Cellular Signalling 23 (2011) 114–124

Contents lists available at ScienceDirect

Cellular Signalling

j ourna l homepage: www.e lsev ie r.com/ locate /ce l l s ig

Ppm1E is an in cellulo AMP-activated protein kinase phosphatase

Martin Voss 1, James Paterson 1, Ian R. Kelsall 2, Cristina Martín-Granados 3, C. James Hastie,Mark W. Peggie, Patricia T.W. Cohen ⁎Medical Research Council Protein Phosphorylation Unit, College of Life Sciences, University of Dundee, Dundee DD1 5EH, Scotland, UK

Abbreviations: AMPK, 5′-AMP-activated protein kindependent protein kinase; CaMKK, Ca2+/calmodulin-deLKB1, protein kinase mutated in Peutz–Jeghers synphosphatase ce:sup>/ce:sup>/Mn2+-dependent; CaMdent protein kinase phosphatase; CaMKP-N, Ca2+-ckinase phosphatase in the nucleus; Ilkap, Integrin linktase; Nerpp-2C, neurite extension related protein phospyruvate dehydrogenase phosphatase catalytic subunit.⁎ Corresponding author. MRC Protein Phosphorylatio

Sir James Black Centre, University of Dundee, Dow StreUK. Tel.: +44 1382 384240; fax: +44 1382 223778.

E-mail address: p.t.w.cohen@dundee.ac.uk (P.T.W. C1 These authors have contributed the major experime2 Present address: Scottish Institute for Cell Signal

University of Dundee, DD1 5EH, UK.3 Present address: College of Life Sciences and Medici

University of Aberdeen, Aberdeen, AB25 2ZD, UK.

0898-6568/$ – see front matter © 2010 Elsevier Inc. Aldoi:10.1016/j.cellsig.2010.08.010

a b s t r a c t

a r t i c l e i n f o

Article history:Received 2 July 2010Accepted 17 August 2010Available online 27 August 2010

Keywords:AMPKProtein serine/threonine phosphatasePpm1EPpm1FType 2 diabetesEnergy metabolism

Activation of 5′-AMP-activated protein kinase (AMPK) is believed to be the mechanism by which thepharmaceuticals, metformin and phenformin, exert their beneficial effects for treatment of type 2 diabetes.These biguanide drugs elevate 5′-AMP, which allosterically activates AMPK and promotes phosphorylation onThr172 of AMPK catalytic α subunits. Although kinases phosphorylating this site have been identified,phosphatases that dephosphorylate it are unknown. The aim of this study is to identify protein phosphatase(s)that dephosphorylate AMPKα-Thr172 within cells. Our initial data indicated that members of the proteinphosphatase ce:sup>/ce:sup>/Mn2+-dependent (PPM) family and not those of the PPP family of proteinserine/threonine phosphatases may be directly or indirectly inhibited by phenformin. Using antibodies raised toindividual Ppm phosphatases that facilitated the assessment of their activities, phenformin stimulation of cellswas found to decrease the ce:sup>/ce:sup>/Mn2+-dependent protein serine/threonine phosphatase activity ofPpm1E and Ppm1F, but not that attributable to other PPM family members, including Ppm1A/PP2Cα. Depletionof Ppm1E, but not Ppm1A, using lentiviral-mediated stable gene silencing, increased AMPKα-Thr172phosphorylation approximately three fold in HEK293 cells. In addition, incubation of cells with lowconcentrations of phenformin and depletion of Ppm1E increased AMPK phosphorylation synergistically.Ppm1E and the closely related Ppm1F interact weakly with AMPK and assays with lysates of cells stablydepleted of Ppm1F suggests that this phosphatase contributes to dephosphorylation of AMPK. The data indicatethat Ppm1E and probably PpM1F are in cellulo AMPK phosphatases and that Ppm1E is a potential anti-diabeticdrug target.

ase; CaMK, Ca2+/calmodulin-pendent protein kinase kinase;drome; Ppm (PP2C), proteinKP, Ca2+-calmodulin-depen-almodulin-dependent proteined kinase associated phospha-phatase, related to PP2C; Pdp,

n Unit, College of Life Sciences,et, Dundee DD1 5EH, Scotland,

ohen).ntal data in the article.ling, College of Life Sciences,

ne (IMS Building), Foresterhill,

l rights reserved.

© 2010 Elsevier Inc. All rights reserved.

1. Introduction

Type 2 diabetes is characterised by elevated blood plasma glucoselevels that are at least partially resistant to lowering by endogenousinsulin. The major pharmaceutical used to treat type 2 diabetes ismetformin, a biguanide, which was shown to activate AMP-activatedprotein kinase (AMPK) in cells and in vivo [1] and is believed to exert

its effect by inhibiting mitochondrial complex I, leading to theelevation of the cellular AMP/ATP ratio [2,3]. The increase in AMPactivates AMPK, which is a major sensor of cellular and whole bodyenergy that switches on catabolic processes, such as glucose uptakeinto muscle and fatty acid oxidation, and turns off anabolic enzymesystems, such as gluconeogenesis and fatty acid synthesis [4–6].AMPK exerts its action on metabolism acutely by phosphorylation ofkey metabolic enzymes and in the longer term by the regulation ofgene expression.

AMPK is a heterotrimer that comprises a catalytic α subunit, aregulatory glycogen binding β subunit and a γ subunit that binds twomolecules of AMP [7]. In mammals, each subunit is encoded by two orthree genes (α1, α2; β1, β2; γ1, γ2, γ3), with splice variants and theuse of alternative promoters further increasing the complexity.Mammalian AMPK is activated by the binding of AMP to the γ subunitand increased phosphorylation of Thr172 in the catalytic loop of thekinase domain of the α-subunit isozymes. The tumour suppressorkinase, LKB1 [8–10], and calmodulin-dependent protein kinasekinase-beta (CaMKKβ) [11,12] phosphorylate Thr172 of theAMPKα-subunit isozymes and one study provides evidence thatboth CaMKKα and CaMKKβ are AMPKα-Thr172 kinases [13]. Thephosphorylation is reversible, but the specific protein phosphatasesthat are responsible for the in vivo dephosphorylation are unknown,

115M. Voss et al. / Cellular Signalling 23 (2011) 114–124

although studies with hepatocytes and in vitro suggest that an okadaicacid insensitive protein phosphatase dephosphorylates AMPK [14,15].

Activation of AMPK is achieved by the binding of AMP to AMPK,which elicits the allosteric changes that activate AMPK and decreasethe action of protein phosphatases on AMPK, permitting increasedphosphorylation of Thr172 by LKB1 and/or CaMKKβ [16,17]. Theincreased phosphorylation, coupled to the activation of AMPK, isbelieved to be a major pharmacological effect accounting for theefficacy of the anti-diabetic drugs metformin and phenformin.Inhibiting the protein phosphates(s) that dephosphorylate AMPKshould also increase the phosphorylation and activation of AMPK, andhere we show Ppm1E as an in cellulo AMPKα-Thr172 phosphatase.

