Biochemical Journal

Research article

N-Myristoylation is essential for protein phosphatases PPM1A and PPM1B to dephosphorylate their physiological substrates in cells

Toko Chida, Masakatsu Ando, Tasuku Matsuki, Yutaro Masu, Yuko Nagaura, Teruko Takano-Yamamoto, Shinri Tamura, Takayasu Kobayashi

Abstract

PPM [metal-dependent protein phosphatase, formerly called PP2C (protein phosphatase 2C)] family members play essential roles in regulating a variety of signalling pathways. While searching for protein phosphatase(s) that act on AMPK (AMP-activated protein kinase), we found that PPM1A and PPM1B are N-myristoylated and that this modification is essential for their ability to dephosphorylate the α subunit of AMPK (AMPKα) in cells. N-Myristoylation was also required for two other functions of PPM1A and PPM1B in cells. Although a non-myristoylated mutation (G2A) of PPM1A and PPM1B prevented membrane association, this relocalization did not likely cause the decreased activity towards AMPKα. In in vitro experiments, the G2A mutants exhibited reduced activities towards AMPKα, but much higher specific activity against an artificial substrate, PNPP (p-nitrophenyl phosphate), compared with the wild-type counterparts. Taken together, the results of the present study suggest that N-myristoylation of PPM1A and PPM1B plays a key role in recognition of their physiological substrates in cells.

  • AMP-activated protein kinase (AMPK)
  • N-myristoylation
  • protein phosphatase 2C (PP2C)
  • metal-dependent protein phosphatase 1A (PPM1A)
  • metal-dependent protein phosphatase 1B (PPM1B)

INTRODUCTION

Protein N-myristoylation is the irreversible covalent linkage of the 14-carbon saturated fatty acid myristic acid to the N-terminal glycine residue of many eukaryotic and viral proteins. N-Myristoylation provides some proteins with a relatively weak membrane anchor and is important for correctly localizing them within the cells [13]. N-Myristoylation also occurs on many proteins that do not appear to associate with membranes. The role of the myristoyl moiety in these proteins is less clear, but may involve protein–protein interaction, stabilizing the protein structure and regulation of the catalytic activity of the enzyme [17]. An example of a protein whose structure is stabilized by a myristoyl group is PKA (cAMP-dependent protein kinase/protein kinase A), in which the myristoyl moiety binds in a deep hydrophobic cleft [4]. Although the non-myristoylated form of PKA is kinetically indistinguishable from the myristoylated enzyme and both are equal in their ability to associate with the regulatory subunit, the myristate seems to increase thermal stability of the kinase, possibly by binding to an adjacent hydrophobic region of the protein. An example of an enzyme whose catalytic activity is regulated by myristoylation is c-Abl tyrosine kinase [5,6]. In c-Abl, the myristoyl group binds to a hydrophobic pocket and induces a conformational change that allows the docking of the SH2 (Src homology 2) and SH3 (Src homology 3) domains on to the kinase domain leading to autoinhibition of the enzyme [5,6]. Although myristoylation is an irreversible modification, its function may be regulated via various ‘myristoyl switches’ which rely on a ligand-dependent conformational change of the protein leading to exposure of the myristoyl moiety sequestered in a hydrophobic pocket to the cytosol [8].

The PPM [metal-dependent protein phosphatase, formerly called PP2C (protein phosphatase 2C)] family is one of two major protein serine/threonine phosphatase families in eukaryotes [9]. In contrast with members of the other major family, the PPP (phosphoprotein phosphatase) family, which function as oligomeric complexes, PPM members do not have regulatory subunits but contain unique domains that are thought to regulate catalytic activity, determine substrate specificity and affect subcellular localization [10]. PPM members are insensitive to any known protein phosphatase inhibitors, including okadaic acid and microcystin-LR. The PPM family includes highly conserved protein phosphatases with 17 distinct genes in the human genome. These proteins play pivotal roles in regulating the stress response, cell-cycle progression, apoptosis, Ca2+ signalling, BMP (bone morphogenetic protein) signalling, metabolism, RNA splicing, mitochondrial function and lipid transfer [1013].

The cDNA of PPM1A (formerly PP2Cα) was first isolated from a rat kidney cDNA library in 1989 [14]. Although PPM1B (formerly PP2Cβ) is closely related to PPM1A (75% identity), knockout mice lacking these isoforms exhibited completely different phenotypes, suggesting that they have different physiological roles in vivo [15,16]. Molecular and cellular biological studies have revealed that PPM1A dephosphorylates p38 MAPK (mitogen-activated protein kinase), MKK (MAPK kinase) 4 and MKK6 in the SAPK (stress-activated protein kinase) cascade, and Smad2/3, proteins in the BMP signalling pathway, whereas PPM1B dephosphorylates TAK1 [TGF (transforming growth factor)-β-activated kinase 1], a member of the SAPK cascade [1722]. However, cellular studies have shown that both PPM1A and PPM1B dephosphorylate IKKβ {IκB [inhibitor of NF-κB (nuclear factor κB)] kinase}, a protein kinase in the NF-κB pathway [23,24]. These observations suggest that PPM1A and PPM1B have both non-redundant and redundant roles in cells. In in vitro experiments, bacterially expressed mammalian PPMA has broad substrate specificity and even can dephosphorylate small molecules such as PNPP (p-nitrophenyl phosphate) [25]. But PP2C purified from rabbit liver (probably a mixture of PPM1A and PPM1B) does not act on this atypical substrate [26], suggesting that the native enzyme is more selective; however, its molecular basis remains unknown.

