Non-catalytic beta- and gamma-subunit isoforms of the 5'-AMP-activated protein kinase.

The mammalian 5′-AMP-activated protein kinase (AMPK) is a heterotrimeric protein consisting of α-, β-, and -subunits. The α-subunit is the catalytic subunit and is related to the yeast Snf1p kinase. In this study, we report the cloning of full-length cDNAs for the non-catalytic β- and -subunits. The rat liver AMPK β-subunit clone predicts a protein of 30,464 Da, which is related to the Sip1p, Sip2p, and Gal83p subfamily of yeast proteins that interact with Snf1p and are involved in glucose regulation of gene expression. The AMPK β-subunit, when expressed in bacteria and in mammalian cells, migrates anomalously on SDS gels at an apparent molecular mass of 40 kDa. Rat and human liver AMPK -subunit clones predict a protein of 37,577 Da (AMPK-), which is related to the yeast Snf4p protein that copurifies with Snf1p and to a larger family of other human AMPK -isoforms. The mRNAs for both AMPK-β and AMPK- are widely expressed in rat tissues, consistent with a broad role for AMPK in cellular regulation. These data reveal a mammalian multisubunit protein kinase strikingly similar to the multisubunit glucose-sensing Snf1 kinase complex. The identification of isoform families for the AMPK subunits indicates the potential diversity of the roles of this highly conserved signaling system in nutrient regulation and utilization in mammalian cells.

Purification of pig and rat liver AMPKs has revealed a heterotrimeric kinase structure consisting of a 63-kDa catalytic ␣-subunit and non-catalytic ␤ (40 kDa)-and ␥ (38 kDa)-subunits (4,6). The AMPK ␣-subunit is 64% identical in its catalytic core to the Saccharomyces cerevisiae Snf1p 2 protein kinase, which is responsible for the glucose derepression response of the SUC1 gene (4,5,7,13). In contrast to AMPK, Snf1p occurs as a heterodimer with Snf4p and does not purify with other identified interacting proteins (4,6). We have recently found that multiple isoforms of AMPK-␣ (␣ 1 and ␣ 2 ), which are products of distinct genes, are present in liver and other tissues (14). The AMPK ␣ 1 -isoform accounts for ϳ90% of total AMPK activity in liver extracts, yet its corresponding mRNA level is low relative to that of the AMPK ␣ 2 -isoform. Preliminary peptide sequencing and limited PCR product analysis of the non-catalytic subunits have indicated that the AMPK ␥-subunit is related to the S. cerevisiae protein Snf4p (CAT3), 3 whereas AMPK-␤ is related to the S. cerevisiae Sip1p/ Sip2p/Gal83p family of proteins. These are known to associate with the Snf1p kinase and to participate in glucose-regulated gene expression (6).
In this study, we report the molecular cloning of full-length cDNAs for the mammalian AMPK ␤and ␥-subunits. These clones have been used to characterize the tissue distribution of subunit mRNA and to express subunit protein in both bacteria and mammalian cells. Knowledge of their complete sequences has also led to the identification of protein isoform families for each of these non-catalytic units.

EXPERIMENTAL PROCEDURES
AMPK Isolation and Peptide Sequencing-Porcine and rat liver AMPKs were isolated by a previously published method (4,6). Peptide sequences derived from the rat liver ␤ (40 kDa)-and ␥ (38 kDa)subunits were obtained after subunit separation by SDS gel electrophoresis, band elution, and in situ protease digestion as described (4,6).
AMPK ␤-Subunit cDNA Isolation-Peptide sequences derived from the AMPK ␤-subunit were used to generate partial-length AMPK ␤-subunit cDNAs by PCR as described (13). One product, a 309-bp cDNA, was used to screen a rat liver ZAPII cDNA library (Stratagene) as described (13). Filters were hybridized with [ 32 P]cDNA, labeled with * This work was supported in part by National Institutes of Health Grant DK35712 (to L. A. W.) and by a grant from the National Heart Foundation (to B. E. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 The abbreviations used are: AMPK, 5Ј-AMP-activated protein kinase; PCR, polymerase chain reaction; bp, base pair(s); MOPAC, mixed oligonucleotide-primed amplification of cDNA; RACE, rapid amplification of cDNA ends; HA, hemagglutinin.
