|Home | About | Journals | Submit | Contact Us | Français|
The atypical AGC kinase Greatwall (Gwl) mediates a pathway that prevents the precocious removal of phosphorylations added to target proteins by M phase-promoting factor (MPF); Gwl is thus essential for M phase entry and maintenance. Gwl itself is activated by M phase-specific phosphorylations that are investigated here. Many phosphorylations are nonessential, being located within a long nonconserved region, any part of which can be deleted without effect. Using mass spectrometry and mutagenesis, we have identified 3 phosphorylation sites (phosphosites) critical to Gwl activation (pT193, pT206, and pS883 in Xenopus laevis) located in evolutionarily conserved domains that differentiate Gwl from related kinases. We propose a model in which the initiating event for Gwl activation is phosphorylation by MPF of the proline-directed sites T193 and T206 in the presumptive activation loop. After this priming step, Gwl can intramolecularly phosphorylate its C-terminal tail at pS883; this site probably plays a role similar to that of the tail/Z motif of other AGC kinases. These events largely (but not completely) explain the full activation of Gwl at M phase.
Greatwall kinase (Gwl) plays a critical role in M phase entry and maintenance in Drosophila melanogaster, Xenopus laevis egg extracts, and human tissue culture cells (2, 4, 41, 44, 45). Gwl inactivates, specifically during M phase, PP2A phosphatase associated with B55-type regulatory subunits. The inactivation of PP2A/B55 prevents the premature removal of M phase-specific phosphorylations added to many target proteins by the cyclin-dependent kinase MPF (M phase-promoting factor; Cdk1/cyclin B) (5, 22, 39). Gwl's role in PP2A/B55 inactivation is indirectly mediated through proteins of the endosulfine family that are phosphorylated by Gwl (10, 23).
The mechanism by which Gwl is itself activated during M phase is only poorly understood, but its hyperphosphorylation is clearly involved. MPF appears to be an upstream kinase important for Gwl activation (45), but the details are obscure and other kinases may also participate.
The evolution of Gwl is of considerable interest. Gwl proteins in flies, frogs, and humans are probably functionally orthologous (2, 44, 45). However, lineages that include species such as Saccharomyces cerevisiae, Caenorhabditis elegans, and plants lack true Gwl kinases (44). In the human kinome, Gwl (MASTL) and its close relatives MAST1 to -4 form an unstudied branch of the AGC kinase family (20). Gwl lacks 6 of the 15 amino acids that distinguish AGC kinases from other eukaryotic protein kinases (17). Remarkably, at the position of the activation loop of other AGCs, Gwl contains an insertion of hundreds of amino acids that is poorly conserved and whose function has yet to be determined.
Here, we describe investigations designed to further our understanding of Gwl's structure and cell cycle regulation. Our results suggest that Gwl follows the main logic of AGC activation, involving phosphorylations both of the presumptive activation loop and of a residue in the enzyme's C-terminal tail (reviewed in references 14 and 27). The activation loop phosphorylations are proline directed and targeted in vitro by cyclin-dependent kinases (CDKs), explaining at least in part how Gwl is turned on at M phase entry. The critical phosphorylation site (phosphosite) in the C-terminal tail can then be generated by intramolecular autophosphorylation. This phosphorylated residue probably interacts with a basic patch on the enzyme's N-terminal lobe to stabilize the active conformation. However, some aspects of Gwl regulation do not conform to the general AGC pattern and remain enigmatic.
The use of antibodies against Xenopus Gwl, Cdc25C, and tyrosine 15 (Y15) of Cdk1 has been previously described (45). MPM-2 antibody against mitotic phosphoepitopes was the kind gift of J. Kuang (M. D. Anderson Cancer Center, Houston, TX) (42).
Cytostatic factor (CSF) extracts were prepared according to the methods in references 25 and 26; frog care was monitored by the Institutional Animal Care and Use Committee at Cornell University. Immunodepletion of CSF extracts was performed as described previously (37, 45) using Affi-prep protein A beads (Bio-Rad Laboratories, Hercules, CA) coated with affinity-purified anti-Greatwall. Mock-depleted extracts were treated with protein A beads alone.
To assess the biological activity of Gwl proteins prepared in Sf9 insect cells infected with baculovirus constructs (see below), the proteins were added to Gwl-depleted CSF extracts and then incubated at 25°C. Aliquots of the extracts were then analyzed by Western blotting with antibodies against Gwl, the interphase-specific inactivating phosphorylation at Y15 on Cdk1, and Cdc25C phosphatase (45). Exogenous, active wild-type (WT) Gwl made in Sf9 cells treated with okadaic acid (OA) is capable of rescuing the effects of Gwl depletion when added at concentrations greater than 0.25 to 0.5 times that of the endogenous Gwl in CSF extracts, while kinase dead (KD) Gwl (with the G41S mutation) is unable to restore the extracts to M phase even at concentrations greater than 10 times that of the endogenous enzyme (Fig. 1A) (47). The results in Fig. 1B display the responsiveness of this assay to unactivated wild-type Gwl made in Sf9 cells that had not been treated with OA. This exogenous Gwl becomes activated and can promote M phase reentry in the extracts, but only slowly and only when added at concentrations 6 times or more higher than the endogenous levels.
