Functional analysis of mutations in the gamma 2 subunit of AMP-activated protein kinase associated with cardiac hypertrophy and Wolff-Parkinson-White syndrome.

Mutations in the gene encoding the gamma(2) subunit of the AMP-activated protein kinase (AMPK) have recently been shown to cause cardiac hypertrophy and ventricular pre-excitation (Wolff-Parkinson-White syndrome). We have examined the effect of four of these mutations on AMPK activity. The mutant gamma(2) polypeptides are all able to form functional complexes following co-expression with either alpha(1)beta(1) or alpha(2)beta(1) in mammalian cells. None of the mutations caused any detectable change in the phosphorylation of threonine 172 within the alpha subunit of AMPK. Consequently, in the absence of an appropriate stimulus the mutant complexes, like the wild-type complex, exist in an inactive form demonstrating that the mutations do not lead to constitutive activation of the kinase. Three of the mutations we studied occur within the cystathionine beta-synthase (CBS) domains of gamma(2). Two of these mutations lead to a marked decrease in AMP dependence, whereas the third reduces AMP sensitivity. These findings suggest that the CBS domains play an important role in AMP-binding within the complex. In contrast, a fourth mutation, which lies between adjacent CBS domains, has no significant effect on AMPK activity in vitro. These results indicate that mutations in gamma(2) have different effects on AMPK function, suggesting that they may lead to abnormal development of the heart through distinct mechanisms.

The AMP-activated protein kinase (AMPK) 1 is the central component of a protein kinase cascade that plays a pivotal role in the regulation of energy metabolism (1). In response to activation following an increase in the AMP/ATP ratio, AMPK phosphorylates a number of downstream targets culminating in the switching off of energy (ATP)-utilizing pathways and the switching on of energy-generating pathways (1,2). Activation of AMPK is complex and involves direct allosteric activation of the enzyme by AMP as well as phosphorylation, catalyzed by an upstream kinase, AMPK kinase (AMPKK) (3,4). AMPK is a heterotrimeric complex, consisting of a catalytic subunit (␣) and two regulatory subunits (␤ and ␥) (5,6) and isoforms of all three subunits have been identified (7)(8)(9). Although there is compelling evidence indicating that formation of the heterotrimeric complex is necessary for significant kinase activity (6,10) the precise role of the regulatory subunits remains unclear. In addition, at present our understanding of the physiological relevance of the different subunit isoforms is very limited.
Several groups have reported recently the identification of six different mutations in the ␥2 subunit from patients with cardiac hypertrophy and associated electrophysiologic abnormalties (11)(12)(13)(14). All six mutations result in amino acid substitutions and are located within the C-terminal half of the protein. The mutations lead to the development of aberrant conduction systems, including pre-excitation, characteristic of Wolff-Parkinson-White syndrome (15) and in all but one case, the mutations also result in severe cardiac hypertrophy. No evidence of cardiac hypertrophy, however, was found in individuals carrying a mutation of arginine to glycine at residue 531 (R531G) (13). At present, the molecular mechanisms by which mutations in ␥ 2 lead to abnormal conduction and hypertrophy are unknown. Interestingly, however, enlarged myocytes from an individual with a mutation in ␥ 2 were found to have vacuoles containing glycogen derivatives, suggesting that the hypertrophy may be a result of increased carbohydrate storage (14). A mechanism involving increased glycogen storage is attractive, because an arginine to glutamine mutation at residue 226 (R226Q) in the ␥ 3 isoform causes excess glycogen accumulation in pig skeletal muscle (16).
Three AMPK ␥ isoforms have been identified and all share a highly conserved C-terminal region of ϳ300 amino acids (9). Within this conserved region are four cystathionine ␤-synthase (CBS) domains (17). CBS domains have been identified in a wide range of proteins, but their function remains unknown. Mutations within the single CBS domain of human cystathionine ␤-synthase, from which the acronym stems, cause homocystinuria (18). Four of the mutations in ␥ 2 , together with the R226Q mutation in the ␥ 3 isoform in pig, occur within the CBS domains. These findings raise the possibility that the CBS domains in AMPK may have an important functional role. It was reported that AMPK activity in skeletal muscle isolated from pigs harboring the R226Q mutation in ␥ 3 was lower than that in the corresponding wild-type pigs (16). In that study, however, total AMPK activity, rather than ␥ 3 -specific activity was measured, making interpretation of the results difficult. In another study (19), the arginine residue in ␥ 1 (Arg 70 ) equivalent to Arg 226 in ␥ 3 was mutated to a glutamine residue in an effort to mimic the ␥ 3 mutation. It was reported that this mutation caused a marked increase in AMPK activity and rendered it largely AMP independent (19).