[␣-32 P]CTP (3000 mCi/mmol; DuPont NEN) by random priming (random primer cDNA labeling system, Life Technologies, Inc.), in 50% formamide, 10 ϫ Denhardt's solution, 1 M NaCl, 50 mM Tris-Cl (pH 7.5), and 100 g/ml salmon sperm DNA at 42°C for 18 h. They were then washed at room temperature 3 ϫ 10 min and then at 55°C for 15 min. Autoradiography was overnight at Ϫ80°C. All plates were lifted in duplicate, and positive plaques were purified through three additional rounds of plating and rescreening.
AMPK ␥-Subunit cDNA Isolation-For the AMPK ␥-subunit, we initially generated a 67-bp cDNA by the MOPAC technique (15). Degenerate PCR primers were synthesized corresponding to the N-and C-terminal sequences of a 17-amino acid rat liver AMPK-␥ peptide (VVDIYSKFDVINLAAEK). The sequence of the sense primer was GCGGATCCGTNGAYATHTA, and the sequence of the antisense primer was CGGAATTCYTTYTCNGCNGCNA. BamHI and EcoRI sites were added to the 5Ј-ends of these primers. The strategy was to create a non-degenerate nucleotide sequence corresponding to the middle portion of the peptide sequence that would be used in library screening. Total rat liver cDNA, prepared with oligo(dT) and random hexamers (pre-amplification kit, Life Technologies, Inc.), was used with PCR to amplify a 67-mer (including primers) oligonucleotide corresponding to a portion of the AMPK-␥ cDNA. The purified PCR product was digested with BamHI and EcoRI and ligated into pBluescript plasmid for transformation of DH5␣ bacteria. Colony hybridization was employed to identify clones of interest; colonies were lifted from replica plates onto nitrocellulose filters. Following bacterial lysis and DNA denaturation (16), filters were probed with a mixture of two 32 P-end-labeled degenerate oligonucleotide probes corresponding to amino acid sequence (KFDVINLA) internal to that of the two PCR primers. These oligonucleotides (oligonucleotide 1, AARTTYGAYGTNATHAAYCTNGC; and oligonucleotide 2, AARTTYGAYGTNATHAAYTTRGC) were added in a ratio of 2 parts oligonucleotide 1 to 1 part oligonucleotide 2 to reflect the degeneracy of the Leu codon. Positive colonies were identified, and plasmid DNA was isolated from each for sequence analysis. One such cDNA was chosen, and the non-degenerate "core" 23-mer oligonucleo-tide sequence was then synthesized for use in library screening (CTC-CAAGTTTGATGTTATCAACC). Screening of ϳ10 6 plaques with this probe, however, did not yield any positive clones.
The non-degenerate 23-mer cDNA was then used in conjugation with degenerate primers constructed from two other peptide sequences to generate a larger AMPK-␥ cDNA by PCR. Both sense and antisense degenerate oligonucleotide primers corresponding to the peptide sequences EELQIG and FPKPEFM were used together with the sense MOPAC-derived non-degenerate sequence to generate all possible PCR products, using rat liver cDNA as template. The largest product (192 bp) obtained was subcloned in pCR-Script (Stratagene) and sequenced. This sequence (see Fig. 4), which actually predicted amino acid sequence corresponding to all three AMPK-␥ peptides used in the PCR strategy, was then used for library screening as described above. Screening of 2 ϫ 10 6 plaques with this larger PCR product yielded several positive clones, which were further characterized (see below); however, none of the rat cDNAs (1-1.3 kilobases) isolated corresponded to a full-length open reading frame. In an effort to extend the sequence to the 5Ј-end of the open reading frame, a primer extension library was constructed using a AMPK-␥-specific antisense primer (Stratagene; ZAPII). Additional screening of this library, while yielding some 5Јextended sequence, did not yield the start Met codon. The application of a 5Ј-RACE strategy with rat liver cDNA was also unsuccessful in attempts at sequence extension, although a 5Ј-RACE product from porcine liver was obtained (data not shown). The most 5Ј rat cDNA sequence (520 bp) was then used to screen a human fetal liver library, which yielded a full-length AMPK-␥ cDNA (see below).
