Mammalian AMP-activated protein kinase subfamily.

The mammalian 5′-AMP-activated protein kinase (AMPK) is related to a growing family of protein kinases in yeast and plants that are regulated by nutritional stress. We find the most prominent expressed form of the hepatic AMPK catalytic subunit (α1) is distinct from the previously cloned kinase subunit (α2). The α1 (548 residues) and α2 (552 residues) isoforms have 90% amino acid sequence identity within the catalytic core but only 61% identity elsewhere. The tissue distribution of the AMPK activity most closely parallels the low abundance 6-kilobase α1 mRNA distribution and the α1 immunoreactivity rather than α2, with substantial amounts in kidney, liver, lung, heart, and brain. Both α1 and α2 isoforms are stimulated by AMP and contain noncatalytic β and γ subunits. The liver α1 isoform accounts for approximately 94% of the enzyme activity measured using the SAMS peptide substrate. The tissue distribution of the α2 immunoreactivity parallels the α2 8.5-kilobase mRNA and is most prominent in skeletal muscle, heart, and liver. Isoforms of the β and γ subunits present in the human genome sequence reveal that the AMPK consists of a family of isoenzymes.

The mammalian 5-AMP-activated protein kinase (AMPK) is related to a growing family of protein kinases in yeast and plants that are regulated by nutritional stress. We find the most prominent expressed form of the hepatic AMPK catalytic subunit (␣ 1 ) is distinct from the previously cloned kinase subunit (␣ 2 ). The ␣ 1 (548 residues) and ␣ 2 (552 residues) isoforms have 90% amino acid sequence identity within the catalytic core but only 61% identity elsewhere. The tissue distribution of the AMPK activity most closely parallels the low abundance 6-kilobase ␣ 1 mRNA distribution and the ␣ 1 immunoreactivity rather than ␣ 2 , with substantial amounts in kidney, liver, lung, heart, and brain. Both ␣ 1 and ␣ 2 isoforms are stimulated by AMP and contain noncatalytic ␤ and ␥ subunits. The liver ␣ 1 isoform accounts for approximately 94% of the enzyme activity measured using the SAMS peptide substrate. The tissue distribution of the ␣ 2 immunoreactivity parallels the ␣ 2 8.5-kilobase mRNA and is most prominent in skeletal muscle, heart, and liver. Isoforms of the ␤ and ␥ subunits present in the human genome sequence reveal that the AMPK consists of a family of isoenzymes.
The 5Ј-AMP-activated protein kinase (AMPK) 1 was initially identified as a protein kinase regulating hydroxymethylglutaryl-CoA reductase (1). Subsequently, the AMPK was shown to phosphorylate hepatic acetyl-CoA carboxylase (2) and adipose hormone-sensitive lipase (3). The AMPK appears to act as a metabolic stress-sensing protein kinase switching off biosynthetic pathways when cellular ATP levels are depleted and when 5Ј-AMP rises in response to fuel limitation and/or hypoxia (4). Partial amino acid sequencing of hepatic AMPK (5,6) revealed that it consists of 3 subunits, the catalytic subunit ␣ (63 kDa), and two noncatalytic subunits, ␤ (40 kDa) and ␥ (38 kDa).
The AMPK is a member of the yeast SNF1 protein kinase subfamily that includes protein kinases present in plants, nematodes, and humans (5)(6)(7)(8)(9). The AMPK catalytic subunit, ␣, has strong sequence identity to the catalytic domain of the yeast protein kinase SNF1, which is involved in the induction of invertase (SUC2) under conditions of nutritional stress due to glucose starvation (10). Both Snf1p and the AMPK require phosphorylation by an activating protein kinase for full activity (11). The sequence relationship between Snf1p and AMPK led to the finding that these enzymes share functional similarities, both phosphorylating and inactivating yeast acetyl-CoA carboxylase (5,11,12). Nevertheless, the AMPK does not complement SNF1 in yeast (11), indicating that their full range of functions are not identical. The noncatalytic ␤ and ␥ subunits of AMPK are also related to proteins that interact with Snf1p; the ␤ subunit is related to the SIP1/SIP2/GAL83 family of transcription regulators and the ␥ subunit to SNF4 (also called CAT3) (6,13).

