Mechanism of Ca2+-mediated Regulation of NDR Protein Kinase through Autophosphorylation and Phosphorylation by an Upstream Kinase*

NDR1 (nuclearDbf2-related) is a serine/threonine protein kinase belonging to subfamily of kinases implicated in the regulation of cell division and morphology. Previously, we demonstrated that the activity of NDR1 is controlled by phosphorylation of two regulatory residues, Ser-281 and Thr-444. Moreover, we found that NDR1 becomes activated through a direct interaction with EF-hand Ca2+-binding proteins of the S100 family. In this work, we characterize this regulatory mechanism in detail. We found that NDR1 autophosphorylates in vitro predominantly on Ser-281 and to a lesser extent on Thr-74 and Thr-444. All of these residues proved to be crucial also for NDR1 activity in vivo; however, in contrast to Ser-281 and Thr-444, Thr-74 seems to be involved only in binding to S100B rather than directly regulating NDR1 activity per se. When we added Ca2+/S100B, we observed an increased autophosphorylation on Ser-281 and Thr-444, resulting in stimulation of NDR1 activity in vitro. Using phosphospecific antibodies, we found that Ser-281 also becomes autophosphorylated in vivo, whereas Thr-444 is targeted predominantly by an as yet unidentified upstream kinase. Significantly, the Ca2+-chelating agent BAPTA-AM suppressed the activity and phosphorylation of NDR1 on both Ser-281 and Thr-444, and specifically, these effects were reversed when we added the sarcoplasmic-endoplasmic reticulum Ca2+ ATPase pump inhibitor thapsigargin.

NDR1 (nuclear Dbf2-related) is a conserved and widely expressed nuclear serine/threonine kinase that belongs to a recently identified subfamily of kinases that play a crucial role in cell division and cell morphogenesis. We originally cloned this kinase from a human fetal brain cDNA library using Caenorhabditis elegans expressed sequence tag clone cm11b8 (1). Later, it was mapped to chromosome 6p21 next to the major histocompatibility complex class I gene cluster (2). Recently, we isolated a second NDR isoform, termed NDR2, which is local-ized on chromosome 12p11 next to the K-ras gene. 1 The two isoforms display an identity of more than 87% at the protein and 77% at the DNA level. In addition to mammalian NDR1 and NDR2, the NDR family of protein kinases comprises orthologous proteins Tricornered (trc gene) from Drosophila melanogaster; Sax-1 from C. elegans; and several closely related kinases such as Warts/Lats kinases from mammals, D. melanogaster, and C. elegans; Cbk1, Dbf2, and Dbf20 from Saccharomyces cerevisiae; Orb6 from Schizosaccharomyces pombe; Ukc1 from Ustilago maydis; and Cot-1 from Neurospora crassa (for review, see Ref. 3). These kinases share 40 -60% amino acid identity within their catalytic domains as well as conserved regions in their regulatory domains. The structural similarity within this family suggests that these kinases might perform related functions even in evolutionary distant organisms.
In fact, recent reports confirmed the involvement of this group of kinases in various aspects of regulation of cell division and cell morphogenesis. For instance, mutations in trc result in splitting of surface projections such as epidermal hairs, shafts of sensory bristles, larval denticles, and the lateral branches of arista (4). The sax-1 mutants display expanded cell bodies and ectopic neurites in several classes of neurons (5). Mutation in cot-1 results in excessive numbers of branched hyphal tips at restrictive temperature, but these tips fail to elongate (6). The cbk1 mutants show profound defects in cell morphogenesis, including changes from ellipsoidal to round morphology, random budding patterns, and abnormal mating projections (7). Likewise, disruption of ukc1 leads to a change of cell shape from elongated to rounded form, to formation of hyphal extensions, and prevents cells from forming filamentous colonies (8). The orb6 mutants lose growth polarity and display altered, spherical morphology with disorganized microtubuli and actin filaments; and moreover, they enter mitosis prematurely (9). On the other hand, cells carrying temperature-sensitive alleles of dbf2 arrest at the end of anaphase with uniform, large budded "dumbbell" morphology (10). Taken together, the functional and structural relatedness within this group of kinases suggests that they may be regulated by similar mechanisms, including, e.g. stimulatory phosphorylation and association with activating proteins, and they may also partially share the same substrates.
