Substrate Specificity and Activity Regulation of Protein Kinase MELK*

Maternal embryonic leucine zipper kinase (MELK) is a protein Ser/Thr kinase that has been implicated in stem cell renewal, cell cycle progression, and pre-mRNA splicing, but its substrates and regulation are not yet known. We show here that MELK has a rather broad substrate specificity and does not appear to require a specific sequence surrounding its (auto)phosphorylation sites. We have mapped no less than 16 autophosphorylation sites including serines, threonines, and a tyrosine residue and show that the phosphorylation of Thr167 and Ser171 is required for the activation of MELK. The expression of MELK activity also requires reducing agents such as dithiothreitol or reduced glutathione. Furthermore, we show that MELK is a Ca2+-binding protein and is inhibited by physiological Ca2+ concentrations. The smallest MELK fragment that was still catalytically active comprises the N-terminal catalytic domain and the flanking ubiquitin-associated domain. A C-terminal fragment of MELK functions as an autoinhibitory domain. Our data show that the activity of MELK is regulated in a complex manner and offer new perspectives for the further elucidation of its biological function.

domain adjacent to the catalytic domain and a C-terminal kinase-associated 1 (KA1) domain. The function of the latter domains are poorly understood. UBA domains are known to bind (poly)ubiquitin and have been suggested to thereby prevent additional ubiquitination and proteosomal degradation of the target protein (9 -12). In some proteins, UBA domains appear to function as dimerization domains (13). Between the UBA and the KA1 domains, MELK contains a TP dipeptide-rich domain that is phosphorylated in mitotically arrested cells and mediates binding to the transcription and splicing factor NIPP1 (14).
Preliminary evidence implicates MELK in various cellular processes. MELK binds tightly to the zinc finger-like protein ZPR9 and causes its nuclear accumulation (15,16). In the nucleus, ZPR9 itself interacts with the transcription factor B-Myb, a regulator of cell proliferation and differentiation, and enhances its transcriptional activity. Another interactor of MELK is the transcription and splicing factor NIPP1, but the binding of NIPP1 requires the phosphorylation of MELK on a specific threonine in its TP dipeptide-rich domain (14). Because wild-type MELK, but not a NIPP1-binding mutant, is a potent inhibitor of pre-mRNA splicing in nuclear extracts and because the MELK-NIPP1 interaction is increased during mitosis, it has been proposed that MELK contributes to the ending of pre-mRNA splicing just before mitosis (14). A third protein ligand of MELK is Cdc25B, a protein-tyrosine phosphatase that triggers mitosis by the activation of protein kinase Cdk1. Davezac et al. (17) reported that the ectopic expression of MELK induces an accumulation of cells in G 2 and that this effect was counteracted by the overexpression of Cdc25B. Intriguingly, MELK is expressed at high levels in both embryonic (18,19) and neural stem cells (20,21), indicating that it may also play a role in stem cell functions of multipotency and self-renewal. Finally, MELK possibly contributes to oncogenesis because its expression is increased in tumor-derived progenitor cells (22) and in cancers of nondifferentiated cells (23).
The MELK ligands ZPR9 (15), NIPP1, 6 and Cdc25B (17) are also MELK substrates, but the significance of their phosphorylation is not clear. A major limitation in studying the role of MELK as a protein kinase is that it is not known what controls its activity and that MELK expressed in mammalian cells seems to be inactive (5). This has prompted us to examine what determines the activity and substrate specificity of MELK, expressed in either bacteria or mammalian cells. We show here that MELK has a rather broad substrate specificity and that its activity is complexly regulated by autophosphorylation, autoinhibition, Ca 2ϩ ions, and reducing agents.

EXPERIMENTAL PROCEDURES
Materials-Mouse monoclonal anti-FLAG antibodies were obtained from Stratagene. Anti-phosphothreonine antibodies were obtained from Zymogen, and anti-phosphotyrosine antibodies were from Santa Cruz Biotechnology. Reduced and oxidized glutathione and PHOS-Select TM gel were purchased from Sigma. 45 CaCl 2 (2.2 mCi/ml, 134 g of Ca 2ϩ /ml) was obtained from Amersham Biosciences.
