Posttranslational Modifications of the 5′-AMP-activated Protein Kinase β1 Subunit*

The AMP-activated protein kinase (AMPK) consists of catalytic α and noncatalytic β and γ subunits and is responsible for acting as a metabolic sensor for AMP levels. There are multiple genes for each subunit and the rat liver AMPK α1 and α2 catalytic subunits are associated with β1 and γ1 noncatalytic subunits. We find that the isolated γ1 subunit is N-terminally acetylated with no other posttranslational modification. The isolated β1 subunit is N-terminally myristoylated. Transfection of COS cells with AMPK subunit cDNAs containing a nonmyristoylatable β1 reduces, but does not eliminate, membrane binding of AMPK heterotrimer. The isolated β1subunit is partially phosphorylated at three sites, Ser24/25, Ser182, and Ser108. The Ser24/25 and Ser108 sites are substoichiometrically phosphorylated and can be autophosphorylatedin vitro. The Ser-Pro site in the sequence LSSS182PPGP is stoichiometrically phosphorylated, and no additional phosphate is incorporated into this site with autophosphorylation. Based on labeling studies in transfected cells, we conclude that α1 Thr172 is a major, although not exclusive, site of both basal and stimulated α1phosphorylation by an upstream AMPK kinase.

The 5Ј-AMP-activated protein kinase (AMPK) 1 isolated from liver consists of three subunits, the catalytic ␣ subunit (548 residues, M r 62,497 ϳ63 kDa) and two noncatalytic subunits, ␤ (M r 30,378 ϳ40 kDa) and ␥ (M r 37,429 ϳ38 kDa) (1)(2)(3). The AMPK phosphorylates a number of protein substrates including key enzymes involved in the control of lipid metabolism, acetyl-CoA carboxylase, HMG-CoA reductase and hormonesensitive lipase (reviewed in Ref. 4). It has been proposed that the AMPK functions in stress responses since it is activated by increasing intracellular AMP resulting from a variety of treatments including arsenite, heat shock (5-7), ischaemia (8), exercise (9), and electrical stimulation of skeletal muscle (10).
There are multiple isoforms of the AMPK subunits present in rat liver (2,11). The AMPK ␣ 1 isoform (2) is encoded by a gene localized to chromosome 5p11 (12), whereas the AMPK ␣ 2 isoform gene is localized to chromosome 1p31 (13). The ␤ 1 and ␥ 1 subunits genes are located on chromosome 12 (12) The AMPK ␣ 1 and ␣ 2 isoforms are 90% identical in their catalytic cores but only 60% identical in their COOH-terminal tails. Previously, we found that the ␤ 1 subunit of the AMPK was multiply phosphorylated in an intramolecular autophosphorylation reaction (14). The ␤ 1 subunit of the AMPK contains an N-terminal consensus sequence for myristoylation with glycine at position 1 and serine (a small uncharged residue) at position 5 (15).
In view of the potential importance of posttranslational modifications of the AMPK subunits in its physiological function, we have characterized the native state of the ␤ subunit. We find that the ␤ subunit purified from rat liver is fully myristoylated and present as a mixture of di-and triphosphorylated species with small amounts of monophospho-␤ subunit. Autophosphorylation of the AMPK in vitro results in up to 6 or more moles of phosphate incorporated per mole of ␤ subunit. The mass of the isolated N-terminally acetylated ␥ subunit is identical to the mass inferred from the cDNA sequence and is not phosphorylated.

EXPERIMENTAL PROCEDURES
AMPK Purification and Assay-The AMPK was purified from rat liver (14) and assayed (16) as described previously using the SAMS peptide substrate and 200 M 5Ј-AMP. The enzyme was diluted in 50 mM Tris⅐HCl, pH 7.5, 0.05% (v/v) Triton X-100, and the reactions were initiated by adding enzyme. The reactions were stopped by withdrawing 30-l aliquots for liquid scintillation counting as described (17). Protein concentration was assayed by the method of Lowry (18).