AMPK (AMP-activated protein kinase) acts as a sensor of cellular energy status [27,28]. AMPK becomes activated in response to increased AMP concentrations, which then induces metabolic pathways that generate ATP, while also repressing ATP consumption that is not essential for short-term survival. AMPK is a heterotrimeric enzyme comprising a catalytic α-subunit and regulatory β- and γ-subunits, each of which contains two or more isoforms encoded by distinct genes. The α-subunits (α1 and α2) contain a kinase domain and are only active after phosphorylation of Thr172 by upstream kinases, including LKB1, CaMKKβ [CaM (calmodulin)-dependent protein kinase kinase β] and TAK1. The β-subunits (β1 and β2) act as scaffolds and are reportedly myristoylated at their N-terminus, which is essential for the membrane localization of the AMPK holoenzyme and to initiate AMPK signalling in response to AMP [7,29]. The γ-subunits (γ1–3) bind AMP, which activates the kinase by promoting Thr172 phosphorylation by upstream kinases, inhibiting dephosphorylation by protein phosphatases and exerting direct allosteric effects. In 1991, Moore et al. [30] proposed that PPM acts as an AMPKα phosphatase on the basis of the finding that okadaic-acid-insensitive protein phosphatase activity was effective in dephosphorylating AMPK in a cell-free assay. Experiments using recombinant proteins expressed in bacteria, as well as siRNA (small interfering RNA), suggested that PPM1A acts on AMPK [31,32]. Recently, however, Voss et al. [33] reported that PPM1E (formerly CaMKP-N/POPX1), but not PPM1A, was responsible for the dephosphorylation of AMPK in HEK (human embryonic kidney)-293 cells. Their conclusion was based on observations that siRNA-mediated knockdown of PPM1E, but not PPM1A, increased AMPKα phosphorylation. However, whether the other PPM family members also regulate AMPKα remains to be elucidated. In addition, the specificity of protein phosphatases for the two isoforms of AMPKα (1 and 2) has not been determined.

In the present study, we provide evidence that among the 11 PPM isoforms tested, PPM1B, in addition to PPM1A and PPM1E, contributes to the dephosphorylation of AMPKα in HeLa cells. Furthermore, we found that PPM1A and PPM1B are N-myristoylated and that the N-myristoyl moiety of PPM1A and PPM1B is essential for their recognition of physiological substrates in cells.

EXPERIMENTAL

Materials

cDNAs encoding mouse PPM1G (PP2Cγ), mouse ILKAP [ILK1 (integrin-linked kinase 1)-associated protein] (PP2Cδ), mouse PPM1D (wip1), mouse PPM1E (CaMKP-N/POPX1), human PPM1F (CaMKP/POPX2), mouse PPM1H (NERPP), human PPM1K (PP2Cκ), human AMPKα1, human AMPKα2, human AMPKβ1, human AMPKγ1 and human LKB1 were cloned using PCR and then subcloned into the pcDNA3 vector. cDNAs for the expression constructs for mouse PPM1A (PP2Cα), mouse PPM1B (PP2Cβ), mouse PPM1L (PP2Cϵ) and mouse PPM1M (PP2Cη) have been described previously [13,21,34,35]. Details of the expression constructs used in the present study are shown in Supplementary Table S1 (at http://www.BiochemJ.org/bj/449/bj4490741add.htm). The modifying enzymes used for DNA manipulation were obtained from New England Biolabs. Lipofectamine™ 2000, Lipofectamine™ RNAiMax and siRNAs were obtained from Invitrogen. Sequences of siRNA used in the present study are shown in Supplementary Table S2 (at http://www.BiochemJ.org/bj/449/bj4490741add.htm). PVDF membranes, ECL (enhanced chemiluminescence) kits and [γ-32P]ATP were obtained from GE Healthcare. S-protein agarose beads and S-HRP [HRP (horseradish peroxidase)-conjugated S-protein] were obtained from Merck. LKBtide (LSNLYHQGKFLQTFCGSPLYRRR, corresponding to amino acids 196–215 of human NUAK2) [36] was obtained from Millipore. RRApTVA peptide was synthesized by Life Technologies. All other reagents were purchased from Wako Pure Chemical Industries.

