Isoform-specific purification and substrate specificity of the 5'-AMP-activated protein kinase.

The 5′-AMP-activated protein kinase (AMPK) mediates several cellular responses to metabolic stress. Rat liver contains at least two isoforms of this enzyme, either α1 or α2 catalytic subunits together with β and γ noncatalytic subunits in a trimeric complex. The α1 isoform is purified using a peptide substrate affinity chromatography column with ADR1 (222-234)P229 (LKKLTRRPSFSAQ), corresponding to the cAMP-dependent protein kinase phosphorylation site in the yeast transcriptional activator of the ADH2 gene, ADR1. This peptide is phosphorylated at Ser230 by AMPK α1 with a Km of 3.8 μM and a Vmax of 4.8 μmol/min/mg compared to the commonly used rat acetyl-CoA carboxylase (73-87)A77R86-87 peptide substrate, HMRSAMSGLHLVKRR, with a Km of 33.3 μM and a Vmax of 8.1 μmol/min/mg. Thus, the AMPK exhibits some overlapping specificity with the cAMP-dependent protein kinase. The rat liver AMPK α1 isoform has a Kcat ∼250-fold higher than the AMPK α2 isoform isolated from rat liver. The AMPK α1 isoform readily phosphorylates peptides corresponding to the reported AMPK phosphorylation sites in rat, chicken, and yeast acetyl-CoA carboxylase and rat hydroxymethylglutaryl-CoA reductase but not phosphorylase kinase. Based on previous peptide substrate specificity studies (Dale, S., Wilson, W. A., Edelman, A. M., and Hardie, G. (1995) FEBS Lett. 361, 191-195) using partially purified enzyme and variants of the peptide AMARAASAAALARRR, it was proposed that the AMPK preferred the phosphorylation site motif Φ(X, β)XXS/TXXXΦ (Φ, hydrophobic; β, basic). In good AMPK α1 peptide substrates, a hydrophobic residue at the P−5 position is conserved but not at the P+4 position. Oxidation of the Met residues in the rat acetyl-CoA carboxylase (73-87)A77R86-87 peptide increased the Km 6-fold and reduced the Vmax to 4% of the reduced peptide.

␥ (38 kDa) subunits. The AMPK phosphorylates a number of key enzymes involved in the control of lipid metabolism, acetyl-CoA carboxylase, hydroxymethylglutaryl (HMG)-CoA reductase, and hormone-sensitive lipase (2). Because it is activated by elevation of intracellular AMP caused by arsenite and heat shock, it is thought to function primarily in stress responses (3,4). The activation of the AMPK by AMP results from three contributing mechanisms: direct allosteric activation of the enzyme; AMP activation of an upstream kinase kinase (5,6); and AMP inhibition of AMPK dephosphorylation (7). Studies on HMG-CoA reductase regulation have reinforced the concept that the AMPK plays a role in stress responses, since mutation of Ser 871 , the AMPK phosphorylation site in HMG-CoA reductase, to Ala blocked phosphorylation by the AMPK and reduction in HMG-CoA reductase activity caused by ATP depletion but did not affect the regulation of HMG-CoA reductase at the transcriptional level (8). In yeast, the snf1p kinase controls invertase transcription in the adaptive response to growth on non-glucose sugars (9) as well as the regulation of acetyl-CoA carboxylase (1, 10), but it is not known whether the AMPK has a similar function.
Recently, we found that multiple isoforms of the AMPK are present in liver (11,12). A distinct gene accounts for ϳ90% of the activity in liver extracts (11). The two isoforms of the AMPK are 90% identical in the catalytic core region and 60% identical in their COOH-terminal tails. Since the AMPK is a multisubstrate protein kinase, its substrate specificity has been of particular interest. Previous studies have shown that an analog of the major AMPK phosphorylation site in rat acetyl-CoA carboxylase, or SAMS peptide, H 73 MRSAMS 79 GLHLVKRR 87 , is a relatively specific substrate for the AMPK. Previous specificity studies performed on the AMPK have used partially purified AMPK likely to have contained a mixture of the ␣ 1 and ␣ 2 isoforms (13)(14)(15). Using synthetic peptides based on the SAMS peptide sequence, it was reported that the enzyme has a requirement for two hydrophobic residues at the P ϩ4 and P Ϫ5 positions and a basic residue at either the P Ϫ2,Ϫ3,Ϫ4 site (14). These findings have recently been extended using substitutions in the SAMS peptide variant, AMARAASAAALARR, with partially purified enzyme (15).
