown). These data indicate that the free catalytic subunits of AMPK possess phosphotransferase activity that can be expressed in mammalian cells.
Activation of the ␣ Subunit by ␤ and ␥ Binding-As isolated from rat liver by substrate affinity chromatography and immu-FIG. 1. Cellular expression and activity of the ␣-1 and ␣-2 isoforms of the AMPK catalytic subunit. A, glutathione-agarose adsorbates were prepared from COS7 cells after transfection with either pEBG-␣1 (␣1), pEBG-␣2 (␣2), or pEBG alone (control, c), and the adsorbed proteins were subjected to SDS-polyacrylamide gel electrophoresis and visualized by immunoblot analysis using an anti-GST antibody. The position of the phosphorylase molecular mass standard (94 kilodaltons (K)) is shown to the left. B, the expressed phosphotransferase activities of both ␣-1 (Alpha-1) and ␣-2 (Alpha-2) from cellular lysates after transfection were determined following adsorption to glutathioneagarose and kinase assay, as described under "Experimental Procedures." Kinase activity is expressed as pmol 32 P transferred to the SAMS peptide/min/ml of lysate adsorbed and is represented as the mean (Ϯ standard deviation) of four independent transfections of each. Phosphotransferase activity has been corrected by subtraction of apparent kinase activity in adsorbates from mock transfected (pEBG only) cells; this value averaged less than 1% of 32 P incorporation in the experimental cell samples. noadsorption, both ␣-1 and ␣-2 catalytic subunits co-purify with ␤ and ␥ noncatalytic subunits (12,16). To document the binding characteristics of the three AMPK subunits, a series of transfection experiments were performed using both the ␣-1 and ␣-2 GST fusion expression constructs and expression plasmids encoding the ␤ and ␥ subunits. These latter plasmids expressed the ␤ and the ␥ subunits tagged on the 5Ј-end with a HA decapeptide. A series of co-transfection experiments in COS7 cells were performed with these plasmids to determine whether the individual ␣ subunit proteins could both bind ␤ and/or ␥ and to determine if this binding had an effect on kinase activity. As shown in Fig. 2A (upper panel), both HA-␤ and HA-␥ fusion proteins can be expressed in COS7 cell lysates when co-expressed with either ␣-1 or ␣-2 GST fusion proteins. Analysis of glutathione-agarose adsorbates following the double transfection with ␣/␤ or ␣/␥ reveals that both HA-tagged noncatalytic subunit proteins can bind to the GST-␣ subunit ( Fig. 2A, lower panel). The extent of binding of HA-␤ and HA-␥ to ␣-1 or ␣-2 appears to be approximately equivalent under these conditions. No immunoblottable HA-␤ or HA-␥ protein was adsorbed to glutathione-agarose in the absence of expressed ␣ subunit (not shown).
As assayed in the glutathione-agarose adsorbed state, coexpression of the ␤ subunit with ␣-1 causes a modest 1.5-fold increase in the SAMS peptide phosphotransferase activity (Fig.  2B). When ␥ is co-expressed with ␣-1, a 2.5-fold increase in the activity of ␣-1 was seen (Fig. 2B). Similar quantitative effects of co-expressed ␤ or ␥ were also seen with ␣-2 (not shown). As shown in the upper panel of Fig. 2B, co-expression of either ␤ or ␥ did not significantly alter the expression of the ␣-1 polypeptide. These data therefore demonstrate that the ␤ or ␥ individually have a modest stimulatory effect on activity upon binding to the ␣ catalytic unit of AMPK.
Triple transient transfection was employed to assess the effects of ␤ and ␥ in combination on ␣ subunit activity. In all instances, 2 g of total DNA was transfected per well, and the amount of each plasmid type was kept constant by the addition, where appropriate, of plasmid lacking the cDNA insert. When co-transfected with either ␣-1 or ␣-2, both HA-␤ and HA-␥ bind to the glutathione-adsorbed GST ␣ fusion protein (Fig. 3, righthand panel). The HA-␤ subunit is expressed as two closely migrating bands at 43 and 46 kDa, whereas HA-␥ is expressed as a single band at ϳ41 kDa as previously reported (17). The total binding of both noncatalytic subunits to ␣-1 or ␣-2 under these conditions is roughly equivalent.