Antibodies

Anti-phospho-AMPKα, anti-AMPKα, anti-phospho-JNK (c-Jun N-terminal kinase), anti-JNK, anti-phospho-p38 MAPK, anti-p38 MAPK, anti-PPM1A and anti-FLAG antibodies were obtained from Cell Signaling Technologies. Anti-HA (haemagglutinin) and anti-Myc antibodies were from Santa Cruz Biotechnology. The anti-PPM1B antibody has been described previously [15].

Cell culture, transfection and S-pull-down assays

HEK-293 and HeLa cells were grown in DMEM (Dulbecco's modified Eagle's medium, Invitrogen) supplemented with 10% (v/v) FBS (fetal bovine serum). Cells were transfected using Lipofectamine™ 2000 (plasmid) or Lipofectamine™ RNAiMax (siRNA) reagent. After a 48-h culture, cells were harvested, washed twice with PBS, and then lysed with ice-cold lysis buffer [50 mM Tris/HCl (pH 7.5), 1% (v/v) Triton X-100, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1 mM sodium orthovanadate, 50 mM sodium fluoride, 10 mM 2-glycerophosphate, 5 mM sodium pyrophosphate and Complete™ protease inhibitor cocktail (Roche)]. For the S-pull-down assay, cell lysates containing 200 μg of proteins were incubated for 0.5 h with S-protein agarose beads (5 μl). After washing the beads five times with lysis buffer, the bound proteins were subjected to SDS/PAGE (10% gel) and then transferred on to PVDF membranes. The membranes were incubated with primary antibody for 1 h at 25°C, followed by an HRP-conjugated secondary antibody for 1 h at 25°C. The membranes were developed using chemiluminescence, and then quantified using an image analyser (LAS4000 mini, GE Healthcare).

In vitro kinase assay

Phosphorylation was carried out in a 50 μl incubation mixture, containing 50 mM Hepes (pH 7.5), 10 mM MgCl2, 0.1% 2-mercaptoethanol, 0.1 mM EGTA, 10 nM microcystin-LR, 100 μM LKBtide and 100 μM [γ-32P]ATP. After 15 min at 30°C, the reaction was terminated by spotting aliquots on to P81 phosphocellulose paper, followed by immediate immersion in 75 mM phosphoric acid. The papers were washed, dried and analysed by Cerenkov counting. One unit of kinase activity was the amount of enzyme that phosphorylated 1 nmol of LKBtide within 1 min.

Metabolic labelling of protein with [3H]myristic acid

HeLa cells were plated in 35 mm dishes, transfected with the expression plasmids indicated using a modified calcium phosphate method [37], and then incubated for 48 h. Cells were washed twice with serum-free DMEM and incubated for 5 h at 37°C in DMEM with 2% (v/v) FBS containing 100 μCi/ml [3H]myristic acid. Cells were washed three times with PBS and lysed with ice-cold lysis buffer. After immunoprecipitation with an anti-FLAG antibody, the samples were subjected to SDS/PAGE and then fluorography with EN3HANCE® reagent (PerkinElmer).

Isolation of microsomal membranes from cultured cells

Cells were scraped into ice-cold PBS and resuspended in buffer B [10 mM Hepes (pH 7.5), 0.25 M sorbitol, 10 mM potassium acetate, 1.5 mM magnesium acetate and Complete™ protease inhibitor cocktail (Roche)]. The suspension was passed through a 23-gauge needle 25 times and centrifuged at 1000 g for 5 min. The resulting supernatant was centrifuged at 16000 g for 20 min. The pellet was resuspended in buffer C [50 mM Hepes (pH 7.5), 0.25 M sorbitol, 70 mM potassium acetate, 2.5 mM magnesium acetate, 5 mM EGTA and Complete™ protease inhibitor cocktail (Roche)], and then centrifuged again at 16000 g for 3 min. The resulting precipitates were resuspended in buffer C to obtain the microsomal fraction.

Generation of 293 Flp-In cell lines

To generate cell lines that stably express PPM1A–FLAG and PPM1B–FLAG in a doxycycline-inducible manner, the Flp-In T-REx system (Invitrogen) was used. cDNAs encoding PPM1A and PPM1B were inserted into the pcDNA5-FRT/TO vector and introduced into Flp-In 293 cells according to the manufacturer's instructions (Invitrogen).

Cell proliferation assay

To evaluate the cell proliferation rate, 293 Flp-In cell lines were seeded in 96-well plates at a concentration of 5×103 cells/well in 100 μl of DMEM. The next day, the medium was changed to DMEM containing 30 ng/ml doxycycline and the cells were incubated for an additional 24 and 48 h. At 1 h after adding 10 μl of WST-8 solution (Dojin) to the culture medium, the absorbance at 450 nm was measured using a microplate reader.