Mutant Gwl proteins (point mutations and deletion mutations) were generated from a Gwl cDNA clone by using the QuikChange site-directed mutagenesis kit (Agilent, Santa Clara, CA). All proteins were expressed in Sf9 insect cells using the Bac-to-Bac system (Invitrogen, Carlsbad, CA) and purified as previously described (45). Active greatwall was produced by treating the infected cells with OA to a final concentration of 100 nM for 12 h before harvesting. The kinase dead allele of Xenopus Gwl used was G41S, previously characterized in references 44 and 45.
All Gwl fusion proteins contained both a His6 tag and a Z tag (the Z subdomain of protein A, which binds to IgG). The Z tag interacts with primary and secondary antibodies during Western blotting, accounting for approximately 50% of the band intensity seen on Western blots of recombinant Gwl proteins that contain it (47). In addition, active M phase Gwl behaves as a diffuse band and usually reacts less efficiently with anti-Gwl than does the unphosphorylated interphase form of Gwl. Because of these issues, the amounts of recombinant Gwl in all preparations were determined by Coomassie blue staining of preparations that had been treated with λ protein phosphatase (LPP). Phosphatase treatment was carried out by incubating Gwl proteins in LPP buffer (50 mM Tris [pH 8.0], 2 mM MnCl2, and 1 mg/ml acetylated bovine serum albumin [BSA]) with 40 U LPP (New England BioLabs, Beverly, MA) at 30°C for 30 min.
mRNAs encoding FLAG-tagged S883E or S883D Gwl proteins were injected into immature G2 phase Xenopus oocytes, which were incubated overnight and then treated (or not) with the hormone progesterone as described previously (43). Untreated injected oocytes remained in G2 phase; no germinal vesicle breakdown (GVBD) was observed, while GVBD occurred in nearly 100% of the oocytes treated with progesterone. Gwl was immunoprecipitated from injected oocytes using anti-FLAG antibody as described previously (43), and the immunoprecipitates were then subjected to standard kinase assays using the model substrate myelin basic protein (MBP).
Standard kinase assays were performed for 10 min at 30°C in 10 μl of kinase buffer (20 mM HEPES, 10 mM MgCl2, 0.1 mg/ml BSA, and 3 mM β-mercaptoethanol) supplemented with 100 μM cold ATP and 1 μCi [γ-32P]ATP (3,000 Ci/mmole) and with 0 to 250 ng of Gwl proteins (diluted into 1 μl of phosphate-buffered saline–50% glycerol) prepared from Sf9 insect cells infected with baculovirus constructs as described above. In early experiments, 1.5 μg of MBP (Active Motif, Carlsbad, CA) was added to the reaction mixture as a model substrate. The kinase reaction was terminated by the addition of SDS sample buffer, and the samples were fractionated by SDS polyacrylamide gel electrophoresis. Radioactive Gwl and MBP bands were identified by autoradiography; the radioactivity incorporated into Gwl and MBP was quantified by cutting the bands out of the dried gel and analyzing them for 32P in a scintillation counter. The results in Fig. 1C show the linear range of this assay for active WT Gwl made in OA-treated Sf9 cells.
In later experiments, 1.5 μg of a fragment of the verified Gwl substrate endosulfine (23) was used instead of MBP in the in vitro kinase assays. This fragment (56KRLQKGQKYFDSGDYNMAKAK76 of Xenopus endosulfine, with the Gwl site underlined) was fused to maltose-binding protein in the vector pMAL-C2x (New England BioLabs). The cloned endosulfine fragment and MBP behaved identically as substrates for all Gwl variants tested.
For tests on phosphorylation of the Gwl C-terminal tail, 8.6 μg of the synthetic peptide RRNNAQHLKVSGFSL887 (500 μM final concentration) was incubated with 100 ng WT or KD Gwl (made in the presence of OA) and radioactive ATP in standard kinase assays. The entire reaction mixture was then spotted on a phosphocellulose filter (Whatman P81). The filter was washed extensively in phosphoric acid, and the radioactivity remaining on the filter was measured in a scintillation counter as described previously (13).
An amount of 250 ng Gwl made in the absence of OA was incubated in kinase buffer containing 100 μM ATP with 25 ng exogenous kinases for 1 h at 30°C; for mass spectrometry (MS) of activated Gwl, the ATP concentration was increased to 1 mM. These kinases include MPF (Cdk1/cyclin B) (45), Cdk2/cyclin A made according to the method in reference 22, Plx1 (see below), recombinant active Aurora A (product number A3483; Sigma-Aldrich, St. Louis, MO), and Nek2 (product number N4787; Sigma-Aldrich). Activity of the exogenous kinases on Gwl was verified both by incorporation of radioactive phosphate from [γ-32P]ATP and by MS analysis. Aliquots corresponding to 90% of the in vitro phosphorylation reaction mixtures were examined after gel electrophoresis by Coomassie blue staining for quantification of Gwl. The remaining 10% of each sample (containing 25 ng of Gwl) was added to standard assays for Gwl kinase activity, some of which contained the CDK inhibitor roscovitine (0.25 mM final concentration) as indicated below.
To determine whether Suc1/Cks/p9 proteins can enhance CDK-driven activation of Gwl, Cks2 (Fitzgerald Industries, Acton, MA) was added at a 2.5× molar ratio to Cdk2/cyclin A prior to their incubation with Gwl made in the absence of OA. The incorporation of radioactivity into recombinant endosulfine was measured either by autoradiography or on phosphocellulose filters. The latter assay accounted for radioactivity incorporated into Gwl by autophosphorylation or CDK's action on Gwl by using parallel reactions performed in the absence of the endosulfine substrate.