Protein Electrospray Mass Spectrometry-Purified AMPK (10 -30 g) was precipitated in 6% (w/v) trichloroacetic acid in the presence of 12.5 g/ml sodium deoxycholate. The protein pellet was washed once in 6% (w/v) trichloroacetic acid then twice in ether/ethanol (80/20), before air-drying, resuspension in 10 l of 30% (v/v) acetic acid and dilution in 10 l acetonitrile for direct infusion into a PE Sciex API III electrospray mass spectrometer. The resultant positive ion spectra were deconvoluted using the hypermass calculation in the supplied software (MacSpec 3.3).
Phosphoprotein Digestion-Peptides for protein sequencing or mass analysis were derived from the rat ␤ 1 subunit of the AMPK separated by SDS-polyacrylamide gel electrophoresis by in situ proteolysis (19). Briefly, Coomassie Blue-stained gel slices were excised and placed in the funnel compartment of a Hewlett Packard G1004B Chemstation where they were washed extensively in water, reduced in 5 funnel volumes (approximately 5 ml) of 10 mM dithiothreitol in 0.2 M Tris⅐HCl buffer, 1 mM EDTA, pH 8.5, at 45°C over 3 h and alkylated with 5 funnel volumes of 1% 4-vinylpyridine in 0.2 M Tris⅐HCl buffer, 1 mM EDTA, pH 8.5, at ambient temperature over 2 h. The gel slices were then destained in 7 funnel volumes of 50 mM ammonium bicarbonate, 50% acetonitrile at 65°C over 3 h and dried in a centrifugal freeze drier. The dry gel slices were allowed to rehydrate completely in digestion buffer (50 mM Tris⅐HCl buffer, 10% acetonitrile, pH.9.3, approximately 30 l per slice from a single lane of a mini-gel) containing achromobacter endoproteinase Lys-C (Wako) with protease to substrate ratio approximately 1:10 (w/w). The rehydrated gel slices were covered with an additional 200 l of digestion buffer and incubated at 37°C overnight. The peptides were recovered by combining the digestion buffer with 1 h washes of the gel slices (in a tube floated on a sonicating water bath) as follows: 200 l of 2% trifluoroacetic acid (v/v); 200 l of 30% acetonitrile, 0.1% trifluoroacetic acid; and finally, 200 l of 60% acetonitrile, 0.1% trifluoroacetic acid. The combined gel eluates were almost dried in a centrifugal freeze drier, reconstituted in 6 l of 100% trifluoroacetic acid, and vortexed. 6 M guanidine hydrochloride (300 l) was added, and the mixture subjected to chromatography on Nucleosil C 18 5 m 300 Å reversed phase material packed in a 1 ϫ 250-mm glass-lined column. The chromatography was performed on a Pharmacia SMART system using a linear 0 -80% acetonitrile, 0.1% trifluoroacetic acid gradient over 120 min at 40 l/min and monitored at 214 nm, and 2 min fractions were collected.
Peptide Sequencing and Phosphate Release-Peptides were sequenced with a Hewlett Packard G1000A protein sequencer utilizing Routine 3.5 Edman degradation chemistry as recommended by the manufacturer. Phosphopeptides were covalently linked to Sequelon AA filters (PerSeptive Biosystems), and phosphate was extracted at each cycle with 3 ϫ 0.5 ml volumes of 90% methanol, 0.015% phosphoric acid as the solvent in the Routine 3.1 PVDF method. Extracted cycles were diverted via valve number RV6 (line 61), collected directly into fraction collector tubes, and counted on a liquid scintillation counter.
Peptide Time of Flight Mass Spectrometry-Peptide mass analysis was performed using a PerSeptive Biosystems Voyager DE mass spectrometer (MALDI-TOF) with the sample crystallized in the presence of ␣-cyano-4-hydroxycinnamic acid as matrix and with the instrument operated in the delayed extraction mode with external calibration against a series of synthetic peptides. Alkaline phosphatase, chymotrypsin, and trypsin digestions were carried out at room temperature in a humidified atmosphere on the MALDI target. Alkaline phosphatase, sequencing grade chymotrypsin (Boehringer Mannheim), and modified sequencing grade trypsin (Promega) were desalted into 25 mM ammonium bicarbonate, 10% acetonitrile on a Pharmacia SMART fast desalting column. Lyophilized aminopeptidase M (Sigma) was reconstituted, and all dilutions were carried out in the same solvent. Peptides in HPLC solvent (0.1% trifluoroacetic acid, variable acetonitrile concentrations) were added to equal volumes of diluted enzyme (a range of enzyme dilutions were tested for each treatment) on the MALDI target and allowed to react for 10 -15 min. At the completion of this time, an equal volume of matrix containing 2% trifluoroacetic acid was added, and the sample was dried. The aminopeptidase M digestion was performed in a tube at 37°C for 24 -48 h, and an aliquot was removed and mixed with matrix prior to mass analysis.
