Mammalian TAK1 Activates Snf1 Protein Kinase in Yeast and Phosphorylates AMP-activated Protein Kinase in Vitro*

The Snf1/AMP-activated protein kinase (AMPK) family is important for metabolic regulation and is highly conserved from yeast to mammals. The upstream kinases are also functionally conserved, and the AMPK kinases LKB1 and Ca2+/calmodulin-dependent protein kinase kinase activate Snf1 in mutant yeast cells lacking the native Snf1-activating kinases, Sak1, Tos3, and Elm1. Here, we exploited the yeast genetic system to identify members of the mammalian AMPK kinase family by their function as Snf1-activating kinases. A mouse embryo cDNA library in a yeast expression vector was used to transform sak1Δ tos3Δ elm1Δ yeast cells. Selection for a Snf+ growth phenotype yielded cDNA plasmids expressing LKB1, Ca2+/calmodulin-dependent protein kinase kinase, and transforming growth factor-β-activated kinase (TAK1), a member of the mitogen-activated protein kinase kinase kinase family. We present genetic and biochemical evidence that TAK1 activates Snf1 protein kinase in vivo and in vitro. We further show that recombinant TAK1, fused to the activation domain of its binding partner TAB1, phosphorylates Thr-172 in the activation loop of the AMPK catalytic domain. Finally, expression of TAK1 and TAB1 in HeLa cells or treatment of cells with cytokines stimulated phosphorylation of Thr-172 of AMPK. These findings indicate that TAK1 is a functional member of the Snf1/AMPK kinase family and support TAK1 as a candidate for an authentic AMPK kinase in mammalian cells.

The Snf1/AMP-activated protein kinase (AMPK) family is important for metabolic regulation and is highly conserved from yeast to mammals. The upstream kinases are also functionally conserved, and the AMPK kinases LKB1 and Ca 2؉ /calmodulindependent protein kinase kinase activate Snf1 in mutant yeast cells lacking the native Snf1-activating kinases, Sak1, Tos3, and Elm1. Here, we exploited the yeast genetic system to identify members of the mammalian AMPK kinase family by their function as Snf1-activating kinases. A mouse embryo cDNA library in a yeast expression vector was used to transform sak1⌬ tos3⌬ elm1⌬ yeast cells. Selection for a Snf ؉ growth phenotype yielded cDNA plasmids expressing LKB1, Ca 2؉ /calmodulin-dependent protein kinase kinase, and transforming growth factor-␤-activated kinase (TAK1), a member of the mitogen-activated protein kinase kinase kinase family. We present genetic and biochemical evidence that TAK1 activates Snf1 protein kinase in vivo and in vitro. We further show that recombinant TAK1, fused to the activation domain of its binding partner TAB1, phosphorylates Thr-172 in the activation loop of the AMPK catalytic domain. Finally, expression of TAK1 and TAB1 in HeLa cells or treatment of cells with cytokines stimulated phosphorylation of Thr-172 of AMPK. These findings indicate that TAK1 is a functional member of the Snf1/AMPK kinase family and support TAK1 as a candidate for an authentic AMPK kinase in mammalian cells.
The Snf1/AMP-activated protein kinase (AMPK) 2 family has major roles in regulation of glucose and lipid metabolism, maintenance of cellular energy homeostasis, and cellular stress responses (reviewed in Refs. 1 and 2). In mammalian cells, reduced energy availability (high cellular AMP:ATP ratio) causes activation of AMPK, which promotes glucose transport and ATP-generating metabolic processes, inhibits ATP-consuming processes, and regulates transcription. AMPK is also regulated by leptin, adiponectin, and ghrelin (3)(4)(5) and has a role in controlling appetite and food intake (5,6). In humans, AMPK is an important therapeutic target for type 2 diabetes (2,7).
In the yeast Saccharomyces cerevisiae, Snf1 protein kinase (8) is the ortholog of AMPK (9,10). Snf1 protein kinase, like AMPK, is heterotrimeric, comprising a catalytic subunit (Snf1/ ␣), and two regulatory subunits (␤ and Snf4/␥). Mutation of SNF1 causes the Snf Ϫ (sucrose-nonfermenting) phenotype, which is characterized by inability to utilize carbon sources that are less preferred than glucose. Like AMPK, Snf1 protein kinase regulates transcription, metabolic enzymes, and transporters in response to stress, particularly carbon stress (reviewed in Refs. 11 and 12).