In contrast with a recent report (40) showing some activation of human Gwl with Polo-like kinase in vitro, we have not observed activation of Xenopus Gwl using several different preparations of Polo-like kinases. These include the constitutively active T201D mutant of Plx1 (isolated by ourselves from baculovirus-infected Sf9 cells not treated with OA) (45), a similar preparation of T201D Plx1 made in the laboratory of James Maller (University of Colorado, Denver, CO), and a commercial preparation of active human Plk1 from Sigma-Aldrich (St. Louis, MO). We checked the activity of these preparations in several ways, including their ability to phosphorylate the known substrate Pin1 (e.g., see Fig. 10A) and a model PLKtide peptide substrate (CKKLGEDQAEEISDDLLEDSLSDEDE; SignalChem, Richmond, BC, Canada) (see Fig. 10B). The Plx1 preparation used in the experiments whose results are shown in Fig. 3B had a specific activity of 6.4 nmol/min/mg when measured with this PLKtide. The activity of the same Plx1 preparation was also demonstrated by its nearly complete phosphorylation of Gwl T416 (located in a reasonable Polo consensus within the nonconserved middle region [NCMR]) (MS data not shown).
Recombinant Greatwall proteins made in Sf9 cells in the absence of OA were incubated with CSF-arrested extracts containing 200 μCi [γ-32P]ATP at 23°C for 1 h, followed by anti-Greatwall antibody immunoprecipitation. The immunoprecipitates were then subjected to two-dimensional tryptic-phosphopeptide mapping as described previously (46). The same mapping procedure was also performed with wild-type active recombinant Gwl (made in Sf9 cells in the presence of OA) incubated in a standard kinase assay in the presence of [γ-32P]ATP.
Mitotic Gwl was immunoprecipitated from CSF extracts by protein A beads coated with Gwl antibody; alternatively, recombinant Gwl was expressed and purified from Sf9 cells, and in some cases activated with Cdk2/cyclin A, as described above. Gwl proteins made by any of these methods were then separated on SDS-polyacrylamide gels, identified by staining with PageBlue Coomassie stain (Fermentas, Glen Burnie, MD), and destained in water. Excised gel bands (~500 ng) were subjected to in-gel digestion by trypsin (or in selected cases with chymotrypsin or endopeptidase GluC) and subsequent extraction as previously described (45). The digest was reconstituted in 10 μl of 2% acetonitrile (ACN) with 0.5% formic acid (FA) for nano-liquid chromatography–electrospray ionization–tandem mass spectrometry (nanoLC–ESI–MS-MS) analysis using an LTQ-Orbitrap Velos (Thermo-Fisher Scientific, San Jose, CA) mass spectrometer equipped with a Plug and Play nano-ion source device (CorSolutions LLC, Ithaca, NY). The nanoLC was carried out with the Dionex UltiMate3000 MDLC (multidimensional LC) system (Dionex, Sunnyvale, CA). The Orbitrap Velos was operated in positive-ion mode with nano spray voltage set at 1.5 kV and source temperature at 275°C. Further details are available upon request.
All MS and MS-MS raw spectra were processed using Proteome Discoverer 1.1 (PD1.1; Thermo-Fisher Scientific, San Jose, CA), and the spectra were searched using in-house licensed Mascot Daemon (version 2.3; Matrix Science, Boston, MA) against NCBInr, taxonomy: Drosophila melanogaster and Xenopus laevis. The database searches were performed with one missed cleavage site by trypsin allowed. The peptide tolerance was set to 10 ppm, and the MS-MS tolerance was set to 0.8 Da for collision-induced dissociation (CID) and 0.05 Da for higher-energy collision-induced dissociation (HCD). Only peptides defined by Mascot probability analysis (www.matrixscience.com/help/scoring_help.html#PBM) as scoring significantly greater than “identity” were considered for the peptide identification and phosphorylation site determinations. All MS-MS spectra for identified phosphorylation peptides were manually inspected, validated, and quantified using PD1.1 and Xcalibur 2.1 software. For certain phosphosites, the relative abundances of the phosphopeptide and unphosphorylated peptide ions at one or two different charged states (doubly charged and triply charged ions) were estimated based on the peak areas of the extracted ion profiles from the nanoLC–MS-MS analysis.
The protein kinase A (PKA) catalytic subunit (PDB ID 3DND) structure was used as the template for modeling; the sequence alignment between Gwl and PKA was determined with the ClustalW2 program (19). The homology model for the kinase domain of Gwl was built using the computational Modeller 9V4 program (http://salilab.org/modeller/9v4/release.html).
Gwl activation during M phase involves its hyperphosphorylation, affecting Gwl's electrophoretic mobility (Fig. 1 and and2)2) and leading to the labeling of more than 20 tryptic phosphopeptides (data not shown). One or more phosphosites in M phase Gwl constitute epitopes for the MPM-2 monoclonal antibody (Fig. 2A). This antibody reacts with many mitotic phosphoproteins, often at motifs targeted by cyclin-dependent kinases (CDKs) that contain phosphoserine or phosphothreonine followed by proline (42).