The NDR group kinases contain all 12 subdomains of the kinase catalytic domain described by Hanks et al. (11). Regarding their kinase domain sequence, they are most closely related to the members of AGC 2 family of protein kinases comprising * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  PKA, 3 PKB, PKCs, PRK, p70 S6K , p90 RSK , and PDK1. A unique feature of NDR group kinases is an insert of about 30 amino acids between subdomains VII and VIII, which, in the case of NDR1, accommodates a nonconsensus nuclear localization signal (1). All members of this group also contain a poorly characterized N-terminal regulatory domain of different lengths. In NDR1, this domain consists of 87 amino acids, and it encompasses a region that is predicted to form an amphiphilic ␣-helix responsible for binding to S100B (12).
In our previous work, we described the basic mechanisms of activation of NDR1 protein kinase (13). We found that NDR1 is potently activated by treatment of cells with the protein phosphatase 2A inhibitor okadaic acid (OA), indicating involvement of phosphorylation of serine and/or threonine residues in the regulation of NDR1 activity. Indeed, we demonstrated that activation of NDR1 involves phosphorylation of two regulatory residues, Ser-281 of the activation segment and Thr-444 from the hydrophobic motif located in the C-terminal region (13). Moreover, we described a mechanism of in vitro activation of NDR1 through a direct interaction with EF-hand Ca 2ϩ -binding protein S100B, which augments NDR1 autophosphorylation (12). Notably, NDR1 also interacts with S100B in vivo, its activity is rapidly stimulated by treatment with the Ca 2ϩ ionophore A23187, and this activation is dependent upon the Nterminally located S100B binding domain. Intriguingly, we observed that overexpression of S100B in several melanoma cell lines leads to hyperactivation of NDR1 and that NDR1 activity can be inhibited in those cells by W7, a cell-permeable inhibitor of CaM and S100 proteins. Altogether, these results point to the involvement of Ca 2ϩ signaling in regulation of NDR1.
In this study, we have examined the mechanism of Ca 2ϩ / S100B-induced autophosphorylation and activation of NDR1 by means of electrospray ionization mass spectrometry (ESI-MS) and phosphospecific antibodies raised against the regulatory phosphorylation sites Ser-281 and Thr-444. We also investigated the mechanism of NDR1 phosphorylation in vivo as well as the dependence of NDR1 phosphorylation events on intracellular Ca 2ϩ . Our findings delineate a Ca 2ϩ -dependent mechanism for activation of NDR1, involving S100B-mediated stimulation of autophosphorylation on Ser-281 and phosphorylation of Thr-444 by an unknown Ca 2ϩ -dependent upstream kinase.

EXPERIMENTAL PROCEDURES
Cell Culture-COS-1 cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Cells were transfected at the subconfluent stage with FuGENE 6 transfection reagent (Roche Molecular Biochemicals) according to the manufacturer's instructions. In some experiments, the cells were treated for 60 min with 1 M OA in 0.1% N,N-dimethylformamide (Alexis Corp.), 50 M BAPTA-AM in 0.1% dimethyl sulfoxide (Sigma), or 20 M thapsigargin in 0.1% dimethyl sulfoxide (Alomone) 24 h after transfection.
Antibodies-Anti-Ser-281P and anti-Thr-444P rabbit polyclonal antisera were raised against the synthetic peptides NRRQLAFS(PO 4 )T-VGTPD for the Ser-281 phosphorylation site and KDWVFINYT-(PO 4 )YKRFEG for the Thr-444 phosphorylation site (Neosystem, Strasbourg, France). The peptides had been conjugated to keyhole limpet hemocyanin. Rabbit injections and bleed collection were done by Strategic Biosolutions. The anti-Ser-281P antiserum was used without further purification, but the anti-Thr-444P antibody was purified on protein A-Sepharose (Amersham Biosciences) followed by antigenic peptide coupled to cyanogen bromide-activated Sepharose (Amersham Biosciences). Antibodies were eluted with 0.1 M glycine, pH 2.5. The 12CA5 HA monoclonal antibody hybridoma supernatant was used for immunodetection and immunoprecipitation of HA-NDR1 variants. A rabbit anti-NDR1_C-term polyclonal antiserum directed against a synthetic peptide TARGAIPSYMKAAK (corresponding to NDR1 amino acids 452-465) has been described previously (1). A rabbit polyclonal antiserum that recognizes S100B was raised against recombinant human S100B and was used without further purification (14).