Plasmid Preparation and DNA Sequencing-Plasmid DNA was prepared using QIAGEN mini-or midi-columns according to the manufacturer's instructions. DNA was sequenced, with vector-or gene-specific primers, using an Applied Biosystems Prism TM ready reaction dye deoxy terminator cycle sequencing kit and cycled in a Perkin-Elmer PCR Thermocycler according to the manufacturers' instructions. Dye terminators were removed from the resulting sequence reactions using a Centri-Step column (Princeton Separations, Inc.). The purified se-

FIG. 1. Nucleotide and deduced amino acid sequences of rat liver AMPK-␤.
Shown is the nucleotide and deduced amino acid sequences of the 1107-bp rat liver AMPK-␤ clone. The underlined peptide sequences correspond to those determined by direct peptide sequencing of the isolated rat liver AMPK-␤ protein, as previously reported (6) and as completed for this study (new sequences include amino acids 55-57, 108 -125, 161-181, and 183-199). Assignment of the start codon is explained under "Results." quencing reactions were then dried in a Speed-Vac and analyzed on an automated DNA sequencer (Applied Biosystems Model 373).
Bacterial Expression of cDNAs-Full-length rat AMPK ␤-subunit cDNA and a partial-length rat AMPK-␥ subunit cDNA (amino acids 33-331; see Fig. 5) were expressed in Escherichia coli using the pET vector system, which introduces polyhistidine (His 6 ) and T7 fusion epitope tag sequences (Novagen), according to the manufacturer's protocols. Bacterial expression was induced with 1.0 mM isopropyl-1-thio-␤-D-galactopyranoside at 37°C for 2 h; expressed protein was detected by both Coomassie Blue staining and immunoblotting with anti-T7 monoclonal antibody (Novagen). The fusion proteins were purified from the inclusion bodies of bacteria by nickel affinity chromatography under denaturing conditions. His 6 -AMPK-␤ or His 6 -AMPK-␥ was solubilized from the inclusion bodies in 6 M urea according to manufacturer's instructions. After sample application, the column was washed extensively with 20 mM Tris-Cl (pH 7.9), 0.5 M NaCl, 20 mM imidazole, and 6 M urea. The His 6 -protein was eluted with the same buffer containing 300 mM imidazole.
Cellular Expression of cDNAs-Full-length rat AMPK-␤ cDNA, a partial-length rat AMPK-␥ cDNA (amino acids 33-331), and full-length human AMPK ␥-subunit cDNA were also expressed in COS-7 cells. cDNAs were cloned into a pMT2 vector in frame with a hemagglutinin (HA) epitope tag (pMT2-HA) (17). Transfection was done using Lipofectamine reagent (Life Technologies, Inc.) according to the manufacturer's general protocol. Cells were plated at 3 ϫ 10 5 /well in 6-well plates (Corning Inc.) in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and penicillin/streptomycin. The following day, the cells were switched to serum-free and antibiotic-free Dulbecco's modified Eagle's medium, and then Lipofectamine-DNA conjugates (2 g of DNA; 10 l of Lipofectamine/well) diluted in the same medium were added. After 5 h of incubation at 37°C, an equal volume of medium containing 20% fetal calf serum was added to each well. The following morning, the medium was switched to the original cell medium. Cells were harvested 48 h after transfection. After washing with phosphatebuffered saline, cells were lysed in buffer containing 50 mM Tris-Cl (pH 7.5), 100 mM NaCl, 50 mM NaF, 5 mM NaPP i , 1 mM EDTA, 2 mM dithiothreitol, and 0.5% Nonidet P-40 with several protease inhibitors (9). For complete lysis, cells were placed on ice for 15 min, followed by scraping and vigorous vortexing (15 s) of the lysate. After clearing of debris by brief centrifugation, this lysate was used for SDS gel electrophoresis and immunoblotting. Blots were probed with an anti-HA monoclonal antibody (derived from the 12CA5 hybridoma line). After secondary probing with a peroxidase-conjugated anti-mouse IgG antibody, blots were developed by ECL (Amersham Corp.).