Purification of proteins expressed in mammalian cells

HEK-293 cells plated in 10 cm dishes were transiently transfected with plasmid constructs encoding PPM1A–FLAG [WT (wild-type) or G2A mutant], PPM1B (WT or G2A) or the AMPK complex (co-expression of FLAG–AMPKα1, AMPKβ1–HA and Myc–AMPKγ1) using a modified calcium phosphate method [37]. After transfection, the cells were cultured for 48 h and then harvested. Cells transfected with the AMPK complex were treated with 10 mM 2-deoxyglucose for 1 h before harvest to activate the AMPK complex. The cells were washed twice with PBS and lysed with ice-cold lysis buffer. Cell lysates were incubated for 2 h with anti-FLAG agarose beads (50 μl, Sigma). The beads were washed three times with lysis buffer and twice with HBS-SG [50 mM Hepes (pH 7.5), 150 mM NaCl, 0.25 M sucrose and 2% (v/v) glycerol] and then the bound proteins were eluted with HBS-SG containing 200 μg/ml 3×FLAG peptide (Sigma). All in vitro experiments were performed using proteins that were expressed in mammalian cells.

Protein phosphatase activity assay

Protein phosphatase activity was determined by measuring the release of Pi from RRApTVA phosphopeptide as a substrate. The standard assay mixture contained 50 mM Hepes (pH 7.5), 0.02% 2-mercaptoethanol, 100 μM RRApTVA phosphopeptide and 0.5 mM MnCl2. The enzyme reaction was carried out at 30°C for 45 min. The reaction was started by adding the enzyme and terminated by adding 100 μl of Biomol Green reagent (Enzo Life Sciences). The absorbance at 620 nm was measured. Pi release was calculated using a standard curve with phosphoric acid that was treated as described above. One unit of phosphatase activity was defined as the amount of enzyme that dephosphorylated 1 nmol of substrate in 1 min. For the kinetic parameter Kmetal, the metal ion (Mn2+) concentration was altered from 0.01 to 1 mM. Kmetal values were determined using the Hanes–Woolf linearization method. PNPP phosphatase activity was measured in a reaction mixture (50 μl) containing 50 mM Hepes (pH 7.5), 0.5 mM MnCl2, 0.02% 2-mercaptoethanol and 20 mM PNPP. The reaction was carried out at 30°C for 45 min and then terminated by adding 0.5 M NaOH, after which the absorbance at 405 nm was measured.

RESULTS

AMPKα is dephosphorylated by PPM1A and PPM1B in cells

To identify the PPM family members that are responsible for dephosphorylating the α isoform of AMPK (α1 and α2) in HeLa cells, we examined the effects of overexpressing PPM members on LKB1-dependent phosphorylation of AMPKα at Thr172. Expressing S-HA–LKB1 alone in HeLa cells, which do not express LKB1 due to abnormal methylation of the promoter region, induced phosphorylation of endogenous AMPKα. However, we could not distinguish the phosphorylation level of AMPKα1 from that of AMPKα2 because a phospho-specific antibody against phospho-Thr172 of AMPKα detected both isoforms (results not shown). Therefore we transfected HeLa cells with expression plasmids for two different AMPK complexes (AMPKβ1–HA, Myc–AMPKγ1, and either S-HA–AMPKα1 or α2) and S-HA–LKB1 and then recovered S-HA–AMPKα from the cell lysates to detect the phosphorylation of AMPKα1 and α2 separately. Under these conditions, both S-HA–AMPKα1 and α2 were phosphorylated in an LKB1-dependent manner, and co-expressing PPM1A and PPM1B caused a large decrease in the phosphorylation of both isoforms with no effect on LKB1 activity (Figures 1a–1c). Furthermore, siRNAs targeting PPM1A or PPM1B increased AMPKα phosphorylation in HeLa cells (Figure 1d). These results indicate that both PPM1A and PPM1B are responsible for the dephosphorylation of AMPKα1 and α2 in cells. In contrast, PPM1E overexpression decreased phosphorylation of AMPKα2, but not AMPKα1 (Figures 1a and 1b).

Figure 1 Identification of protein phosphatases acting on AMPKα at Thr172 in cells