AMPK Subunit Mutagenesis and Cellular Expression-Two mutant forms of AMPK subunits were used in these experiments. To generate an inactive AMPK ␣ 1 subunit, threonine 172 (a major activating ␣ phosphorylation site) was mutated to alanine as described previously (20). This mutation rendered AMPK (when expressed as a heterotrimer) nearly inactive (2% the activity of wild-type) and also created a dominant negative inhibitor of wild-type AMPK heterotrimer (data not shown), similar to the inactivating ␣ 1 K45R mutant (20). The ␤ 1 subunit glycine 2 was mutated to alanine to remove the myristoylation site. For cell-labeling experiments with 32 P i, COS-7 cells were triply transfected (␣ 1 ,␤ 1 ,␥ 1 ) with either wild-type GST-tagged ␣ 1 or GST-tagged T172A-␣ 1 (both in pEBG vector) in association with HA (hemagglutinin)-tagged ␤ 1 and ␥ 1 in pMT2 vector (20). Cells were washed two times with phosphate-and serum-free Dulbecco's modified Eagle's medium 48 h after transfection and then incubated for 2 h with the same medium containing 32 P i (0.2 mCi/ml). One set was exposed for an additional hour to sodium azide (10 mM) to activate the transfected kinase. Both control and stimulated cells were harvested in a Triton-X-containing buffer, and cell lysates were adsorbed to glutathione-agarose beads (20). Adsorbed proteins were eluted in SDS sample buffer and separated by SDS-polyacrylamide gel electrophoresis (9% gel). 32 P-labeled AMPK subunits were visualized on a Molecular Dynamics PhosphorImager.
For studies of AMPK subcellular distribution, COS-7 cells were triply transfected with GST-tagged wild-type ␣ (in pEBG), HA-tagged ␥ 1 (in pMT2), and one of three ␤ 1 constructs, all in pBK-CMV vector (Stratagene): wild-type ␤ 1 , G2A-␤ 1 , or HA-tagged (N-terminal) ␤ 1 . Cells were washed with ice-cold phosphate-buffered saline 48 h after transfection and collected by scraping and brief low-speed centrifugation. Cells were then briefly sonicated at setting 3 (100% duty; continuous) for two 10-s periods on ice employing a Heat Systems Model W225 sonicator equipped with a microtip. The sonication buffer employed was identical to that used for Triton X-100 lysis but with the omission of the detergent (20). Lysates were then centrifuged for 1 h at 150,000 ϫ g to generate a soluble (S) and membrane (M) fraction. The latter was resuspended in a volume equal to that of the soluble fraction in 1 ϫ SDS sample buffer. These samples were analyzed by immunoblotting using ␣ 1 -specific, ␤ 1 -specific, and anti-HA antibodies with development by horseradish peroxidase-labeled secondary antibodies and enhanced chemiluminescence (Amersham).

Mass Spectral Analysis of AMPK Subunits-
The observed mass of the rat liver ␥ subunit was M r 37,439 (Fig. 1A), corresponding closely with the expected M r of 37,429 for the Nterminally acetylated ␥ 1 subunit based on the cDNA-derived sequence (11). These results indicate that none of the ␥ 1 subunit side chains are posttranslationally modified in the isolated enzyme. We were not successful in obtaining a mass signal for the ␣ subunit of the AMPK by electrospray mass spectrometry.