Here, we took advantage of this conservation of the Snf1/ AMPK pathway and exploited the yeast genetic system in an effort to identify new members of the AMPK kinase family. Given that yeast, a simple unicellular organism, has three Snf1 protein kinase kinases, it seems likely that mammals have multiple AMPK kinases. The heterologous function of LKB1 and CaMKK in yeast provides the basis for a convenient and powerful genetic selection for mammalian AMPK kinases: the restoration of the Snf ϩ growth phenotype in sak1⌬ tos3⌬ elm1⌬ mutant yeast. The power of this selection lies not only in its simplicity but also in its sensitivity. The Snf1 pathway is robust, and very little activity is required for growth; for example, expression of LKB1 alone restores growth despite causing only a modest elevation of Snf1 catalytic activity (16). This sensitivity is important because the catalytic subunit of an AMPK kinase must function in yeast without other mammalian proteins, either alone or in association with yeast orthologs.
We present genetic and biochemical evidence that TAK1 phosphorylates Thr-210 of Snf1 and functions as a Snf1-activating kinase. We further show that TAK1 phosphorylates Thr-172 of the AMPK catalytic subunit in vitro and that expression of TAK1 and TAB1 stimulates phosphorylation of AMPK in HeLa cells. These findings suggest that TAK1 is a member of the Snf1/AMPK kinase family.
Selection for Mammalian Snf1-activating Kinases in Yeast-DNA of a two-hybrid library prepared from mouse 17-day embryo cDNAs in a yeast expression plasmid vector carrying the LEU2 marker (Clontech catalog number 638846) was used to transform (36) yeast strain MCY5138 (see Fig. 2). A total of 5 ϫ 10 6 transformants were selected on 500 plates of SC solid medium containing 2% glucose and lacking leucine. Colonies from each plate were resuspended in SC medium and transferred to a fresh plate of SC-leucine solid medium containing 2% raffinose plus the respiratory inhibitor antimycin A (1 g/ml). Growth on this medium requires activation of Snf1 protein kinase; in control experiments, colonies expressing LKB1 appeared in 3-7 days. After 5-7 days, two colonies from each plate were picked and retested for growth. Plasmid DNAs were rescued by passage through bacteria, retested by transformation of MCY5138, and sequenced. One plasmid was saved from each plate.

Assay of Phosphorylation of Recombinant Snf1 and AMPK Catalytic Domains-Glutathione S-transferase (GST) fusions
to the mutant Snf1 catalytic domains Snf1KD-K84R and Snf1KD-T210A were expressed in bacteria and purified as described (16). His-tagged AMPK-KD-WT and AMPK-KD-T172A catalytic domains were expressed in bacteria and purified using AKTA fast protein liquid chromatography on chelating HiTrap resin (Amersham Biosciences). Bound proteins were eluted with a linear gradient as described by the manufacturer. Cultures of MCY5138 expressing HA-TAK1 and/or LexA-TAB1 were grown in SC with 2% glucose, collected by filtration, incubated in 0.05% glucose for 30 min, and collected by filtration. HA-tagged proteins were immunoprecipitated from extracts (200 g) with anti-HA antibody as described (13). Kinases were assayed for phosphorylation of GST-Snf1KD (3 g) or AMPK-KD (0.5 g) substrates using [␥-32 P]ATP as described (16). His-tagged recombinant human TAK1-TAB1 fusion protein (100 ng; Upstate catalog number 14-600) was incubated with substrates and cold ATP.
Analysis of Phosphorylation of AMPK in HeLa Cells-HeLa cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine. Cells were transfected with DNAs (8 g/6-cm dish) using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's instructions. TNF-␣ and IL-1␤ were purchased from R&D Systems. Cells were lysed by the addition of ice-cold lysis buffer as described (20), except without prior rinsing. Lysates were collected immediately and clarified by brief centrifugation in the cold.

