AMP-activated Protein Kinase Up-regulates Mitogen-activated Protein (MAP) Kinase-interacting Serine/Threonine Kinase 1a-dependent Phosphorylation of Eukaryotic Translation Initiation Factor 4E*

AMP-activated protein kinase (AMPK) is a molecular energy sensor that acts to sustain cellular energy balance. Although AMPK is implicated in the regulation of a multitude of ATP-dependent cellular processes, exactly how these processes are controlled by AMPK as well as the identity of AMPK targets and pathways continues to evolve. Here we identify MAP kinase-interacting serine/threonine protein kinase 1a (MNK1a) as a novel AMPK target. Specifically, we show AMPK-dependent Ser353 phosphorylation of the human MNK1a isoform in cell-free and cellular systems. We show that AMPK and MNK1a physically interact and that in vivo MNK1a-Ser353 phosphorylation requires T-loop phosphorylation, in good agreement with a recently proposed structural regulatory model of MNK1a. Our data suggest a physiological role for MNK1a-Ser353 phosphorylation in regulation of the MNK1a kinase, which correlates with increased eIF4E phosphorylation in vitro and in vivo.

AMP-activated protein kinase (AMPK) is a molecular energy sensor that acts to sustain cellular energy balance. Although AMPK is implicated in the regulation of a multitude of ATP-dependent cellular processes, exactly how these processes are controlled by AMPK as well as the identity of AMPK targets and pathways continues to evolve. Here we identify MAP kinaseinteracting serine/threonine protein kinase 1a (MNK1a) as a novel AMPK target. Specifically, we show AMPK-dependent Ser 353 phosphorylation of the human MNK1a isoform in cellfree and cellular systems. We show that AMPK and MNK1a physically interact and that in vivo MNK1a-Ser 353 phosphorylation requires T-loop phosphorylation, in good agreement with a recently proposed structural regulatory model of MNK1a. Our data suggest a physiological role for MNK1a-Ser 353 phosphorylation in regulation of the MNK1a kinase, which correlates with increased eIF4E phosphorylation in vitro and in vivo.
Balancing catabolic and anabolic processes is fundamental to energy homeostasis and metabolic adaptation. The ␣␤␥ heterotrimeric AMP-activated protein kinase (AMPK) 3 promotes ATP production and limits ATP consumption (1,2). Shifts in the cellular AMP:ATP ratio are detected through the ␥ subunit, which cooperatively binds AMP. Consequential conformational changes within the heterotrimer stimulate AMPK kinase activity. Activation of AMPK further involves Thr 172 phosphorylation in the activation loop of the ␣ subunit kinase domain (3). Additional (auto-) phosphorylation events are known to regulate AMPK (4 -6). Through its established roles in lipid, glucose, and protein metabolism, AMPK functions at the crossroads of energy metabolism and basic cellular processes including cell proliferation, growth, and survival (7)(8)(9). For example, AMPK-mediated regulation of acetyl CoA carboxylases ACC1 and ACC2 accelerates fatty acid synthesis and inhibits fatty acid oxidation, respectively (10). AMPK is involved in control of cell growth via phosphorylation of regulatory-associated protein of mTOR (mechanistic target of rapamycin, Raptor) and tuberous sclerosis 2 (TSC2, tumor suppressor) (11,12).
Despite its well known role as energy sensor, a lack of knowledge on the identity of direct AMPK targets and thus of pathways involved in these processes persists. A substantial number of functional screening approaches have been developed to identify AMPK downstream targets and interaction partners (13)(14)(15)(16)(17). High-density protein microarrays enable the rapid identification of potentially novel human kinase substrates at a proteomic scale (18). Using this strategy, we identified MAP kinase-interacting serine/threonine protein kinase 1 (MNK1) as a putative novel AMPK target (19). The human MKNK1 gene encodes two splice variants, MNK1a and MNK1b, which differ in their C-terminal sequences. In contrast to other MNKs, MNK1a has low basal activity and is highly induced upon activation (20). MNK1a is activated by mitogen-and stress-activated protein kinases (M/SAPK), ERK (extracellular signalingregulated kinase), and P38, via phosphorylation of two threonine residues (Thr 209 and Thr 214 ) within the activation/ T-loop (21,22). MNK1 and MNK2 isoforms phosphorylate the eukaryotic translation initiation factor and mRNA cap-binding protein 4E (eIF4E) on serine 209 (23)(24)(25). Although the exact biological relevance of this phosphorylation event is still under debate, cellular eIF4E plays an important role in the regulation of mRNA translation, in which interaction with the 5Ј-cap structure of mRNA appears pivotal (26). Deregulation of eIF4E has been linked to tumorigenesis (26,27); inhibition of protein translation, e.g. via MNK, is being considered for cancer treatment (20).