The identification of mutations in ␥ 2 that lead to severe heart defects in humans has important implications for our understanding of the pathogenesis and treatment of cardiac hypertrophy and ventricular pre-excitation. The lack of information regarding the impact of these mutations on AMPK activity prompted us to carry out our current study. We report here the effect of four different mutations in ␥ 2 , three that occur within the CBS domains and one that lies between adjacent domains, on the activity of AMPK in mammalian cells. None of the mutations in ␥ 2 that we studied cause a detectable increase in the phosphorylation of threonine 172 within the ␣ subunit of AMPK (Thr 172 ) compared with wild-type ␥ 2 complexes. Thus, in the absence of an appropriate stimulus to promote phosphorylation of Thr 172 , the mutant complexes, like the wild-type complex, exist in a relatively inactive form. These results demonstrate that the ␥ 2 mutations do not lead to constitutive activation of the kinase. Two of the three mutations occurring within the CBS domains cause a marked decrease in AMP activation of the kinase, whereas a third mutation in the CBS domain leads to a doubling of the A 0.5 for AMP. In contrast, the leucine insertion mutation that occurs between CBS domains 1 and 2 has no significant effect on AMP sensitivity in vitro.
These results suggest that the pathways by which mutations in ␥ 2 lead to the pathogenesis of heart disease involve different mechanisms.

EXPERIMENTAL PROCEDURES
Mutagenesis of ␥ 2 Subunit-Oligonucleotides spanning the sequence to be mutated, and including the appropriate base changes, were used to amplify cDNA encoding AMPK ␥ 2 (9) using a PCR-based strategy as described previously (20). A sequence encoding a FLAG epitope tag (amino acid sequence DYKDDDDK) was inserted immediately after the initiating methionine codon to facilitate detection of the ␥ 2 subunit. All cDNAs were cloned into pCDNA3 (Invitrogen) and sequenced completely on both strands to ensure their authenticity.
Mammalian Cell Expression-Plasmid DNA was prepared using a Qiagen maxi-prep kit according to the manufacturer's instructions. CCL13 cells were co-transfected with cDNA (10 g/plasmid) encoding either ␣ 1 or ␣ 2 , and ␤ 1 and ␥ 2 (all cDNAs cloned into pCDNA3) by calcium phosphate precipitation (21). The ␣ 1 and ␣ 2 cDNAs encode a Myc epitope tag (amino acid sequence EQKLISEEDL) at the N terminus (20). Cells were harvested 72 h post-transfection by one of two methods. In the first method (rapid lysis), the culture medium was removed and 1 ml of ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, 50 mM NaF, 5 mM sodium pyrophosphate, 1 mM EDTA, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 10% (v/v) glycerol, 1% (v/v) Triton X-100) immediately added to the cells. Insoluble material was removed by centrifugation (13,000 ϫ g, 1 min at 4°C) and the supernatant fraction used for subsequent analysis. In some cases, cells were submitted to hyperosmotic stress, by addition of sorbitol (final concentration of 0.6 M) to the culture medium and incubated for 30 min immediately prior to lysis. In the second method, cells were allowed to become anoxic thereby leading to activation of AMPK in response to an increase in the AMP/ATP ratio. In this method, following removal of the culture media the cells were briefly rinsed with phosphate-buffered saline. Cells were removed from the plate in 1 ml of phosphate-buffered saline containing 2 mM EDTA and collected by centrifugation. The resulting cell pellet was re-suspended in 1 ml of lysis buffer and insoluble material was removed by centrifugation.