Bacterial Expression of Human GST-fused NDR1-BL21-DE3 Escherichia coli strain (Novagen) was transformed with the pGEX-2T_NDR1 wild-type or mutant plasmids and the pRep4 plasmid bearing LacI q repressor (Qiagen). Mid-logarithmic phase cells were induced with 0.5 mM isopropyl ␤-D-thiogalactopyranoside for 4 h at 30°C. Bacteria were lysed by a French press in presence of 1 mg/ml lysozyme, and the fusion proteins were purified on glutathione-agarose (Amersham Biosciences) as described. Recombinant proteins were assayed for kinase activity as described below, and autophosphorylation was determined after SDS-PAGE separation either by Cerenkov counting or by exposure to a PhosphorImager screen followed by analysis with Image-Quant software (Molecular Dynamics).
GST-NDR1 Kinase Assay-1 g of purified recombinant GST-NDR1 wild-type and mutants (without further treatment or autophosphorylated for 2 h in the presence or absence of 1 mM CaCl 2 and 10 M bovine S100B (Sigma)) were assayed in a 20-l reaction containing 20 mM Tris-HCl pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, 100 M [␥-32 P]ATP (ϳ2,000 cpm/pmol; Amersham Biosciences), and 1 mM NDR1 substrate peptide (KKRNRRLSVA). After a 30-min incubation at 30°C, reactions were stopped with 50 mM EDTA, and 10-l reaction solutions were spotted onto 2-cm 2 squares of P-81 phosphocellulose paper (Whatman). These were subsequently washed 4 ϫ 5 min and 3 ϫ 20 min in 1% phosphoric acid and once in acetone before counting in a liquid scintillation counter. One unit of NDR1 activity was defined as the amount that catalyzed the phosphorylation of 1 nmol of peptide substrate in 1 min.
Western Blotting-To detect GST-NDR1 or HA-NDR1, samples were resolved by 10% SDS-PAGE and transferred to polyvinylidene difluoride membranes (Immobilon-P; Millipore). Membranes were blocked in TBST (50 mM Tris-HCl pH 7.5, 150 mM NaCl, and 0.05% Tween 20) containing 5% skimmed milk powder and were then probed for 2 h with 4 g/ml anti-NDR1_C-term or 1:100 12CA5 monoclonal antibody supernatant. Bound antibodies were detected with corresponding horseradish peroxidase-linked secondary antibodies and ECL (Amersham Biosciences). For detection of phosphorylated NDR1, either 1 g of GST-NDR1 or HA-NDR1 immunoprecipitated from 100 g of COS-1 detergent extracts as described earlier in this paper were separated on 10% SDS-PAGE and transferred to polyvinylidene difluoride membranes. Membranes were blocked in TBST containing 5% skimmed milk powder and incubated for 2 h at room temperature or overnight at 4°C with 1:1,000 anti-Ser-281P antiserum in the presence of 50 g/ml competing dephosphopeptide or 1:500 anti-Thr-444P purified antibody. Both antibodies were detected with horseradish peroxidase-conjugated donkey anti-rabbit Ig antibody (Amersham Biosciences) and ECL. To detect immunoprecipitated S100B, samples were separated on 18% SDS-PAGE and transferred to polyvinylidene difluoride membrane. The membrane was blocked for 2 h with 5% bovine serum albumin and 1% fetal calf serum in TBST and was then probed with anti-S100B antiserum diluted 1:1,000 in the same buffer. Bound antibody was then detected with corresponding horseradish peroxidase-linked secondary antibody in TBST and ECL.