We have reconstituted a significant fraction of the Gwl activation process in vitro by treating interphase Gwl (made from baculovirus constructs in Sf9 cultured cells) with CDKs (Fig. 3) (also see references 40 and 45). Consistent with the general lack of CDK specificity in vitro (15), Cdk1-cyclin B (MPF), Cdk2-cyclin E, and Cdk2-cyclin A are equally efficient in activating Gwl (Fig. 3A and data not shown). CDK-treated interphase Gwl attains at most only ~25% of the kinase activity of M phase Gwl (made in Sf9 cells arrested with the phosphatase inhibitor okadaic acid [OA]; see references 45 and 47). However, following the prescient suggestion of a reviewer, we added the CDK-associated protein Cks2 (Suc1/p9) to the in vitro activation reaction mixture. This factor was previously reported to promote CDK phosphorylation of other G2/M regulators, such as Cdc25 (29). The addition of Cks2 enhanced CDK activation of Gwl by ~4 times, producing an enzyme with activity more comparable to that of Gwl made in the presence of OA (Fig. 3C and D). Gwl activation may therefore not require upstream kinases other than CDKs. Indeed, in our hands, the M phase kinases Plx1, Aurora A, and Nek2 do not contribute to Gwl activation alone or in synergism with CDKs (Fig. 3B); the same is true of PKA and p42 mitogen-activated protein kinase (MAPK) (data not shown).
Gwl undergoes autophosphorylation; that is, active preparations of wild-type Gwl (either endogenous protein purified from M phase extracts or recombinant enzyme made in OA-treated Sf9 cells) become labeled by radioactive ATP. This labeling represents autophosphorylation rather than the action of a contaminating kinase because kinase dead (KD) Gwl does not incorporate 32P (45) (Fig. 3A and see Fig. 4A) and because a linear correlation exists between the phosphorylation of mutant forms of Gwl and their activities on exogenous substrates (see below). At least 15 tryptic peptides become labeled by Gwl autophosphorylation (data not shown); as explained below, most of these sites are not critical to Gwl activation, but at least one of the sites does play a key role.
To determine whether Gwl autophosphorylation is intra- or intermolecular, we performed two experiments in which a small amount of active, WT Gwl was added to an excess of inactive Gwl (Fig. 4). 32P was incorporated almost exclusively into active Gwl, showing that the autophosphorylation is primarily intramolecular. 32P labeling of active Gwl displays first-order kinetics typical of unimolecular reactions (Fig. 1C and D), supporting the intramolecular autophosphorylation model.
Gwl's highly conserved kinase domain is split by a poorly conserved ~500-amino-acid insertion (residues 223 to 717 in Xenopus Gwl) (Fig. 5 and and6).6). To establish whether this nonconserved middle region (NCMR) is critical for Gwl activity, we made truncated proteins, including Gwl(Δ241-549), Gwl(Δ400-699), and Gwl(Δ300-650), that collectively remove almost all of the NCMR (Fig. 7A and data not shown). The truncated proteins, as well as WT and KD Gwl, were expressed in Sf9 cells that were either arrested in M phase by OA or untreated and thus mostly in interphase.
The purified proteins were adjusted to the same molar concentration and assayed for kinase activity (Fig. 7B). The results in Fig. 1C show the assay's dynamic range. The model substrates myelin basic protein (MBP) and endosulfine yielded similar results. The interphase enzymes displayed little or no kinase activity. Surprisingly, when expressed in OA-treated cells, Gwl(Δ241-549), Gwl(Δ400-699), and Gwl(Δ300-650), as well as proteins with smaller NCMR deletions, exhibit wild-type levels of autophosphorylation and kinase activity (Fig. 7B and C and data not shown; summarized in Fig. 5). Apparently, no specific sequence in the NCMR has an essential role.
We also tested the biological function of the truncated Gwl proteins in Xenopus egg extracts. Depletion of endogenous Gwl from CSF extracts leads to a loss of mitotic status that can be prevented by adding active exogenous Gwl (45). The results in Fig. 1A show the dynamic range of this functional assay. All deletion mutant Gwl proteins just described (made in OA-treated cells) can maintain Gwl-depleted CSF extracts in M phase (Fig. 7D), verifying their biological activity.
Another indication that the NCMR's particular amino acid sequence is nonessential is the finding that Drosophila Gwl, which has no sequence homology with frog Gwl in the NCMR, can substitute for the Xenopus enzyme in CSF extracts (data not shown). Fly and frog Gwl are thus functionally orthologous, a conclusion earlier implied by phenotypic similarities (44, 45, 47). Drosophila Gwl's autophosphorylation is less pronounced than that of Xenopus Gwl (Fig. 2B), suggesting that some autophosphorylation sites in the frog enzyme are nonessential.
The NCMR is still likely to be important for Gwl regulation, even if particular NCMR sequences are not. For example, Gwl(Δ241-699), in which almost the entire NCMR is removed (Fig. 5), is inactive in both the kinase and extract assays (Fig. 7B to D). Interestingly, however, interphase Gwl with half the NCMR deleted is more easily activated in extracts than the WT protein (Fig. 1B, right). We suggest possible explanations for these paradoxical results in Discussion.
To investigate how phosphorylation mitotically activates Gwl, we mutated individually all conserved S and T residues outside the NCMR to nonphosphorylatable A. We altered selected MS-validated sites within the NCMR to serve as controls, but we did not comprehensively analyze all NCMR phosphosites because proteins collectively lacking almost the whole NCMR remain functional. Gwl's mitotic regulation must therefore rely on phosphorylations elsewhere, either within the kinase domain or in the short conserved regions adjacent to the kinase domain (blue in Fig. 5) that distinguish Gwl from other AGC kinases.