AMPK Activity-AMPK activity was determined in immune complexes isolated by immunoprecipitation from cell lysates using an anti-Myc monoclonal antibody (clone 9E10 (22)). Cell lysates (0.1 ml for ␣ 1 containing complexes and 0.5 ml for ␣ 2 complexes) were incubated for 2 h at 4°C with 10 l of a 50% (w/v) slurry of anti-Myc antibody bound to protein G-Sepharose. Immune complexes were precipitated by centrifugation at 6000 ϫ g for 1 min and washed 3 times with lysis buffer prior to assay. Kinase activity was measured by the incorporation of 32 P-radiolabeled phosphate into the SAMS peptide as described previously (23), in the presence or absence of varying concentrations of AMP as described in the figure legends. Activities are calculated as picomole of phosphate incorporated into the SAMS peptide per minute per mg of total protein.
AMP dependence curves were generated using the equation , where s is the relative stimu-lation, b is the basal activity, and A 0.5 is the concentration of AMP giving half-maximal stimulation, using Graphpad Prism software.
Western Blot Analysis-Proteins present in the immune complexes were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane. The membrane was blocked by incubation in 10 mM Tris-HCl, pH 7.4, 0.5 M NaCl, 0.5% Tween 20, 5% low fat milk powder for 1 h at room temperature. Membranes were probed with an anti-Myc (22), anti-AMPK␤ (6), anti-FLAG M2 (Sigma), or anti-AMPK P-Thr 172 (Cell Signaling Technologies) antibody for 2-18 h at 4°C in blocking buffer and then washed extensively with 10 mM Tris-HCl, pH 7.4, 0.5 M NaCl, 0.5% Tween 20. The blots were incubated for 1 h at room temperature with either goat anti-mouse IgG (for anti-Myc and anti-FLAG blots) or donkey anti-rabbit IgG (AMPK␤ and AMPK phospho-Thr 172 ), followed by extensive washing. Blots were developed using enhanced chemiluminescence (Roche Molecular Biochemicals).
Yeast Complementation and Two-hybrid Analysis-A mutation in Snf4 cDNA was introduced by PCR to alter arginine residue 294 to glycine. snf4⌬ mutant yeast (strain MCY2634) were transformed with either wild-type or mutated Snf4 in pGAD424 (Clontech) and selected by growth on media lacking tryptophan. Transformants were patched onto solid media containing either 2% glucose, 2% raffinose or 5% glycerol, and growth was determined after 5 days incubation at 30°C. Interactions using the two-hybrid system were performed following transformation of SFY526 yeast (Clontech) with Snf1 cDNA in pGBT9 (Clontech) and either wild-type or mutated Snf4 in pGAD424. Transformants were grown to mid-log phase in selective media (lacking tryptophan and leucine) containing either 2% glucose or 2% raffinose as the sole carbon source. ␤-Galactosidase activity was determined in permeabilized cells as described previously (6) and expressed in Miller units (24).
Statistical Analyses-Statistical significance was determined using a two-tailed unpaired Student's t test.

RESULTS
In an attempt to elucidate the effect of mutations within the ␥ 2 subunit of AMPK we determined AMPK activity following co-expression of wild-type and mutant forms of ␥ 2 with either ␣ 1 ␤ 1 or ␣ 2 ␤ 1 in mammalian cells. The locations of these mutations within ␥ 2 are shown in Fig. 1A. Alignment of the amino acid sequences of the CBS domains of ␥ 2 (19) reveals that three of the mutations, R302Q, H383R, and R531G, occur at, or adjacent, to the equivalent position of the R226Q mutation in pig ␥ 3 (Fig. 1B). The fourth mutation generates an insertion of a leucine residue (L ins) between residues 350 and 351 in the region linking CBS domains 1 and 2.