We have reexamined the specificity of the AMPK in the light of the presence of multiple isoforms of the enzyme. In this paper, we report the purification of the AMPK isoforms from rat liver and compare their substrate specificities. Using a wide range of peptide substrates, we confirm the importance of a hydrophobic residue at the P Ϫ5 position, but not at the P ϩ4 position. Surprisingly, synthetic peptides corresponding to the cAMP-dependent protein kinase phosphorylation site, Ser 230 , in the yeast transcriptional factor ADR1 are phosphorylated with K m values 10-fold lower than that of the SAMS peptide and can be used for selective substrate affinity purification of the liver ␣ 1 isoform. The AMPK ␣ 2 isoform isolated from rat liver has a very low K cat compared to ␣ 1 using protein or peptide substrates.

EXPERIMENTAL PROCEDURES
Enzyme Purification from Porcine and Rat Liver-The porcine liver AMPK was purified by a modification of the procedure reported for rat liver (16) and the same buffers. Porcine liver (1 kg) was homogenized in a Waring blender with 4 volumes of buffer. A 2.5-7.0% (w/v) polyethylene glycol 6000 fraction was prepared, and the resultant fraction was batched onto 1500 ml of DEAE-cellulose (Whatman) and eluted with buffer containing 0.25 M NaCl (2000 ml). The eluate was chromatographed on 150 ml of Blue Sepharose (Pharmacia Biotech Inc.) and eluted with buffer containing 1 M NaCl. The enzyme fraction was concentrated and desalted by 10% (w/v) polyethylene glycol 6000 precipitation prior to affinity chromatography. The peptide substrate affinity column was washed with the same buffer containing 0.1% (v/v) Triton X-100 and 0.5 M NaCl, and the AMPK was eluted with buffer containing 2 M NaCl and 30% (v/v) ethylene glycol. Protein concentration was assayed by the method of Lowry et al. (17) using detergentcompatible protein assay reagents (Bio-Rad).
The rat AMPK was purified (250 g of liver) by a modification of the above procedure used for porcine liver preparations. DEAE-cellulose (500 ml) was used, the eluate was precipitated with 40% saturated (NH 4 ) 2 SO 4 , and the resultant pellet was resuspended in 400 ml of buffer and chromatographed on 100 ml of Blue Sepharose. A further wash was introduced to the peptide substrate affinity step using buffer containing 0.1% (v/v) Triton X-100 and 2 M NaCl immediately prior to elution with 2 M NaCl and 30% (v/v) ethylene glycol.
AMPK Assay-AMPK assays were performed using the SAMS peptide substrate, HMRSAMSGLHLVKRR-amide (13). 5Ј-AMP was included at either 20 or 70 M as indicated. The enzyme was diluted in buffer (50 mM Tris-HCl (pH 7.5), 0.05% (v/v) Triton X-100) prior to assay, and reactions were initiated by adding 10 l of enzyme to the reaction mixture. The phosphopeptide was isolated by withdrawing 30-l aliquots and applying them to Whatman P-81 papers (18).