Preparation of Recombinant MELK (Mutants)-Wild-type MELK and the indicated MELK mutants and fragments were cloned in the pET16b vector, in-frame with the polyhistidine tag. The His-tagged proteins were purified on Ni 2ϩ -Sepharose TM 6 Fast Flow (Amersham Biosciences). The constructs encoding FLAG-tagged MELK (fragments) have been described previously (14). Point mutations were made according to the QuikChange site-directed mutagenesis protocol of Stratagene, using the appropriate primers and templates. The sequences of the DNA constructs were verified by DNA sequencing.
Cell Cultures and Immunoprecipitations-COS-1 and HEK293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum. 48 h after transfection, the cells were washed twice with ice-cold phosphate-buffered saline and lysed in 50 mM Tris at pH 7.5, 0.3 M NaCl, 0.5% (v/v) Triton X-100, 0.5 mM phenylmethanesulfonyl fluoride, 0.5 mM benzamidine, and 5 M leupeptin. After sonication, the cell lysates were cleared by centrifugation (10 min at 10,000 ϫ g), and the supernatants were used for immunoprecipitations. The cleared cell lysates were incubated with anti-FLAG antibodies coupled to protein G-Sepharose (Amersham Biosciences) for 3 h at 10°C. After one wash with Tris-buffered saline supplemented with 0.2 M LiCl and three washes with Tris-buffered saline plus 0.1% (v/v) Nonidet P-40, the beads were resuspended in 25 mM Tris at pH 7.5 and used for kinase assays and immunoblotting with anti-FLAG antibodies.
Protein Kinase Assays-The kinase activities of MELK (mutants) and fragments were determined with the indicated substrates for 1 h at 30°C in a buffer containing 25 mM Tris at pH 7.5, 0.1 mM [␥-32 P]ATP, and 2 mM magnesium acetate. Reactions with peptide substrates were stopped by transfer of the assay mixture to P81 papers and washing with 75 mM orthophosphoric acid. Reactions with myelin basic protein (MBP) as substrate were run on SDS-PAGE and analyzed by autoradiography.
Peptide Chip-The Trial PepChip kinase array (Pepscan Systems B.V.) contains 2 ϫ 192 peptides, divided in 2 ϫ 4 subarrays of 6 ϫ 4 spots. The incubation was done as described in the manual. Briefly, the PepChip kinase slide was incubated for 2 h at 30°C in the following reaction mixture: 50 mM Hepes at pH 7.4, 20 mM MgCl 2 , 0.02 mg/ml bovine serum albumin, 0.01% Brij-35, 20 mM DTT, 10 M [␥-33 P]ATP, and 0.5 M MELK. After the incubation, the chip was washed once with phosphate-buffered saline supplemented with 1% (v/v) Triton X-100, washed twice with 2 M NaCl plus 1% (v/v) Triton X-100 and then rinsed twice with Milli Q water before drying. The dry PepChip kinase array was analyzed on a phosphorimaging device (Storm 640; Molecular Dynamics). 45 Ca 2ϩ Overlay and Ca 2ϩ Buffer-The indicated amounts of MELK or NIPP1 were spotted on a polyvinylidene difluoride membrane (Hybond-P; Amersham Biosciences), and the 45 Ca 2ϩ binding was carried out as described previously (24). The free Ca 2ϩ concentration in bath solutions was calculated with the CaBuf program (Droogmans G., KULeuven), taking into consideration the values of pH, temperature, ionic strength, association constants, and Mg 2ϩ concentration.
Mapping of Autophosphorylation Sites of MELK-Phospho-amino acid analysis by thin-layer chromatography was done as described previously (25). MELK-(1-340) was incubated for 3 h at 30°C in a buffer containing 25 mM Tris at pH 7.5, 0.1 mM [␥-32 P]ATP, and 2 mM magnesium acetate to allow complete autophosphorylation.