2. Materials and methods

2.1. Assay of protein phosphatase activity

Cell lysates were prepared in 30 mM Tris–HCl pH 7.5, 150 mMNaCl, 0.1 mM EGTA, 5% (v/v) glycerol, 0.1% 2-mercaptoethanol,‘Complete’ protease inhibitor (Roche Diagnostics, Mannheim,Germany) and 1% 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS). Protein phosphatase activities in lysatesor in immunopellets of specific protein phosphatases were assayed bythe release of [32P]orthophosphate from 10 μM [32P]-labelled phos-phorylase a [18], 2 μM [32P]-labelled casein or 1 μM [32P]-labelledGST-AMPK α1 (see Supplementary file) in the presence or absence ofprotein phosphatase inhibitors or activators. Bacterially expressedprotein phosphatase activity was assayed in 50 mM Tris–HCl pH 7.5,0.1 mMEGTA, 0.1% (v/v) 2-mercaptoethanol, 0.01% (w/v) Brij-35with1 mM synthetic phospho-peptide RRAT(p)VA as substrate at 37 °C for10 min in the presence or absence of Mg2+/Mn2+, with the release ofphosphate being measured by Biomol Green (Biomol ResearchLaboratories, Plymouth Meeting, PA, USA).

2.2. Immunological analyses

Antibodies were raised to unique peptide sequences of each Ppmanalysed (see Supplementary file) and affinity purified. For assay ofphosphatase activity, antibodies were non-covalently coupled toprotein G sepharose. Antibody binding to blots was analysed using theOdyssey infrared imaging system (Li-Cor Biosciences, Cambridge, UK).

2.3. Lentivirus-delivered stable gene silencing

Specific depletion of Ppm1A, Ppm1E and Ppm1F in HEK293 cellswas achieved using lentivirus-delivered shRNAi according to themanufacturer's protocol (Sigma-Aldrich, Dorset, UK). Control linesexpressed the puromycin resistance marker alone (P) or togetherwith a scrambled shRNA sequence (non-target, NT). Selection for twodays in media containing puromycin ensured that the lentiviralsequences (expressing the puromycin resistance marker and theshRNAi) were present and active in all cells in the culture [19]. Thecells were then allowed to recover for one day in non-selective mediabefore examination or further treatments over a period of two weeks.

2.4. Analysis of AMPK phosphorylation

Following treatment of cells with lentivirus, cells were lysed in50 mMTris–HCl pH 7.5, 1 mM EGTA, 4 mMEDTA, 270 mM sucrose, 1%(v/v) Triton X-100, 1 mM sodium orthovanadate, 10 mM sodium-β-glycerophosphate, 50 mM sodium fluoride, 5 mM sodium pyrophos-phate, 0.1% (v/v) 2-mercaptoethanol, ‘Complete’ protease inhibitorand 1 μM microcystin-LR. The protein concentrations of the lysateswere adjusted to 2 mg/ml and 20 μg lysate protein was analysed bySDS-PAGE and immunoblotting using anti-AMPK-pThr172 antibodiesand fluorescent dye-coupled secondary antibodies, followed by

quantification on the Odyssey system. Blots were also probed withanti-AMPK α1 or anti-AMPK α2 and/or anti-GAPDH to control forequal loading of the samples.

For assay of the phosphorylation of AMPK-pThr172 in HEK293 celllysates, cells were lysed in 50 mM Tris–HCl pH 7.5, 100 mM KCl,10 mM NaCl, 1 mM EGTA, 4 mM EDTA, 270 mM sucrose, 1% (v/v)Triton X-100, 0.1% (v/v) 2-mercaptoethanol, and ‘EDTA free’ proteaseinhibitor. Additionally, 5 mM N-ethylmaleimide (NEM) was includedin the lysis buffer where stated and 2-mercaptoethanol was omittedwhen NEM was added. Following additions to the lysates thatproduced an increase in both the ATP/AMP ratio and the free Mg2+

concentration followed by a 30 min incubation at 37 °C, the lysateswere examined by immunoblotting for AMPK-Thr172 phosphoryla-tion, which was quantified as described above.

2.5. Expression constructs

Human Ppm1E cDNA (NM_014906.3, encoding 755 amino acids)and human Ppm1F cDNA (NM_014634, encoding 454 amino acids)were subcloned into pGEX-6 and the bacterially expressed proteinswere affinity purified with cleavage of glutathione-S-transferase tag(GE-Healthcare, Little Chalfont, UK). Ppm1A/PP2Cαwas also expressedin E. coli without an epitope tag and purified as described previously[15]. Ppm1E and Ppm1Fwith a FLAG epitope tag at theN- or C-terminuswere subcloned into pCMV5 for expression into mammalian cells.

2.6. Statistical analysis

Error bars show the standard error of the mean (SEM). Statisticalanalyses were performed using Student's t test.

3. Results

3.1. Effects of phenformin on protein serine/threonine phosphataseactivities in human cells

Protein serine/threonine phosphatase activities were examined inHEK293 cell lysates after stimulation of the cells with phenformin, ananalogue of metformin, that promotes phosphorylation of AMPK(Fig. 1A). Assessment of the okadaic acid sensitive PPP activitiesattributable to PP1 complexes (Fig. 1B) and PP2A-like complexes(which although predominantly PP2A/Ppp2 may also include Ppp4,Ppp5 and Ppp6) showed no differences between activities inuntreated and phenformin treated cells (Fig. 1C). In contrast assaysof the Mn2+/Mn2+-dependent, okadaic acid insensitive, phosphataseactivities, attributable to the PPM-like phosphatases, which includePpm1A/PP2Cα [20], showed a statistically significant decrease in theactivities in phenformin treated HEK293 cells compared with those inuntreated cells (Fig. 1D). A similar (~20%) statistically significantdecrease in PPM-like phosphatase activities was observed in HeLacells stimulated with phenformin (Fig. 1E) and in HEK293 cellsstimulated by metformin (Fig. 1F). The results suggest that phenfor-min and metformin, either directly or indirectly cause inhibition ofone or more PPM-like phosphatases. The inhibition is not due to aneffect on LKB1, the predominant kinase phosphorylating AMPKα2 inresponse to phenformin [21], because, while it is expressed atappreciable levels in HEK293 cells, LKB1 is very low or absent inHeLa cells [8,9,22].