Mass Spectrometry-10 g of untreated or autophosphorylated GST-NDR1 (2 h at 30°C, either in presence or absence of 1 mM CaCl 2 and 10 M bovine S100B homodimer) was separated by 10% SDS-PAGE, stained with Coomassie Blue R-350 (Sigma), and excised from the gel for further processing. Gel slices were sequentially washed 3 ϫ 8 min in acetonitrile and 2 ϫ 12 min in 25 mM NH 4 HCO 3 . Afterward, the gel slices were reduced for 1 h at 57°C in 10 mM dithiothreitol followed by alkylation for 45 min at room temperature in 55 mM iodoacetamide. Upon repeated washing in acetonitrile/NH 4 HCO 3 , gel-bound GST-NDR1 was digested with modified trypsin (sequencing grade; Promega) overnight at 37°C. The cleaved peptides were then extracted by two 15-min sonication steps in H 2 O:acetonitrile:formic acid (20:70:10) and H 2 O:methanol:formic acid (93:2:5). All experiments were performed on an API 300 triple-quadrupole mass spectrometer (PE-Sciex, Toronto, Canada) equipped with a NanoESI source (Protana, Odense, Denmark). For precursor ion scans of m/z Ϫ79, samples were injected by off-line nanoelectrospray according to Wilm and Mann (15). The instrument was operated in negative ion mode; a small percentage of ammonia was added in the spraying needle during sample preparation for better sensitivity in the detection of m/z Ϫ79. For LC-MS sequence analysis, phosphopeptides were further separated by high performance liquid chromatography (HPLC) interfaced with the API 300 mass spectrometer. The Rheos 4000 chromatograph was equipped with a 1 ϫ 250-mm Vydac C 8 column (Hesperia, Canada). The HPLC column was equilibrated in 95% solvent A (2% CH 3 CN, 0.05% trifluoroacetic acid in H 2 O), 5% solvent B (80% CH 3 CN, 0.045% trifluoroacetic acid in H 2 O), and a linear gradient was developed from 5 to 50% of solvent B in 60 min at a flow rate of 180 l/min After the column, the flow was split with a small percentage (about 5%) being directed to the mass spectrometer.

Mapping of in Vitro Autophosphorylation
Residues of NDR1-We observed that NDR kinase becomes potently autophosphorylated on serine and threonine residues in vitro, and that the autophosphorylation as well as NDR1 kinase activity markedly increase upon incubation of NDR1 with the EF-hand Ca 2ϩ -binding proteins of the S100 family, in particular S100B (1,12). These results prompted us to analyze the NDR1 autophosphorylation sites by means of nanospray ECI-MS/MS analysis. 10 g of purified GST-NDR1 was left untreated (Fig. 1A) or autophosphorylated either in the absence (Fig. 1B) or presence (Fig. 1C) of purified bovine S100B. After tryptic cleavage of the differently treated GST-NDR1 samples, resultant mixtures of peptides were introduced into a triple-quadrupole mass spectrometer and analyzed for phosphorylated peptides by precursor ion scanning for m/z Ϫ79 in negative ion mode. This procedure detects each peptide in the mixture which, upon fragmentation in the collision cell, liberates a species of m/z Ϫ79, representing a negative fragment ion PO 3 Ϫ characteristic of phosphorylated peptides (16,17).
When nontreated GST-NDR1 was analyzed by precursor ion scan of m/z Ϫ79, only weak phosphopeptide signals corresponding to peptides P2 and P4 could been detected (Fig. 1, A and D). With the GST-NDR1 samples autophosphorylated either in the absence or presence of Ca 2ϩ /S100B, however, the abundance of phosphorylated peptides increased markedly (Fig. 1, B and C). The observed m/z value of most of these signals could be assigned in both samples to four NDR1-derived phosphopeptides P1-P4 which, upon liberation of the phosphate group (79.97 Da), give rise to expected tryptic (or in one case chymotryptic) NDR1 dephosphopeptides (Fig. 1D). To confirm the identity of these peptides, we performed a LC-MS analysis of GST-NDR1 autophosphorylated with [␥-32 P]ATP (data not shown). After collection of the phosphopeptide-enriched radioactive fractions, we again analyzed the samples by ESI-MS using an m/z Ϫ79 precursor ion scanning method. Having found the same species as in the analysis of the unlabeled GST-NDR1, we subsequently performed a low energy MS/MS analysis in positive ion mode. We detected ion species corresponding to the phosphopeptides listed in Fig. 1D, including the phosphorylated residues, thereby confirming identity of the NDR1 in vitro autophosphorylation sites.
The four phosphopeptides derived from three regions of the NDR1 polypeptide. The first three phosphopeptides correspond to sites identified previously also in NDR1 activated in vivo by OA. P1 and P2 correspond to phosphopeptides encompassing Ser-281 of the activation segment and P3 represents a phosphopeptide containing the residue Thr-444 from hydrophobic motif located in the C-terminal region of NDR1. Significantly, the phosphopeptide P4 proved to be an as yet unidentified phosphoresidue Thr-74 from the N-terminal part of NDR1 which has been defined previously as an S100 binding region (residues 62-83; Ref. 12).