The autophosphorylation sites of 32 P-labeled MELK-(1-340) were determined by phosphopeptide sequencing as well as by mass spectrometry. For this purpose, 32 P-labeled MELK-(1-340) was first subjected to a reduction for 45 min at 60°C with 10 mM DTT and then to an alkylation for 45 min at room temperature in the dark with 35 mM iodoacetamide. The remaining iodoacetamide was neutralized by an incubation with 15 mM DTT for 45 min at room temperature in the dark. Subsequently, the proteins were precipitated with 10% trichloroacetic acid, and the aceton-washed pellet was digested overnight with 5 g/l trypsin in 200 mM ammonium bicarbonate plus 0.1% RapiGest (Waters). The resulting peptides were separated on a RPC C2/C18 SC2.1/10 column (SMART System; Amersham Biosciences), equilibrated in 0.1% trifluoroacetic acid, and eluted with a linear gradient of 0 -70% acetonitrile in 0.1% trifluoroacetic acid. The radioactive fractions were either analyzed by a 492 Procise (Applied Biosystems) amino acid sequencer, operated in the pulsed liquid mode, or by electrospray mass spectrometry on an Applied Biosystems API 3000 tandem mass  DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 spectrometer. The latter was equipped with a Protana nanospray source. Identification of the phosphopeptides and determination of the phosphorylation sites were performed using the neutral loss scan mode and the product ion scan mode, respectively.

Substrate Specificity and Regulation of MELK
The autophosphorylation sites of full-length MELK-(1-651) were also determined by mass spectrometry, using a different protocol. Wildtype MELK was digested in-gel with trypsin (5 g/ml) in ammonium bicarbonate plus 0.1% N-octyl-glucoside as described previously (5). Digests were diluted to 0.1 ml with PHOS-Select TM wash/bind buffer (0.25 M acetic acid, 30% acetonitrile) and supplemented with 30 l of a 10% (v/v) PHOS-Select TM gel, equilibrated in the wash/bind buffer. After an incubation for 20 min, the beads were captured in an Eppendorf gel loader tip to form a mini-column. The beads were washed three times with 30 l wash/bind buffer and eluted with two times 25 l of 0.4 M ammonium hydroxide. The eluate was dried under vacuum and resuspended in 5% formic acid for analysis by matrix-assisted laser desorption ionization time-of-flight mass spectrometry on an Applied Biosystems 4700 Proteomics Analyser using ␣-cyano-4-hydroxy cinnamic acid (5 mg/ml ϩ 10 mM ammonium phosphate in 50% acetonitrile, 0.1% trifluoroacetic acid) as matrix. Electrospray liquid chromatography-mass spectrometry was performed on a Dionex Ultimate capillary high pressure liquid chromatography system coupled to an Applied Biosystems 4000 Q TRAP mass spectrometer. The peptides were separated on a PepMap C 18 column equilibrated in 0.1% formic acid/water and developed with a discontinuous acetonitrile gradient at 350 l/min.

MELK Has a Broad Substrate
Specificity-Consistent with a previous report by Lizcano et al. (5), bacterially expressed human MELK was spontaneously active and phosphorylated the AMARA peptide, a classical substrate of the subfamily of AMPK-related protein kinases (Fig.  1A). Another widely used substrate for these kinases, the SAMS peptide, was an even better in vitro substrate of MELK. Surprisingly, MELK also phosphorylated a broad range of structurally unrelated proteins, such as MBP, Histone H1, and the splicing factors CDC5L, NIPP1, and SAP155 (not shown). To obtain some insights into the substrate determinants of MELK, we screened a peptide chip for candidate substrates (Fig. 1B). The used chip contained two identical sets of 192 peptides that all comprise established in vivo phosphorylation sites. Numerous peptides on this chip were phosphorylated by MELK. The three best MELK substrates were peptides derived from casein, eukaryotic initiation factor 2␣, and lamin B1. These peptides did not show any structural similarity and were also very different from the AMARA and SAMS peptides (Fig.  1C) and from the sequences surrounding the established MELK phosphorylation sites of ZPR9, CDC25B, and NIPP1 (not shown). Using phospho-epitope-specific antibodies, we have confirmed that Ser 51 of full-length, recombinant eukaryotic initiation factor 2␣ is phosphorylated by MELK (not shown). Collectively, our data indicate that MELK has a rather broad substrate specificity in vitro.