In the absence of an appropriate stimulus AMPK exists in a relatively inactive form. Numerous studies have shown that activation of AMPK requires the phosphorylation of threonine 172 within the ␣ subunit (Thr 172 ) catalyzed by an upstream kinase, AMPKK (1). We have previously reported that AMPK is activated by anoxia in cultured cells, which can occur as a consequence of the method used for harvesting the cells (25). AMPK activity in anoxic cells was readily detected in anti-Myc immunoprecipitates for all the ␣1 complexes measured in the presence of 200 M AMP ( Fig. 2A), although the activity of the R302Q, H383R, and R531G mutants was approximately half that of the wild-type enzyme. Similarly, following expression with ␣ 2 , AMPK activity was present in all the ␥ 2 complexes and again the activity of the R302Q, H383R, and R531G mutants was reduced relative to the wild-type complex (Fig. 2B). The lower activity of the ␣ 2 complexes compared with the ␣ 1 complexes is consistent with the results of a previous study examining the activity of ␥ 1 -containing complexes (20). Western blot analysis of the AMPK subunits present in the immunoprecipitates revealed similar levels of expression of the complexes indicating that the changes in activity were not because of differences in the level of expression, or association, of the different complexes. Furthermore, there was no obvious difference in the phosphorylation of Thr 172 , as judged by Western blotting, indicating that the reduction in activity of the mutant complexes is not because of decreased phosphorylation. As ex- The location of the four mutations examined in this study are represented schematically in A. The four CBS domains within the C-terminal half of ␥ 2 are shown as shaded boxes. In B the amino acid sequences of the CBS domains of human ␥ 2 , in which the R302Q (CBS1), H383R (CBS2), and R531G (CBS4) mutations are located, are shown aligned. The CBS domain of pig ␥ 3 (CBS1), which contains the R226Q mutation, and the single CBS domain of human cystathionine ␤-synthase, which contains the D444N mutation, are also shown aligned. Amino acid identities in the CBS domains of the ␥ isoforms are shaded in black and conservative substitutions in gray. The residues at which the mutations occur are denoted by arrows above the sequences.
FIG. 2. Effect of ␥ 2 mutations on AMPK activity in anoxic cells. Wild-type ␥ 2 (WT) or mutant ␥ 2 (R302Q, H383R, R531G, and L ins) were co-expressed with either ␣ 1 ␤ 1 or ␣ 2 ␤ 1 in CCL13 cells and AMPK complexes were immunoprecipitated from lysates prepared from anoxic cells. AMPK activity of ␣ 1 -containing complexes (A) and ␣ 2 -containing complexes (B) was determined in immune complexes using the SAMS peptide assay. Values shown are the mean Ϯ S.E. from four independent experiments and are expressed as picomole/min/mg. Background activity present in immune complexes isolated from untransfected cells (UT) is also shown. Western blot analysis of proteins within the immune complexes is shown below the graphs. Blots were probed with anti-Myc (present on the ␣ subunit), anti-phosphothreonine 172 (P-Thr172), anti-FLAG (present on the ␥ subunit), or anti-␤ subunit antibodies. In each case a representative blot is shown. pected, virtually no AMPK activity or expression was detectable in immunoprecipitates isolated from untransfected cells.
In contrast to anoxic cells, AMPK isolated from rapidly lysed cells is relatively inactive because the intracellular ratio of AMP/ATP remains low (25). As can be seen in Fig. 3, the activity of ␣ 1 -containing AMPK, measured in the presence of 200 M AMP, isolated from rapidly lysed cells is very low in either wild-type or mutant ␥ 2 complexes. Similar results were obtained for ␣ 2 -containing complexes (results not shown). Consistent with the low activity of the complexes, no phosphorylation of Thr 172 was detected in immunoprecipitates isolated from rapidly lysed cells, despite similar levels of expression of the complexes determined by Western blot analysis (Fig. 3). These results indicate that the ␥ 2 mutations do not have any significant effect on the phosphorylation and activation of AMPK by AMPKK under basal conditions. We next investigated the effect of the mutations on the allosteric activation of the kinase by AMP. Fig. 4, A and B, shows the activity of ␣ 1 (Fig. 4A) and ␣ 2 (Fig. 4B) complexes isolated from anoxic cells measured in vitro in the presence or absence of 200 M AMPK. Qualitatively similar results were obtained for both isoforms, although the degree of stimulation by AMP was slightly greater for the ␣ 1 complexes compared with the ␣ 2 complexes. Based on these results, the activity of  Table I. As can be seen, the R302Q, H383R, and R531G mutations all cause a significant reduction in AMP stimulation measured at 200 M AMP. This is particularly evident with the R531G mutation, which almost completely abolishes activation by AMP. In contrast, the L ins mutation did not cause a significant difference in AMP stimulation measured at 200 M AMP. It is noteworthy that following expression with ␣ 1 , the activity of the R531G mutant is increased relative to the other complexes when measured in the absence of AMP (Fig. 4A). This is unlikely to be because of an increase in the phosphorylation state of Thr 172 because we did not detect any change in phosphorylation of this residue between the different complexes (see Fig. 2A). One potential explanation for the increased activity is that the R531G mutation reduces the inhibitory effect of ATP on the kinase. Previous studies have shown that the allosteric activation of AMPK is antagonized by ATP (4) suggesting that both AMP and ATP bind at the same site. It is possible, therefore, that mutations that affect AMP-binding could also affect ATP binding. If this were the case for the R531G mutation it could account for the increased activity observed in the absence of AMP. We have not been able to address this possibility directly because of the fact that measurement of AMPK utilizes radiolabeled ATP in the assay, which raises a number of technical problems associated with varying ATP concentrations.