Reduction of Autophosphorylation Rate and Kinase Activity in GST Phosphorylation Site Mutants-To confirm the in vitro autophosphorylation of NDR1 on the residues Thr-74, Ser-281, and Thr-444 and to estimate the influence of these residues on the autophosphorylation-induced kinase activity of NDR1, we constructed a series of GST-NDR1 alanine mutants at these positions (as well as a GST-NDR1 kinase-dead alanine mutant in the catalytic site residue Lys-118). As shown in Fig. 2, B and C, wild-type GST-NDR1 became efficiently autophosphorylated in vitro (up to 0.4 pmol of phosphorus/pmol of GST-NDR1). The addition of S100B homodimers brought about an approximate 2-fold increase in the autophosphorylation rate accompanied by about a 5-fold increase in kinase activity. Because the kinasedead mutant did not show any detectable autophosphorylation, the NDR1 kinase activity must account for the observed effects. As shown in Fig. 2, A and B, the S281A mutant displayed a markedly reduced autophosphorylation rate (irrespective of the addition of S100B), thereby confirming this residue as the major in vitro autophosphorylation site of NDR1. As expected, the kinase activity of S281A decreased similarly and was almost undetectable (Fig. 2C). Although not comparable with S281A, the T444A point mutation also led to a significant decrease of both autophosphorylation rate and kinase activity, indicating that this residue is also susceptible to autophosphorylation in vitro. The smaller effects of T444A, compared with the S281A mutation, may be readily explained by the unfavorable, hydrophobic peptide sequence upstream of this residue with only two basic amino acids located at positions Ϫ8 and Ϫ10 (Table I). Finally, the T74A mutant displayed only a negligible reduction in autophosphorylation rate and kinase activity in absence of Ca 2ϩ /S100B. Although the sequence Nterminal to this site in fact matches the basic NDR1 consensus sequence fairly well, according to the quantitative analyses by LC-MS, the phosphorylation at this site always accounted for less than 5% of the total phosphate incorporation (data not shown). After the addition of Ca 2ϩ /S100B, however, the impact of this point mutation approached that of T444A, and the stimulatory effect of Ca 2ϩ /S100B on autophosphorylation rate and kinase activity was diminished strongly compared with the other GST-NDR1 variants. These facts indicate that Thr-74 represents only a minor autophosphorylation site, and the low responsiveness to the addition of Ca 2ϩ /S100B most likely reflects the inability of T74A mutant to interact with this protein.
Mechanism of NDR1 Activation by Ca 2ϩ /S100B-To examine the effect of S100B on NDR1 phosphorylation status, we raised rabbit polyclonal antibodies directed against phosphoepitopes of the two known activation phosphorylation sites Ser-281 and Thr-444. As shown in Fig. 2D, these antibodies were specific for the phosphorylated NDR1 because corresponding phosphorylation point mutants S281A, T444A, and kinase-dead were not recognized. Using these reagents, we found that Ca 2ϩ /S100B enhances phosphorylation on both Ser-281 and Thr-444 (similar to the general increase in autophosphorylation as presented in Fig. 2, A and B) without an apparent preference for one of the residues (Fig. 2D, middle and  bottom panels). This synergy between the phosphorylation of both activation segment and hydrophobic motif sites may provide a simple explanation for why NDR1 activity increases more robustly than does the overall autophosphorylation after the addition of Ca 2ϩ /S100B. Nevertheless, it is striking that although the S281A mutation entirely abolishes the autophosphorylation on Thr-444, the T444A mutant displayed an almost normal autophosphorylation on Ser-281. Apart from the fact that Thr-444 only was autophosphorylated to a minor extent, these data indicate that the impact of the activation segment phosphoresidue on the NDR1 kinase activity was higher than the impact of the hydrophobic motif residue or that these residues are phosphorylated in a sequential manner as is known for most of other AGC family protein kinases (18). Both alternatives may apply for an intramolecular mechanism of the NDR1 autophosphorylation as defined previously (1). Finally,

L-R-R-S-A-H-A-R-K-E-T-E-F
Ser-181 sequence

K-N-K-D-W-V-F-I-N-Y-T-Y-K
a The NDR1 consensus sequence has been identified by testing a peptide library as described previously (12). b X stands for any amino acid. Phosphate acceptor residues are in bold. Single-letter codes used for amino acids. the T74A point mutation abolished the S100B-mediated increase of phosphorylation on both Ser-281 and Thr-444 residues without affecting the intrinsic NDR1 autophosphorylation, again stressing the crucial role of this residue in the interaction with S100B.