3.2. Analyses of the effects of phenformin on individual members in thePPM family of protein phosphatases in cell cultures and in vitro

In order to examine the effects of phenformin on individualphosphatases in the PPM family, antibodies were raised to the first 11known members of the PPM family [23], which now comprises 17phosphatases [24,25]. Immunoblots, probed with these antibodies,

116 M. Voss et al. / Cellular Signalling 23 (2011) 114–124

identified bands of the expected molecular sizes but showed nodifference in the level of Ppm1A, Ppm1B1, Ppm1B2, Ppm1D, Ppm1F,Ppm1G and Ilkap in HEK293 cell lysates in the presence and absenceof phenformin (Fig. 2A). In order to examine the activities of thesephosphatases, the specific antibodies non-covalently coupled toprotein G sepharose, were used to individually immunopellet eachphosphatase from HEK293 cell lysates untreated or treated withphenformin. The phosphatase activity, assayed using phosphorylatedcasein as the substrate and associated with each Ppm enzyme, isshown in Fig. 2B. The activities of Ppm1A were very similar inimmunopellets obtained using two different anti-Ppm1A antibodiesand no decrease occurred following phenformin treatment. Ppm1B2,

Fig. 1. Analysis of protein serine/threonine phosphatase activities in cells either untreated,treated with 10 mM phenformin for 1 h or 2 mM metformin for 3 h. (A) Levels ofphosphorylation of AMPKat Thr172 in four untreated and four treatedHEK293 cell lysates.Ppp5, a phosphatase in the PPP family inhibited by okadaic acid, is used as a control forsample loading. Molecular masses of the proteins are indicated in kDa to the left of thepanel. (B) Phosphatase activities of PP1 complexes in lysates of untreated and phenformintreated HEK293 cells using phosphorylase a as substrate in the presence of 4 nM okadaicacid. (C) Phosphatase activities of PP2A-like complexes in lysates of untreated andphenformin treated HEK293 cells using phosphorylase a as substrate in the presence of200 nM inhibitor-2. (D and E) Mg2+-dependent, okadaic acid insensitive, PPM-likephosphatase activities in lysates of untreated and phenformin treated HEK293 (D) andHeLa (E) cells using phosphorylated casein as substrate in the presence of 10 mMmagnesium acetate and 5 μM okadaic acid. Data are the mean±SEM of four independentsamplesmeasured in triplicate (B and C),five independent samplesmeasured in duplicate(D), and four independent samplesmeasured induplicate (E). Themeanactivitydifferencebetween untreated and phenformin treated cells (*) is statistically significant (pb0.001).(F) Mg2+-dependent, okadaic acid insensitive, PPM-like phosphatase activities in HEK293cell lysatesuntreated and treatedwith2 mMmetformin.Data are themean±SEMof threeindependent samples. The mean activity difference between untreated and metformintreated cells (*) is statistically significant (pb0.001).

Ppm1G and Pdp1 showed appreciable protein phosphatase activitythat was similar in cells with and without incubation withphenformin, while others exhibited lower activities and no activitywas found associated with Ppm1D and Pdp2 immunopellets.Although Ppm1E could not be detected on immunoblots of celllysates with the initial antibodies raised to Ppm1E peptides, Ppm1Eactivity could be detected in immunopellets of these antibodies(Fig. 2B). Interestingly, this initial data suggested that the activity ofPpm1E present in untreated cells was almost completely ablated incells treated with phenformin and the activity of Ppm1F inphenformin treated cells was decreased to ~60% of the level inuntreated cells.

Ppm1E and Ppm1F are Mg2+/Mn2+-dependent protein phospha-tases that share 60% identity in their phosphatase domains and aremoreclosely related to each other than they are to the other members of thePPM family [23]. Unfortunately, several more batches of partiallyhydrolysed casein substrate gave high background phosphatase activitywith the consequence that measurements of low PPM activities wereunreliable. Further assays of Ppm1E and Ppm1F were thereforeperformed using GST-AMPKα1 as substrate labelled with [γ32P]-ATPand CaMKKα. The results show that Ppm1A activity was similar in cellswith and without phenformin stimulation (Fig. 2C). In contrast, theactivities of Ppm1E, immunopelleted with two different antibodies, andPpm1F were significantly decreased in phenformin stimulated com-pared with untreated cells, essentially confirming the results obtainedwith casein substrate. Although the initial aim in labellingGST-AMPKα1was to obtain a specific substrate, the incorporation of 10 mol [32P] permol GST-AMPKα1 indicates many Ser or Thr residues are likely to bephosphorylated in addition to AMPKα-Thr172. The GST-AMPKα1 istherefore acting as a non-specific substrate similar to casein.

In vitro activities of bacterially expressed Ppm1A and Ppm1B wereactivated by Mg2+ but not Mn2+, whereas Ppm1E and Ppm1F wereactivated by the addition of either metal ion, although the highestactivities were observed in the presence of Mg2+ (Fig. 3A). Ppm1A,Ppm1B, Ppm1E and Ppm1F were essentially unaffected by in vitroincubationwith up to 10 mMphenformin, AMP or ATP, indicating thatthese molecules do not directly bind and inhibit Ppm1E and Ppm1F(Fig. 3B–D). The AMPK activator A-769662 similarly did not affect theactivities of the PPM phosphatases at concentrations up to 10 μM(Fig. 3E).

3.3. Characterisation of the Ppm1E isoforms

Since the initial antibodies raised to Ppm1E did not detect theendogenous phosphatase on immunoblots of cell lysates, severalfurther anti-Ppm1E antibodies were generated and examined forspecificity using both N- and C-terminally FLAG-tagged Ppm1Eexpressed in HEK293 cells. Two anti-Ppm1E antibodies, anti-Ppm1E

Fig. 2. Investigation of the expression levels and activities of PPM family protein phosphatases in HEK293 cells untreated and incubatedwith 10 mMphenformin for 1 h. (A) Cell lysateswereimmunoblotted using antibodies raised against the Ppm enzymes indicated. Two representative samples are shown for each treatment. The observedmolecularmass in kDa for each proteinphosphatase is indicated on the left and themolecularmass predicted from the cDNA is indicated in parenthesis on the right. Anti-Ppm1D antibodies detected several bands including one atthe expected size of 66 kDa, which was not analysed further. (B) Ppm specific antibodies non-covalently coupled to protein G sepharose were used to immunopellet the individualphosphatases indicated from HEK293 cell lysates. Anti-Ppm1A/PP2Cα protein [Ab1, [15]], anti-Ppm1A/PP2Cα peptide (Ab2) and anti-Ppm1E(739–752) were employed in these studies;details of other antibodies are given in the Supplementary file. The immunopellets werewashed and protein phosphatase activity wasmeasured using 32P-labelled casein as substrate in thepresenceof10 mMmagnesiumacetate and5 μMokadaic acid.Data are themeanactivities (permgof lysateprotein)±SEMforassaysperformed intriplicateon immunopellets fromlysatesofcells untreated and stimulated with 10 mM phenformin for 1 h. (C) Anti-Ppm1A/PP2Cα peptide (Ab2), anti-Ppm1E(717–728), anti-Ppm1E(739–752) and anti-Ppm1F antibodies non-covalently coupled to protein G sepharose were used to immunopellet the individual phosphatases indicated from 0.2–1 mg protein in cell lysates. The immunopellets were washed andprotein phosphatase activitywasmeasuredusing 32P-labelledGST-AMPKas substrate in thepresence of 10 mMmagnesiumacetate and5 μMokadaic acid. Data are themeanactivities±SEMfor assaysperformedon three independent immunopellets for eachantibody from lysatesof cells untreatedand stimulatedwith10 mMphenformin for 1 h.Differentbatchesof antibodywereemployed in experiments B and C.