To investigate in more detail the effect of the mutations on AMP activation, ␣ 1 complexes were assayed over a range of AMP concentrations (Fig. 4C) Table I). At 500 M AMP the activity of the H383R mutant is similar to that of the wild-type (Fig. 4C), indicating that although the AMP sensitivity is reduced, the AMP dependence is not significantly affected. The R531G mutation almost completely abolishes AMP stimulation and we were unable to determine accurately the A 0.5 for AMP for this complex. These results strengthen the hypothesis that the CBS domains play an important role in the allosteric activation of the kinase by AMP, but do not reveal the mechanism by which the L ins mutation alters AMPK function.
We recently reported that AMPK is activated by two distinct pathways, one that involves changes in the AMP/ATP ratio and one that is independent of this ratio (26). Hyperosmotic stress activates AMPK, concomitant with an increase in the phosphorylation of Thr 172 , without any detectable change in the AMP/ ATP ratio (26). The finding that two of the ␥ 2 mutations have a marked effect on AMP dependence prompted us to investigate the activity of the ␥ 2 complexes following hyperosmotic shock. Fig. 5 shows that hyperosmotic stress (0.6 M sorbitol) increases ␣1 activity in both the wild-type and mutant complexes relative to control cells. None of the mutations had any detectable effect on Thr 172 phosphorylation (data not shown). The effect of the mutations becomes evident, however, when comparing activity from hyperosmotically stressed cells in the presence and absence of AMP. The activity of the R302Q and R531G mutants is increased relative to wild-type when assayed in the absence of AMP, whereas the activity of the H383R and L ins mutants is the same as wild-type. The increased activity of the R302Q and R531G mutants may be because of a reduction in inhibition by ATP as described previously. In the presence of AMP, the activity of the R302Q, H383R, and R531G mutants is approximately half that of the wild-type and L ins complexes, similar to the situation observed in anoxic cells.
AMPK is structurally and functionally related to the SNF1 kinase in Saccharomyces cerevisiae (1). Yeast contain a single ␣ subunit (Snf1), a single ␥ subunit (Snf4), and three isoforms of the ␤ subunit (Sip1, Sip2, and Gal83). Like the AMPK ␥ subunit isoforms, Snf4 contains four CBS domains. Alignment of the CBS domains in ␥ 2 and Snf4 reveals that arginine 531 in ␥ 2 is conserved in Snf4 (arginine 294), whereas the residues equivalent to arginine 302 and histidine 383 are not conserved (19). We mutated arginine 294 to glycine in Snf4 and transformed a yeast snf4 deletion strain with wild-type or mutant Snf4. Both wild-type Snf4 and Snf4R294G were able to rescue the growth of snf4 yeast on glycerol and raffinose media (Fig.  6A), indicating that the mutated Snf4 is capable of forming a functionally competent SNF1 complex in vivo. Snf4 interacts with Snf1, and this interaction is markedly increased in glucose-deprived cells (27). We examined the interaction of wildtype and mutant Snf4 with Snf1 in the two-hybrid system. The interaction of both wild-type and mutant Snf4 was low in cells grown in glucose and in both cases this interaction was in-  creased in cells grown in raffinose (Fig. 6B). These results indicate that the R294G mutation does not have a marked effect on the glucose-regulated interaction of Snf4 and Snf1. DISCUSSION In this study we have examined the effect of four mutations within the ␥ 2 subunit on the activity of AMPK. These mutations were first identified in individuals with abnormal cardiac function, including pre-excitation (Wolff-Parkinson-White syndrome) and hypertrophy (11)(12)(13). None of the mutations caused constitutive activation of AMPK as judged either by direct measurement of activity or by the phosphorylation state of Thr 172 . A number of previous studies have shown that conditions that lead to an increase in the intracellular AMP/ATP ratio promote phosphorylation and activation of AMPK (e.g. Refs. 4,26,28,and 29). More recently, we reported activation of AMPK by stimuli that do not alter the AMP/ATP ratio and showed that this AMP-independent mechanism also involves phosphorylation of Thr 172 (25,26). In the current study we have used either anoxia or hyperosmotic stress (incubation with 0.6 M sorbitol for 30 min) as stimuli to activate the kinase.