Mass Spectrometry Analysis-Matrix-assisted laser desorption ionization mass spectrometry was performed (Voyager DE mass spectrometer; Perceptive Biosystems Inc.) utilizing delayed extraction technology. Purified samples (reversed phase HPLC) were dried on the sample stage in the presence of the matrix, ␣-cyano-4-hydroxycinnamic acid. Carboxypeptidase Y digests were carried out using the Perceptive Biosystems Inc. COOH-terminal sequencing reagents on the sample stage of the mass spectrometer according to the manufacturer's instructions. Tryptic digests were carried out using modified Promega sequencing grade trypsin diluted in 50 mM NH 4 HCO 3 , 10% (v/v) acetonitrile and mixed in equal proportions with the HPLC fraction on the sample stage of the mass spectrometer. The reaction was allowed to proceed at room temperature for 15 min before quenching with matrix and drying prior to analysis. Electrospray mass spectrometry was performed on a Perkin-Elmer Sciex API 111, triple quadripole mass spectrometer. Peptides and proteins were infused at a rate of 10 l/min in 0.05% (v/v) trifluoroacetic acid, 50% (v/v) acetonitrile. Proteins were desalted by precipitation in 7% (v/v) trichloroacetic acid, and the pellets were washed twice in 80% diethyl ether, 20% ethanol and dissolved in 10 -20 l of 30% (v/v) acetic acid together with an equal volume of acetonitrile.
Peptide Synthesis and Affinity Column Preparation-Peptides were synthesized (Applied Biosystems 433A peptide synthesizer) as described previously (18). All peptides were purified by cation-exchange chromatography followed by reversed phase chromatography (19) and analyzed by quantitative amino acid analysis using a Beckman 6300 amino acid analyzer. Electrospray mass spectrometry was also used to characterize some of the peptides (Sciex API 111, Perkin-Elmer). For methionine oxidation, the peptides (ϳ3 mM) were solubilized in 6 M guanidine-HCl, and the pH was adjusted to neutral. Oxidation was performed overnight in the presence of 0.1% (v/v) H 2 O 2 . As a control, peptides were incubated under the same conditions with ␤-mercaptoethanol to ensure that they were fully reduced. The peptides were then acidified by addition of 0.1% (v/v) trifluoroacetic acid, subjected to C 18 chromatography on a Sep-Pak column, and dried under vacuum. Dried peptides were dissolved in 50 mM Tris-HCl buffer, pH 7.5. The peptide substrate affinity column (available from G. McMurray, St. Vincent's Institute of Medical Research, Fitzroy, Australia) was prepared by coupling the ADR1(222-234)P 229 peptide, LKKLTRRPSFSAQ, to a Pharmacia HiTrap N-hydroxysuccinamide ester-activated Superose column. This resin contains a 6-aminohexanoic acid spacer arm. The conditions of coupling were as described by the manufacturer with 10 mg of peptide per 5 ml of column contents; peptide coupling was mon-itored by reversed phase HPLC.
Partial Purification of AMPK ␣ 2 -The unbound fraction from the substrate affinity column containing AMPK ␣ 2 and residual AMPK ␣ 1 (see text) was immunodepleted of AMPK ␣ 1 using either 10 g of affinity-purified anti-AMPK ␣ 1 -antibody coupled to protein A-Sepharose or anti-AMPK ␣ 1 antibody coupled to CNBr-activated Sepharose (Pharmacia) (Fig. 1C). Incubations were performed overnight at 4°C, and the AMPK ␣ 1 -immunodepleted fraction was collected by centrifugation. Recombinant AMPK ␣ 2 , obtained by expression as a glutathione S-transferase fusion product expressed in Escherichia coli 2 and quantitated by amino acid analysis, was used as a standard for measuring the amount of AMPK ␣ 2 by comparative immunoblotting and densitometry (Fig. 1D).