MELK Contains an Autoinhibitory Domain-The catalytic domain of MELK is located in its N terminus. The C-terminal 60% of MELK comprises consecutively a UBA domain, a TP dipeptide-rich domain, and a KA1 domain ( Fig. 2A). To examine whether the latter domains affect the catalytic activity of MELK, we generated various MELK mutants (Fig. 2B). Deletion of the KA1 domain or the KA1 plus TP-rich domains increased the K cat /K m ratio 2-4-fold, using either the SAMS peptide or MBP as substrates (Fig. 2C). However, the additional deletion of the UBA domain, as in MELK-(1-266), resulted in an inactive enzyme, indicating that the UBA domain is essential for the expression of catalytic activity. Consistent with this notion, we found that the deletion of the UBA domain from full-length MELK also resulted in an inactive kinase. Likewise, MELK was completely inactive following the fortuitous mutation of the three first residues (D283K/D284K/D285K) of the UBA domain.
Our observation that the N-terminal half of MELK is more active than the full-length protein (Fig. 2) suggested that the C-terminal half of MELK functions as an autoinhibitory domain. Accordingly, we found that the catalytic activity of full-length MELK (not shown) or MELK-(1-340) was inhibited by the addition of MELK-(326 -651) (Fig. 3A). With MBP as substrate, a nearly complete inhibition was obtained with 15 M MELK-(326 -651) (Fig. 3A). Using smaller MELK fragments, we could delineate an inhibitory fragment to residues 326 -530, which largely corresponds to the TP-rich domain ( Fig. 2A). MELK-(602-651), essentially comprising the KA1 domain, was not inhibitory. The latter finding was unexpected because the deletion of the KA1 domain increased the catalytic efficiency of MELK (Fig. 2C). Lineweaver-Burk plots showed that the inhibition of MELK-(1-340) by MELK-(326 -601) was largely accounted for by a decreased V max (Fig. 3B), consistent with the increased K cat for MELK-(1-340) as compared with the K cat of wild-type MELK (Fig. 2C). The inhibition of MELK-(1-340) by MELK-(326 -651), MELK-(326 -601), and MELK-(326 -530) was also detected when the SAMS peptide was used as substrate, but the maximal extent of inhibition was less pronounced with this substrate and amounted to only about 65% (not shown).
Autophosphorylation of MELK-Using immunoblotting with phospho-epitope-specific antibodies and phospho-amino acid analysis by thin-layer chromatography, we found that bacterially expressed MELK was phosphorylated on serine, threonine, and tyrosine residues and that the level of phosphorylation was further augmented by incubation of the purified enzyme with MgATP (not shown). The phosphorylation on Ser and Thr could be reversed by incubation with the catalytic subunit of protein phosphatase-1, whereas the phosphorylation on tyrosine was reversed by incubation with a protein-tyrosine phosphatase from Yersinia enterocolitica (not shown). The phosphorylation of MELK is likely to be an autocatalytic process because the inactive variant MELK-D150A (14), which is mutated in the essential DFG triplet of kinase subdomain VII, was not phosphorylated at all.
We have initially used mass spectrometry to map the autophosphorylation sites of bacterially expressed wild-type MELK and MELK- . This resulted in the identification of 14 autophosphorylation sites, distributed all over the polypeptide chain, with the exclusion of the UBA and KA1 domains (TABLE ONE). Surprisingly, two independent mass spectrometric analyses did not identify tyrosine-phosphorylated peptides, which was in contrast with the data of immunoblot analysis. Because MELK contains a tyrosine in its catalytic loop (Tyr 163 ) that is not conserved in the catalytic loop of the other members of the subfam-ily of AMPK-activated protein kinases (Fig. 4A) and because MELK is also the only member of this subfamily that is known to be autophosphorylated on tyrosine, we have examined whether it is perhaps Tyr 163 that is autophosphorylated by MELK. In accordance with this view, we found that the autophosphorylation on tyrosine was largely lost in MELK-Y163F, despite the fact that this mutant was still catalytically active toward an exogenous substrate (Fig. 4B). We have also used N-terminal sequencing by Edman degradation to map autophosphorylation sites in the catalytic loop. A tryptic peptide that was derived from 32 P-labeled MELK-(1-340), was phosphorylated on both Thr 167 and Ser 171 (not shown). The latter site was not identified by mass spectrometry, possibly because Ser 171 was phosphorylated substoichiometrically (see "Discussion"). In conclusion, the total number of autophosphorylation sites that we mapped amounted to 16 (Fig. 4C).