Anoxia causes a marked increase in the AMP/ATP ratio (25), whereas hyperosmotic stress has no effect on this ratio (25,26). Using these different methods, we did not find any evidence to suggest that the ␥ 2 mutations cause increased phosphorylation of Thr 172 relative to wild-type. Rather, changes in Thr 172 phosphorylation in response to either anoxia or hyperosmotic stress were remarkably similar in both the wild-type and mutant complexes. Our results differ from those of a previous study (19) where it was reported that mutation of arginine to glutamine at residue 70 (R70Q) within the ␥ 1 isoform, which is equivalent to arginine 302 in ␥ 2 and arginine 226 in ␥ 3 , caused a marked activation of AMPK. In that study, phosphorylation of Thr 172 within the ␣1 subunit of the R70Q ␥1 mutant complex was significantly increased relative to the wild-type complex. The reason for the difference between our results and those of Hamilton et al. (19) remains unclear and further studies will be required to resolve these apparent discrepancies.
Whereas we did not find any difference in phosphorylation of AMPK, three of the mutations did have a significant effect on kinase activity in vitro. AMPK is routinely assayed in the presence or absence of 200 M AMP and at this concentration the R302Q, H383R, and R531G mutations all caused a significant reduction in the degree of stimulation by AMP. These results are consistent with the effect of the R70Q mutation in ␥ 1 , which reduced AMP activation by approximately half (19). Determination of the kinetic parameters involved in AMP activation demonstrated that the R302Q and R531G mutations cause a marked reduction in the AMP dependence of the kinase. The H383R mutation, however, increases the A 0.5 for AMP, but has no significant effect on overall AMP dependence. At high concentrations of AMP (500 M) the activity of the H383R complex is similar to wild-type. The concentrations of AMP required for half-maximal stimulation of the ␥ 2 complexes are significantly higher than those previously reported for ␥ 1containing AMPK complexes. Immunoprecipitation of ␣ 1 and ␣ 2 complexes from rat liver, in which ␥1 accounts for over 90% of the total AMPK activity (9), yielded A 0.5 values for AMP of 12 Ϯ 3 and 22 Ϯ 3 M, respectively (30). We reported similar values for ␣1 (5.7 Ϯ 2 M) and ␣ 2 (16 Ϯ 3.5 M) complexes following co-expression with ␤ 1 ␥ 1 in mammalian cells (20). Our results indicate, therefore, that ␥ 2 -containing complexes are less sensitive to AMP than the corresponding ␥ 1 complexes. This finding may have important implications when considering the physiological role of the different AMPK complexes. The relatively high concentration of AMP required to activate ␥ 2 complexes would suggest that these complexes could respond to a greater range of AMP concentrations compared with ␥ 1 . This could be important in tissues such as heart, where large fluctuations in adenine nucleotides may occur. For instance, following 12 min of cardiac ischemia, AMP increased from around 1 to 143 M, whereas ATP levels fell well below millimolar concentrations (31). Under these conditions, ␥ 1 complexes would be predicted to be maximally active, whereas ␥ 2 complexes would still be within a responsive range. Mutations in ␥ 2 that affect AMP activation of the kinase could have a significant effect on activity across this concentration range.