Anti-AMPK Antibodies and Immunoblotting-Anti-␣ 1 and anti-␣ 2 antibodies were prepared using synthetic peptides corresponding to the predicted amino acid sequences for residues 339 -358 from ␣ 1 and 352-366 from ␣ 2 , respectively, as described previously (11). Protein fractions were analyzed by SDS-PAGE (20), transferred to nitrocellulose (Schleicher & Schuell, Dassal, Germany), and probed with 4 g/ml affinity-purified ␣ 1 or ␣ 2 antibodies. Primary antibody was detected using anti-rabbit IgG antibody conjugated to horseradish peroxidase (DAKO, Carpinteria, CA) and 0.032% Acetyl-CoA Carboxylase Purification and Phosphorylation-Hepatic acetyl-CoA carboxylase was prepared from four female Sprague-Dawley rats after a 48-h fast and refeeding for a further 48 h, using a modification of the method reported by Tipper and Witters (21). The livers were homogenized in 5 volumes of the AMPK homogenization buffer. The acetyl-CoA carboxylase was then purified by 40% saturated (NH 4 ) 2 SO 4 precipitation followed by chromatography on monomeric avidin-Sepharose (Pierce). Acetyl-CoA carboxylase (3.9 pmol) was incubated with 1.95 pmol of AMPK ␣ 1 or AMPK ␣ 2 (␣ 1 immunodepleted) in the presence of 20 M 5Ј-AMP and 20 M [␥-32 P]ATP (10,000 cpm/pmol) as described previously (1). [ 32 P]Phosphate transfer was determined following SDS-PAGE and liquid scintillation counting of the gel bands.

RESULTS AND DISCUSSION
AMPK Purification-We were unsuccessful in obtaining purified AMPK from rat liver using the procedure reported previously (16). Since protein kinases can sometimes be purified by peptide substrate affinity chromatography (22), we screened a selected library of synthetic peptides, including analogs of proteins not known to be substrates for the AMPK. Initially, only partially purified enzyme (purified to the Blue Sepharose step) was used so that we could assess the specificity of the peptides for the AMPK. The peptide sequences tested included the SAMS peptide and peptides derived from the myosin light chains, ADR1, glycogen synthase, and phospholemman (23), a myocardial protein substrate for protein kinase C and cAMPdependent protein kinase (Table I). The phospholemman peptides tested were poor substrates and were not investigated further. The glycogen synthase peptide PLSRTLSVAAKK was phosphorylated in an AMP-dependent manner at ϳ40% of the rate of the SAMS peptide, but this peptide is an excellent substrate for a number of protein kinases (24) and was not investigated further. The myosin light chain peptides tested were phosphorylated with rates ϳ15% of the SAMS peptide. Surprisingly, the ADR1 peptides, ADR1(225-234) and ADR1-(222-234)P 229 (25), were phosphorylated at rates of ϳ50% of the SAMS peptide with an 8-fold lower apparent K m of ϳ4 M compared with 33 M for the SAMS peptide.
When the ADR1(222-234)P 229 peptide was coupled to a Pharmacia HiTrap column, the AMPK was bound avidly and required 2 M NaCl plus 30% (v/v) ethylene glycol for elution. Because the enzyme bound so tightly, it was possible to load the enzyme in buffer containing 0.5 M NaCl, an important step in the purification. A balance sheet of the AMPK purification is given in Table II. The resultant purified AMPK consisted of a 63-kDa ␣-catalytic subunit (which proved to be the ␣ 1 subunit, see below) and 40-kDa (␤) and 38-kDa (␥) noncatalytic subunits (26) (Fig. 1A). The AMPK was not evident on SDS-PAGE by protein stain until the final step of purification (Fig. 1A). Western blot analysis with anti-␣ 1 antibody suggested the presence of some minor proteolytic fragments in the purified enzyme (Fig. 1B). In some preparations, the AMPK was associated with high molecular mass material that corresponded to non-muscle myosin, as assessed by tryptic peptide sequencing (data not shown). Typically, an apparent purification of ϳ11,000-fold with a yield of 15% and a recovery of at least 90 g (range, 90 -200 g) of enzyme was obtained from 250 g of rat liver (Table II). Approximately 1 kg of porcine liver was required to yield a similar amount of purified enzyme. The fold purification may be an overestimate due to the presence of uncharacterized inhibitory material in the early fractions. Only the ␣ 1 isoform of the AMPK and not the ␣ 2 isoform binds to the peptide affinity column. The reason for this selectivity is not known, but it allowed us to prepare a partially purified fraction of AMPK ␣ 2 using the peptide affinity column effluent and antibody to deplete residual AMPK ␣ 1 (Fig. 1C). The specific activity of the purified rat AMPK (␣ 1 isoform) ranged from 8 to 27 mol/ min/mg without any obvious difference in the SDS-PAGE profile. This may reflect differences in the state of activation of the isolated enzyme due to phosphorylation by the upstream kinase kinase (3,5). Using electrospray mass spectrometry, we have observed ions corresponding to the ␤ and ␥ subunits but not the ␣ subunit; therefore, we have been unable to assess the state of endogenous phosphorylation of the ␣ subunit directly for different preparations.