EXPERIMENTAL PROCEDURES
Peptide Sequencing-Peptides were derived from rat and porcine ␣ 1 subunit of the AMPK, by in situ proteolysis (5), and sequenced on either an Applied Biosystems 471A Protein Sequencer or a Hewlett Packard G1000A Protein Sequencer.
Tissue Distribution-Activity Studies-A 35% saturated ammonium sulfate fraction was prepared for each tissue, following homogenization in AMPK homogenization buffer (HB, 50 mM Tris-HCl, pH 8.5, 250 mM sucrose, 5 mM sodium pyrophosphate, 50 mM sodium fluoride, 1 mM EGTA, 1 mM EDTA, 1 mM dithiothreitol, 1 mM benzamidine, 1 g/ml soybean trypsin inhibitor, and 0.2 mM phenylmethylsulfonyl fluoride). The resultant pellet was resuspended in 5 ml of HB and assayed for protein concentration (14). The AMPK was assayed as described previously (5) with the following modifications: a final reaction volume of 120 l was used, enzyme aliquots (30 l) containing 1 g of protein prediluted in 50 mM Tris-HCl, pH 7.5, and 0.05% (v/v) Triton X-100 were used to initiate the reaction. Three aliquots (30 l) were removed after 2, 4, and 6 min. Reactions were performed in duplicate Ϯ 5Ј-AMP (200 M), with a minus peptide substrate control. The specific activity of the enzyme was determined using linear rates of phosphorylation with the specific synthetic peptide substrate SAMS (15). The AMPK was purified from rat or porcine liver as described previously using substrate affinity chromatography (5).
Isolation of AMPK cDNA-A radiolabeled cDNA (774 base pairs) encoding porcine AMPK ␣ 1 was used to screen a rat hypothalamus Zap II cDNA library (Stratagene) according to the manufacturer's instructions. Positives were plaque-purified on subsequent rounds of screening, and phagemid from positive clones were rescued with helper phage (Stratagene). Screening of 7 ϫ 10 6 plaques yielded three unique clones, the largest consisting of an open reading frame, corresponding to AMPK ␣ 1 (2-549). The AMPK ␣ 1 5Ј end was isolated using a 5Ј-rapid amplification cDNA ends kit (Life Technologies, Inc.) with an ␣ 1 -specific primer to residues 41-48 and rat liver cDNA. 2 Human AMPK ␣ 1 (14 -270) was isolated from fetal human liver cDNA primed with sense and antisense partially degenerate oligonucleotides to ␣ 1 peptide sequence by reverse transcription-polymerase chain reaction. 3 Human AMPK ␣ 1 , * This work was supported by the National Heart Foundation (Australia) and the National Health and Medical Research Council (to B. E. K.) and by National Institutes of Health Grant DK35712 (to L. A. W.). 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 /EMBL Data Bank with accession number(s) U40819.
§ Present address: Telethon Institute of Genetics and Medicine, San Raffaele International Biomedical Science Park, Via Olgettina, 58, Milano 20132, Italia.
residues 291-448, is a partial length human liver cDNA clone obtained from the Lawrence Livermore National Laboratory (clone 78297, accession number T50799).
Northern Blotting-A rat multiple tissue Northern (MTN) blot (Clontech) containing 2 g of poly(A) ϩ RNA of individual tissues was probed with 32 P-labeled rat AMPK ␣ 1 and ␣ 2 cDNAs according to the instructions supplied.
Production of Anti-AMPK Antibodies-Polyclonal antibodies to AMPK ␣ 1 and ␣ 2 were prepared as follows. Peptides based on the predicted amino acid sequences of AMPK ␣ 1 for residues 339 -358 (DFYLATSPPDSFLDDHHLTR) and AMPK ␣ 2 for residues 352-366 (MDDSAMHIPPGLKPH) were synthesized and coupled to keyhole limpet hemocyanin (Sigma, H-2133) via a cysteine residue added to the N terminus of the peptide using the heterobifunctional reagent, N-succinimidyl-3-(2-pyridyldithio)propionate (Pharmacia, Uppsala, Sweden). New Zealand White rabbits were immunized with 2 mg of peptide conjugate initially in 50% (v/v) Freund's complete adjuvant and in 50% (v/v) Freund's incomplete adjuvant for subsequent immunizations. Rabbits were boosted fortnightly with 2 mg of peptide conjugate and bled 7 days after booster injections. Anti-AMPK ␣ 1 and ␣ 2 peptide antibodies were purified by peptide affinity chromatography.