We tested 33 sites by preparing mutant proteins from OA-treated Sf9 cells and analyzing their kinase activities in vitro (Fig. 8A to D). The Gwl mutants with mutations S39A, S89A, S98A, S101A, S119A, T193A, T206A, S212A, T748A, T868A, and S883A exhibited significantly reduced autophosphorylation and kinase activities toward endosulfine or MBP substrates. Among this set of 33 mutants, a linear correlation exists between the phosphorylation of exogenous substrates and of Gwl itself (r = 0.892, P < 1 × 10−6), verifying that Gwl labeling represents autophosphorylation rather than a contaminating kinase (Fig. 8D). The regression line's slope is less than 1, probably because a few mutations, such as S98A and S119A, disrupt substrate recognition but not intrinsic enzyme activity.
We also tested the mutant proteins for biological activity in extracts (Fig. 8E and data not shown). The results of these functional analyses were in general consistent with those of the in vitro kinase assays. With the exception of the protein with the S119A mutation, all of the mutant proteins with disrupted kinase activity were also unable to sustain M phase in Gwl-depleted CSF extracts. S119A is probably a borderline case in which the mutant enzyme's reduced activity is still sufficient to cross the threshold needed to sustain M phase in the extracts, between 25 and 50% of the activity of endogenous WT Gwl (Fig. 1A).
We also altered 9 potentially important sites to D or E to mimic phosphorylated S or T, respectively. When expressed in OA-treated cells, most of these mutants had reduced activity in vitro, with the exception of the hyperactive S101D (Fig. 9A and B). Importantly, however, proteins containing the T193E, T206E, and S883D mutations, which are phosphomimetic mutations of the sites most critical to Gwl regulation (see below), are functional in extracts (Fig. 9C) and have only mild reductions in kinase activity in vitro (Fig. 9A and B). The inactivity of the corresponding original T193A, T206A, and S883A mutants thus probably reflects the inability of these sites to be phosphorylated rather than specific structural requirements for T or S at these positions.
Because one purpose of making the phosphomimetic mutants was the generation of constitutively active Gwl, we also expressed combinations of these mutations in Sf9 cells not treated with OA. None of the many variations we tried exhibited full constitutive activity. However, the interphase S101D T193E double mutant has about 10 to 15% of M phase WT Gwl's kinase activity (normalized for protein amount), even though this double mutant remains hypophosphorylated (Fig. 9D). The S101D T193E mutant made in the absence of OA cannot rescue the M phase status of Gwl-depleted CSF extracts except in considerable excess (Fig. 9E). The limited constitutive activity of this mutant enzyme is stimulated in vitro by CDK treatment (data not shown); CDK targeting of sites other than S101 and T193 is thus needed for full Gwl activation.
We conducted several MS experiments to identify the M phase phosphosites in frog and fly Gwl. In some experiments, endogenous mitotic Gwl was immunoprecipitated from Xenopus CSF extracts; in others, recombinant Gwl was prepared by expression in OA-treated Sf9 cells. Finally, because CDKs can activate Gwl in vitro (Fig. 3), recombinant Gwl expressed in Sf9 cells in the absence of OA (and thus lacking M phase phosphorylations) was treated with MPF or Cdk2/cyclin A.
Gwl prepared by these methods (Fig. 2C) was digested with trypsin or, to increase coverage, with chymotrypsin or endopeptidase GluC. MS was used to find phosphorylated peptides (Fig. 6). Peptides representing 89% of Gwl were seen at least once, although individual experiments covered only 45 to 80% of the protein. MS identified 23 phosphoamino acids in fly Gwl and >50 phosphoamino acids in frog Gwl, of which ~25 were seen consistently. These numbers roughly correspond with the number of phosphopeptides seen by two-dimensional tryptic mapping (not shown), indicating that MS accounts for most M phase phosphorylation sites. With the exception of phosphotyrosine at positions 727 and 751 in Xenopus Gwl, all of the phosphorylated sites are S or T. More than half are proline directed, that is, S/T P motifs probably targeted by CDKs. Except for T725, all of the phosphosites we identified in conserved parts of Gwl (that is, outside the NCMR) were previously documented as M phase-specific sites in previous studies of human cells (7, 8, 28) (results summarized in Fig. 6). Phosphorylations of Y727 or Y751 are unlikely to play critical roles in Gwl regulation, because fly Gwl has F at the Y727 cognate position, while all AGC kinases have Y at the 751 cognate position.
Only a small subset of MS-detected phosphorylations intersects with the critical phosphorylation sites identified by the mutagenesis experiments described above; these are phosphorylated T193 (pT193), pT206, pS212, pT748, and pS883 in Xenopus Gwl. We estimated the phosphate occupancy of these sites by comparing the peak areas for all phosphorylated and nonphosphorylated forms of the corresponding peptides (Table 1). The phosphate occupancy of all of these sites is higher in active frog Gwl from OA-treated cells than in inactive Gwl from untreated cells. T193 and S883 are likely nearly stoichiometrically occupied by phosphate in active samples; the results presented in Table 1 probably underestimate occupancies because MS detection efficiencies for phosphopeptides are often >10 times lower than for nonphosphorylated forms (12). The phosphate occupancy of pT206 also appears to be relatively high (>25%), while those for pS212 and pT748 are more modest. Similar results were observed for occupancy of the corresponding sites in Drosophila Gwl made in OA-treated cells (Table 1).
T193 and T206 are proline-directed sites that are probably recognized by CDKs. Indeed, MPF can target both sites in vitro when either a fragment of Gwl containing these sites (Fig. 10A) or full-length inactive Gwl (Table 1) is used as a substrate. A fragment containing T193 but with the T206A mutation is a better substrate than a fragment with T193A and T206, indicating that T193 is more efficiently targeted by MPF (Fig. 10A).