Here we show that MNK1a is a genuine AMPK target in vitro and in vivo. MNK1a phosphorylation at Ser 353 by AMPK increases its kinase activity toward eIF4E phosphorylation. The relevance of our findings for human metabolic and neoplastic conditions is discussed.

Results
MNK1a Is a Novel AMPK Target in Vitro-The online tool Scansite predicted the presence of a putative AMPK phosphorylation motif LQRNSSTMDL in MNK1a; this domain is absent in MNK1b. The putative AMPK target site in this motif, Ser 353 , is evolutionarily highly conserved among mammals and lower vertebrates (Fig. 1A). To validate the initial phospho-protein microarray finding, cell-free in vitro kinase (IVK) assays were performed using recombinant AMPK and MNK1a. Mass spectrometric analysis of in vitro phosphorylated recombinant MNK1a WT indicated Ser 353 as the primary phospho-site in the MNK1a-specific tryptic peptide sequence NSSTMDLTL-FAAEAIALNR (Ser 353 underlined; localization probability: 74%; Fig. 1A). To further probe the specificity of this phosphoevent in vitro, the GST tag was also proteolytically removed from recombinant GST-MNK1a to exclude interference by GST tag phosphorylation (28). Both GST-MNK1a WT and MNK1a WT protein were phosphorylated specifically and only in the presence of AMPK (Fig. 1B). The absence of radiolabeled MNK1 WT protein in AMPK-free control reactions suggested that MNK1a WT phosphorylation in vitro was AMPK-dependent and that MNK1a WT protein did not show auto-phosphorylation under these conditions. A commercially available AMPK substrate antiserum, which detects a common AMPKphosphorylated LXRXXpS/pT motif, recognized in vitro phosphorylated MNK1a WT , indicating that MNK1a phosphorylation occurred at an AMPK substrate-like motif (Fig. 1C). The absence of detectable MNK1a WT phosphorylation using a kinase-dead AMPK further corroborated the notion that MNK1 phosphorylation is dependent on the kinase activity of AMPK in vitro (Fig. 1D) To definitively identify the phosphorylated MNK1a residue, a recombinant MNK1a Ser 353 to alanine mutant (MNK1a S353A ) was tested in cell-free assays. To exclude the possibility that the neighboring Ser 352 residue was phosphorylated by AMPK under these conditions, we also generated a MNK1a S352A mutant. Although Mnk1a WT and MNK1a S352A were clearly detected by autoradiography, the MNK1a S353A mutant protein was not radiolabeled under these conditions (Fig. 1E). Analogously, comparative analysis of MNK1a WT and MNK1a S353A showed that MNK1a S353A , in contrast to MNK1a WT , was not detected by the AMPK substrate antiserum (Fig. 1F). These combined data strongly support the idea that AMPK phosphorylates MNK1a WT at Ser 353 in vitro. Cellular MNK1a Is Phosphorylated at Ser 353 upon AMPK Activation-To establish that MNK1a phosphorylation at Ser 353 occurs in living cells, we first compared MNK1a expression levels in different normal diploid and cancer cells. The osteosarcoma cell line U2OS was found to express endogenous MNK1a at a very low level, whereas its AMPK level was relatively high (Fig. 1G). The functional interaction between AMPK and MNK1a was studied in these cells using retrovirally expressed 2PY-MNK1a. Cells were treated with the AMPKactivating compound 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR); AMPK activity was assessed by pACC1 measurement. Although relatively low under control conditions (i.e. non-stimulated or serum-starved), AICAR dramatically increased 2PY-MNK1a WT phosphorylation at an AMPK substrate-like motif (Fig. 1H). Importantly, in contrast to immunoprecipitated (IP) 2PY-MNK1a WT , IP 2PY-MNK1a S353A mutant protein did not show any detectable phosphorylation (Fig. 1H). These data are congruent with cellular AMPK-mediated phosphorylation of MNK1a at Ser 353 .