Genetic Selection for Mammalian Snf1-activating Kinases in
Yeast-The sak1⌬ tos3⌬ elm1⌬ mutant yeast strain lacks all three native Snf1 protein kinase kinases and therefore exhibits the Snf Ϫ (sucrose-nonfermenting) phenotype, which is characterized by the ability to utilize glucose but not alternative carbon sources. To identify mammalian Snf1-activating kinases, and thus candidates for AMPK kinases, we selected mammalian cDNAs that allow sak1⌬ tos3⌬ elm1⌬ cells to grow on raffinose, as shown schematically in Fig. 2. We used a library of mouse 17-day embryo cDNAs fused to the Gal4 activation domain (GAD) in a yeast expression vector to transform sak1⌬ tos3⌬ elm1⌬ yeast cells and then selected for growth on raffinose. We recovered the cDNA plasmid from Snf ϩ colonies by passage through Escherichia coli, retransformed sak1⌬ tos3⌬ elm1⌬ yeast cells to confirm that the cDNA conferred a Snf ϩ phenotype, and identified the cDNA by sequencing. In a screen of 5 ϫ 10 6 transformants, we recovered 49 cDNA clones expressing LKB1, five expressing CaMKK␤, and six expressing TAK1, also known as MAPKKK7 (23). This selection also yielded seven cDNAs expressing transcription factors, which were not characterized further, but none expressing CaMKK␣, which we previously showed to function in yeast (Ref. 16; see Fig. 3B). The recovery of LKB1 and CaMKK␤, which are both known AMPK kinases, validates this approach. TAK1 is thus a candidate for a Snf1-activating kinase and potentially an AMPK kinase.
Growth Phenotype Conferred by TAK1 Requires Snf1 Protein Kinase-We first sought to confirm that the ability of TAK1 to confer growth on raffinose requires Snf1 protein kinase. A cDNA plasmid expressing GAD-TAK1 was used to transform snf1⌬ mutant cells. The transformants did not grow on raffinose (Fig. 3A), indicating that TAK1 requires Snf1 protein kinase to confer a Snf ϩ phenotype and does not function by  Cultures expressing the indicated proteins were spotted with serial 5-fold dilutions on selective SC medium containing 2% glucose, 2% raffinose and the respiratory inhibitor antimycin A (1 g/ml), or 2% glycerol plus 3% ethanol. Plates were incubated at 30°C and were photographed after 3 days, or after 6 days for raffinose plates. A, snf1⌬10 cells expressed GAD-Snf1, GAD-TAK1, or GAD (vector). B and C, sak1⌬ tos3⌬ elm1⌬ cells expressed HA-CaMKK␣, HA-TAK1, HA-TAK1K63W, HA-Sak1, or HA (vector) and either LexA-TAB1 or LexA (vector). Additional transformants expressing TAK1 and TAB1 and transformants expressing TAK1K63W from three independent mutant plasmids were tested on raffinose with similar results.
bypassing Snf1. In control experiments, expression of Snf1 in the mutant cells restored growth, as expected.
The cDNA clones recovered from the library expressed TAK1 with GAD fused to its N terminus. To exclude the possibility that this fusion protein had aberrant function, we expressed full-length TAK1, tagged with a triple-HA epitope at its N terminus, from the yeast ADH1 promoter of vector pWS93. Expression of this HA-tagged TAK1 allowed sak1⌬ tos3⌬ elm1⌬ cells to grow on raffinose (Fig. 3B) and on glycerol plus ethanol (Fig. 3C); HA-TAK1 was used in all subsequent experiments. TAB1, a TAK1-binding protein identified in the two-hybrid system, increases the catalytic activity of TAK1 (33); however, TAK1 acts independently of TAB1 in some signaling pathways in mammalian cells (44). Coexpression of LexA-TAB1 from the ADH1 promoter did not improve growth of sak1⌬ tos3⌬ elm1⌬ cells on raffinose (Fig. 3B), although some improvement was noted on glycerol-ethanol (Fig. 3C); expression was confirmed by immunoblot analysis (data not shown and Fig. 4C). In addition, TAK1, with or without TAB1, did not allow raffinose utilization by snf1⌬ cells expressing mutant Snf1T210A with the activation loop Thr-210 replaced by Ala, as predicted if TAK1 functions by phosphorylating Thr-210 (data not shown).