117M. Voss et al. / Cellular Signalling 23 (2011) 114–124

(739–752) and anti-Ppm1E(495–755), as well as anti-FLAG anti-bodies specifically recognised epitope tagged Ppm1E (83 kDa+1 kDaFLAG tag) migrating at ~120 kDa (Fig. 4A). A cleaved form migratingat ~90 kDa was detected in the N- and not in the C-terminally FLAG-tagged Ppm1E using anti-FLAG, anti-Ppm1E(506–526) and anti-Ppm1E(495–755) antibodies indicating that cleavage at the C-terminus of the ~120 kDa species forms an ~90 kDa species,consistent with previous studies [26]. The slightly slower migrationof the N-terminally tagged Ppm1E compared with the C-terminallytagged Ppm1E at ~90 kDa can be explained by the loss of the FLAG tagonly in the latter. Additional cleaved forms migrating at ~75 kDa

and ~65 kDa, detected with anti-Ppm1E(495–755), must be cleavedat both the N- and C-termini, because they are not detected with anti-FLAG antibodies. The endogenous Ppm1Emigrating at ~120 kDa is onlydetected by immunoblotting with the anti-Ppm1E(495–755) antibo-dies, which also detect endogenous Ppm1E fragments migrating at ~75and ~65 kDa, but not an endogenous ~90 kDa form. Note thatendogenous Ppm1E migrating at ~120 kDa is not detected in the celllysate (Con) with the anti-Ppm1E(506–526) or anti-Ppm1E(739–752)antibodies probably because the sensitivities of the anti-peptideantibodies are less than that of the antibodies to the protein fragmentPpm1E(495–755). Endogenous Ppm1E migrating at ~75 kDa

Fig. 3. Analysis of the activities of bacterially expressed Ppm1E and Ppm1F. In vitro assays were performed by incubation of bacterially expressed Ppm1A(1–382) (0.5 μg/ml), Ppm1B(1–387) (1.5 μg/ml), Ppm1E(1–755) (5 μg/ml) and Ppm1F(1–454) (2.5 μg/ml) with the synthetic phospho-peptide substrate RRAT(p)VA at 37 °C for 10 min, with the release ofphosphate beingmeasured by Biomol Green reagent. (A) Stimulation of Ppm activities byMg2+ andMg2+. The effects of (B) 0–10 mMphenformin, (C) 0–10 mMAMP, (D) 0–10 mMATP and (E) 0–10 μM AMPK activator A-769662 on Ppm1A, Ppm1B, Ppm1E and Ppm1F in the presence of 10 mM Mg2+ are presented.

118 M. Voss et al. / Cellular Signalling 23 (2011) 114–124

and ~65 kDa is not detected with the anti-peptide antibodies becausecleavage of the C-terminal region of Ppm1E removes the peptides towhich these antibodies are raised.

3.4. Investigation of the interaction of Ppm1E and Ppm1F with AMPK

Specific immunoadsorption from cell lysates of endogenousPpm1E, migrating at 75 kDa, revealed that AMPKα2 could be detectedin Ppm1E immunopellets and was absent from immunopellets ofPpm1A and pre-immune immunoglobulin (Fig. 4B). However, Ppm1Ewas only very weakly detected in the immunopellets of AMPKα2.Endogenous Ppm1F could be observed in the immunopellets of

AMPKα1 and AMPKα2 (Fig. 4C), and AMPKα2 migrating at 62 kDacould sometimes be observed in immunopellets of Ppm1F (Fig. 4B–D).

The regulation of AMPK and related protein kinases by poly-ubiquitylation has recently been described [27] and therefore welooked for the presence of high molecular mass polyubiquitylatedAMPK in Ppm1F immunopellets (Fig. 4D). Strong AMPKα1 andAMPKα2-specific bands were observed migrating with molecularmasses of ~75 kDa to N150 kDa and these bands were also specificallydetected with anti-ubiquitin antibodies (Fig. 4D). Furthermore, whenthe immunoadsorption of Ppm1F and its interactors was performed inthe presence of the proteasome inhibitors, lactacystin and MG-132,and N-ethylmaleimide, which inhibits deubiquitylation, the presence

Fig. 4. Analysis of Ppm1E and Ppm1F interaction with AMPK. (A) Immunoblotting of control HEK293 cell lysates (Con) and lysates from HEK293 cells expressing Ppm1E N-terminallyFLAG-tagged (N-tag) and C-terminally FLAG-tagged Ppm1E (C-tag) with anti-FLAG antibodies and several different anti-Ppm1E antibodies as indicated. (B) Immmunoadsorption ofendogenousPpm1EandPpm1F fromHEK293cell lysates, followedbyblottingof the immunopellets (IP) shows thepresenceofAMPKα2 in thepellets. Ppm1Aandpre-immune IgG servedas a control to verify the specificity. Immmunoadsorption of endogenous AMPKα2 shows weak co-immunoadsorption of endogenous Ppm1E (75 kDa) and endogenous Ppm1F. Theantibodies used for immunoadsorption and detection of Ppm1E were anti-Ppm1E(495–755). (C) Immmunoadsorption of endogenous AMPKα1 and AMPKα2 from HEK293 cell lysates,followed by blotting of the immunopellets (IP) reveals co-immuno adsorption of endogenous Ppm1F (51–55 kDa). (D) Immmunoadsorption of endogenous Ppm1F from HEK293 celllysates, prepared in the presence of 5 mM NEM and/or 5 μM MG132 and 20 μM lactacystin, followed by blotting the immunopellets with anti-Ppm1F, anti-AMPKα1, anti-AMPKα2 andanti-ubiquitin antibodies. Note the presence of high molecular bands N150 kDa, which are detected with the latter three antibodies.

119M. Voss et al. / Cellular Signalling 23 (2011) 114–124

of AMPKα2 high molecular mass bands (N150 kDa) was slightly morepronounced (Fig. 4D right-hand two blots, lane 5). The results suggestthat Ppm1F may preferentially interact with the polyubiquitylatedAMPKα1 and AMPKα2.