The impact of co-expression of both ␤ and ␥ on the catalytic activity of the ␣ subunit is large in contrast to the relatively small effects of individual noncatalytic ␤ or ␥ subunits (Fig. 3, left-hand panel). ␣-1 kinase activity in these experiments is stimulated about 110-fold by co-expression of both ␤ and ␥, whereas ␣-2 activity is stimulated about 50-fold under the same conditions of transfection. This striking increase in kinase activity on triple transfection ranged from 50-to Ͼ500fold in other experiments (see, for example, Fig. 6). Thus, ␤ and ␥ have a pronounced synergistic effect on the expressed activity of both isoforms of the ␣ subunit.
To assess possible mechanisms underlying this marked increase in SAMS peptide phosphotransferase activity, the glutathione-agarose adsorbates from these experiments were analyzed for GST-␣ content by immunoblotting. As shown in Fig.  4A, as analyzed with an anti-GST antibody, co-expression of ␤ and ␥ result in increased recovery (9 -9.6-fold, as determined by quantitative densitometry) of ␣-1 or ␣-2 GST fusion protein in the adsorbate. This same fold increase in ␣-1 and ␣-2 expression was also observed on immunoblot analysis of the cell lysates prior to glutathione-agarose adsorption, indicating that co-expression of ␤/␥ subunits increases the overall level of cellular ␣ subunit expresssion under these conditions (data not shown). However, increased ␣ subunit expression does not entirely account for the increased in expressed kinase activity. When SAMS peptide kinase activity was corrected for expressed ␣ protein in the same glutathione-agarose adsorbate (determined by immunoblotting and scanning densitometry), 12.3-(␣-1) and 5-fold (␣-2) increases in kinase activity due to  subunit co-expression are still evident (Fig. 4B). Thus, ␤/␥ appear to have two effects on the ␣ subunit: to increase its level of expression and to increase its specific activity (activity per mass of ␣ subunit).
Interactions of the ␤ and ␥ Subunits-Experiments shown above indicate that ␤ and ␥ can bind individually to either of the ␣ subunit isoforms. To determine whether ␤ and ␥ can bind to one another, another series of co-transfection experiments with ␣/␤/␥ in all combinations were performed in COS7 cells. In these experiments, the HA epitope tag of the ␤ subunit was replaced with a c-myc epitope, so that each of the subunits had a unique fusion epitope sequence. Cell lysates were then immunoprecipitated with either anti-c-myc or anti-HA antibodies; these immunoprecipitates were then separated by SDSpolyacrylamide gel electrophoresis and immunoblotted. As shown in Fig. 5, HA-tagged ␥ can be immunoprecipitated with anti-c-myc antibody either when expressed with c-myc-␤ alone (Fig. 5, lane 5) or with both ␣-1 and c-myc-␤ (Fig. 5, lane 6). Similarly, c-myc-tagged ␤ can be precipitated with anti-HA antibody either when expressed with HA-␥ alone (Fig. 5, lane  11) or with both ␣-1 and HA-␥. Similar data were seen for co-expression of ␤ and ␥ with ␣-2 (not shown). These data indicate that the heterotrimeric kinase complex results not only from interactions between ␣-␤ and ␣-␥ but also from ␤-␥ binding.
Inhibitors of AMPK Activity-Lys 45 in subdomain II of the ␣ catalytic subunit (␣-1 and ␣-2) of the AMPK corresponds to the invariant lysine involved in the ATP binding by protein kinase (26). Mutations of this lysine typically result in loss of kinase activity (26). Site-directed mutagenesis of this lysine to arginine on the ␣-1 subunit was employed to create a catalytically inactive ␣ subunit in order to assess its ability to bind ␤ and ␥ and to serve as a potential inhibitor of kinase activity. This mutated form of ␣-1 (␣-1 Lys 3 Arg) was then co-transfected with ␤ and ␥ into COS7 cells with or without the wild type form of ␣-1. When the kinase activities of these adsorbates were examined, no kinase activity of the Lys 3 Arg mutant could be detected even in the presence of ␤/␥ (Fig. 6A). As shown in Fig.  6B, the binding of ␤/␥ to the glutathione-adsorbed ␣-1 subunit is similar in the wild type and in the Lys 3 Arg mutant. Thus, an active kinase is not necessary for ␤/␥ binding. Because the Lys 3 Arg ␣-1 is still capable of binding ␤ and ␥ and may also contain the activating phosphorylation site for the AMPK kinase (Thr 172 ) (16), it is conceivable that the mutant could act to inhibit the expressed activity of wild type ␣-1 in either the absence or the presence of ␤/␥. Indeed, when co-transfected with ␤, ␥, and wild type ␣-1, the Lys 3 Arg mutant markedly diminishes (by 90%) expressed ␣-1 activity (as compared with its activity when co-transfected with equivalent amounts of pEBG vector) (Fig. 6A). Co-expression of wild type and Lys 3 Arg mutant did not alter the expression of total ␣ subunit protein, as determined by immunoblotting either with anti-GST or anti-␣-1 antibodies (not shown). Because these immunologic reagents cannot distinguish the two fusion protein products, we cannot assess the individual expression of wild type and Lys 3 Arg mutant by this analysis. In other experiments (not shown), the ␣-1 Lys 3 Arg mutant also inhibits the activity of the co-expressed ␣-2 catalytic subunit.