Purification of AMPK ␣ 2 -Affinity-purified AMPK ␣ 2 antibody (2 mg) was coupled to CNBr-activated Sepharose 4B (Pharmacia, Uppsala, Sweden) according to the manufacturer's instructions. The unbound fraction from the substrate affinity column was applied directly to the AMPK ␣ 2 antibody column, washed with 5 volumes of phosphate-buffered saline, and eluted with 200 mM glycine buffer, pH 2.5, and immediately neutralized.

RESULTS
In the course of sequencing the porcine AMPK, we found that the amino acid sequence of some peptides derived from the pig liver AMPK ␣ subunit did not match those deduced from the rat liver cDNA sequence (7,8). Therefore, the rat liver AMPK catalytic subunit, ␣, was purified, and peptides accounting for 40% of the protein were sequenced (222/548 residues, Fig. 1). Eight of the sixteen peptides contained mismatched residues with the reported AMPK cDNA sequence, but did match the pig liver enzyme sequence. Using reverse transcription-polymerase chain reaction and cDNA library screening, we obtained cDNA sequence of the rat hypothalamus enzyme that accounted for all of the peptide sequences of the purified rat liver AMPK catalytic subunit containing mismatches (Fig. 1). The cDNA sequence of this AMPK catalytic subunit has been named ␣ 1 , since it corresponds to the purified enzyme and is clearly derived from a gene different from the previously cloned ␣ sequence (now referred to as ␣ 2 ). The ␣ 1 isoform of the AMPK catalytic subunit accounts for approximately 94% or more of the SAMS peptide phosphotransferase activity of rat liver and is therefore the predominant expressed hepatic isoform. Despite sequencing multiple preparations of the AMPK catalytic subunit from both pig and rat liver, no peptides were obtained that matched the ␣ 2 isoform sequence.
Within the catalytic cores of the ␣ 1 and ␣ 2 isoforms, there is 90% amino acid identity but only 61% identity outside the catalytic core (Fig. 1). Strong homology between the ␣ 1 and ␣ 2 sequences in the vicinity of the substrate binding groove, inferred from the protein kinase crystal structure for positions P Ϫ5 to P ϩ5 (17), suggest that the substrate specificities will be related. The substrate anchoring loop (also called the lip or activation loop) contains an insert Phe-Leu 170 for ␣ 1 , ␣ 2 , and Snf1p that may provide a hydrophobic anchor for a P ϩ3 or P ϩ4 hydrophobic residue in the peptide substrate. There is also Glu 100 (Glu 127 in cAMP-dependent protein kinase) and Asp 103 available for a P Ϫ3 basic residue specificity determinant for the ␣ 1 , ␣ 2 , and Snf1p. Both isoforms contain a Thr 172 residue equivalent to Thr 197 in the cAMP-dependent protein kinase, which is likely to be phosphorylated and necessary for optimal activity. Since the major differences in the ␣ 1 and ␣ 2 sequences FIG. 1. Comparison of AMPK ␣ 1 , ␣ 2 , and peptide sequence. The alignment of rat AMPK ␣ 1 cDNA-derived amino acid sequence (rAlpha 1) to human AMPK ␣ 1 cDNA-derived amino acid sequence (hAlpha 1), rat ␣ 2 cDNA-derived amino acid sequence (rAlpha 2) (8), and rat and porcine ␣ 1 peptide sequences (rat aa and porcine aa) is shown. Amino acids that could not be identified confidently are represented with an X. The sequences, aligned with the Pileup program (GCG, University of Wisconsin and Ref. 26), were formatted with the residues identical with the AMPK ␣ 1 cDNA-derived sequence being shaded.
occur in their COOH-terminal tails, they may interact with different proteins within this region.