The avidity of the enzyme for the peptide bound to the Pharmacia HiTrap resin was greater than could be expected from the free peptide binding to the enzyme (K m 3 M). It seems reasonable that the enhanced binding is due in part to the aminohexanoic acid linker between the peptide and the resin. In the case of the cAMP-dependent protein kinase, there is a hydrophobic pocket between the D and G helices that is responsible for high affinity binding of the peptide inhibitor PKI(5-22) at the P Ϫ11 position. Since the ADR1(222-234)P 229 peptide, LKKLTRRPSFSAQ, is linked through the amine residues on its NH 2 terminus or Lys residues, it is possible that the hydrophobic linker group has been fortuitously juxtaposed to a related hydrophobic pocket on the AMPK.
Peptide Substrate Specificity of AMPK Isoforms-Initially, we assessed whether purified ␣ 1 isoform-phosphorylated synthetic peptides corresponding to reported phosphorylation sites for the AMPK in acetyl-CoA carboxylase, HMG-CoA reductase, and phosphorylase kinase. The SAMS peptide rat acetyl-CoA carboxylase (73-87)A 77 ,R 86 -87 was phosphorylated with a K m of 33 M and a V max of 8.1 mol/min/mg compared with a K m of 59 M reported previously (14) (Table III). The corresponding chicken acetyl-CoA carboxylase peptide cACoAC (74 -88)R 87-88 was phosphorylated with a similar K m of 43 M but a higher V max of 23 mol/min/mg. Increasing the length of the cACoAC (74 -88) peptide to 25 residues, cACoAC (74 -98) reduced the K m from 43 M to 12 M but also decreased the V max from 23 to 4 mol/min/mg (Table III). The other chicken and yeast acetyl-CoA carboxylase peptides were also tested as substrates, with yeast acetyl-CoA carboxylase (1151-1165) R 1164 -1165 being the better substrate, with a K m of ϳ10 M and a V max comparable with that of the SAMS peptide. Interestingly, the phosphorylase kinase ␣-subunit peptide (1014 -1023), FRRLSISTES, was not a substrate, nor was the analog FRRLSISES, for either AMPK ␣ 1 or AMPK ␣ 2 . Despite an earlier report that phosphorylase kinase was a substrate for the AMPK (27), albeit weaker than for the cAMP-dependent protein kinase, the phosphorylase kinase peptide was not a substrate. Thus, the reported phosphorylation of the phosphorylase kinase ␣-subunit may have been due to a contaminating protein kinase or phosphorylation of an alternative site on phosphorylase kinase. The peptides cACoAC (1209 -1223)R 1222-3 and cACoAC (1194 -1208)R 1207-8 were phosphorylated with K m values somewhat higher than those of the SAMS peptide and with similar V max values (Table III), but yeast acetyl-CoA carboxylase (1136 -1150)R 1149 -50 was not a substrate for the AMPK. These results indicate that the AMPK is capable of phosphorylating a range

5Ј-AMP-activated Protein Kinase
of peptides corresponding to its reported phosphorylation site sequences, but not phosphorylase kinase.