To be able to study the biological context and relevance of MNK1a phosphorylation at Ser 353 , we generated a polyclonal antiserum specifically recognizing Ser 353 -phosphorylated MNK1a. Both phosphorylated recombinant GST-MNK1a WT and proteolytically released MNK1a WT were detected by the MNK1a-Ser(P) 353 antiserum (Fig. 1I). The finding that recombinant GST-MNK1a S353A was no longer detectable by the MNK1a-Ser(P) 353 antiserum under similar assay conditions corroborated the specificity of the newly generated MNK1a-Ser(P) 353 antiserum (Fig. 1J). To obtain direct evidence that MNK1a-Ser 353 phosphorylation occurs in living cells, U2OS cells expressing MNK1a WT or MNK1a S353A were treated with AICAR. As expected, AICAR increased AMPK-Thr(P) 172 (pAMPK) and ACC1-pS79 (pACC1) in both cell types (Fig. 1K). Importantly, AMPK activation correlated well with increased MNK1a-Ser(P) 353 signal, and the MNK1a S353A mutant was not recognized by the MNK1a-Ser(P) 353 antiserum (Fig. 1K).
MNK1a-Ser 353 Phosphorylation Is AMPK-dependent and Involves Direct Physical Interaction-To obtain genetic evidence for the AMPK dependence of MNK1a-Ser 353 phosphorylation, 2PY-MNK1a-transduced mouse embryonic fibroblasts (MEFs) derived from wild type (AMPK WT ) or AMPK␣1/␣2 double null mutant mice (AMPK dKO ) were treated with either AICAR or A769662, the latter being an AMPK activator with a distinct mode of action (29,30). Both AICAR and A769662 enhanced pAMPK and pACC1 phosphorylation, which correlated well with enhanced MNK1a-Ser 353 phosphorylation in AMPK WT , but not in AMPK dKO MEFs ( Fig. 2A). These data convincingly support the notion that MNK1a is phosphorylated at Ser 353 in an AMPK-dependent fashion in vivo. We then asked whether AMPK physically interacts with MNK1a in living cells. Endogenous AMPK was IP from 2PY-MNK1a WT U2OS cells. The interaction between MNK1a was evident under basal conditions and increased dramatically upon AMPK activation by AICAR or glucose deprivation (Fig. 2B). Accordingly, AMPK-MNK1a complex formation is induced in response to AMPK-activating stimuli.
Metabolic Stress-induced MNK1a-Ser 353 Phosphorylation Controls Its Kinase Activity toward eIF4E-To chart the consequences of AMPK-mediated MNK1a-Ser 353 phosphorylation status as a function of time, 2PY-MNK1a WT U2OS cells were glucose-deprived and AMPK and MNK1 phosphorylation was monitored. MNK1a-Ser(P) 353 and pACC1 were both induced upon glucose deprivation; both phosphorylation events correlated well with AMPK activation (Fig. 2C). Relevantly, 1 h after replenishment of glucose, cells down-regulated MNK1-Ser(P) 353 and pACC1, in parallel with pAMPK (Fig. 2C). These results are strongly suggestive of direct dynamic control of reversible MNK1a phosphorylation at Ser 353 by AMPK.
AMPK activation aims to conserve cellular ATP by tuning energy-consuming processes such as mRNA translation to metabolic conditions. MNK1 is known to phosphorylate the eukaryotic initiation factor 4E (eIF4E) at Ser 209 , a rate-limiting component of the translation apparatus (31). To explore the potential functional relevance of AMPK-mediated MNK1a-Ser 353 phosphorylation in regard to eIF4E-Ser 209 phosphorylation (peIF4E), we compared eIF4E phosphorylation status in the U2OS model transduced with MNK1a S353A or MNK1a S3535D mutants. Cells expressing MNK1a S353D showed almost 2-fold increased eIF4E phosphorylation at basal levels when compared with cells expressing MNK1a WT or MNK1a S353A (Fig. 2D). Combined, these data suggest AMPKdependent stimulation of MNK1a toward its downstream target eIF4E, via MNK1a-Ser 353 phosphorylation.