MELK did not phosphorylate the SAMS peptide when the phosphorylatable residue (Ser 79 ) was replaced by a tyrosine (not shown), indicating that MELK is not a true dual specificity protein kinase and can only phosphorylate tyrosine(s) autocatalytically. Also, the autophosphorylation rate of MELK was independent of its dilution (not shown), suggesting that the autophosphorylation represents an intramolecular rather than an intermolecular process. In further agreement with this view, we found that wild-type MELK neither phosphorylated the recombinant (inactive) catalytic domain of MELK nor a synthetic peptide (residues 159 -173) comprising the catalytic loop of MELK and including the autophosphorylation sites Tyr 163 , Thr 167 , and Ser 171 (not shown).
We have subsequently explored the role of autophosphorylation of the catalytic domain of MELK by site-directed mutagenesis (Fig. 5). MELK-(1-340)-T167A was inactive, strongly suggesting that autophosphorylation on this site is essential for activity. However, MELK-(1-340)-T167D was active, indicating that this replacement mimicked the phosphorylation of Thr 167 . Interestingly, both MELK-(1-340)-S171A and MELK-(1-340)-S171D were inactive, suggesting that phosphoryl-  ation of Ser 171 is essential for activity but is not mimicked by an acidic residue. Mutation of Thr 56 or Ser 253 into either an aspartic acid or an alanine did not measurably affect the activity of MELK-(1-340). Mutation of Tyr 163 into either a phenylalanine or an aspartic acid residue also had no effect on the activity of MELK- (1-340).
The Expression of MELK Activity Requires Reducing Agents-The ability of MELK to autophosphorylate and to phosphorylate an exogenous substrate was completely dependent on the presence of reducing agents (Fig. 6A). A maximal stimulation of the protein kinase activity was obtained with 5-10 mM DTT. GSH was a less potent activator than was DTT, and GSSG did not activate MELK at all. The requirement for reducing agents suggested that MELK is inactivated by the covalent modification of one or several cysteines. The involved cysteine(s) are at least partially located in the catalytic and/or UBA domains because MELK-(1-340) was also fully DTT-dependent (Fig. 6B). Accordingly, the combined mutation of all nine cysteines yielded a MELK-(1-340) variant that no longer required reducing agents to be active (Fig. 6B). However, MELK remained DTT-dependent following the separate mutation (C29V, C70V, C89A, C154A, C168A, C169A, C204A, C286A, or C339A) of each of the nine cysteines that are present in the catalytic and UBA domains (not shown), indicating that MELK-(1-340) contains at least two cysteines that can be modified and thereby cause a complete inactivation of the kinase. Gel filtration chromatography revealed that MELK is a monomer, both in the absence and presence of DTT, indi-  cating that its activation by reducing agents is not the result of the disruption of one or more interchain disulfide bonds (not illustrated).
MELK Is Inhibited by the Binding of Ca 2ϩ -Following the serendipitous observation that MELK was more active in the presence of the Ca 2ϩ chelator EGTA, we have explored whether the activity of MELK is perhaps controlled by Ca 2ϩ . Using a Ca 2ϩ -EGTA buffer to generate various concentrations of free Ca 2ϩ , we found that the kinase activities of MELK-(1-651) and MELK-(1-340) were inhibited in a dosedependent manner by Ca 2ϩ . Using MBP as substrate, we detected a nearly complete inhibition at free Ca 2ϩ concentrations of around 1 M (Fig. 7A). This roughly corresponds to the Ca 2ϩ concentration that is detected following the stimulation of cells with a maximally effective concentration of a Ca 2ϩ agonist. Interestingly, higher Ca 2ϩ concentrations were somewhat less inhibitory. We obtained similar data with the SAMS peptide as kinase substrate, except that the maximal inhibition, at 1 M Ca 2ϩ , amounted to only around 65% (not illustrated). Also, supraphysiological free Ca 2ϩ concentrations (2 M) were not inhibitory with the peptide substrate.