(a) HeLa cells were co-transfected with expression plasmids encoding S-HA–AMPKα1, AMPKβ1–HA, Myc–AMPKγ1, S-HA–LKB1 (except lane 1) and PPM1A–FLAG, PPM1B–FLAG, FLAG–PPM1E, FLAG–PPM1F, PPM1G–FLAG, PPM1H–FLAG, PPM1L–FLAG, PPM1M–FLAG, ILKAP–FLAG, PPM1K–FLAG or PPM1D–FLAG. S-HA–AMPKα1 was recovered at 48 h post-transfection using S-protein agarose beads and then immunoblotted with a phospho-specific antibody for AMPKα Thr172 (pT172) or an anti-HA antibody (total protein). The quantification of phospho-Thr172 signal relative to the total protein level is shown as the mean±S.D. for three independent experiments. Whole-cell lysates were also analysed using an anti-FLAG antibody to determine the expression level of the PPM1 family. (b) Same as (a), except that S-HA–AMPKα2 was used instead of S-HA–AMPKα1. (c) HeLa cells were transfected with the plasmids indicated and S-HA–LKB1 was recovered from cell lysates (200 μg of protein) using S-protein agarose beads. An in vitro kinase assay was performed using LKBtide as the substrate. (d) HeLa cells were transfected with siRNA targeted against PPM1A (α1 and α2) or PPM1B (β1 and β2), or a control siRNA for 48 h. The cells were harvested and the cell lysates were subjected to immunoblotting using the antibodies indicated. Thr172 phosphorylation relative to the total AMPKα levels is shown as the mean±S.E.M. for three independent experiments. cont., control.

PPM1A and PPM1B are N-myristoylated

The β-subunit of AMPK is N-myristoylated and this modification was reported to be involved in regulating the key steps of AMPK activation [7,29]. Although the N-termini of PPM1A and PPM1B do not completely match the consensus sequence for N-myristoyl transferase (MGXXXS/T, where X is any amino acid residue), a portion of the N-terminal sequence was same as that of c-Src, a well-known N-myristoylated protein (Figure 2a) [38]. These facts motivated us to examine whether PPM1A and PPM1B are also N-myristoylated. HeLa cells expressing PPM1A–FLAG or PPM1B–FLAG were metabolically labelled with [3H]myristic acid and the incorporation of radioactivity into these proteins was analysed. As shown in Figure 2(b), [3H]myristic acid was incorporated into WT but not the G2A mutant, indicating that these proteins were N-myristoylated.

Figure 2 PPM1A and PPM1B are N-myristoylated

(a) Sequence alignment of the N-terminus of human AMPKβ, c-Src, PPM1A and PPM1B. (b) HeLa cells were transfected with expression plasmids encoding PPM1A–FLAG or PPM1B–FLAG using a modified calcium phosphate method and incubated for 48 h. The cells were washed twice with serum-free DMEM and incubated for 5 h at 37°C in DMEM with 2% (v/v) FBS containing 100 μCi/ml [3H]myristic acid. Labelled proteins were recovered by immunoprecipitation and the samples were subjected to SDS/PAGE, followed by fluorography. The bottom panel shows a gel stained with Coomassie Brilliant Blue (CBB). IgG, immunoglobulin used for immunoprecipitation. (c) HEK-293 cells were transiently transfected with the plasmids indicated using a modified calcium phosphate method and incubated for 48 h. The cells were resuspended in buffer B, passed through a 23-gauge needle 25 times, and centrifuged at 1000 g for 5 min at 4°C. The resulting supernatant was centrifuged at 16000 g for 20 min at 4°C to obtain the cytosolic (supernatant) and membranous (pellet) fractions. The fractions were immunoblotted with an anti-FLAG antibody. PPM1L (membrane) and ΔN-PPM1L (cytosol) were used as controls.

Since N-myristoylation has been shown to facilitate protein–membrane interactions [3], we examined whether the G2A mutation affected the subcellular localization of PPM1A and PPM1B. When extracts of HEK-293 cells expressing these proteins were separated into cytosolic and membrane fractions, WT PPM1A and PPM1B were predominantly recovered in the cytosolic fraction. However, a small, but significant, amount of these proteins were recovered in the membrane fraction (Figure 2c). Distribution to the membrane fraction markedly decreased in cells expressing the G2A mutants (Figure 2c). These results suggest that PPM1A and PPM1B weakly associate at the membrane and N-myristoylation mediates the membrane localization of these isoforms.

PPM1A and PPM1B must be N-myristoylated to dephosphorylate their physiological substrates in cells

To investigate the possible role of N-myristoylation in the activities of PPM1A and PPM1B, we examined whether the G2A mutants could dephosphorylate the α-subunit of AMPK in cells. In contrast with ectopically expressed WT PPM1A and PPM1B, which caused a large decrease in AMPKα phosphorylation, the G2A mutants were unable to reduce AMPKα phosphorylation. Essentially there was no significant alteration in the expression levels of AMPKα/β/γ, and LKB1 was observed upon expression of either WT or the mutant PPM1A/PPM1B in cells (Figure 3a).