MNK1-Thr 209 /Thr 214 are phosphorylated within the activation T-loop by the M/SAPKs ERK and P38 (21,22). The C terminus of MNK1a, harboring Ser 353 , is thought to play a dual role in the regulation of MNK1a kinase activity: the C terminus acts repressive in the basal state while allowing full stimulation upon ERK/P38-dependent phosphorylation of Thr 209 /Thr 214 (21). To explore the potential functional interdependence of Thr 209 /Thr 214 and Ser 353 phosphorylation, we studied specific MNK1a phospho-mutants in U2OS cells in combination with pharmacological stimulation rather than metabolic stress, as this allows separation of upstream signaling events. AICAR treatment, alone or in combination with TPA (M/SAPK activator), consistently and highly induced pACC1, whereas TPA treatment alone only modestly induced pACC1 (Fig. 2E). TPA, with or without AICAR, massively increased MNK1-Thr(P) 209 / Thr(P) 214 (MNK1a-pTpT) (Fig. 2E); these observations are in good agreement with reported upstream signaling events connecting M/SAPKs to MNK1a (22). Combination treatment with TPA and AICAR resulted in increased MNK1a-Ser(P) 353 , but only in the case of MNK1a WT and MNK1a K3 M (K78M; kinase-dead). Reduced Ser 353 phosphorylation in U2OS MNK1a K3M cells may be explained by consistently lower expression of the MNK1a K3M mutant (Fig. 2E). As expected, the T209A/T214A (MNK1a T23A2 ) mutant was no longer phosphorylated at the mutated phospho-sites, and neither IP MNK1a T23A2 nor MNK1a K3M mutant kinases phosphorylated eIF4E (Fig. 2E). Remarkably, MNK1a T23A2 was no longer phosphorylated at Ser 353 , whereas conversely, Ser 353 mutation (S3 A or S3 D) did not block the ability of Thr 209 /Thr 214 to become phosphorylated (Fig. 2E). IP MNK1a WT from AICAR only-treated cells led to only marginally increased peIF4E, in support of the need for regulatory co-signaling by M/SAPKs (Fig. 2E). In contrast, and in good agreement with the observations above (cf. Fig. 2D), MNK1a S353D consistently displayed enhanced peIF4E under AICAR only-or TPA only-treated conditions; moreover, eIF4E phosphorylation by MNK1a S353D was already enhanced under non-stimulated conditions (Fig. 2E). Of note, pACC1 was not altered by MNK1a mutation. The combined data suggest that MNK1a selectively acts in the signaling downstream of AMPK, leading to increased eIF4E phosphorylation, and point toward a hierarchical order of regulatory events within MNK1a: M/SAPK-mediated MNK1a-Thr 209 / Thr 214 phosphorylation precedes AMPK-dependent MNK1a-Ser 353 phosphorylation.

Discussion
In this study, we report that MNK1a is a direct AMPK target. We provide biochemical and genetic evidence that MNK1a-Ser 353 is phosphorylated in vitro and in vivo in an AMPK-dependent manner and that AMPK and MNK1a physically interact. Finally, we show that MNK1a-Ser(P) 353 correlates with increased eIF4E phosphorylation in vitro and in vivo.
Members of the MAPK-activated protein kinase (MK) family fulfill multiple biological roles in cellular responses to extracellular cues (e.g. mitogenic stimulation and cellular stress). MNK1a, like other MAPKAPKs (MAP kinase-activated protein kinases), is tightly regulated by M/SAPKs, including ERK and P38 (32,33). The most well characterized biochemical connection in regard to MNK1 is eIF4E-Ser 209 phosphorylation (24), although the exact relevance of MNK1-mediated eIF4E phosphorylation remains unclear (34). MNK1a mutation appears to affect phosphorylation of eIF4E, not of ACC1, suggesting a selective role for MNK1a as a signaling intermediate in the regulation of protein synthesis, downstream of AMPK.
Based on a previously proposed regulatory model of MNK1a, our data suggest a physiological role for Ser 353 phosphorylation in regulating MNK1a activity. MNK1a is activated upon T-loop phosphorylation, but also undergoes a conformational change to a more "open" structure (21). Our MNK1a mutant analysis indicating that Thr 209 /Thr 214 phosphorylation is a prerequisite for Ser 353 phosphorylation appears congruent with this phosphoevent being dependent on conformational change. Interestingly, an ␣ helical structure from the first part of the unique MNK1a . F, as in C. IVK analyses were performed with cold ATP. IB detection was done using the indicated antisera. G, MNK1a and AMPK protein levels in HeLa cervical adenocarcinoma, U2OS, MEF, TIG3 human primary fibroblasts, HEK293T human embryonic kidney, and HL-1 immortal murine cardiomyocytes. IB analyses of total cell extracts using the indicated antisera were performed. H, U2OS cells expressing 2PY-MNK1a WT or MNK1a S353A (control: empty vector (ev)) were treated as indicated (c: control conditions; ss: serum-starved; AI: ssϩAICAR). MNK1a was IP using PY antiserum; IB analysis of immunoprecipitation and input material was performed using the indicated antisera. I, as in C. IVK analyses using purified recombinant GST-MNK1a WT protein, active AMPK, and cold ATP were performed. Protease-mediated removal of the GST tag was performed as indicated in the figure. IB of phosphorylated MNK1a were performed using the indicated antisera. J, as in H. IVK analyses using purified GST-MNK1a WT or GST-MNK1a S353A with active AMPK and cold ATP were performed. IB was done using the indicated antisera. K, U2OS cells expressing 2PY-MNK1a WT , 2PY-MNK1a S353A , or transduced with empty vector (ev) were treated as indicated (serum-starved (ss) or ssϩAICAR (AI)).