The above data prompted us to examine whether MELK is a Ca 2ϩbinding protein. Using 45 Ca 2ϩ overlays, we indeed found that both fulllength MELK and MELK-(1-340) and MELK-(1-260) bind Ca 2ϩ . In contrast, a similar concentration of the unrelated protein NIPP1 did not bind Ca 2ϩ (Fig. 7B).

MELK Expressed in Mammalian Cells Has Similar Properties-
The previous data were all obtained with bacterially expressed MELK constructs. To verify whether MELK has similar properties when expressed in mammalian cells, we transfected HEK293T cells and COS-1 cells (not shown) with an expression vector for MELK or MELK mutants with an N-terminal FLAG tag. The fusion proteins were immunoprecipitated from the cell lysates with anti-FLAG antibodies. In Fig. 8 it is shown that, similar to what we observed for bacterially expressed MELK, FLAG-MELK expressed in HEK293T cells was completely dependent on DTT for activity. Also, the SAMS kinase activity of FLAG-MELK was inhibited about 60% by 1 M Ca 2ϩ but not by higher concentrations of Ca 2ϩ . Furthermore, the activity of FLAG-MELK was not affected by mutation of Thr 167 into an aspartic acid but was lost when Thr 167 was replaced by an alanine. Collectively, these data indicate that MELK expressed in mammalian cells has similar properties as bacterially expressed MELK.

DISCUSSION
Substrate Specificity of MELK-Most protein kinases have a rather restricted substrate specificity that is largely determined by the nature of the residues that surround their phosphorylation sites (26). However, MELK phosphorylated an unusually large fraction of the tested (poly)peptides, and the sequences flanking the phosphorylation sites did not conform to a consensus site (Fig. 1). Neither could a consensus sequence be delineated for the 16 mapped autophosphorylation sites (TABLE ONE). Our data therefore suggest that MELK has a rather broad substrate specificity in vitro. This does not necessarily imply, however, that MELK has also numerous substrates in vivo. For example, it cannot be excluded that MELK interacts in vivo with proteins that function as substrate specifiers, similarly to the well established regulation of protein phosphatase-1 (27).
Functions of the Noncatalytic Domains-The minimal MELK fragment that was catalytically active consists of the catalytic domain and the UBA domain (Fig. 2). Also, point mutations within the UBA domain of full-length MELK yielded a completely inactive enzyme. These findings were unexpected because UBA domains have thus far only been implicated in (poly)ubiquitin binding (9 -12) and in the dimerization of proteins (13). Our data represent the first reported evidence that a UBA domain is required for the expression of catalytic activity by a protein kinase. Because UBA domains are also present in other members of the AMPK-related kinases, e.g. the MARKs (6), it will be important to examine whether they fulfill the same essential function in these kinases as well. The UBA domain is a small domain of about 40 residues that folds into a characteristic three-helix bundle (11). It is difficult to predict exactly how this domain contributes to the expression of kinase activity. The UBA domain does not appear to be involved in the dimerization of MELK because gel filtration studies indicated that MELK is a monomer  (not shown). It is possible that the UBA domain has a structural function and is important for the correct folding of the catalytic domain.
The C-terminal half of MELK, comprising the TP dipeptide-rich and KA1 domains, clearly functions as an autoinhibitory domain (Figs. 2 and  3). Using smaller fragments we could map the major inhibitory region to the TP dipeptide-rich domain. The KA1 domain alone was not inhibitory, but the TP dipeptide-rich domain was also a less potent inhibitor than the entire C-terminal half of the kinase, suggesting that the KA1 domain also contributes to autoinhibition in intact MELK. This conclusion is in accordance with a recent report on the yeast AMPK-related family members Kin1 and Kin2, which also have a C-terminal KA1 domain that was found to be autoinhibitory and to interact with the N terminus of the kinases (28).
Regulation by (Auto)Phosphorylation-MELK is a unique member of the AMPK-related protein kinases in that it does not require an upstream protein kinase for activation and is fully activated by autophosphorylation (Ref. 5 and this work). By various experimental approaches we have identified no less than 16 autophosphorylation sites, i.e. five in the catalytic domain and nine in the TP-rich domain and its flanking sequences. Surprisingly, two of these phosphorylation sites (Tyr 163 and Ser 171 ) were not identified by mass spectrometry, possibly because MELK was phosphorylated substoichiometrically on these sites. Other groups have also noted that the detection of all phosphorylation sites in a protein remains a most challenging analytical task, even when sophisticated techniques of mass spectrometry are adopted (29,30). In any case, our findings nicely illustrate the importance of using multiple, independent approaches for the mapping of phosphorylation sites.