DNA Sequence Analysis and DNA Sequences-DNA sequences were analyzed using MacVector ® and the Genetics Computer Group software package. Sequences were compared to the data base using BLAST and Genetics Computer Group software; amino acid alignments were made using the Genetics Computer Group Pileup program. Sequences were formatted using an Excel ® macro (18). The DNA sequences in this report have been deposited in the GenBank TM Data Bank with the following accession numbers: U42411, rat liver AMPK-␤; U42413, rat liver AMPK-␥; and U42412, human fetal liver AMPK-␥.
Materials-[␣-32 P]CTP for cDNA probe labeling was purchased from DuPont NEN. Nitrocellulose for Northern blots and library screening was purchased from MSI and Schleicher & Schuell. Gene-specific primers for PCR and DNA sequence analysis were synthesized by Midland Certified Reagent Co. (Midland, TX).

AMPK-␤ cDNA Isolation and
Characterization-PCR amplification of pig and rat liver cDNAs with degenerate oligonucleotides corresponding to selected AMPK-␤ peptide sequences yielded two major PCR products (6). One product, a rat 309-bp partial-length cDNA, was used to screen a rat liver cDNA library, yielding a 1107-bp clone, shown in Fig. 1. The screening PCR probe corresponded to nucleotides 279 -588 of this sequence. Shown in Fig. 1 (underlining) are the endoproteinase Lys-C peptide sequences obtained by direct sequencing of the purified rat liver AMPK-␤ protein. These account for 60% of the cDNA-derived sequence and provide strong evidence that the identity of the clone is correct.
This clone contains an open reading frame encoding for a 270-amino acid peptide, which contains all of the 15 independent (some overlapping) peptide sequences obtained from extensive sequence analysis of the purified protein. The translational start methionine codon is assigned from the typical Kozak sequence present for a initiation codon (19) and the lack of any other upstream in-frame methionine codons. While no in-frame stop codon is present in this 5Ј-upstream sequence, a human expressed sequence tag cDNA (GenBank TM accession number T78033; see below) in the data base contains such a stop codon preceding the same assigned start methionine codon. This reading frame, however, predicts a protein of 30,464 Da, well below the estimated molecular mass of 40 kDa evident on SDS gel electrophoresis (4,6).
To clarify the size of the protein product that could be synthesized from this cDNA, the AMPK-␤ clone was expressed both in bacteria and mammalian cells. As shown in Fig. 2, in both expression systems, the protein product migrates at a higher than predicted molecular mass. When purified as a His 6 -tagged fusion protein from E. coli, the isolated protein migrates on SDS gels with an apparent molecular mass of ϳ43,000 Da (the same as the ovalbumin standard). This corresponds to a AMPK-␤ polypeptide product of 40 kDa with an additional 3 kDa of fusion tag sequence derived from the pET vector. When expressed in mammalian cells from an HAtagged expression vector, two polypeptides are evident, with the major product corresponding to a 40-kDa species (after correction for the size of the HA epitope tag). A second product of 42-43 kDa is also evident using this expression system. Taken together, these data demonstrate that the protein product of this AMPK-␤ migrates on SDS-polyacrylamide gel electrophoresis with an anomalously high molecular mass.
Comparison of the rat liver AMPK-␤ sequence to the data base reveals that it is highly homologous to three yeast proteins (Sip1p, Sip2p, and Gal83p) and to two recently cloned human expressed sequence tag cDNA sequences (Fig. 3). This alignment, as gapped according to the sequence of the S. cerevisiae protein Sip1p (20), is most striking at the C terminus of AMPK-␤ and these yeast proteins.