Experiments above have shown that an active ␣-1 subunit is not necessary for the binding to the noncatalytic subunits, ␤ and ␥. We hypothesized that the noncatalytic region of the ␣-1 protein was responsible for the binding to ␤ and ␥ and thus might serve as an inhibitor of kinase activity by competition for available ␤/␥ subunits. Therefore, the catalytic domain of the ␣-1 subunit was removed, and the carboxyl terminus (␣-1-C; amino acid residues 262-548) was expressed as a GST-fusion protein in the presence of wild type ␣-1/␤/␥. Co-expression of ␣-1-C with wild type ␣-1 leads to a 50% decrease in kinase activity (Fig. 7A) and similar diminution in the expressed level of the wild type protein (about 50% as determined by scanning densitometry) (Fig. 7B, upper panel). It should be noted in this panel that an ␣-1-specific antibody raised against an ␣-1 peptide (amino acids 339 -358) was used for analysis, because the anti-GST antibody recognized a nonspecific protein that migrated to the same location as the truncated ␣-1 subunit. As determined by anti-HA immunoblotting, the total amount of ␤ plus ␥ subunits bound to glutathione-adsorbed proteins (wild type ␣-1, ␣-1-C, or both) remains constant, as expected (Fig. 7B, lower panel). Thus, ␣-1-C can bind both the ␤ and ␥ subunits and can also serve as an inhibitor of kinase activity under these co-transfection conditions, although it appears that its major effect is to diminish expression of the wild type ␣-1 subunit.

DISCUSSION
The AMPK, as purified from pig or rat liver, is a heterotrimeric enzyme consisting of an ␣, ␤, and ␥ subunits (12). Two isoforms of the ␣ subunit, ␣-1 and ␣-2, have recently been identified, although the former appears to account for about 90% of expressed SAMS peptide kinase activity in rat liver (16). Sequence analysis of these ␣ subunits has revealed that each contains the consensus domain present in all serine/threoninedirected protein kinases, indicating that ␣ is the catalytic subunit (12,15,16,18,26). However, efforts by others to express an active ␣ subunit (␣-2) have been unsuccessful in bacteria, yeast, and mammalian systems (18,19). This present report establishes that both the ␣-1 and ␣-2 subunits indeed contain phosphotransferase activity against the SAMS peptide, which represents one of the sites of phosphorylation of acetyl-CoA carboxylase by AMPK (9). The present study did not reveal any catalytic differences between these two ␣ subunit isoforms. However, we have previously shown that the ␣-2 isoform present in rat liver does not bind to an ␣-1 peptide substrate affinity column, possibly because the ␣-2-containing AMPK enzyme isolated from rat liver is not in its active phosphorylated form (16).