ADR1 Peptide Phosphorylation-The ADR1 peptide ADR1(222-234)P 229 , LKKLTRRPSFSAQ, was phosphorylated with an apparent K m of 3.7 M. This is ϳ8-fold lower than the K m for the SAMS peptide and 3-fold lower than the K m for the AMARA peptide (AMARAASAAALARRR) recently reported (15), The parent ADR1(225-234) peptide was phosphorylated with an apparent K m of 5 M, indicating that the Pro substitution at residue 229 was not important. Since the ADR1 peptide contained three potential phosphorylation sites, we determined the site of phosphorylation. Following cleavage of the phospho-ADR1 peptide with chymotrypsin, the phosphopeptide mass was 1127.84, measured by electrospray mass spectrometry, indicating that chymotrypsin had cleaved the Leu-Thr bond to give the monophospho form of the peptide TRRPSFSAQ. Further, the yield of phenylthiohydantoin-derivatives at cycles 1 and 7 was as expected for Thr and Ser, respectively, but the recovery of phenylthiohydantoin-Ser at cycle 5 was poor. Since phosphoserine undergoes near complete ␤-elimination during the Edman sequencing cycle, this is consistent with Ser 230 as the site of phosphorylation. It is apparent that the presence of phospho-Ser 230 prevented the chymotryptic cleavage at the Phe-Ser bond. The phospho-ADR1 peptide LKKLTRRPS(P)F-SAQ (mass, 1609.4; theoretical, 1610.8) was also digested to completion with carboxypeptidase Y and was analyzed by time of flight mass spectrometry (Fig. 2B). Following digestion, the residual phosphopeptide corresponding to the peptide LKKLTRRPS(P)F had a mass of 1324.4 (theoretical, 1325.7). The same starting material was digested with trypsin. Following digestion, the residual phosphopeptide corresponding to the peptide RPS(P)FSAQ had a mass of 871.6 (theoretical, 871.3) (Fig. 2C). These results demonstrate that neither the Thr 226 nor the Ser 232 is phosphorylated, in agreement with the Edman sequencing result.

TABLE III
Comparison of kinetics of peptide phosphorylation Peptide phosphorylation was assayed as described under "Experimental Procedures" ϮAMP (70 M). The phosphorylation of the peptides was determined using purified enzyme, and the V max values were corrected using the SAMS peptide as an internal standard. Phosphorylation sites in these peptides are shown in boldface, and substituted residues are underlined.
that the presence of a hydrophobic residue at the P ϩ4 position was important for phosphorylation by the AMPK (14), based on substitutions in the SAMS peptide. However, in the case of the ADR1 peptide, a Gln is present at this position; it is therefore clear that a hydrophobic residue at the P ϩ4 position is not obligatory. The ADR1 peptide does, however, have an Arg at position P Ϫ2 and a Leu at position P Ϫ5 consistent with the requirement of a basic residue and hydrophobic residue at these positions, respectively, as was assessed with analogs of the SAMS peptide (14). Truncation of the cACoAC peptide to Met 78 abolished its capacity to act as a substrate. This is consistent with the requirement for NH 2 -terminal residues, including a hydrophobic residue at the P Ϫ5 position. Substitution of the chicken acetyl-CoA carboxylase peptide (74 -88)R 87-88 with Ala and Ile at positions P Ϫ2 and P Ϫ1 , respectively, reduced the V max for peptide phosphorylation slightly without having a marked effect on the apparent K m . The chicken sequence has a Pro at position P Ϫ3 , and substitution of this with Ala had no effect on the kinetics of peptide phosphorylation.