AMPK-␥ cDNA-Using the MOPAC procedure and other PCR amplification protocols, a 192-bp cDNA corresponding to rat liver AMPK-␥ sequence was obtained and used for library screening. Despite exhaustive attempts, only a partial-length rat liver cDNA of ϳ1.3 kilobases could be obtained from this library; this sequence did not contain either a start methionine codon or all the peptide sequences obtained from the purified protein. Attempts to extend this sequence to the 5Ј-end by the use of a primer extension library and 5Ј-RACE only succeeded in adding ϳ200 nucleotides to this sequence without identification of the start codon. A partial-length rat cDNA was then used to screen a human fetal liver library, which did yield the full-length clone shown in Fig. 4. This clone contains deduced amino acid sequence corresponding to all of 22 independent (some overlapping) peptide sequences obtained from the purified rat and porcine liver AMPKs-␥ (shown by underlining in Fig. 4), confirming clonal identity.
A typical Kozak translation initiation sequence surrounds the assigned methionine start codon; this start is also in frame with a 5Ј-upstream stop codon. The assigned start methionine codon is followed by an open reading frame predicting a protein of 331 amino acids and of 37,546 Da, which corresponds to the molecular mass observed on SDS gel electrophoresis of the protein as purified from rat and porcine liver (4,6). Expression of a truncated rat AMPK-␥ cDNA (amino acids 33-331) and the full-length human AMPK-␥ (331 amino acids) in COS-7 cells yields products consistent with the molecular mass predicted for each cDNA (34,081 and 37,577 Da, respectively) (Fig. 2). The rat liver AMPK-␥ product expressed in bacteria also displayed the molecular mass predicted by the cDNA (data not FIG. 3. Yeast and human homologs of rat liver AMPK-␤. Shown are alignments of the rat liver AMPK-␤ deduced amino acid sequence with S. cerevisiae proteins Sip1p, Sip2p, and Gal83p and with two partial-length human cDNAs in the data base (GenBank TM accession numbers T78033 and F11147), as matched by BLAST searching. Sequences were aligned with the Pileup program of the Genetics Computer Group software package and are gapped with reference to the Sip1p sequence. Residues identical to AMPK-␤ are boxed. The NCBI data base also contains four other human sequences that are either very similar or identical to the two human sequences indicated (data not shown); these sequences are accessible as R14746, H06094, R20494, and R25722.
shown). Thus, unlike AMPK-␤, there is no anomalous migration of the protein product of AMPK-␥ cDNA.
Comparison of the human and rat liver AMPK-␥ amino acid sequences to the data base yields a significant alignment of this protein with S. cerevisiae Snf4p (Fig. 5). In addition, our hu-man (and rat; data not shown) full-length cDNA also aligns with several other human partial-length expressed sequence tag cDNA sequences from brain, breast, placenta, liver, and heart, recently reported in the data base (Fig. 6). Inspection of these sequences reveals that there are multiple isoforms of the human AMPK-␥ protein. There are likely also similar AMPK ␥-isoform families expressed in rat and pig. This latter expectation is based on sequence analysis of 14 other MOPAC-derived partial AMPK-␥ cDNA sequences, as identified on colony hybridization of the AMPK-␥ MOPAC products with 32 P-labeled degenerate oligonucleotides (see "Experimental Procedures"). These products showed at least two reproducible patterns of nucleotide heterogeneity within the non-degenerate core (data not shown).
AMPK-␤ and AMPK-␥ mRNA Distribution-We (13,14) and others (5) have previously demonstrated that the catalytic AMPK ␣-subunit is widely expressed in several rat tissues. As shown in Fig. 7, the AMPK-␤ and AMPK-␥ sequences have a similar wide tissue expression. Two species of AMPK-␥ mRNA of 2.7 and 1.9 kilobase pairs are evident in total mRNA preparations; only the latter is present in poly(A) ϩ RNA from rat liver, suggesting that the larger mRNA is an unprocessed precursor. Only a single mRNA species for AMPK-␤ of 1.9 kilobases is evident. Both AMPK-␥ and AMPK-␤ mRNAs are highly expressed in kidney, white adipose tissue, lung, and spleen, while AMPK-␥ mRNA appears to be more highly expressed in heart and brain. While detectable, the mRNA level for each subunit is relatively lower in skeletal muscle, lactating mammary gland, and liver. In other studies (data not shown), we have also found high concentrations of mRNA for both subunits in the rat Fao hepatoma cell and the Syrian hamster insulin-secreting HIT cell, cell lines that both express substantial levels of AMPK activity.