TAK1 Activates Snf1 Protein Kinase in Vivo-To determine whether TAK1 activates Snf1 protein kinase in vivo, we assayed Snf1 catalytic activity in sak1⌬ tos3⌬ elm1⌬ mutant cells expressing HA-TAK1. Cells were grown to mid-log phase in glucose and then shifted to medium containing 0.05% glucose for 30 min, a condition that results in activation of Snf1 in wildtype cells. Cell extracts were prepared, and phosphorylation of a synthetic peptide substrate, the SAMS peptide, by partially purified Snf1 protein kinase was determined. The presence of HA-TAK1 in the mutant cells resulted in the activation of Snf1 to levels similar to those caused by CaMKK␣ (Fig. 4A), which is roughly 2-fold reduced relative to wild type (16). Coexpression of LexA-TAB1 with HA-TAK1 did not substantially increase activation of Snf1 (Fig. 4A), consistent with the growth phenotypes (Fig. 3). Amounts of Snf1 protein were similar in all assays, and coexpression of TAB1 did not result in elevated levels of TAK1, although TAK1 appeared to stabilize TAB1, as judged by immunoblot analysis (Fig. 4C). Together with growth assays, these data suggest that in yeast cells, TAK1 functions as a Snf1activating kinase and does so largely independently of TAB1. We cannot exclude the possibility that a native yeast protein functionally substitutes for TAB1, but no yeast sequence homolog is evident.
TAK1 Catalytic Activity Is Required for Activation of Snf1 Protein Kinase-To determine whether the effects of TAK1 in yeast cells were due to the catalytic activity of TAK1, we introduced a mutation altering Lys-63 to Trp, which was previously shown to abolish catalytic activity of TAK1 (23). The kinasedead mutant protein, TAK1K63W, was expressed ( Fig. 4D) but did not confer growth on raffinose, indicating that catalytic activity and not some other property of the protein is required (Fig. 3C). In accord with this result, TAK1K63W did not activate Snf1 protein kinase activity in vivo, as judged by phosphorylation of the SAMS peptide (Fig. 4B). Thus, the function of TAK1 as a Snf1-activating kinase in yeast requires its catalytic activity.

TAK1-TAB1 Phosphorylates the Activation Loop Thr-210 of Recombinant Snf1
Catalytic Domain-We next assayed the ability of TAK1 purified from yeast cells to phosphorylate the kinase domain of Snf1 (Snf1KD) in vitro. We used as substrates two bacterially expressed, inactive forms of the Snf1 catalytic domain, GST-Snf1KD-K84R, which has a substitution of the conserved Lys of the ATP-binding site, and GST-Snf1KD-T210A, which is mutant for the activation loop Thr-210. HA-TAK1 was expressed in sak1⌬ tos3⌬ elm1⌬ cells in combination with LexA-TAB1 or LexA, immunoprecipitated with anti-HA antibody, and incubated with different substrates and [␥-32 P]ATP. TAB1 has been reported to stimulate the autophosphorylation of TAK1 (45,46). The presence of TAB1 increased the phosphorylation of Snf1KD substrates (Fig. 5A) but did not increase the recovery of TAK1 (Fig. 5C); longer exposure revealed very weak phosphorylation of Snf1KD by TAK1 in the absence of TAB1 (Fig. 5A, lower panel). These results stand in contrast to the minimal effect of TAB1 on activation of Snf1 by TAK1 in yeast cells; however, the substrate in vivo was fulllength Snf1 protein, presumably in the context of the heterotrimeric Snf1 protein kinase complex.
Both Snf1KD substrates were phosphorylated by TAK1 with TAB1, although Snf1KD-T210A was phosphorylated less strongly than was Snf1KD-K84R (Fig. 5A), and Coomas- A and B, partially purified Snf1 was assayed for phosphorylation of the SAMS peptide. Extracts were prepared from two independent transformants, and each extract was assayed twice, with dilutions. Values are averages of four assays. For TAK1K63W, three transformants carrying independent mutant plasmids were used. C and D, fractions assayed above were subjected to immunoblot analysis with anti-Snf1, anti-HA, and anti-LexA antibodies. Lane numbers correspond to assays numbered in panels A and B. SEPTEMBER 1, 2006 • VOLUME 281 • NUMBER 35

TAK1 Activates Snf1 in Yeast
sie Blue staining confirmed that both substrates were present at similar levels (Fig. 5B). These findings suggest that TAK1 recognizes Thr-210, as well as other residues. To directly assess the phosphorylation of Thr-210, we carried out immunoblot analysis with anti-phospho-Thr-172-AMPK-specific antibody, which cross-reacts with phospho-Thr-210 of Snf1. This antibody detected Snf1KD-K84R, but not Snf1KD-T210A, indicating that TAK1 phosphorylates Thr-210 in the activation loop (Fig. 5C).