The ␣ 2 8.5-kb mRNA is most abundant in skeletal muscle with lower levels in liver, heart, and kidney as reported recently (8,18). In contrast, there were very low levels of the ␣ 1 6-kb mRNA in all tissues examined except testis, where a low level of 2.4-kb mRNA was observed ( Fig. 2A). A testis-specific kinase related to Snf1p has been reported (19), but the corresponding transcript is 1.6 kb and may not be related to the 2.4-kb transcript seen here. The low levels of ␣ 1 mRNA explains why ␣ 1 was more difficult to clone than the ␣ 2 isoform (Fig. 2B). Northern blot analysis of the ␤ and ␥ subunits revealed a complex pattern of expression. The ␤ subunit mRNA was least abundant with similar levels across a range of tissues except brain, whereas the ␥ subunit mRNA was abundant in heart, lung, skeletal muscle, liver, and kidney. 4 An earlier report on the tissue distribution of the AMPK activity had claimed that it was predominantly a liver enzyme (15). In view of the mRNA distribution of the ␣ 1 and ␤ subunits, we reassessed the tissue distribution of the AMPK activity. The kidney contained the highest specific activity with similar levels in the liver, lung, and heart (Fig. 3) and little, if any, activity in skeletal muscle. It is clear that the AMPK activity has a wider tissue distribution than appreciated heretofore (15), and this closely parallels the distribution of ␣ 1 mRNA and not that of ␣ 2 mRNA. Using peptide-specific antisera to ␣ 1 (residues 339 -358) and ␣ 2 (residues 352-366), we found that the ␣ 2 immunoreactivity was predominant in the heart, liver, and skeletal muscle (Fig. 2E) where there is also the highest concentrations of ␣ 2 mRNA. In contrast, the ␣ 1 immunoreactivity is widely distributed (Fig. 2D) as is the less abundant ␣ 1 mRNA. The antibody to ␣ 2 recognized a minor component in the purified ␣ 1 preparation (Fig. 2E, lane 1), but sufficient amounts of this have not been obtained to determine whether it represents weak cross-reactivity with a form of ␣ 1 , an additional isoform of the AMPK or a low level contaminant of the ␣ 1 preparation by the ␣ 2 isoform. The antibody to ␣ 2 does not immunoprecipitate ␣ 1 activity from affinity-purified ␣ 1 AMPK. Both ␣ 1 and ␣ 2 migrate on SDS-PAGE at approximately 63 kDa (Fig. 2, D and  E). Unexpectedly, we found that the liver ␣ 2 immunoreactivity was not bound by the peptide substrate affinity column. This column specifically binds the ␣ 1 isoform. Using immune precipitation of the effluent from the peptide substrate affinity column with ␣ 2 specific antibody, we found that the ␣ 2 isoenzyme contained ␤ and ␥ subunits (Fig. 4) and catalyzed the phosphorylation of the SAMS peptide. Immune precipitates of ␣ 1 and ␣ 2 showed variable activation by 5Ј-AMP ranging from 2-3-and 3-4-fold, respectively. There was also an approximate 60-kDa band recognized by the ␣ 1 -specific antibody in tissue extracts from heart and lung (Fig. 2D). This band is not present in the purified liver enzyme, and its relationship to the ␣ 1 isoform is not yet known.
The proportion of SAMS peptide phosphotransferase activity bound to the peptide affinity column with a single pass varied (ranged 90 -92%, n ϭ 7, and 74 -86%, n ϭ 6 rat liver preparations). With recycling, approximately 94% of the activity was bound to the column. The residual activity was attributable to ␣ 2 isoform activity based on immunoprecipitation with the ␣ 2 -specific antibody. However, the amount of protein immunoprecipitated based on Coomassie Blue staining (Fig. 4) indicated that there was substantially more ␣ 2 protein than was expected from only 6% of the total SAMS peptide activity. The specific activity of the ␣ 2 isoform is not yet known in the absence of bound antibody. Based on the ␣ 2 cDNA sequence, Carling et al. (7) reported that a peptide specific antibody immunoprecipitated virtually all of the partially purified AMPK activity from liver. The peptide used in their experiments, PGLKPHPERMPPLI, contains 8/15 residues that are identical (underlined) between ␣ 1 and ␣ 2 so it seems reasonable that their polyclonal antisera may recognize both isoforms. In this event, their immunoprecipitation data are consistent with our results.