Figure 3 The G2A mutants of PPM1A and PPM1B are unable to dephosphorylate their physiological substrates in cells

(a) HeLa cells were transiently transfected with the indicated plasmids and incubated for 48 h. S-HA–AMPKα1 was recovered from cell lysates using S-protein agarose beads. The samples were subjected to immunoblotting with anti-phospho-AMPKα (pT172) and anti-HA antibodies. Whole-cell lysates (WCL) were also analysed to determine the cellular expression level of each protein. EV, empty vector. (b) 293 Flp-In cell lines expressing the indicated constructs were seeded into 12-well plates and incubated for 24 h (EV, empty vector). The medium was replaced with serum-free DMEM containing doxycycline (30 ng/ml) and then incubated for an additional 24 h. After treating the cells with TNFα (20 ng/ml) for 10 min, cell lysates were prepared and subjected to immunoblotting using the antibodies indicated. (c) 293 Flp-In cell lines expressing the indicated constructs were seeded into 96-well plates and incubated for 24 h (EV, empty vector). The medium was replaced with DMEM containing doxycycline (30 ng/ml) and then the cells were incubated for 24 h or 48 h. After adding WST-8 solution to the culture medium, the absorbance at 450 nm was measured. The results are expressed as the ratio of the absorbance at 24 h and 48 h to that obtained at 0 h. The results are shown as the means±S.D. for three independent experiments.

We then determined whether N-myristoylation is required for the general function of these two isoforms. Since PPM1A and PPM1B reportedly regulate the SAPK cascade [17,21,22], we examined the effects of overexpressing PPM1A and PPM1B with or without the G2A mutation on TNFα (tumour necrosis factor α)-dependent phosphorylation of p38 MAPK and JNK, the most downstream components of the SAPK cascade. While ectopic expression of both WT PPM1A and PPM1B in cells caused a large decrease in TNFα-enhanced phosphorylation of both JNK and p38 MAPK, the G2A mutants had little effect (Figure 3b).

Ofek et al. [39] reported that PPM1A inhibited the proliferation of HEK-293 cells through activation of the p53 protein [39]. Therefore we tested the effects of expressing PPM1A or its G2A mutants on HEK-293 cell proliferation. As shown in Figure 3(c), expressing WT PPM1A inhibited cell growth at 48 h after the addition of doxycycline. In contrast, the proliferation rate of cells expressing the G2A mutant was similar to control cells.

Taken together, these results suggest that N-myristoylation is indispensable for the ability of PPM1A and PPM1B to dephosphorylate not only AMPK but also the other endogenous substrates involved in the stress response and cell growth regulation.

N-Myristoylation is required for PPM1A and PPM1B to recognize their physiological substrates in cells

Since mutating the N-myristoylation site in PPM1A and PPM1B affected the membrane association of these proteins, we suspected that proper intracellular localization mediated by N-myristoylation was required for PPM1A and PPM1B to dephosphorylate their physiological substrates. It was previously reported that membrane association of the AMPK complex mediated by N-myristoylation of the β-subunit is required for the metabolic stress-sensing function of the AMPK complex [7]. Therefore we asked whether co-localization of the AMPK complex and PPM1A or PPM1B at the membrane was required for these phosphatases to dephosphorylate AMPKα. To address this question, we examined whether PPM1A or PPM1B could dephosphorylate AMPKα that forms a complex with a non-myristoylated mutant of AMPKβ (G2A). LKB1 and AMPKβ-S (WT or G2A) were co-expressed with PPM1A or PPM1B in HeLa cells and phosphorylation of endogenous AMPKα complexed with AMPKβ-S (WT or G2A) was analysed after the AMPK complex was recovered by S-tag-mediated pull down of AMPKβ-S. PPM1A and PPM1B similarly dephosphorylated AMPKα that was complexed with either WT or the G2A version of AMPKβ (Figure 4a). These results indicated that both PPM1A and PPM1B could dephosphorylate AMPKα, irrespective of its subcellular localization, suggesting that PPM1A and PPM1B do not have to co-localize with the AMPK complex at the membrane to dephosphorylate AMPKα. This assumption is supported by evidence that PPM1E, which has a disrupted N-myristoylation site, is also able to dephosphorylate AMPKα in cells.

Figure 4 N-Myristoylation is an essential determinant of substrate specificity for PPM1A and PPM1B