To confirm that TAK1, and not a coprecipitating kinase, is responsible for phosphorylation of Thr-210, we incubated recombinant human TAK1-TAB1 fusion protein (TAK1 residues 1-303 fused to TAB1 residues 437-end) (47) with the Snf1KD substrates and cold ATP. Immunoblot analysis with phospho-Thr-172-AMPK antibody detected Snf1KD-K84R but not Snf1KD-T210A (Fig. 5D). These biochemical studies indicate that TAK1 phosphorylates the activation loop Thr-210 of Snf1 in vitro and, together with genetic evidence, suggest that TAK1 functions directly as a Snf1-activating kinase in yeast cells in vivo.
TAK1 Restores Normal Cell Morphology in elm1⌬ Cells-The Snf1 protein kinase kinases Elm1 and Sak1 have other roles in the cell besides activation of Snf1, and we tested TAK1 for the ability to provide these distinct functions. The elm1 mutation is named for the elongated morphology of the mutant cells, which results from defects in cell cycle progression and has no apparent relationship to Snf1 (48 -51). Expression of TAK1, with or without TAB1, in elm1⌬ cells restored normal cell morphology, whereas expression of TAK1K63W with TAB1 had no effect (Fig. 6A). Thus, TAK1 functionally substitutes for Elm1 with respect to this phenotype. These findings suggest that TAK1 and Elm1 phosphorylate substrate(s) that are not efficiently recognized by Sak1 or Tos3. Neither LKB1 nor CaMKK substituted for Elm1 in this regard (16).
The sak1⌬ mutation prevents the nuclear enrichment of Snf1 protein kinase containing the Gal83 ␤ subunit (Snf1-Gal83) in response to glucose limitation (37). Although activation of Snf1 is required for this nuclear enrichment (37), evidence suggests that Sak1 also functions in another capacity, besides activating Snf1, to promote the nuclear enrichment of Snf1-Gal83 (16). To test whether TAK1 provides this function, we expressed TAK1 and TAB1 in sak1⌬ cells carrying a centromeric plasmid expressing Gal83-GFP from its native promoter. Cells were grown to mid-log phase in medium containing 2% glucose and shifted to 0.05% glucose for 10 min. Microscopic observation revealed no nuclear enrichment of Gal83-GFP (Fig.  (vector) were grown on selective SCϩ2% glucose and were imaged by differential interference contrast (DIC ). B, sak1⌬ cells expressing Gal83-GFP, HA-TAK1, and LexA-TAB1 were grown in selective SCϩ2% glucose and shifted to 0.05% glucose for 10 min. Nuclei were stained with 4Ј,6-diamidino-2-phenylindole (DAPI). GFP fluorescence, 4Ј,6-diamidino-2-phenylindole staining, and differential interference contrast are shown. Cells were viewed using a Nikon Eclipse E800 fluorescence microscope, and images were taken with an Orca100 (Hamamatsu) camera by using Open Lab (Improvision) software. 6B). Together, these findings suggest that with respect to function, TAK1 is more closely related to Elm1 than to Sak1.
Recombinant TAK1-TAB1 Fusion Protein Phosphorylates Thr-172 of AMPK Catalytic Domain-The above evidence that TAK1 functions in vivo and in vitro as a Snf1-activating kinase supports TAK1 as a candidate for an AMPK kinase. To examine whether TAK1 phosphorylates AMPK on the activation loop Thr-172 in vitro, we expressed in bacteria the wild-type AMPK catalytic domain, AMPK-KD-WT, and a mutant version containing a replacement of Thr-172 with Ala, AMPK-KD-T172A.
We then assayed phosphorylation of AMPK-KD substrates by TAK1. HA-TAK1 was expressed in sak1⌬ tos3⌬ elm1⌬ cells in combination with LexA or LexA-TAB1, immunoprecipitated, and incubated with both versions of AMPK-KD and [␥-32 P]ATP. The presence of TAB1 increased the phosphorylation of both substrates (Fig. 7A), without increasing the recovery of TAK1 (Fig. 7B). A mock immunoprecipitation with no cell extract gave results similar to the control sample with HA and LexA-TAB1 ( Fig. 7A and data not shown). TAK1, when coexpressed with TAB1, phosphorylated AMPK-KD-WT more strongly than AMPK-KD-T172A, and close inspection revealed a doublet in the case of the wild-type substrate, suggesting that Thr-172 is one of the sites recognized. We were unable to assay phosphorylation using phospho-Thr-172-AMPK antibody, however, because AMPK-KD migrated close to immunoglobulin.