One major finding of the present investigation is that the ␤ and ␥ subunits synergistically increase the expression of AMPK activity for both ␣ subunit isoforms in transfected mammalian cells. ␤ or ␥ alone are very weak activators of kinase activity (1.5-2.5-fold), but co-expression of ␤ and ␥ with ␣ leads to a minimum of 50 -110-fold increases in expressed kinase activity. The positive regulatory effects of ␤ and ␥ on AMPK activity have previously been predicted, based on their homologies to Snf1p protein kinase interacting proteins in yeast (14,17,23). These proteins may serve as adaptors that promote the activity of Snf1p toward specific targets (23). As analyzed in yeast mutants, some data suggest that these proteins may facilitate interaction of Snf1p with unique and different targets. Mutations of Gal83p can abolish most of the Snf1p kinase activity detectable in immune complexes precipitated with anti-Snf1p antibody (23). A Sip2p⌬ gal 83⌬ mutant shows reduced Snf1 protein kinase activity that is restored following expression of either Sip2p or Gal83p LexA fusion proteins in the mutant strain (23). Analysis of SNF4 mutants in yeast suggests that Snf4p also positively regulates the activity of its associated catalytic subunit, Snf1p (20,21). However, although both the Snf4p and Sip1/Sip2/Gal83p proteins have been shown to interact with the Snf1p kinase polypeptide in the yeast two-hybrid system and by immunoprecipitation (20 -24), these studies of yeast mutants (and their complementation) do not establish the direct positive regulation of kinase activity of the catalytic unit by either of these Snf1p-interacting proteins. In the present study, we have successfully reassembled the AMPK heterotrimeric complex and isolated it by glutathione-agarose adsorption. These adsorbates demonstrate increased kinase specific activity, even when corrected for the amount of adsorbed catalytic subunit, demonstrating directly the positive regulation of the ␣ subunit by the ␤ and ␥ subunits. A parallel activating interaction for Snf1p and its ␤/␥ homolog proteins in yeast remains to be demonstrated. It should be noted, however, that on isolation of SNF1 kinase from yeast by nickel affinity chromatography, only Snf4p is co-purified with the catalytic Snf1p polypeptide, suggesting perhaps that only a heterodimer is required for optimal activity of this kinase (12). In addition to kinase activation, co-expression of ␤ and ␥ leads to an increase in the cellular expression of the ␣ subunit. Although several mechanisms might account for this, it seems most likely that the stabilization of the heterotrimeric complex by co-expression of the noncatalytic units might diminish protein turnover, allowing for higher levels of expression. This mechanism has previously been demonstrated in other multimeric protein complexes, including other protein kinases (27,28).
Another major finding revealed by the present studies is the ability of the ␤ and ␥ subunits to bind not only to the ␣ subunit but also to each other. This was not anticipated based on analogy to the yeast system, because no interactions between Snf4p and Sip1/Sip2/Gal83p proteins have been reported. The ability of ␤ and ␥ to bind to each other coupled with the independent binding of each to the ␣ subunit may contribute to increased stabilization of the heterotrimer. This ␤/␥ interaction could also contribute directly to the marked enhancement of kinase specific activity by altering the interactions of each with the ␣ subunit (resulting in a different conformation for the catalytic unit) or by increasing the interactions of ␤ and/or ␥ with kinase protein substrates. The demonstration of ␤/␥ binding in the absence of ␣ also raises the question as to whether this heterodimer might have some cellular function independent of its ability to stimulate kinase activity. Although our results indicate that heterodimers of ␣/␤ and ␣/␥ can be formed, thus far, we have not observed these heterodimers in liver extracts; only the heterotrimeric complex has been detected.
The ␤ and ␥-1 subunits used herein appear to have nearly equal affinity for either the ␣-1 or ␣-2 subunit and to have similar effects on kinase activity catalyzed by each. We have previously identified a ␥ isoform family consisting of at least three distinct ␥ proteins, and it seems likely that a ␤ isoform family exists as well (17). This raises the possibility that ␣-1 and ␣-2 could each associate preferentially with unique ␤/␥ isoforms in vivo. The ␤ and ␥-1 subunits employed in the present study have been shown by direct peptide sequence analysis to be the noncatalytic subunits that associate with ␣-1 (14,17). Whether the same or other isoforms associate with ␣-2 remains to be determined. If unique noncatalytic subunits associate with each ␣ isoform, this could increase the diversity of differential interactions between the protein target and enzyme subunits.
Based on mutagenesis of the ␣-1 subunit, we have been able to generate two inhibitors of expressed enzyme activity. Mutation of Lys 45 to arginine, as expected, abolishes ␣-1 protein kinase activity; however, the ability to bind ␤/␥ in ␣-1 Lys 3 Arg mutant is fully preserved and thus does not require an active catalytic subunit. Co-expression of this mutant with wild type ␣-1 (and ␣-2) led to a marked decrease in expressed AMPK activity. This could result from competition for binding of ␤/␥ between the wild type ␣ subunit and the Lys 3 Arg catalytically incompetent mutant or competition of mutant with the wild type ␣-1 for AMPK kinase binding required for activation of ␣-1 subunit via phosphorylation. It is also possible that because ␤/␥ binding appears to increase the expression of the ␣ subunit, perhaps through alterations in protein turnover, competition for ␤/␥ binding by the ␣-1 Lys 3 Arg mutant might increase the turnover of the free wild type ␣ subunit. Less likely as an explanation for the inhibitory properties of ␣-1 Lys 3 Arg mutant are direct dominant negative effects of the ␣-1 Lys 3 Arg mutant on the wild type ␣ by the formation of ␣ heterodimers, because ␣ subunit dimerization has not been demonstrable to date.