(a) HeLa cells were transiently transfected with S-HA–LKB1, AMPKβ1-S (WT or G2A) and increasing amounts of PPM1A–FLAG or PPM1B–FLAG and then incubated for 48 h. The AMPK complex was recovered by S-tag-mediated pull down of AMPKβ1-S, and phosphorylation of endogenous AMPKα in the complex was analysed using an anti-phospho-Thr172 antibody. The membrane was reprobed with anti-AMPKα and S-HRP to confirm the amount of AMPKα, S-HA–LKB1 and AMPKβ1-S. Whole-cell lysates (WCL) were also subjected to immunoblotting with an anti-FLAG antibody. In the absence of PPM1A or PPM1B expression, AMPKα complexed with the mutant β-subunit was more highly phosphorylated than AMPKα complexed with the WT β-subunit, as was previously reported [29]. The efficiency of AMPKα dephosphorylation was quantified and is shown on the right-hand side. (b) A phosphatase assay was performed in a reaction mixture (50 μl) containing 50 mM Hepes (pH 7.5), 0.5 mM MnCl2, 0.02% 2-mercaptoethanol, 100 μM RRApTVA peptide and 8 μg/ml protein phosphatase. The enzyme reaction was carried out at 30°C for 45 min and then terminated by adding Biomol Green Reagent, after which the absorbance at 650 nm was measured using a plate reader. The results are shown as the means±S.E.M. for three independent experiments. (c) Phosphatase activity against RRApTVA phosphopeptide was measured with increasing concentrations of MnCl2. The results are expressed as a ratio of the activity to the highest activity. (d) The AMPK complex was incubated with PPM1A (WT), PPM1A (G2A), PPM1B (WT) or PPM1B (G2A) in a reaction mixture (50 μl) containing 50 mM Hepes (pH 7.5), 0.5 mM MnCl2 and 0.02% 2-mercaptoethanol at 30°C for 20 min. The amount of protein phosphatases added to the reaction was equalized for their ability to dephosphorylate RRApTVA peptide (0.15, 0.3, 0.45 and 0.6 units/ml). At the end of the reaction, the samples were analysed by immunoblotting using anti-pAMPKα (phospho-Thr172) and anti-AMPKα antibodies. (e) A phosphatase assay was performed in a reaction mixture (50 μl) containing 50 mM Hepes (pH 7.5), 0.5 mM MnCl2, 0.02% 2-mercaptoethanol, 20 mM PNPP and 8 μg/ml protein phosphatase. The reaction was carried out at 30°C for 45 min and then terminated by adding 0.5 M NaOH, after which the absorbance at 405 nm was measured using a plate reader. The results are shown as the means±S.E.M. for three independent experiments. (f) Two orthogonal views of the molecular-surface representation of PPM1A. Gly2 is indicated in red. Amino acid residues composing the active site (Asp38, Asp60, Thr128, Asp146, Asp199, Asp239 and Asp282) are coloured in yellow, and hydrophobic residues (leucine, valine and isoleucine) are shown in cyan. The catalytic metals in the catalytic pocket are difficult to see in this projection. The structure is displayed using PyMOL (http://www.pymol.org). The PDB code for PPM1A is 3FXJ.

To test the possibility that N-myristoylation of PPM1A and PPM1B directly affects their activities, we expressed WT and the G2A mutant of PPM1A–FLAG and PPM1B–FLAG in HEK-293 cells, immunoprecipitated these proteins with an anti-FLAG antibody, and then measured the enzymatic activities of these proteins towards various substrates in vitro. When the RRApTVA phosphopeptide, which is widely used for in vitro phosphatase assays, was used as the substrate, the G2A mutants of PPM1A and PPM1B showed a slight and 50% increase in specific activity compared with the WT enzymes respectively (Figure 4b). Furthermore, there was no apparent difference in Mn2+-dependency [Figure 4c, Kmetal values for PPM1A WT, PPM1A (G2A), PPM1B WT and PPM1B (G2A) were 0.011, 0.019, 0.015 and 0.011 mM respectively]. These results suggest that the non-myristoylated forms of PPM1A and PPM1B have catalytic activities towards the phosphopeptide that are similar to the WT enzymes. We then analysed the dephosphorylation activity towards an AMPK complex that had been activated with 2-deoxyglucose and then affinity-purified from HEK-293 cells. Both WT PPM1A and PPM1B efficiently dephosphorylated AMPKα, whereas the G2A mutants had much lower activities towards AMPKα, although the amounts of the phosphatases in the reaction mixture were adjusted to obtain equal activities towards the RRApTVA peptide (Figure 4d). Interestingly, when an artificial substrate, PNPP, was used, both PPM1A (G2A) and PPM1B (G2A) had much higher specific activities compared with the WT enzymes (Figure 4e). These results indicated that PPM1A and PPM1B showed lower enzymatic activity towards the physiological substrate (AMPK), but had higher specific activity against the artificial substrate (PNPP) in the absence of the N-myristoyl group. Taken together, these results suggest that the N-myristoyl group is essential for the recognition of physiological substrates by PPM1A and PPM1B in cells.