3.5. Effect of lentiviral-mediated Ppm1E and Ppm1F shRNAi on AMPKphosphorylation

Lentiviruses expressing three different small hairpin inhibitoryRNAs (shRNAi) for both Ppm1E and Ppm1F were used to infectHEK293 cell cultures, which were then examined for depletion ofPpm1E and Ppm1F. Lentiviruses expressing the puromycin resistance

marker (P) with or without a scrambled shRNA sequence (non-target,NT) were employed as controls. The HEK293 cells showing the largestdecrease of Ppm1E and Ppm1F compared with the controls werechosen for detailed studies. Specific depletion of Ppm1A, Ppm1E andPpm1F was observed (Fig. 5A–C), without affecting the level of thePpm in NT and Cs controls (see Fig. 5 legend for details) and otherlentiviral treated samples. Depletion of Ppm1A was 96% with respectto the Cs and NT controls, while depletion of Ppm1F was 80–83%compared with the two controls (Fig. 5A and C). Depletion of Ppm1E(~120 kDa band) was 41–45% with respect to the NT and Cs controls,while Ppm1E (~75 kDa band) depletion was 38–41% with respect tothe same controls (Fig. 5B). A robust increase in AMPKα-Thr172

Fig. 5. Depletion of Ppm1A, Ppm1E and Ppm1F using lentiviral-mediated shRNAi and its effect on AMPK phosphorylation. (A–C) Immunoblots of lysates from HEK293 cells withlentiviral delivered stable depletion of Ppm1A (A), Ppm1E (B) and Ppm1F (C) compared with lentiviral delivered scrambled non-target (NT) controls. Glyceraldehyde 3-phosphatedehydrogenase (GAPDH) was analysed on the same blot and used as a control for sample loading. (D) HEK293 cells with stable depletions of Ppm1A, Ppm1E and Ppm1F wereincubated in serum free medium for 18 h and the lysates were examined for phosphorylation of AMPKα-Thr172. Immunoblotting was performed using fluorescence dye-coupledsecondary antibodies and the Odyssey imaging system. The controls are uninfected HEK293 cells cultured in the presence of fetal bovine serum (Cs); uninfected HEK293 cellscultured in serum free medium for 18 h (C), and virally infected control cells (NT) cultured in serum free medium for 18 h. In the right-hand panel HEK293 cells in serum freemedium were stimulated with 5 mM phenformin for 60 min, prior to immunoblotting of the lysates.

120 M. Voss et al. / Cellular Signalling 23 (2011) 114–124

phosphorylation in lysates of cells depleted of Ppm1E is observed,whilelysates prepared similarly from cells infectedwith NT control lentivirus,Ppm1A and Ppm1F lentiviruses show no increase in AMPKα-Thr172phosphorylation (Fig. 5D left-hand panel). The same results wereobtainedwithat least twodifferent shRNAi constructs for eachPpm.Asapositive control, cells were stimulated with phenformin and the lysatesshowed a similar increase in AMPKα-Thr172 phosphorylation to thatseen by depletion of Ppm1E (Fig. 5D right-hand panel).

3.6. Analysis of combined Ppm1E shRNAi and phenformin treatments onAMPK phosphorylation

The combined effects of Ppm1E depletion and stimulation of thecells with increasing concentrations of phenformin in the presence(Fig. 6A) and in the absence of fetal bovine serum (Fig. 6C) are shownwith blots from one representative experiment. Average values forAMPKα-Thr172 phosphorylation were determined from six

121M. Voss et al. / Cellular Signalling 23 (2011) 114–124

experiments (Fig. 6B and D). In the absence of phenformin, depletionof Ppm1A did not cause any increase in AMPKα-Thr172 phosphor-ylation compared with that in NT control cells, while depletion ofPpm1E resulted in an ~2-fold statistically significant increase inAMPKα-Thr172 phosphorylation compared with NT control andPpm1A depleted cells (Fig. 6B and D). Treatment of NT control cellswith 5 mM phenformin produced a 2.5-fold increase in AMPKα-Thr172 phosphorylation. Interestingly, stimulating cells depleted ofPpm1E with low concentrations of phenformin revealed a strikingsynergistic effect of these two treatments on AMPKα-Thr172phosphorylation. In Fig. 6B, NT control HEK293 cells stimulated with0.05 mM phenformin showed a 1.2-fold increase in AMPKα-Thr172phosphorylation, while Ppm1E depleted cells stimulated with0.05 mM phenformin showed a 2.9-fold increase in AMPKα-Thr172phosphorylation, compared with the level in unstimulated NT controlcells. If the effects of phenformin and Ppm1E depletion were additive,a 1.9–2.2-fold increase would be expected. A similar synergistic effectwas noted from the data in Fig. 6D, indicating that phenformin andPpm1E depletion are affecting AMPK phosphorylation by differentprocesses and that Ppm1E may dephosphorylate AMPKα-Thr172 invivo.

3.7. Effects of depletion of Ppm1F and inhibition of deubiquitylation onAMPK phosphorylation

An assay was designed using cell lysates prepared in the absenceand presence of the deubiquitylase inhibitor, N-ethylmaleimide.Additions to the lysates that produced a 1.5 mM ATP/1 mM AMPratio (assuming no equilibration) and 2 mM free Mg2+ concentrationresulted in a 3-fold rise in AMPKα-Thr172 phosphorylation in controland Ppm1F depleted cells in the absence of N-ethylmaleimide (Fig. 7).Higher free Mg2+ concentrations resulted in a reduction of the AMPKphosphorylation and several different ATP/AMP ratios did notenhance the phosphorylation in control cells (data not shown). Inthe presence of N-ethylmaleimide, the phosphorylation of AMPK wasblocked almost completely in lysates from control cells, whereas inPpm1F depleted, N-ethylmaleimide treated lysates, AMPKα-Thr172phosphorylation was only partially decreased. The higher level ofAMPKα-Thr172 phosphorylation in Ppm1F depleted cell lysatescontaining N-ethylmaleimide compared with controls supports theinvolvement of Ppm1F in the dephosphorylation of AMPK. Thedependence of the results on N-ethylmaleimide indicates thatdeubiquitylation is essential for AMPK phosphorylation and isconsistent with the hypothesis that Ppm1F may specifically interactwith the polyubiquitylated AMPKα1 and AMPKα2.