Phosphorylation of Oxidized SAMS Peptide-The effect of oxidation on SAMS peptide phosphorylation was investigated because the SAMS peptide contains 2 Met residues. The SAMS peptide was exquisitely sensitive to oxidation (Table IV) with the K m increasing 6-fold and the V max reduced to ϳ4%. In general, oxidation of peptides with Met at the P Ϫ5 position reduced their rates of phosphorylation; however, the V max was not always most affected, because peptide yeast acetyl-CoA  The kinetics of phosphorylation of SAMS, HMG-CoAR a (861-876), and ADR1(222-234)P 229 peptides by AMPK ␣ 1 and AMPK ␣ 2 were compared. AMPK ␣ 1 and AMPK ␣ 2 were purified from rat liver as described under "Experimental Procedures." Phosphorylation sites in these peptides are shown in boldface, and substituted residues are underlined. Velocities were calculated for the catalytic subunit of both isoforms, expressed as K cat , and the AMPK ␣ 1 holoenzyme, expressed as V max . V max values were not determined for the ␣ 2 isoform because the concentration of the ␣ 2 holoenzyme was not measured, only that of the catalytic subunit.

5Ј-AMP-activated Protein Kinase
carboxylase (1151-1165)R 1164 -1165 , GMNRAVSVSDLSYRR, had an increased K m from 9.7 to 181 M following oxidation. The cACoAC (1194 -1208)R 1207-8 , PTLNRMSFASNLNRR had only a 2-fold increase in its K m following oxidation and a negligible increase in the V max . These results indicate that Met at the P Ϫ1 position is not as critical for substrate recognition as the P Ϫ5 position.
Comparison of AMPK ␣ 1 and ␣ 2 Specificities-It was of interest to compare the specificities of the AMPK ␣ 1 and ␣ 2 isoenzymes. The SAMS peptide, rat acetyl-CoA carboxylase (73-87)A 77 R 86 -7 , ADR1 (222-234)P 229 , and HMG-CoA reductase (861-876) were all phosphorylated by AMPK ␣ 2 but with greatly reduced velocities (Table V). The apparent K m values for rat acetyl-CoA carboxylase (73-87)A 77 R 86 -87 and HMG-CoA reductase (861-876) were ϳ3-fold higher than with AMPK ␣ 1 ; in contrast, the apparent K m for ADR1 (222-234)P 229 was the same for both isoenzymes. Despite having essentially the same K m for the ADR1 (222-234)P 229 , the affinity column is highly selective for the ␣ 1 isoform. This may occur with a mechanism where the rate constants for dissociation or association of the ␣ 2 /peptide complex are higher or lower than for the ␣ 1 /peptide complex but subsequent steps in the reaction path have a more dominant effect on the K m . The large differences in the velocities observed between the isoenzymes with peptide substrates were also seen with purified acetyl-CoA carboxylase as substrate (Fig. 3). As shown in Fig. 3, both isoforms of rat liver acetyl-CoA carboxylase (ACoAC-280 kDa and ACoAC-265 kDa, respectively) are phosphorylated by ␣ 1 (28). These results suggest that the AMPK ␣ 2 isoform may have a specificity similar to that of AMPK ␣ 1 but is in an inactive form when isolated from the liver. Analysis of the ␣ 2 by immunoblotting (Fig. 1D) reveals that it is not proteolyzed. Further, it seems unlikely that it is inactivated by postextraction dephosphorylation because the ␣ 1 isoform is active in the extract. Since there are such large differences in the velocities of the AMPK ␣ 1 and ␣ 2 isoforms isolated from rat liver, then the earlier specificity studies (14,15) using partially purified enzyme containing both isoenzymes predominantly reflected the activity of the AMPK ␣ 1 isoform.
The AMPK ␣ 1 isoform is likely to be important in the control of hepatic lipid metabolism but it seems more likely that the AMPK ␣ 2 isoform is under separate control. The low velocity of the ␣ 2 isoform is consistent with its being in an inactive form, possibly dephosphorylated on the activation loop. It is not yet known whether both isoforms of the AMPK share the same upstream-activating kinases. While our results demonstrate that the AMPK shares a higher level of peptide substrate specificity overlap with the cAMP-dependent protein kinase than was previously recognized, they are nevertheless clearly different. The amino acid substitutions that ensure very high affinity binding of PKI had no effect on recognition of the SAMS peptide substrate by AMPK.