In contrast, there were multiple species of the ␤ 1 subunit (Fig. 1B). The form with M r of 30,552 accounted for approximately 10% of the material corresponded to the mass of the myristoylated form of the protein with a single phosphate group. The M r expected of the unmodified species was 30,264 based on the cDNA-derived sequence reported by Woods (NCBI Seq. ID 1185269) (3). Our earlier reported cDNA sequence (11) (NCBI Seq. ID 1335858) was found to contain two errors, Glu for Gly at position 26 and a Pro deletion at position 137. These errors were confirmed by either direct Edman sequencing for position 26 or by mass spectrometry for position 137 (results not shown) and confirm the correct ␤ 1 cDNA sequence (3). The two most prominent species corresponded to the di-(M r 30,635) and triphosphorylated (M r 30,722) forms of the myristoylated protein (Fig. 1B). Depending on the enzyme preparation, the relative amounts of the di-and triphosphorylated species varied but were always much more prominent than the monophosphorylated form.
␤ Subunit Phosphorylation Site and N Terminus Analysis-␤ subunit was digested with endoproteinase Lys-C and the resultant peptides were chromatographed on reversed phase HPLC (Fig. 2). Each fraction was screened using MALDI-TOF mass spectrometry. The fractions labeled A-D contained peptides with masses that did not match the predicted Lys-C digest patterns that were further examined. Residue numbering is based on the protein sequence inferred from the cDNA sequence (3) with the inferred methionine at position 1 being absent from the mature protein.
Fraction A contained a phosphopeptide of M r 1242 that reduced to M r 1161 (Table I) on phosphatase treatment, which is consistent with a single phosphate addition to the predicted Lys-C fragment 19 TPRRDS 24 S 25 GGTK 29 . The corresponding dephospho-form of this peptide was observed in the adjacent fraction. Attempts to elucidate the site of phosphorylation by mass fragmentation techniques were unsuccessful. This peptide is also autophosphorylated in vitro with [␥-32 P]ATP. When the corresponding 32 P-labeled monophosphorylated peptide was subjected to [ 32 P]phosphate release sequencing, most radioactivity was released at cycle 6, the primary phosphorylation site, but with significant phosphate release at cycle 7 (Fig.  3). Thus either Ser 24 or Ser 25 (but not both) are phosphorylated with a preference for Ser 24 .
Fraction B contained a phosphopeptide of M r 3478 that re-duced to M r 3399 (Table I) on phosphatase treatment, which is consistent with a single phosphate addition to the predicted Lys-C fragment 173 BSDVSELSSS 182 PPGPYHQEPYISKPEE-RFK 201 (where B is pyridylethyl cysteine). Chymotryptic digestion of the phosphopeptide yielded a peptide fragment of M r 1710 (Table I) consistent with the phosphate being located in the fragment 173 BSDVSELSSSPPGPY 187 . Aminopeptidase M digestion of this fraction resulted in a mass spectral ladder (Table I) containing fragments consistent with phosphorylation on Ser 182 . The corresponding dephospho-form of this peptide was not observed, which is consistent with Ser 182 being stoichiometrically phosphorylated. Furthermore, autophosphorylation in vitro did not result in any incorporation into this site, and no dephosphorylated ␤ 1 subunit was detected in the isolated enzyme (Fig. 1B).
Fraction C contained a phosphopeptide of M r 2862 that reduced to M r 2782 (Table I) on phosphatase treatment, which is consistent with a single phosphate addition to the predicted Lys-C peptide 103 LPLTRSQNNFVAILDLPEGEHQYK 126 . The corresponding dephospho-form was also present in fraction C. Tryptic digestion of this fraction (Table I) (Table I), which is N-terminally blocked by Edman sequencing. This mass is consistent with the predicted N-terminal Lys-C peptide myristoyl-2 GNTSSERAALERQAGHK 18 and a tryptic fragment of M r 961 (Table I) corresponding to the peptide myristoyl-2 GNTSSER 8 and confirming the structural assignment.