Our results provide further evidence for a role of the CBS domains in the allosteric regulation of AMPK by AMP. CBS domains occur in many proteins, and although the function of these domains is unknown it appears that they play a regulatory role, rather than a catalytic one. In human cystathionine ␤-synthase mutation of aspartic acid to asparagine at residue 444 (D444N) abolishes activation by S-adenosylmethionine, without affecting the basal catalytic activity (18). Intriguingly, this mutation lies in the equivalent position to the R302Q and H383R mutations within the CBS domain (see Fig. 1). These FIG. 6. Examination of an R294G substitution on Snf4 function in yeast. Arginine residue 294 in Snf4 (equivalent to Arg 531 in ␥ 2 ) was mutated to glycine. A, snf4 yeast were transformed with wild-type (WT) Snf4, mutant (R294G) Snf4, or empty vector (pGAD424), plated on selective media containing either 2% glucose, 5% glycerol or 2% raffinose as the sole carbon source and incubated for 5 days at 30°C. B, yeast strain SFY526 was co-transformed with Snf1 (in pGBT9) and either wild-type (WT) Snf4 or mutant (R294G) Snf4 (in pGAD424). Interactions were determined by ␤-galactosidase liquid assays in permeabilized cells. Activities shown are the average values from duplicate assays that varied by less than 10% and are plotted as Miller units. findings strongly suggest that residues at or near this position in the CBS domain have important functions in binding adenosine derivatives. Previously, we presented evidence that the ␥ 1 subunit binds AMP (9), although we did not attempt to identify the residues involved in nucleotide binding. Because the ␥ isoforms each contain four copies of the CBS domains, it is possible that they bind more than one AMP molecule per subunit, and this could explain our finding that mutations within CBS1, -2, and -4 all have an effect on AMP activation. The varying effects on AMP activation produced by mutations in the different CBS domains may indicate that the four domains are not functionally equivalent in terms of AMP binding. Alternatively, it is possible that the differences may be because of the specific nature of the mutations. The R302Q and R531G mutations alter a basic arginine residue of an uncharged residue. It seems reasonable to predict that this could have a marked effect on ionic interactions, either within the AMPK complex or with AMP itself. On the other hand, substitution of histidine by arginine in the H383R mutation is a more conservative change and may have less of an effect on ionic interactions. Understanding the precise effect of these mutations at the molecular level will require detailed structural information from x-ray crystallographic studies.
An interesting observation that emerges from our study is that the phosphorylation of Thr 172 appears to be unaffected by the ␥ 2 mutations, despite the fact that some of these mutations bring about large changes in the AMP dependence of the kinase. Previous studies have shown that AMP makes AMPK a better substrate for phosphorylation by the upstream kinase in the cascade, AMPKK (30,32). As a consequence of this it might have been predicted that the R302Q and R531G mutations would effect phosphorylation of Thr 172 , because they clearly reduce AMP dependence. That this is not the case suggests that either the effect of AMP on AMPK activity and phosphorylation are distinct, or that direct activation of the upstream kinase by AMP (32) overrides the substrate-mediated effect of AMP on phosphorylation.
Recently, two further mutations in ␥ 2 were identified in individuals with Wolff-Parkinson-White syndrome and associated cardiac hypertrophy (14). Based on results obtained with Snf4, the yeast equivalent of the ␥ subunit (1), it was reported that these mutations, T400N and N488I, caused constitutive activation of AMPK (14), although direct measurement of AMPK activity was not reported in this study. Introduction of these equivalent mutations into Snf4 increased the interaction of Snf4 with Snf1, the yeast equivalent of the ␣ subunit, in yeast grown in the presence of glucose as determined by twohybrid analysis. We introduced the equivalent of the R531G mutation into Snf4, but did not observe any change in the interaction of Snf1 and Snf4 in cells grown in the presence or absence of glucose. Therefore, if there is an effect of the equivalent ␥ 2 mutations within Snf4 on its interaction with Snf1 it does not appear to be a common feature of all the mutations. The T400N mutation occurs toward the end of CBS domain 2, whereas the N488I mutation lies between CBS domains 3 and 4. Because these mutations do not align with any of the four mutations we examined it remains possible that they could have a different effect on AMPK activity and it will be important to determine directly the effect of the T400N and N488I mutations on AMPK activity.