4. Discussion

4.1. Ppm1E is an AMPK phosphatase and Ppm1F may contribute to thedephosphorylation of AMPK

Our initial studies suggested that treatmentof cellswithphenformin,a knownactivator of AMPK, directly or indirectly inhibited the activity ofone or more okadaic insensitive phosphatases. Several members of thePPP family of protein phosphatases, including PP1 and PP2A, arepotently inhibited by okadaic acid but PPM family members are notknown to be sensitive to this inhibitor [24]. Development of specificantibodies enabled assay of a range of PPM family phosphatases andshowed that the activities of Ppm1E and Ppm1F, but not other Ppmphosphatases,were substantiallydecreased in the lysates of phenformintreated cells, suggesting they were potential candidates for AMPKphosphatases. Employing lentiviral-mediated shRNAi to deplete Ppm1Ein HEK293 cell cultures, we found that the phosphorylation of AMPKα-Thr172 was significantly increased compared with controls and cellsdepleted in Ppm1A, previously proposed as an AMPK phosphatase[15,28]. Statistically significant results were obtained with cells stably

depleted in Ppm1E in the presence or absence of fetal bovine serum.Although stable depletion of Ppm1F did not show a statisticallysignificant increase in AMPKα-Thr172 phosphorylation in cell cultures,assays of AMPK phosphorylation cell lysates, in the presence of adeubiquitylation inhibitor and following additions to increase the ATP:AMP ratio, indicated that Ppm1F may contribute to the dephosphory-lation of AMPK.

Heterotrimeric AMPKα2-containing complexes are enriched in thenucleus and have a greater dependence on AMP compared withAMPKα1-containing complexes [29]. Studies inmice lacking theα1 orα2 isozyme indicate that AMPK complexes containing theα2 isozymeplay a more predominant role in controlling glucose homeostasis andinsulin sensitivity [30]. Our studies showed that endogenous AMPKα2was observed in the immunopellets of endogenous Ppm1E, consistentwith AMPKα2 specifically interacting with Ppm1E. EndogenousPpm1F was detected in the immunopellets of both endogenousAMPKα1 and AMPKα2, and polyubiquitylated AMPKα1 and AMPKα2were detected in the immunopellets of endogenous Ppm1F. Previousstudies have shown that AMPK can be polyubiquitylated [27] and ourdata suggest that Ppm1F may predominantly interact with poly-ubiquitylated AMPK. Our observation that a deubiquitylation inhibitorcompletely blocks phosphorylation of AMPK indicates that deubiqui-tylation is essential for AMPKα-Thr172 phosphorylation and thatpolyubiquitylation may play a role in directly or indirectly bindingAMPK phosphatase(s).

Ppm1E and Ppm1F were first identified as Ca2+/calmodulin-dependent protein kinase phosphatases, and like the initialPPM family member identified (Ppm1A/PP2Cα) they are activatedby Mg2+and/or Mn2+ and insensitive to inhibitors such as calyculin Aand okadaic acid that inhibit the majority of PPP family phosphatases[23,31]. Ppm1E and Ppm1F were able in vitro to dephosphorylatephospho-threonine residues involved in the activation of cytoplasmicCaMKI, nuclear CaMKIV, and CaMKII, which is found in both thenucleus and the cytoplasm [31–33]. Ppm1F (CaMKP) is cytoplasmicand ubiquitously expressed, while Ppm1E (CaMKP-N) is highlyabundant in the brain and present in both the nucleus and thecytoplasm [26,34,35]. Our studies show that Ppm1E is present inHEK293 cells, but may not be easily detected in cell lines and tissueswhere it is present at lower levels than in the brain, because it iscleaved into several different sized fragments, at least one of which isthought to be generated in vivo. Full-length Ppm1E/CaMKP-N (83 kDa,migrating at ~120 kDa on SDS-PAGE) is present in the nucleus andcontains two nuclear localisation signals near the C-terminus, both ofwhich play a role in determining its subcellular localisation [35].Cleavage of the C-terminal region, which encompasses the twonuclear localisation sequences (NLS), yields a major form of epitopetagged Ppm1E/CaMKP-N (61 kDamigrating at ~90 kDa on SDS-PAGE)that is present in the cytoplasm [26]. Our studies show that FLAG-tagged Ppm1E is cleaved to a fragment migrating at ~90 kDa, butendogenous Ppm1E appears to be cleaved to a fragment migrating at~75 kDa.

AMPK is phosphorylated by the tumour suppressor LKB1 incomplex with its two accessory proteins STRAD and MO25 [8–10]and by the Ca2+-activated protein kinase, CaMKKβ [11,12,36](Fig. 8A). Neither of these two kinases is known to be activated byphosphorylation, LKB being constitutionally active and CaMKKβ beingactivated by Ca2+. Ppm1E and Ppm1F are therefore unlikely to beprotein phosphatases acting on kinases upstream of AMPK and theirpresence in AMPKα1 and α2 immunopellets indicates that they actdirectly on AMPK.

4.2. Ppm1E shRNAi and phenformin synergistically increase thephosphorylation of AMPK

Interestingly, we observed that depletion of Ppm1E by shRNAiacts synergistically with phenformin-mediated increase of AMPK

122 M. Voss et al. / Cellular Signalling 23 (2011) 114–124

phosphorylation in HEK293 cells, particularly at low phenforminconcentrations, indicating that the two processes operate by differentmechanisms. In agreement, with these studies, bacterially expressedPpm1E and Ppm1F are not directly inhibited by phenformin or AMPpointing to an indirect effect of phenformin on Ppm1E and Ppm1F.We

have not observed any changes in the total level of Ppm1E and Ppm1Ffollowing phenformin treatment indicating that the phosphatases donot appear to undergo significant net degradation. However, it isplausible that binding of AMP to AMPK causes changes in thestructure and ubiquitylation of AMPK, leading to dissociation of the

Fig. 8. Model for the role of Ppm1E and Ppm1F in the dephosphorylation and activation ofAMPK. (A) AMPK is phosphorylated on Thr172 of the catalyticα subunit and activated bythe kinases, LKB1 (in complex with MO25 and STRAD) in response to an increase in theAMP:ATP ratio and/or CaMKKβ in response to a rise in Ca2+ ions. Ppm1E (and probablyPpm1F) dephosphorylates AMPKα-Thr172, inactivatingAMPK. Phenformin stimulation ofcells causes a rise in the AMP:ATP ratio, and binding of AMP to the AMPK γ subunit maycause an allosteric change inAMPK that facilitates decrease of Ppm1Ephosphatase activitygiving rise to net phosphorylation of AMPKα-Thr172. (B) Model of the mechanism bywhich ubiquitylation may control the interaction of Ppm1F with AMPK. Polyubiquitin,with ubiquitin Lys29–Lys33 linkages, covalently attaches to theα subunit of AMPK througha Lys residue [27]. The carboxy-terminal region of the AMPKα subunit may contain aubiquitin-associated domain (UBA), which interacts with and stabilises the polyubiquitin.Our data indicate that Ppm1F (and possibly also Ppm1E) interacts with polyubiquitylatedAMPK and the presence of the phosphatase prevents the AMPKα-Thr172 from beingphosphorylated. A rise in the AMP:ATP ratio or phenformin may allow a deubiquitylatingenzyme to cleave the polyubiquitin thereby releasing or inhibiting the phosphatase andallowing LKB1 to phosphorylate AMPKα-Thr172 and activate AMPK.