Site-directed mutagenesis indicated that Thr 167 and Ser 171 , located between the DFG and APE motifs in the activation loop or T-loop, need to be autophosphorylated for MELK to be active as a protein kinase (Fig.  5). These sites are conserved in all other AMPK-related protein kinases (Fig. 4A), and the site corresponding to Thr 167 has been shown to be phosphorylated by protein kinase LKB1 (5). It is not yet known whether the site corresponding to Ser 171 is also phosphorylated in other AMPKrelated protein kinases. Intriguingly, the phosphorylation of Ser 171 could not be mimicked by an acidic residue.
MELK also autophosphorylates its T-loop on Tyr163 (Fig. 4B), a site that is not conserved in other AMPK-related protein kinases (Fig. 4A). In this respect, MELK is similar to the dual specificity tyrosine phosphorylation-regulated protein kinases (DYRKs), because they also autophosphorylate their activation loop on a tyrosine but phosphorylate their substrates on serine and threonine (31). It has recently been demonstrated that the tyrosine autophosphorylation by DYRKs is mediated by a translational intermediate and represents an essential step in the maturation of the kinase (31). Once DYRKs are fully translated and released from the ribosome, the transitional tyrosine kinase activity is lost, and they function as protein Ser/Thr kinases. This mechanism most likely does not apply to MELK, however, because autophosphorylation on Tyr 163 is not essential for its activation (Fig. 5), and mature, bacterially expressed MELK also autophosphorylates on Tyr (not illustrated).
We have not yet explored the role of autophosphorylation of nine residues in the C-terminal, autoinhibitory domain (Fig. 4C). An enticing hypothesis is that these autophosphorylations decrease the inhibitory potency of this domain and thereby contribute to the activation of the kinase. It should be pointed out, however, that the autoinhibitory domain also contains numerous consensus phosphorylation sites for other protein kinases (14). For example, the TP-rich domain harbors 10 consensus phosphorylation sites for proline-directed protein kinases and the phosphorylation of one of these sites, Thr 478 , in mitotically arrested cells mediates the recruitment of the transcription and splicing factor NIPP1. It has not yet been investigated whether the binding of NIPP1 affects the activity of MELK.
Regulation by Reducing Agents-In addition to its regulation by autoinhibition and autophoshorylation, the N-terminal half of MELK is controlled by oxidation of at least two cysteines (Fig. 6). Importantly, the dependence on reducing agents also applies to MELK that is expressed in mammalian cells and is probably the major reason for previous failures to detect kinase activities associated with MELK in mammalian cell lysates. There are many types of cysteine modification (32)(33)(34), and we currently do not know which one pertains to MELK.
Regulation by Ca 2ϩ -Finally, MELK is inhibited by free Ca 2ϩ concentrations that are detected after the addition of optimal concentrations of a Ca 2ϩ agonist (Fig. 7). Remarkably, surpraphysiological Ca 2ϩ concentrations were less inhibitory, which is possibly accounted for by the existence of a second, lower affinity binding site for Ca 2ϩ that opposes the effect of the binding of Ca 2ϩ to the inhibitory, high affinity binding site.
MELK is the only member of the AMPK-related kinases that is known to be directly regulated by Ca 2ϩ . Overlay assays suggested that Ca 2ϩ binds to the catalytic domain of MELK (Fig. 7B), but its primary structure does not contain an established Ca 2ϩ -binding motif. Therefore, additional work is needed to map the Ca 2ϩ -binding motif of MELK in more detail.
In conclusion, our initial failure to detect a MELK-associated kinase activity has led us to examine in some detail the determinants for the activity of MELK as a protein kinase. As a result of these studies we have shown here that MELK is complexly regulated by autoinhibition, (auto)phosphorylation, reducing agents, and free Ca 2ϩ . These data will be helpful in studying the physiological regulation and functions of MELK.