DISCUSSION
Mammalian AMPK, as isolated from rat and porcine liver, contains three polypeptide subunits termed AMPK-␣, AMPK-␤, and AMPK-␥. The ␣-subunit contains the kinase catalytic domain sequence and is highly homologous to several members of the SNF1 kinase family (4,6,7,13,14). There are multiple isoforms of the ␣-subunit, with ␣ 1 being responsible for ϳ90% of the AMPK activity detected in liver extracts (14). The present report, based on very extensive peptide sequence and on predicted amino acid sequence from cDNA clones, establishes that full-length AMPK ␤and ␥-subunits are likewise homologous to two classes of proteins in S. cerevisiae. This extends information previously available from limited peptide sequence analysis and from smaller PCR-derived cDNAs (6). The present work further demonstrates, both by cDNA cloning and by direct peptide sequencing, which isoforms of AMPK ␤and ␥-subunits interact with the catalytic ␣ 1 -subunit in liver. This work also establishes that these non-catalytic subunits, like the ␣-subunit isoforms, have a wider tissue distribution, as evidenced by mRNA content of several rat tissues, than expected from the AMPK activity distribution previously reported (13,21).
The AMPK ␤-subunit is a mammalian homolog of a class of proteins in yeast, represented by Sip1p/Sip2p/Gal83p. The GAL83 gene product is known to affect glucose repression of the GAL genes (22). All of these proteins have been shown to interact with the Snf1p protein kinase either in the two-hybrid system or by immunoprecipitation (20,23). It has been proposed that these proteins serve as adaptors that promote the activity of Snf1p toward specific targets (23). Based on analysis of yeast mutants, it has been suggested that these proteins may facilitate interaction of Snf1p with unique and different targets. Of interest is the demonstration of a highly conserved domain of ϳ80 amino acids in the C terminus of Sip1p/Sip2p/ Gal83p, termed the ASC domain (association with Snf1p complex) (23). As studied in the two-hybrid system, the ASC domain of both Sip1p and Sip2p interacts strongly with Snf1p (23). However, the interaction of Sip2p with Snf1p is not entirely lost on deletion of this domain, suggesting that the ASC domain is not solely responsible for this protein-protein interaction. A putative ASC domain is also highly conserved in the C terminus of rat liver AMPK-␤ (amino acids 203-270), suggesting that this region may be responsible, in part, for binding to the AMPK ␣-subunit.
AMPK-␤, like Sip2p and Gal83p, is phosphorylated in vitro when associated with a catalytic subunit (AMPK-␣ or Snf1p, respectively) (4,6,23). Mutations of Gal83p can abolish most of the Snf1p kinase activity detectable in immune complexes, precipitated with anti-Snf1p antibody (23). A Sip2p⌬ gal83⌬ mutant shows reduced Snf1 protein kinase activity, which is restored following expression of either Sip2p-or Gal83p-LexA fusion proteins in the mutant strain (23). Taken together, these data suggest the possibility that AMPK-␤ may also serve as an adaptor molecule for the catalytic AMPK ␣-subunit and will positively regulate AMPK activity. This possibility is being tested experimentally.