The present work makes plain that there is an isoenzyme family of AMPK ␣ catalytic subunits, increasing the complexity of activity analysis. This also raises the question of what function the ␣ 2 isoform has and whether ␣ 2 associates with a specific subset of ␤ and ␥ subunits. A significant fraction of the ␣ 2 isoform mRNA has a 142-base pair out-of-frame deletion within its catalytic domain that would encode a truncated, nonfunctional protein (8,18). The close sequence relationship between the ␣ 1 isoforms from pig, rat, and human ( Fig. 1) FIG. 2. Distribution of rat AMPK␣ isoforms: mRNA and protein. A rat multiple tissue Northern (MTN) blot (Clontech) containing 2 g of poly(A) ϩ RNA of individual tissues was successively probed and is shown in A-C. A, MTN hybridized with a 2.3-kb rat ␣ 1 cDNA.The blot was washed with 2 ϫ SSC, 0.5% SDS at 42°C for 3 ϫ 20 min. B, MTN hybridized with a 1.6-kb rat ␣ 2 cDNA. The blot was washed with 2 ϫ SSC, 0.5% SDS at 42°C and 0.1 ϫ SSC, 0.5% SDS at 65°C for 30 min. C, MTN hybridized with a 2-kb ␤-actin cDNA, the blot was washed with 2 ϫ SSC, 0.5% SDS at 42°C for 2 ϫ 20 min. RNA markers are shown in kilobases. A rat multiple tissue Western blot containing 100 g of protein/lane is shown in D and E. D was probed with affinity-purified ␣ 1 peptide serum, and E was probed with affinity-purified ␣ 2 peptide serum. The ␣ 1 control lane represents 0.1 g of purified rat liver AMPK (5), and ␣ 2 represents 0.1 g of expressed AMPK ␣ 2 (J. Dyck, unpublished data). Protein markers are shown in kDa. Legend: H, heart; B, brain; Sp, spleen; Lu, lung; Li, liver; Sk, skeletal muscle; K, kidney; and T, testis. means that there is strong conservation across species. Previously, it was reported that human liver does not contain AMPK mRNA (20); however, it is now clear that ␣ 2 mRNA was being probed for and not the dominant ␣ 1 isoform mRNA. The gene encoding the human liver AMPK catalytic subunit reported on chromosome 1 (20,21) is therefore the gene for the ␣ 2 isoform, whereas the gene for the ␣ 1 isoform is located on chromosome 5. 2 Recent genome sequencing has revealed multiple isoforms of the noncatalytic ␥ and ␤ subunits of the AMPK. There appear to be at least three isoforms of the ␥ subunit in brain with ␥ 2 and ␥ 3 present, distinct from the rat liver ␥ 1 isoform. Human brain also contains multiple ␤ subunit isoforms distinct from the rat liver ␤ 1 isoform. The accession numbers for putative AMPK ␤ and ␥ subunit isoforms are: ␥ 2 , M78939; ␥ 3 , R35524; ␤ 2 , R20494; ␤ 3 , R14746. Thus, a potentially large subfamily of heterotrimeric AMPKs, based on various combinations of all three AMPK subunits, may be present.
The structural relationships between the AMPK and SNF1 kinase, as well as the presence of multiple isoforms, brings into focus a vista of questions concerning the diverse physiological roles of this new subfamily of protein kinases. Whereas the AMPK regulates lipid metabolism in hepatocytes under conditions of metabolic stress, its role in other tissues, including the heart and kidney, are unknown. Recent studies by Kudo et al. (22) have shown that the AMPK is activated during cardiac ischemia, and the activation persists during reperfusion, possibly contributing to the ischemia-driven decoupling of metabolism and cardiac mechanical function. Regulation of cardiac acetyl-CoA carboxylase by AMPK plays an important role in the switching of cardiac metabolism between the use of glucose and fatty acids as oxidative fuel (23). In the ␤ cell of the pancreas, where AMPK subunits are highly expressed in islet cells, 5 glucose availability rapidly regulates acetyl-CoA carboxylase through changes in AMPK-directed phosphorylation, suggesting strongly a role for AMPK in stimulus-secretion coupling for insulin release (24). In addition to these metabolic roles, members of the SNF1 protein kinase subfamily appear to play important roles in development, with the par-1 gene of Caenorhabditis elegans playing an essential role in embryogenesis (25).