In our current study we have been unable to detect a direct effect of the L ins mutation on AMPK activity. It is possible that this mutation has a subtle effect on activity that was not detected in the in vitro assays. Alternatively, this mutation may affect a function of AMPK that was not measured in our system, such as substrate binding. Clearly, further studies are required to investigate these possibilities. The remaining three mutations alter AMP activation of the kinase, such that under conditions that stimulate the kinase cascade via an increase in the AMP/ATP ratio, e.g. anoxia, these mutations would be predicted to cause a reduction in AMPK activity relative to wild-type. Conversely, in response to stimuli that promote phosphorylation of Thr 172 without altering the AMP/ATP ratio, e.g. hyperosmotic stress, the R302Q and R531G mutations appear to be activating when assayed in vitro in the absence of AMP. These results are somewhat paradoxical at first sight, but may be because of changes not only in AMP-binding but also in ATP-binding in these mutant complexes. The allosteric activation of AMPK is antagonized by ATP (4) and the simplest hypothesis is that AMP and ATP compete for the same allosteric site within the kinase. If this is the case it is likely that mutations that affect AMP binding will also affect ATP binding. Our results fit a model in which the basic residues (arginine or histidine) that are mutated within CBS domains 1, 2, and 4 of ␥ 2 (see Fig. 1B) are involved in nucleotide binding. In this model, substitution of these residues reduces the affinity for both AMP and ATP, although the effects need not necessarily be equivalent for either the different mutations or the effect of the mutations on AMP and ATP. Based on this model, the effect of the individual mutations on AMPK activity would depend on the ratio of AMP/ATP within the cell, or in the in vitro assay. Although we have been able to study the effect of altering AMP within the assay we have not been able to determine the effect of changing ATP concentrations because of technical difficulties. Clearly, however, the important challenge is to determine the effect of the mutations on AMPK activity in vivo. The generation of animal models expressing the various ␥ 2 mutations will facilitate a clearer understanding of these complex issues.
To our knowledge our results provide the first direct demonstration of the effects of naturally occurring mutations within ␥ 2 on AMPK activity. These findings should facilitate further studies aimed at defining the molecular mechanisms by which mutations in ␥ 2 lead to heart disease. A key question that arises from our study is how would a reduction in AMP dependence or AMP sensitivity of ␥ 2 -containing complexes lead to the observed phenotype of pre-excitation and aberrant atrioventricular conduction, combined with cardiac hypertrophy? We have previously shown that in rat, ␥ 2 accounts for only a small proportion (ϳ10%) of total AMPK activity in most tissues, including heart (9). Although there have been no comparable studies in humans, given the close similarity between the rat and human AMPK complexes, it seems likely that this will also be the case in humans. Combined with the observation that mutations in ␥ 2 are inherited in a dominant manner (11,12), these findings indicate that a relatively small reduction in total AMPK activity within the cell can lead to a severe pathophysiological condition. Taken together with the knowledge that the ␥ 2 mutations cause a cardiac-specific defect, and do not cause detectable abnormalities in other tissues, these characteristics indicate that ␥ 2 -containing AMPK complexes must play a unique role in heart development, which cannot be compensated for by the other ␥ isoforms. AMPK has been implicated in the regulation of gene and protein expression (33)(34)(35)(36)(37) and it is possible that ␥ 2 -containing complexes could play a role in the regulation of cardiac-specific proteins that are required for normal development of the heart conductance system. The ␥ 2 polypeptide has an N-terminal extension of ϳ300 amino acids that is not present in either the ␥ 1 or ␥ 3 isoforms (9) and it is conceivable that this is involved in determining substrate recognition. Combined with phosphorylation of cardiac-specific proteins required for heart development, this could account for the isoform and tissue specificity of the ␥ 2 mutations. We are currently exploring these scenarios by generating animal models with genetically altered ␥ 2 subunits.