Coexpression of TAK1 and TAB1 in HeLa Cells Increases AMPK Phosphorylation-These findings suggest TAK1 as a candidate for an AMPK kinase in mammalian cells. To explore this possibility, we transfected HeLa cells, which do not express the major AMPK kinase LKB1 (52), with combinations of plasmids for transient expression of TAK1, TAK1K63W, and TAB1 from the vector pCMV-FLAG2. At 30 h after transfection, cells were subjected to serum-free medium for 4 h. Cell lysates, prepared by a rapid lysis procedure, were subjected to immunoblot analysis with anti-phospho-Thr-172-specific antibody, and blots were reprobed with anti-AMPK␣, anti-TAK1, and anti-FLAG antibodies (Fig. 8A). Expression of TAK1 alone had little or no effect, but coexpression of TAK1 and TAB1 led to increased phosphorylation of Thr-172, whereas levels of AMPK catalytic subunit remained constant. Kinase-dead TAK1K63W, with TAB1, did not increase phosphorylation of Thr-172. TAK1K63W was expressed at lower levels than TAK1, which was not the case in yeast (Fig. 4D), suggesting that in HeLa cells, the kinase-dead protein is either less stable or deleterious. Similar results were observed when cells were transferred to serum-free medium for 14 h at 12, 18, or 24 h after transfection (Fig. 8A). Coexpression of TAK1 and TAB1 also increased   SEPTEMBER 1, 2006 • VOLUME 281 • NUMBER 35 phosphorylation of AMPK in HeLa cells treated with 0.5 M sorbitol or 1 mM hydrogen peroxide for 15 min (Fig. 8B). Both treatments are known to activate AMPK, but hyperosmotic stress does not alter the AMP:ATP ratio (7) and has been shown to stimulate TAK1 activity (25,53). Together, these findings indicate that coexpression of TAK1 and TAB1 stimulates phosphorylation of AMPK in HeLa cells. Although TAK1 is capable of phosphorylating AMPK in vitro, it remains possible that the effects of TAK1 on AMPK in vivo are indirect. Several cytokines stimulate TAK1 activity, including TGF-␤, TNF-␣, and IL-1 (23,25,26,53). Exposure to TGF-␤ induces phosphorylation of AMPK on Thr-172 in HeLa and HepG2 cells (54,55). To test the effects of TNF-␣ and IL-1, we subjected HeLa cells to serum-free medium for 14 h, treated with cytokine, and analyzed cell lysates by immunoblotting. In both cases, we detected a modest increase in Thr-172 phosphorylation between 2 and 10 min (Fig. 8C), consistent with the possibility that native TAK1 phosphorylates AMPK.

DISCUSSION
Taking advantage of the conservation of the Snf1/AMPK pathway, we have used the yeast system to identify putative AMPK kinases by their function as Snf1-activating kinases. This genetic selection yielded two authentic AMPK kinases, LKB1 and CaMKK␤, and a new candidate, TAK1. The utility of this genetic approach is that it is based on function. Although LKB1 and CaMKK are homologous to the three yeast Snf1 protein kinase kinases, TAK1 was not identified as a candidate AMPK kinase on the basis of sequence similarity. In this study, we used a mouse embryo cDNA library, which may not represent the entire repertoire of AMPK kinases. Different libraries from other developmental stages or from specific tissues may yield additional AMPK kinases. Such kinases are potentially useful therapeutic targets in the AMPK pathway.
We present genetic and biochemical evidence that validates TAK1 as a Snf1-activating kinase. We further show that recombinant TAK1-TAB1 phosphorylates AMPK on Thr-172 in vitro and that overexpression of TAK1 and TAB1 stimulates phosphorylation of AMPK in HeLa cells. The stimulatory effects of TGF-␤ (54), TNF-␣, and IL-1 on phosphorylation of AMPK in HeLa cells are also in accord with the possibility that TAK1 phosphorylates AMPK. Together, these findings support TAK1 as a candidate for an authentic AMPK kinase in mammalian cells.
We also note that other work has connected TAB1 with AMPK. TAB1 interacts with the ␣2 isoform of the catalytic subunit of AMPK in mouse heart, and activation of AMPK promoted the association of p38 MAPK with TAB1 in ischemic heart (56); however, TAK1 was not implicated, and evidence suggests that TAK1 and TAB1 have some independent roles (44).
Further studies are required to validate TAK1 as an AMPK kinase in mammalian cells, and the regulation of AMPK by cytokines warrants further investigation in different cell types. It will also be interesting to address the possibility that TAK1 phosphorylates other members of the AMPK-related protein kinase family, as does LKB1 (57).