C-terminal region was predicted to suppress basal T-loop phosphorylation and activity (21); Ser 353 is located at the start of this region. Based on existing and current data, we propose a functional extension to the MNK1a activation model: T-loop phosphoryla-tion by canonical M/SAPK pathway activity precedes AMPK-mediated phosphorylation of MNK1a at Ser 353 (Fig. 2F).
Our observations suggest that MNK1a-Ser 353 phosphorylation by AMPK is not required for MNK1a activation per se, but affects and/or possibly directs kinase activity, upon a conformational change induced by M/SAPK. MNK1-Ser(P) 353 may serve to fine-tune MNK1-eIF4E-mediated protein synthesis (mRNA translation) via AMPK. This finding has potential therapeutic implications for chronic or acute human conditions, including metabolic syndrome-associated disorders and cancer. Abnormal regulation of protein synthesis is known to drive tumor cell proliferation and survival. Importantly, both AMPK and MNK1a have been independently put forward as potential druggable targets (35,36).
Mass Spectrometry-GST-MNK1a and AMPK proteins were incubated with or without ATP in vitro as described above. Samples were separated by SDS-PAGE; MNK1a bands were excised and in-gel digested as described (40). Extracted tryptic peptides were desalted using homemade C18 columns, resuspended in 10% formic acid, and analyzed by reversed phase nanoLC-MS/MS (Ultimate 3000 and Orbitrap Elite; Thermo Scientific). Peptides were trapped on a precolumn for 10 min (Acclaim PepMap100, C18, 5 m, 100 Å, 300-m inner diameter ϫ 5 mm, Thermo Scientific) in Buffer A (0.1% formic acid in water) and separated on an analytical column (Acclaim Pep-Map100, C18, 5 m, 100 Å, 75-m inner diameter ϫ 25 cm) using a 70-min gradient (0 -10 min, 5% buffer B (80% acetonitrile, 0.1% formic acid); 10 -45 min, 10 -45% buffer B; 45-47 min, 45-99% buffer B; 47-53 min, 99% buffer B; 53-70 min, 5% buffer B) at 250 nl/min. The mass spectrometer was operated in data-dependent mode with a 20-s dynamic exclusion range. Full-scan MS spectra were acquired in the Orbitrap (range: m/z 350 -1500) with a resolution of 120,000 and an automatic gain control of 1E6 ions. Collision-induced dissociation fragmentation in the ion trap was performed on the top five precursors of each full scan employing a collision energy of 35%. Analysis of raw data was done by using MaxQuant (version 1.4.1.2 (41)). The spectra were searched against the human Swiss-Prot database version 06/2014 using the Andromeda search engine using default mass tolerance settings (42). Trypsin was set as the protease (two missed cleavages allowed). Fixed modification: carbamidomethylation (Cys); variable modifications: oxidation (Met), phosphorylation (Ser, Thr, Tyr), and N-terminal protein acetylation. False discovery rate was set to 0.01 for peptides, proteins, and modification sites. The minimum peptide score for modified peptides was set to 40, and the minimum peptide length was seven amino acids (aa).
eIF4E Kinase Assay-350 g of U2OS 2PY-MNK1a cell lysates were incubated with PY antibody (2 h, 4°C); 40 l of protein G slurry were added for 1.5 h (4°C). Beads were pelleted (5 min, 800 ϫ g) and washed twice with mild lysis buffer, twice with 0.5 M LiCl, and twice with kinase buffer as described previously (21). Beads were re-suspended in 50 l of kinase buffer and incubated (15 min, 37°C) with 1 g of recombinant GST-eIF4E. The reaction was terminated by adding SDS-PAGE sample buffer (5 min, 95°C). Samples were analyzed by SDS-PAGE (9% gels) and autoradiography. 200 g of cell lysates were processed in parallel with immunoprecipitation antibody as controls.