AMPK-␤ appears to migrate anomalously on SDS gels, with the polypeptide migrating at a molecular mass ϳ10 kDa larger than the size predicted from the cDNA. This slower migration is evident for both the bacterially expressed His 6 fusion protein and the protein expressed in COS-7 cells. These observations suggest that higher orders of structure are responsible for the anomalous migration on SDS-polyacrylamide gel electrophoresis. The AMPK ␤-subunit is autophosphorylated in vitro (4,6); this suggests that the two AMPK-␤ bands expressed on transfection of mammalian cells with AMPK-␤ cDNA may result from a similar post-translational modification giving rise to smaller mobility shifts. Interestingly, this aberrant migratory behavior of AMPK-␤ is similar to that of its yeast homolog, Gal83p. The LexA fusion protein(s) of Gal83p, as expressed in yeast, also migrate at greater than the expected molecular mass and display more than one band on SDS gels, consistent with the known phosphorylation of Gal83p by Snf1p (23).
Rat and human liver AMPKs-␥ are mammalian homologs of S. cerevisiae Snf4p (CAT3) (24 -26). Snf4p was shown to interact with the Snf1p protein in the first reported use of the two-hybrid system and also coimmunoprecipitates with it (26). Indeed, on isolation of the Snf1p kinase from yeast, Snf4p, but not the other Snf1p-interacting proteins, copurifies in a 1:1 stoichiometry with the Snf1p polypeptide (4). Analysis of SNF4 mutants in yeast suggests that Snf4p also positively regulates the activity of its associated catalytic subunit, Snf1p (24,27). By analogy, our prediction is that AMPK-␥ will also have such a positive influence on the AMPK ␣-subunit.
Examination of the data base reveals that, in addition to the homology of AMPK-␥ to Snf4p, there are two or three different human proteins highly homologous or identical to our human and rat liver AMPK-␥ sequences. However, some of these data base sequences, as predicted from expressed sequence tag cDNAs in brain, heart, breast, and placenta, are distinct from each other and from our clones; some, for example, have a C-terminal extension. This indicates that there is a mammalian isoform family of potential AMPK ␥-subunits, each perhaps with different tissue expression and regulatory roles. We propose that these different ␥-isoforms be designated ␥ 1 , ␥ 2 , ␥ 3 . . . . ␥ n , as their full-length sequences are delineated. We have designated the rat liver/human liver AMPK-␥ sequence reported herein as AMPK-␥ 1 . The isoform diversity of both the ␣and ␥-subunits of AMPK underscores the need for complete characterization of the translation products of the enzyme, as isolated from various sources, in order to properly identify the relevant isoforms that make up the heterotrimeric complex. To what extent various species of AMPK heterotrimers with varying composition of individual subunits could exist in vivo is not yet known.
AMPK was first recognized as a protein kinase active on enzymes of lipid metabolism (acetyl-CoA carboxylase, hydroxymethylglutaryl-CoA reductase, and hormone-sensitive lipase) (1-3). However, as has been observed for the AMPK ␣-subunit (5,13,14), the AMPK ␤and ␥ 1 -subunits have wider tissue distribution than might be expected for a protein active only in the regulation of lipid metabolism. While mRNAs for each are detectable in "classic" lipogenic tissues like liver, white adipose tissue, and lactating mammary gland, high concentrations of mRNA in non-lipogenic tissues like heart, brain, spleen, and lung, for example, suggest that these proteins have roles that extend beyond the regulation of fatty acid and sterol metabolism. Of note are the relatively low amounts of AMPK-␤ and AMPK-␥ mRNAs in skeletal muscle; this observation is consistent with the relatively low levels of AMPK activity reported by others in this tissue and with the failure of skeletal muscle AMPK-␣ to immunoprecipitate with detectable ␤and ␥-subunits (28).
The wide tissue distribution of the mRNAs for all three subunits for AMPK raises the question of other potential roles for AMPK beyond lipogenic regulation. The striking homology of all three subunits to yeast proteins that are involved in nutrient (glucose) responses raises the possibility that the three mammalian proteins may be involved in glucose (or other nutrient) regulation of gene expression in mammalian tissues or in other adaptive responses to a changing nutrient environment. We (9) and others (12) have presented evidence that AMPK may be a important "metabolic sensor" linked to oxidative fuel choice in the heart and to glucose sensing in the pancreatic beta cell, perhaps being important for insulin secretion. There is every reason to believe that further study of the AMPK subunits may shed light on multiple aspects of cellular regulation.