Effect of ␤ Subunit Myristoylation on Subcellular Distribution-Mutation of the Gly 2 myristoylation site or addition of a 12-amino acid HA-tag to the N terminus of the ␤ 1 subunit (removing Gly 2 as a target for myristoylation) alters the subcellular distribution of expressed AMPK subunits in COS-7 cells. As shown in Fig. 4 (upper panel) under the selected conditions of triple transfection of AMPK subunit cDNAs, cell lysis, and centrifugation, wild-type ␤ 1 protein is almost entirely recovered in a membrane fraction (M), as are its associated ␣ 1 (middle panel) and ␥ 1 (lower panel) subunits. In expression of AMPK heterotrimers containing either the G2A ␤ 1 mutant or with the addition of an HA-tag to the N terminus of ␤ 1 , a substantial fraction of ␤ 1 and ␥ 1 are recovered in the soluble cell fraction along with a small amount of ␣ 1 . It should be noted that immunostaining of the expressed ␤ 1 subunit reveals more than one protein band, particularly with the G2A mutant or HA-tagged construct (upper panel). This could reflect molecular heterogeneity due to other posttranslational modifications, such as phosphorylation. Taken together, these data indicate that the Gly 2 myristoylation site on ␤ 1 , in part, influences the subcellular localization of the transfected/expressed AMPK, but other factors must be involved. ␤ Subunit Phosphorylation in Intact Cells-To study the phosphorylation of the ␤ 1 subunit in intact cells and the relative importance of AMPK autophosphorylation, AMPK heterotrimer was expressed in 32 P i -labeled COS-7 cells using heterotrimers with wild-type ␣ 1 or an inactive ␣ 1 subunit. The latter was generated by site-directed mutagenesis of Thr 172 , initially predicted to be the major site of activating phosphorylation by an upstream kinase kinase, based on kinase structure (19). As recently confirmed by peptide sequencing, Thr 172 is indeed the site phosphorylated by such an AMPK kinase (21). The Thr 172 to Ala mutant has Ͻ2% of the activity of the wild-type subunit when expressed in the AMPK heterotrimer, indicating the importance of Thr 172 in regulation of activity (data not shown). As shown in Fig. 5, both wild-type ␣ 1 and ␤ 1 in AMPK heterotrimer isolated from labeled cells by glutathione adsorption are phosphorylated in control cells, and the 32 P content of each increases on treatment of cells with sodium azide. No phosphorylation of the ␥ 1 subunit is observed. However, expression of inactive AMPK heterotrimers containing the T172A ␣ 1 reveals reduced phosphorylation of both ␣ 1 and ␤ 1 subunits in the basal state and resulted in no change in their 32 P content on azide treatment. Taken together, these data indicate that Thr 172 is the major, though not exclusive site, of both basal and azide-activated ␣ 1 phosphorylation. Some ␤ 1 phosphorylation depends on the activity of the ␣ 1 catalytic subunit, but non-␣ 1 protein kinases active on ␤ 1 are likely to account for some of the residual phosphorylation observed in the presence of inactive ␣ 1 .

DISCUSSION
The N terminus of ␤ 1 (MGNTSSERAA . . . ) contains a penultimate glycine (Gly 2 ) in an appropriate context of downstream sequence (MGNXXS) of other known myristoylation sites (15). In some cases myristoylation is responsible for targeting proteins to the membrane, but may also facilitate protein/protein interactions (22). Myristoylation of proteins in the src family (23), the myristoylated alanine-rich protein kinase C substrate (MARCKS) (24), ADP-ribosylation factor 1 (25) and recoverin (26) has been shown to facilitate membrane binding, but for calcineurin-B and the cAMP-dependent protein kinase catalytic subunit, myristoylation does not affect membrane association (27).  Procedures") and are loaded to reflect the same percentage of total fraction protein. The upper panel is blotted with a ␤ 1 -specific antibody raised against bacterially expressed 6x-His-tagged ␤ 1 protein, and the middle panel is blotted with an ␣ 1 -specific anti-peptide antibody. The lower panel is blotted with an anti-HA monoclonal antibody; the ␥ 1 subunit is the fastest migrating band observed. When co-transfected with HA-tagged ␤ 1 , both ␤ 1 and ␥ 1 are thus detected (lower panel, two lanes on the right side). In these lanes, ␤ 1 immunoreactivity migrates more slowly than in other lanes due to the addition of the 3-kDa HA epitope.