S883 is not proline directed, and a fragment including this site is not phosphorylated in vitro by MPF (Fig. 10A). Indeed, the region surrounding S883 does not conform to the consensus sequence for any known kinase. We thus tested the idea that Gwl itself can target S883 through autophosphorylation. As shown in Table 1, pS883 is produced when inactive WT frog Gwl (made in the absence of OA) is partially activated in vitro with MPF but not when KD Gwl is used instead. Furthermore, WT Gwl (but not KD Gwl, Plx1, or Cdk1/cyclin A) can utilize a short peptide containing S883 as an exogenous substrate (Fig. 10B), and MS analysis demonstrates that pS883 is the only site Gwl phosphorylates in this peptide (data not shown). The S883-containing peptide is phosphorylated ~200-fold less efficiently than a similarly sized peptide containing the Gwl target site in endosulfine (data not shown), consistent with experiments on other kinases using peptides containing autophosphorylation sites as exogenous substrates (11).
The common core mechanism of AGC kinase activation involves phosphorylation at three conserved sites: in the activation loop, in the hydrophobic motif (HF) at the end of the C-terminal tail, and in the middle of the tail (called the turn motif in PKA or the tail/Z site in growth factor-stimulated AGC kinases) (reviewed in references 14, 27, and 30) (see Fig. 11). Activation loop phosphorylation (often by phosphoinositide-dependent protein kinase 1 [PDK1]) increases phosphoryl transfer catalysis and, in some cases, substrate access to the active-site cleft (1). HF phosphorylations (added either autocatalytically or by exogenous kinases such as mTOR or phosphomimicked by acidic amino acids) promote the enzyme's active conformation. The phosphorylated HF becomes anchored to the “HF pocket” in the kinase's N-terminal lobe (reviewed in references 27 and 30) (also see Fig. 12). This binding stabilizes the α-C helix that forms one side of the pocket and, thus, enhances phosphotransferase activity. Phosphorylations of the activation loop and the HF have synergistic effects because they promote mutually stabilizing interactions on the α-C helix (14).
The third phosphorylation (added autocatalytically or by other kinases), near the middle of the C-terminal tail, helps the tail to loop around the kinase core. This third phosphorylation can occur at either of two adjacent sites (the tail/Z site and the turn motif) that promote looping by distinct mechanisms. In growth factor-stimulated AGC kinases, phosphorylation of the tail/Z site allows it to bind to a positively charged patch on the N-terminal lobe. In PKA, turn motif phosphorylation causes a bend in the tail due to interactions of this pS338 with nearby residues (Fig. 11 and and12).12). The looping facilitated by both mechanisms allows the phosphorylated HF to locate its binding pocket in the N-terminal lobe (14, 17).
After the manuscript's original submission, another paper addressing Gwl regulation appeared in print (40). The authors concluded that phosphorylation of a single site in human Gwl (corresponding to S883 in Xenopus) is essential for activation, that this site is the direct target of both MPF and Plk1, and that this phosphosite plays a role in Gwl analogous to that of the tail/Z site in other AGC kinases. We agree that S883 is critical and probably functions similarly to the tail/Z site, but otherwise we view Gwl regulation differently. If S883 was the sole essential phosphosite, then Gwl with a S883D or S883E phosphomimetic substitution should exhibit constitutive activity when made either in interphase cells in the absence of OA or in immature Xenopus oocytes. However, mutant enzymes synthesized by either method are inactive (Fig. 9D and E and and13),13), even though the same proteins acquire biologically relevant kinase activity in OA-treated cells (Fig. 9A to C) or in eggs induced to mature by progesterone (Fig. 13). Other events in addition to S883 phosphorylation must therefore occur to activate Gwl during M phase.
In the model we describe below and depict in Fig. 14, Gwl activation is initiated by CDK phosphorylation of T193 and T206, which act as activation loop phosphorylations. In contrast with the other report (40), we do not think that either MPF or Plk1 directly targets S883; instead, CDK-primed Gwl autophosphorylates S883.
Although Gwl is phosphorylated during M phase at many sites, our results focus attention on five candidates: pT193, pT206, pS212, pT748, and pS883 (Xenopus numbering). These residues are located in evolutionarily conserved regions characteristic of Gwl but not of other AGC kinases (blue in Fig. 5). In fact, these regions discriminate Gwl from the closely related Rim15p kinase in S. cerevisiae, which has the same substrate specificity (it targets a yeast homolog of endosulfine) but is regulated by a different mechanism (38). Mutation of any of these five sites to A severely compromises Gwl function (Fig. 8). MS has detected all five phosphosites in D. melanogaster, X. laevis, and Homo sapiens Gwl (Fig. 6) (7, 8, 28, 40).
The data implicating pT193, pT206, and pS883 are particularly strong. Enzymes made in OA-treated cells with the phosphomimetic mutations T193E, T206E, and S883D rescue Gwl activity in extracts (Fig. 9C), arguing that inactivity of the corresponding nonphosphorylatable mutants does not reflect structural requirements for T or S in those positions. The phosphate occupancies at T193, T206, and S883 are very high in active WT frog or fly Gwl (Table 1). Finally, the equivalents of pT193 and pT206 are known to be highly enriched in human M phase cells (8), while the human equivalent of pS883 is highly correlated with Gwl activity, as shown by a phosphospecific antibody (40).