Fig. 7. Effect of NEM on AMPK phosphorylation. HEK293 cells were treatedwith a lentiviral-vector-basedsiRNAtoknockdownPpm1F.Untreatedcellswereusedas acontrol.HEK293celllysates were incubated with (+) and without (−) ATP (1.5 mM ATP, 1 mM AMP) in thepresence of 2 mM free Mg2+. Immunoblotting of AMPK phospho-threonine 172 wasperformed using fluorescence dye-coupled secondary antibodies and the Odyssey infraredimaging system fromLi-Cor Bioscience. A representative blot is shown in the lower panel andalthough the bands shown are on the same blot the lanes are not all contiguous. The values ofthephosphorylation intensitywerenormalised to the valueof the samples incubatedwithoutATP andMg2+ (=100%). The bar chart in the top panel showsmean values+/−SEM for sixexperiments. The statistical significance for Ppm1F depletion versus control NT values inlysates prepared in the presence of NEMwas *pb0.05.

123M. Voss et al. / Cellular Signalling 23 (2011) 114–124

associated phosphatases with alteration of their activities (Fig. 8B)and/or that covalent modification decreases the phosphatase activi-ties. In HeLa cells, which do not express LKB1, and in mouseembryonic fibroblasts with a deletion of LKB1, AMPKα2 complexesare not phosphorylated and activated in response to phenforminstimulation and the activity of the α1 containing complexes isdecreased although not completely abolished [8,21]. Nevertheless, theallosteric changes in AMPK structure caused by the binding of AMPmay occur independently of phosphorylation [17,37,38] and thesemay alter ubiquitylation and interaction with a protein phosphatase,modifying its activity. The effect of incubating cells with phenforminon the different endogenous PPM family phosphatases has not beeninvestigated previously.

Phenformin and drugs developed to activate AMPK in intact cells[5-amino-4-imidazolecarboxamide riboside (AICAR) and A-769662]inhibit Ppm1A/PP2Cα by substrate mediated mechanisms in vitro[15,28,39], but Ppm1A/PP2Cα has not been demonstrated todephosphorylate AMPK-pThr172 in vivo. We did not observephosphorylation and activation of AMPK following depletion of

Fig. 6. Effect of Ppm1Edepletion andphenformin stimulation onAMPKphosphorylation inHEK29serum,with lentiviral deliveredstabledepletionofPpm1AandPpm1Ecomparedwith controls. Cellof the lysates for phosphorylationof AMPKα-Thr172.Glyceraldehyde 3-phosphate dehydrogenaseloading. Immunoblotting was performed using fluorescence dye-coupled secondary antibodies anHEK293 cells cultured in the presence of fetal bovine serum with no phenformin (Cs/0), 0.05 mMinfected control cells (NT). (B) The values of the AMPKα-Thr172 phosphorylation intensity in (A) acontrol (Cs/0) cells. The error bars show+/−SEMfor the six independent experiments. The statisticperformedas in (A) except thatHEK293 cellswere incubated in the absence of fetal bovine serumoof serum. (D)Thevaluesof theAMPKα-Thr172phosphorylation intensity in (C) andfive similar indThe error bars show+/−SEM for the six independent experiments. The statistical significance fo

Ppm1A/PP2Cα as we did subsequent to depletion of Ppm1E,indicating that Ppm1E and probably Ppm1F are more likely to bespecific AMPK phosphatases than Ppm1A/PP2Cα. The synergisticeffect of depleting Ppm1E and phenformin treatment of cells on AMPK

3 cells. (A) Immunoblots of lysates fromHEK293 cells, cultured in the presence of fetal bovineswere stimulatedwithvarying concentrationsofphenformin for60 min,prior toexamination(GAPDH)was analysed on the sameblot as AMPKα-pThr172 andused as a control for sampled the Odyssey infrared imaging system from LI-COR Bioscience. The controls are uninfectedphenformin (Cs/0.05), 0.5 mM phenformin (Cs/0.5), 5 mM phenformin (Cs/5) and virally

ndfive similar independent experimentswere normalised to the value of the non-stimulatedal significance for Ppm1Eversus theNTvalues are *pb0.05; **pb0.001. (C)Experimentswerevernight and stimulatedwith varying concentrations of phenformin for 60 min in the absenceependent experimentswerenormalised to thevalueof thenon-stimulated control (C/0) cells.r Ppm1E depletion versus the NT values are *pb0.05; **pb0.002; ***pb0.0001.

124 M. Voss et al. / Cellular Signalling 23 (2011) 114–124

activation suggests that Ppm1E would be a valuable anti-diabetictarget. In addition, a combination of a specific inhibitor of Ppm1Ewithlow levels of metformin or phenformin may also be useful, becausethis may eliminate the detrimental effects that high doses of thebiguanide drugs are believed to exert on mitochondrial function.

5. Conclusions

Using lentiviral delivered stable gene silencing in HEK293 cellcultures, we have demonstrated that depletion of Ppm1E, but notPpm1A/PP2Cα, increases AMPKα-Thr172 phosphorylation approxi-mately 3 fold. In addition, depletion of Ppm1E and phenforminstimulation of cells increases phosphorylation of AMPKα-Thr172synergistically. These data demonstrate for the first time that Ppm1Eis an AMPK phosphatase in cell cultures and a potential pharmaceu-tical target for the treatment of type 2 diabetes. Ppm1E and the relatedPpm1F interact weakly with AMPK and assays with lysates of cellsstably depleted of Ppm1F suggest that this phosphatase contributes todephosphorylation of AMPK. These assays also indicate that AMPKdeubiquitylation is essential for AMPKα-Thr172 phosphorylation,providing evidence that polyubiquitylation of AMPKmay play a role inbinding the AMPK phosphatases.

Supplementarymaterials related to this article can be found onlineat doi:10.1016/j.cellsig.2010.08.010.

Acknowledgements

We thank Grahame Hardie for useful discussions and AMPKantibodies, and Rachel Toth for assistancewith DNA cloning. Theworkwas funded by the Medical Research Council UK, MRC TechnologyDevelopment GAP Fund A853-0078 and pharmaceutical companiessupporting the Division of Signal Transduction Therapy (AstraZeneca,Boehringer Ingelheim, GlaxoSmithKline, Merck Serono and Pfizer). J.P.was the recipient of a postgraduate studentship from the MedicalResearch Council UK.