DISCUSSION

In the present study we have provided evidence that N-myristoylation is required for PPM1A and PPM1B to function in cells. Importantly, N-myristoylation is essential for the recognition of physiological substrates by PPM1A and PPM1B in cells, but dispensable for their catalytic activity itself. This conclusion is based on the following observations. First, a non-myristoylated mutant (G2A) of PPM1A and PPM1B could not dephosphorylate endogenous substrates, including AMPKα, p53 and components of the SAPK cascade, in cells (Figure 3). Secondly, in in vitro experiments, the G2A mutants exhibited reduced activities towards AMPKα, whereas they exhibited higher specific activity against an artificial substrate (PNPP) than the WT protein (Figures 4d and 4e). Finally, the G2A mutation did not affect Mn2+-dependency of PPM1A and PPM1B (Figure 4c). There are two possible mechanisms by which N-myristoylation is involved in substrate recognition: (i) the myristoyl group may increase the affinity of the enzyme for its substrate by interacting directly with the substrate protein and (ii) the myristoyl group may interact with the intramolecule-binding pocket and induce conformational changes in the enzyme that affect its affinity for the substrate protein. Considering our observation that G2A mutants exhibited higher specific activity against an artificial substrate (PNPP) than the WT protein, it is unlikely that myristoylation directly associates with the substrate. Rather, the myristoyl group may preserve the conformation of the catalytic pocket by interacting with hydrophobic residues near the active site, which allows only the physiological substrate proteins to access the catalytic site. In the absence of the modification, the structure around the active site may be destabilized, so that the artificial substrates with low molecular mass readily approach the catalytic site. This conclusion may be reinforced by the fact that Gly2 is located at the entrance of the catalytic pocket (Figure 4f) [40]. It remains uncertain whether N-myristoylation is involved in regulation of PPM1A and PPM1B function in cells, but displacement of the myristoyl moiety from the hydrophobic pocket by an as yet unidentified mechanism might lead to change of substrate recognition of these isoforms.

In the present study we also provide evidence that PPM1A and PPM1B redundantly dephosphorylate the α-subunit of AMPK. Although these isoforms are closely related (75% identity), they were shown to be distinctly localized in cells [15,19,41], raising the possibility that they might act on AMPK heterotrimeric complexes at different subcellular locations within cells. Recently, Voss et al. [33] reported that PPM1E is a protein phosphatase that acts on AMPK. In our assay, PPM1E decreased phosphorylation only of AMPKα2 and had little effect on AMPKα1 (Figures 1a and 1b). Considering the observation by Voss et al. [33] that endogenous PPM1E specifically interacted with AMPKα2, PPM1E may act as an AMPKα2-containing-complex-specific phosphatase. In the same paper [33], the authors claimed that PPM1A was not involved in the dephosphorylation of AMPK on the basis of the finding that PPM1A-specific shRNA (short hairpin RNA) did not increase AMPKα phosphorylation. Although the reason for the discrepancy between their observations and the findings of the present study is not known, there might be differences in relative PPM1A and PPM1E expression in the cell lines used in the knockdown studies, resulting in different responses with the PPM1A-specific siRNA.

PPM1A and PPM1B expressed in Escherichia coli have been generally used by researchers for in vitro experiments. However, it should be noted that bacteria do not have an N-myristoyltransferase gene [3] and therefore are not capable of modifying PPM1A and PPM1B. Therefore using recombinant PPM1A and PPM1B produced in mammalian or other eukaryotic cells is highly recommended for in vitro experiments.

During the preparation of the present paper, Tsugama et al. [42] reported that WT PP2C74, a PPM family protein phosphatase in Arabidopsis, but not its mutant with an N-terminal glycine-to-alanine substitution, localizes to the plasma membrane in Arabidopsis and interacts with SnRK1 (SNF1-related kinase 1), the plant orthologue of mammalian AMPK [42]. Using in silico analysis, they found that the N-terminal glycine residue of PP2C74 could be a consensus site for myristoylation and proposed that its myristoylation is required for the plasma membrane localization of SnRK1 and its association with SnRK1. Although they did not demonstrate that PP2C74 is myristoylated or show that its activity is altered upon myristoylation in cells, these observations raise the possibility that myristoylation is a conserved mechanism to functionally regulate PPM family members in a variety of organisms.

AUTHOR CONTRIBUTION

Takayasu Kobayashi and Shinri Tamura designed the overall experiments and wrote the paper. Toko Chida executed the major parts of this work. Masakatsu Ando, Tasuku Matsuki and Yutaro Masu helped with the initial experiments. Yuko Nagaura and Teruko Takano-Yamamoto contributed to a comprehensive interpretation of the results and data analyses.

FUNDING

This work was supported, in part, by a grant-in-aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Abbreviations: AMPK, AMP-activated protein kinase; BMP, bone morphogenetic protein; DMEM, Dulbecco’s modified Eagle’s medium; FBS, fetal bovine serum; HA, haemagglutinin; HEK, human embyonic kidney; HRP, horseradish peroxidase; ILKAP, ILK1 (integrin-linked kinase 1)-associated protein; JNK, c-Jun N-terminal kinase; MAPK, mitogen-activated protein kinase; MKK, MAPK kinase; NF-κB, nuclear factor κB; PKA, cAMP-dependent protein kinase/protein kinase A; PNPP, p-nitrophenyl phosphate; PP, protein phosphatase; PPM, metal-dependent protein phosphatase; SAPK, stress-activated protein kinase; S-HRP, HRP-conjugated S-protein; siRNA, small interfering RNA; SnRK1, SNF1-related kinase 1; TAK1, TGF (transforming growth factor)-β-activated kinase 1; TNFα, tumour necrosis factor α; WT, wild-type

References

View Abstract