The roles of pS212 and pT748 are less clear. Because the phosphomimetics S212D and T748E display little or no function (Fig. 9A to C), the inactivity of S212A and T748A (Fig. 8) might be explained by intolerance of Gwl's structure for amino acid substitutions at these positions. pS212 is difficult to quantify because it is found in a multiply phosphorylated peptide, while the phosphate occupancy of pT748 in active Gwl is only about 2% (Table 1).
Our results demonstrate the importance of phosphorylations at T193, T206, and S883 for Gwl activation. However, because we have yet to find a combination of phosphomimetic mutations with high constitutive activity, other sites probably play subsidiary roles. As just mentioned, pS212 and pT748 are potential contributors. Other conserved sites (S215, T221, T725, T729, and S886) are consistently phosphorylated during M phase (Fig. 6). Although S/T-to-A mutations of these latter sites have no obvious consequences, the functional tests in Fig. 8 may have missed small effects of specific mutations. Several nonconserved residues within the NCMR are also phosphorylated during M phase in fly, frog, or human Gwls. None of these phosphosites is needed individually, but they could be redundant such that Gwl activity might require multiple NCMR phosphorylations.
Finally, alanine mutations at S39, S89, S98, S101, and T868 strongly disrupt Gwl function (Fig. 8), although evidence for the phosphorylation of these sites is lacking. The absence of such evidence is particularly surprising for S101, because the S101D mutation confers limited constitutive activity to Gwl, particularly in the double mutant S101D T193E (Fig. 9D and E). S101D thus appears to stabilize a partially active conformation of Gwl for unknown structural reasons that may or may not reflect actual S101 phosphorylation.
The remainder of this discussion focuses on T193, T206, and S883 phosphorylations as key regulatory steps in Gwl activation, but it must be remembered that these events are unlikely to represent the entire story.
We propose that CDKs prime Gwl at the critical sites T193 and T206, consistent with our findings that (i) Gwl is specifically activated during M phase, (ii) the CDK inhibitors roscovitine and p21Cip prevent Gwl activation in frog oocytes and egg extracts (47), (iii) CDKs potentiate interphase Gwl's enzymatic activity in vitro (Fig. 3) (see also references 40 and 45), (iv) both T193 and T206 are evolutionarily conserved S/T P motifs (the minimal CDK consensus), and (v) MPF phosphorylates both T193 and T206 in vitro (Fig. 10A and Table 1). It is highly likely that MPF is the kinase targeting T193 and T206 during M phase proper, but a different CDK such as Cdk1-cyclin A or a related enzyme such as Cdk1-speedy/ringo (6, 33) could target these sites during M phase entry prior to full MPF mobilization. In this way, Gwl could act not only downstream of MPF (by preventing the premature dephosphorylation of MPF effectors), but also upstream as part of the MPF triggering mechanism (by helping control MPF regulators such as Cdc25 phosphatase and Myt1/Wee1 kinases) (reviewed in reference 31).
S883 is the other site critical for Gwl activation identified in our studies. We disagree with the recent proposal that S883 is directly targeted by CDKs (40). In contrast with almost all known CDK substrates, S883 is not followed by proline, and we have failed to observe CDK phosphorylation of a fragment or peptide containing S883 (Fig. 10A and B). We believe instead that pS883 results from Gwl autophosphorylation primed by prior CDK phosphorylation at T193 and T206. This model explains why CDKs cause the appearance of pS883 when added to purified WT but not KD interphase Xenopus Gwl in vitro (Table 1). pS883 autophosphorylation also explains the general correlation seen between this site's phosphorylation and Gwl activity (40). Direct evidence for this hypothesis is the ability of active but not inactive Gwl to phosphorylate a peptide containing S883 (Fig. 10B). Although the effect is clear, this peptide is only a poor exogenous substrate for Gwl, consistent with the finding that Gwl autophosphorylation is normally intramolecular (Fig. 4). Several precedents exist for autophosphorylation of S883-analogous sites. In PKA and some PKC isotypes, the turn motif/tail/Z site is autophosphorylated intramolecularly (3, 16, 27, 32); in fact, bacterially synthesized PKA autophosphorylates the turn motif (S338) to a phosphate occupancy of >98% (36). Interestingly, the sequences preceding pS883 in Gwl (880LKVpS883) and pS338 of PKA (335IRVpS338) are quite similar.
CDKs, together with Cks2 (Suc1/p9), produce substantial activation of interphase Gwl (Fig. 3). Suc1 is needed for M phase entry of egg extracts, a requirement ascribed to Suc1's ability to enhance CDK phosphorylation of G2/M regulators, such as Cdc25, Myt1, and Wee1 (29). Our results indicate that Gwl should be added to this target list. At present, we see no compelling reason to invoke roles for upstream kinases other than CDKs in Gwl activation, although we cannot exclude the possibility. For example, in our hands, abrogation of the function of several kinases with cell cycle involvement, including Mos, MEK, MAPK, CaMKII, and p90Rsk, does not interfere with Gwl activation (data not shown) (47). Studies of drug-treated Xenopus oocytes suggest that neither phosphatidylinositol 3-kinase (PI-3K), Jun N-terminal protein kinase (JNK), mTOR1, S6K, nor PKBβ is required to turn Gwl on during maturation (24, 35). The results shown in Fig. 3B provide no evidence for participation of the mitotic kinases Plx1, Aurora, or Nek2; similar experiments (not shown) also appear to exclude PKA and p42 MAPK.