References

[1] G. Zhou,R.Myers, Y. Li, Y. Chen,X. Shen, J. Fenyk-Melody,M.Wu, J. Ventre, T.Doebber,N.Fujii, N. Musi, M.F. Hirshman, L.J. Goodyear, D.E. Moller, J. Clin. Invest. 108 (2001) 1167.

[2] M.R. Owen, E. Doran, A.P. Halestrap, Biochem. J. 348 (2000) 607.[3] M.Y. El-Mir, V. Nogueira, E. Fontaine, N. Averet, M. Rigoulet, X. Leverve, J. Biol.

Chem. 275 (2000) 223.[4] B.B. Kahn, T. Alquier, D. Carling, D.G. Hardie, Cell Metab. 1 (2005) 15.[5] D. Carling, Novartis Found. Symp. 286 (2007) 72.[6] D.G. Hardie, Int. J. Obes. 32 (Suppl 4) (2008) S7 (Lond).

[7] D.G. Hardie, J. Cell Sci. 117 (2004) 5479.[8] S.A. Hawley, J. Boudeau, J.L. Reid, K.J. Mustard, L. Udd, T.P. Makela, D.R. Alessi, D.G.

Hardie, J. Biol. 2 (2003) 28.[9] A. Woods, S.R. Johnstone, K. Dickerson, F.C. Leiper, L.G. Fryer, D. Neumann, U.

Schlattner, T. Wallimann, M. Carlson, D. Carling, Curr. Biol. 13 (2003) 2004.[10] R.J. Shaw, M. Kosmatka, N. Bardeesy, R.L. Hurley, L.A. Witters, R.A. DePinho, L.C.

Cantley, Proc. Natl Acad. Sci. USA 101 (2004) 3329.[11] S.A. Hawley, D.A. Pan, K.J. Mustard, L. Ross, J. Bain, A.M. Edelman, B.G. Frenguelli,

D.G. Hardie, Cell Metab. 2 (2005) 9.[12] A. Woods, K. Dickerson, R. Heath, S.P. Hong, M. Momcilovic, S.R. Johnstone, M.

Carlson, D. Carling, Cell Metab. 2 (2005) 21.[13] R.L. Hurley, K.A. Anderson, J.M. Franzone, B.E. Kemp, A.R. Means, L.A. Witters,

J. Biol. Chem. 280 (2005) 29060.[14] F. Moore, J. Weekes, D.G. Hardie, Eur. J. Biochem. 199 (1991) 691.[15] S.P. Davies, N.R. Helps, P.T.W. Cohen, D.G. Hardie, FEBS Lett. 377 (1995) 421.[16] D. Carling, M.J. Sanders, A. Woods, Int. J. Obes. 32 (Suppl 4) (2008) S55 (Lond).[17] M. Suter, U. Riek, R. Tuerk, U. Schlattner, T. Wallimann, D. Neumann, J. Biol. Chem.

281 (2006) 32207.[18] C. MacKintosh, G. Moorhead, in: D.G. Hardie (Ed.), Protein Phosphorylation, a

Practical Approach, Oxford University Press, Oxford, 1999, p. 153.[19] S.A. Stewart, D.M. Dykxhoorn, D. Palliser, H. Mizuno, E.Y. Yu, D.S. An, D.M. Sabatini,

I.S. Chen, W.C. Hahn, P.A. Sharp, R.A. Weinberg, C.D. Novina, RNA 9 (2003) 493.[20] A.K. Das, N.R. Helps, P.T.W. Cohen, D. Barford, EMBO J. 15 (1996) 6798.[21] K. Sakamoto, A. McCarthy, D. Smith, K.A. Green, D. Grahame Hardie, A. Ashworth,

D.R. Alessi, EMBO J. 24 (2005) 1810.[22] M. Tiainen, A. Ylikorkala, T.P. Makela, Proc. Natl Acad. Sci. USA 96 (1999) 9248.[23] P.T.W. Cohen, in: J. Arino, D.R. Alexander (Eds.), Topics in Current Genetics:

Protein Phosphatases, Springer-Verlag, Berlin Heidelberg, 2004, p. 1.[24] P.T.W. Cohen, in: R.A. Bradshaw, E.A. Dennis (Eds.), Handbook of Cell Signaling,

Elsevier Inc. Academic Press, Oxford, 2009, p. 659.[25] T. Lammers, S. Lavi, Crit. Rev. Biochem. Mol. Biol. 42 (2007) 437.[26] T. Kitani, S. Okuno, Y. Nakamura, H. Tokuno, M. Takeuchi, H. Fujisawa,

J. Neurochem. 96 (2006) 374.[27] A.K. Al-Hakim, A. Zagorska, L. Chapman, M. Deak, M. Peggie, D.R. Alessi, Biochem. J.

411 (2008) 249.[28] O. Göransson, A. McBride, S.A. Hawley, F.A. Ross, N. Shpiro, M. Foretz, B. Viollet, D.G.

Hardie, K. Sakamoto, J. Biol. Chem. 282 (2007) 32549.[29] I. Salt, J.W. Celler, S.A. Hawley, A. Prescott, A. Woods, D. Carling, D.G. Hardie,

Biochem. J. 334 (1998) 177.[30] B. Viollet, F. Andreelli, S.B. Jorgensen, C. Perrin, D. Flamez, J. Mu, J.F. Wojtaszewski,

F.C. Schuit, M. Birnbaum, E. Richter, R. Burcelin, S. Vaulont, Biochem. Soc. Trans. 31(2003) 216.

[31] T. Kitani, A. Ishida, S. Okuno, M. Takeuchi, I. Kameshita, H. Fujisawa, J. Biochem.125 (1999) 1022 (Tokyo).

[32] M. Takeuchi, A. Ishida, I. Kameshita, T. Kitani, S. Okuno, H. Fujisawa, J. Biochem.130 (2001) 833 (Tokyo).

[33] B.P. Harvey, S.S. Banga, H.L. Ozer, J. Biol. Chem. 279 (2004) 24889.[34] T. Kitani, S. Okuno, M. Takeuchi, H. Fujisawa, J. Neurochem. 86 (2003) 77.[35] M. Takeuchi, T. Taniguchi, H. Fujisawa, J. Biochem. 136 (2004) 183.[36] S.A. Hawley, M.A. Selbert, E.G. Goldstein, A.M. Edelman, D. Carling, D.G. Hardie,

J. Biol. Chem. 270 (1995) 27186.[37] D.G. Hardie, I.P. Salt, S.A. Hawley, S.P. Davies, Biochem. J. 338 (1999) 717.[38] M.J. Sanders, P.O. Grondin, B.D. Hegarty, M.A. Snowden, D. Carling, Biochem. J. 403

(2007) 139.[39] M.J. Sanders, Z.S. Ali, B.D. Hegarty, R. Heath, M.A. Snowden, D. Carling, J. Biol.

Chem. 282 (2007) 32539.