Recurring basic amino acids and palmitoylation may also effect membrane binding either alone (27) or with myristoyl groups (28). A number of evenly spaced basic amino acids that could potentially interact with acidic phospholipids in membrane bilayers have been described for the src proteins (23) and there is a polybasic sequence in the N terminus of ␤ 1 (Arg 8 , Arg 13 , Lys 18 , Arg 20 , Arg 21 ). The ␤ 1 subunit is not modified by palmitoylation. In other examples the myristoyl-dependent membrane association may also be modulated by protein phosphorylation (24) or calcium binding (26) and play a role in signal-induced protein localization. The present studies indicate that the presence/absence of the penultimate Gly 2 myristoylation site has a modest influence on the intracellular localization of AMPK heterotrimer, although other factors also seem to be involved. More studies are needed to examine the role/localization of particulate AMPK in the action of this kinase. It is not known if the presence of the myristate group contributes to the stability of the AMPK heterotrimer as it does for the cAMP-dependent protein kinase (29).
Extensive in vitro phosphorylation of noncatalytic subunits of protein kinases has been observed for a number of protein kinases including phosphorylase kinase and CK2 (30, 31) but the physiological significance of these phosphorylation events is uncertain. In the AMPK we have found three phosphorylation sites in the native ␤ 1 subunit, whereas no phosphorylation of the ␥ subunit has been observed. Of the three ␤ 1 phosphorylation sites, Ser 24/25 and Ser 108 are part of the intramolecular autophosphorylation cascade that are phosphorylated in vivo.
There are (at least) three additional sites present in the hyperphosphorylated state following in vitro autophosphorylation that we have not characterized. We have observed these on two-dimensional phosphopeptide maps (32) of tryptic peptides derived from in gel digests of autophosphorylated ␤ 1 subunit. Phosphorylation at Ser 182 seems likely to result from phosphorylation by a separate protein kinase with a Ser-Pro specificity. This site appears to be stoichiometrically phosphorylated in the native enzyme as no additional [ 32 P]phosphate is incorporated into this site; the corresponding dephosphorylated peptide is not detected nor is any dephosphorylated ␤ 1 subunit detected.
We were unable to obtain electrospray mass spectrometry data on the native ␣ subunit to determine its state of phosphorylation. Tryptic phosphopeptide mapping of in vitro and celllabeled ␣ 1 subunits indicate that they are multiply phosphorylated (data not shown). The phosphorylation site Thr 172 in the activation loop that is phosphorylated by the AMPK kinase has been recently characterized (21). Our intact cell studies with expression of AMPK heterotrimers indicate that ␣ 1 Thr 172 is critical for enzyme activity and that its absence diminishes phosphorylation of both the ␣ 1 and ␤ 1 subunits. Furthermore, upon activation of AMPK activity by the uncoupling of oxidative phosphorylation with sodium azide, no increase in 32 P content in either ␣ 1 or ␤ 1 is observed when the T172A mutant is incorporated into the heterotrimer. These data indicate that Thr 172 is the critical activating phosphorylation site under these conditions and that ␤ 1 subunit phosphorylation, in part, occurs through an intramolecular autophosphorylation. However, there is residual phosphorylation of both ␣ 1 and ␤ 1 subunits, even in the absence of an active catalytic subunit (we have been unable to detect any endogenous ␣ 1 subunit in COS-7 cells), suggesting that there may be distinct AMPK kinases active on both the ␣ 1 and ␤ 1 subunits.
FIG. 5. 32 P-Labeling of AMPK ␣ 1 and ␤ 1 in intact cells. Shown is a PhosphorImager scan obtained after SDS gel electrophoresis (9% gel) of glutathione-agarose-adsorbed AMPK heterotrimers isolated from 32 P-labeled cells (each isolated in parallel from duplicate wells). Cells were triply transfected either with wild-type AMPK subunits (␣ 1 /␤/␥; labeled ␣1 wt) or with wild-type ␤ 1 and ␥ 1 with the mutant inactive T172A ␣ 1 construct (labeled ␣1 mut). Both forms of ␣ 1 subunit were epitope-tagged with GST in the pEBG vector to allow facile isolation of AMPK heterotrimer. 48 h after transfection, cells were labeled with 32 P i and then exposed to either control vehicle (control) or sodium azide (10 mM) for 2 h prior to harvesting (see "Experimental Procedures"). The arrows indicate the migration of Coomassie Blue-stained standards of each subunit.