Vigneron et al. have recently reported activation of human Gwl with the Polo-like kinase Plk1 and suggest that Plk1 (as well as CDKs) can directly target the human equivalent of S883 (40). We regard the idea that Polo-like kinases are critical Gwl activators as unsettled. First, S883 is unlikely to be directly phosphorylated by Plk1. S883 does not reside in an obvious Plk1 consensus in frog or human Gwl (Plk1's selectivity for E/N/D/Q at the −2 position relative to the S/T is very strong ), nor does Plx1 target a fragment or peptide containing Gwl S883 in vitro (Fig. 10). Second, we have repeatedly failed to activate interphase frog Gwl with any of several preparations of human Plk1 or frog Plx1. One potential explanation for the differing results is that Plk1 phosphorylates human Gwl in vitro at several locations within the NCMR that are not found in frog Gwl and that have never been detected as phosphosites even in M phase human Gwl (39). Plk1 targeting of these nonconserved sites might partially activate human Gwl in vitro yet not reflect a fundamental aspect of Gwl regulation in vivo. Third, hypermorphic mutations of gwl in Drosophila are dominant enhancers of polo loss-of-function mutations (2). The direction of this genetic interaction is exactly the opposite of expectations if Polo were a Gwl activator. Finally, in starfish, Gwl can be activated during oocyte maturation in the absence of Polo activity (18).
Our current working model (Fig. 14) has strong precedents in PKA's activation mechanism. The initial step is activation loop phosphorylation by an exogenous kinase: PDK1 for PKA (30) and MPF for Gwl. The pT193 and pT206 Gwl presumptive activation loop phosphorylations are in fact found in positions analogous to the pT197 activation loop site in PKA (Fig. 11). Primed Gwl can now intramolecularly autophosphorylate pS883, which is located in the same relative position in the C-terminal tail as PKA's turn motif (S338) (Fig. 11 and and12).12). In PKA, S338 is autophosphorylated intramolecularly to >98% occupancy (36). Our depiction of Gwl pS883 phosphorylation in Fig. 14 is based on structural analyses indicating that PKA's tail is flexible enough that the turn motif can access the active site within the same molecule (32).
Other phosphorylations, particularly within the NCMR, probably play subsidiary or redundant functions in Gwl activation. One possibility, depicted in Fig. 14, is that the NCMR in interphase Gwl blocks substrates from the active site while M phase NCMR phosphorylations relieve this inhibition. A smaller NCMR could require fewer phosphorylations, explaining the easier activation of interphase Gwl with a deletion of part of the NCMR relative to the WT enzyme (Fig. 1B). However, the evidence for such function of the NCMR is at present only fragmentary. Several NCMR phosphorylations observed by MS (Fig. 6) are proline directed and could result from CDK action, but others are not proline directed and could be the targets of exogenous kinases or, perhaps more likely, autophosphorylation.
The steps subsequent to the phosphorylations involved in Gwl activation raise a curious paradox. Gwl's N-terminal lobe appears to contain a “pocket” like those that interact with the phosphorylated hydrophobic motif (HF) of other AGC kinases (Fig. 12). Gwl shares the residues within such pockets that make the strongest contacts with the HF, and mutations of the corresponding amino acids in human Gwl to A abrogate enzyme function (40). The paradox is that Gwl's C-terminal tail does not appear to be long enough to wrap around the N-terminal lobe and contact the HF binding pocket in the same polypeptide (Fig. 12).
Gwl's C terminus has very weak similarities with classical HF motifs, so we have entertained the notion that Gwl might multimerize during M phase, with the pS883 phosphorylated tail of one subunit inserting into the HF binding pocket of another. However, we have to date seen no evidence for Gwl multimerization (data not shown). Vigneron et al. have proposed a provocative alternative explanation for Gwl's retention of an HF binding pocket but lack of an HF: perhaps other AGC kinases can help activate Gwl by inserting their phosphorylated HFs into Gwl's pocket (40). We regard this hypothesis as very attractive, but we have not been able to verify that HF motif phosphopeptides elevate Gwl function.
What does the pS883 phosphorylation in Gwl's tail do if it does not act as an HF? Most likely, pS883 plays a role similar to that of the tail/Z site found in the analogous position of other AGC kinases, ~10 to 20 residues upstream of the C-terminal HF (Fig. 11 and and12).12). As mentioned previously, the phosphorylated tail/Z site interacts with a patch of basic amino acids on the N-terminal lobe, helping to deliver the HF to the HF binding pocket (14, 30). Gwl has 3 or 4 of the N-terminal-lobe amino acids defining the basic patch (R40, K45, K64, and possibly N104 in frog Gwl ), and mutations in these basic residues disrupt Gwl function (40). However, the putative contact between pS883 and the basic patch (depicted in Fig. 14) could not accelerate binding of the HF to the pocket as in other AGC kinases, because Gwl's tail is too short. Instead, the pS883/basic patch interaction would need to promote Gwl activation in some other fashion not easily discerned from known AGC structures. Gwl's activation thus still presents puzzles and paradoxes worthy of future research.
James Maller (University of Colorado School of Medicine, Aurora, CO) has been remarkably generous with reagents, advice, laboratory space, and comments on the manuscript. We also thank J. Kuang (M. D. Anderson Cancer Center, Houston, TX) for the gift of MPM-2 antibody.
This study was supported by NIH grant GM048430 to M.L.G.
Published ahead of print 21 February 2012