Ser-281 Is the Major Autophosphorylation Residue, whereas Thr-444 Is Targeted by an Upstream Kinase in Vivo-To evaluate the relative influence of the individual phosphorylation sites on kinase activity of NDR1 in vivo, we constructed a series of mammalian expression vectors with HA-tagged NDR1 wildtype, kinase-dead, and alanine mutants in the three identified phosphorylation sites. As shown in Fig. 3A, point mutations in the two established in vivo regulatory phosphorylation sites Ser-281 and Thr-444 led to an expected decline of NDR1 kinase activity in both untreated and 1 M OA-treated COS-1 cells (again with some preference for the S281A mutant). In contrast to the GST-NDR1 kinase assays, however, we also observed a drastic reduction of HA-NDR1 kinase activity after mutating Thr-74. This indicates that the Thr-74-dependent, S100B-mediated autophosphorylation may play a crucial role in vivo.
To examine this possibility, we analyzed extracts from transfected COS-1 cells again with the anti-Ser-281P and anti-Thr-444P antisera. As shown in Fig. 3B, the Ser-281 residue was constitutively phosphorylated in vivo (and this phosphorylation was enhanced further after treatment with 1 M OA), whereas the Thr-444 site only became modified after OA treatment. Intriguingly, we found that the phosphorylation of Ser-281 depends entirely on the activity of NDR1 because the kinasedead K118A mutant did not display any phosphorylation on this position. This fact implies that Ser-281 is an autophosphorylation residue also in vivo. Significantly, autophosphorylation accounts for the major portion of NDR1 phosphorylation in vivo because wild-type NDR1 incorporated markedly higher amounts of radioactive orthophosphate than its kinase-dead counterpart in metabolically labeled COS-1 cells upon treatment with OA (data not shown). Furthermore, mutation of Thr-444 did not impair the phosphorylation of Ser-281, which again points to the low importance of this residue for NDR1 autophosphorylation (Fig. 3B, middle panel). For the phosphorylation of Thr-444, however, we found that this residue became modified also after abolishing the kinase activity of NDR1 in kinase-dead mutant K118A (Fig. 3B, bottom panel). This fact strongly suggests that, at least in OA-treated cells, this site must be targeted by an as yet unidentified upstream kinase. Finally, mutation of residue Thr-74 led to a marked reduction of both Ser-281 and Thr-444 phosphorylation (both with and without OA treatment), which explains the somewhat unexpected kinase activity data observed for this mutant. This implies that Thr-74 is not only involved in the S100B-mediated autophosphorylation of NDR1, but it probably participates also in targeting the upstream kinase to NDR1.
Phosphorylation on both Ser-281 and Thr-444 Occurs in a Ca 2ϩ -dependent Manner-Because our current and previous work indicated that Ca 2ϩ /S100B is essential both for NDR1 activity in vivo and for an efficient NDR1 autophosphorylation on Ser-281 and Thr-444 in vitro (Fig. 2D) and because the S100B-NDR1 interaction is known to be fully Ca 2ϩ -dependent (12), we sought to examine the role of intracellular Ca 2ϩ in the activation of NDR1. For this purpose, we treated transfected COS-1 cells with the membrane-permeable agent BAPTA-AM, which is freely taken up into cells, where it is hydrolyzed by cytosolic esterases and trapped intracellularly as active, membrane-impermeable Ca 2ϩ chelator BAPTA (19). As shown in Fig. 4A, 50 M BAPTA-AM dramatically reduced the OA-stimulated NDR1 activity almost to the basal activity level of the untreated cells. Likewise, examination of the phosphorylation status of NDR1 demonstrated that phosphorylation on both Ser-281 and Thr-444 declined almost to the base line (Fig. 4B). Notably, both NDR1 activity and the phosphorylation of Ser-281 and Thr-444 were rescued by coincubation of BAPTA-AM with 20 M thapsigargin, a sesquiterpene lactone capable of increasing cytoplasmic Ca 2ϩ by inhibition of the sarcoplasmicendoplasmic reticulum Ca 2ϩ ATPase pumps, causing liberation of intracellular Ca 2ϩ stores (20). These results confirm the Ca 2ϩ specificity of the observed BAPTA-AM effects and combined with the experiments showed in Fig. 3, allow us to conclude that NDR1 is regulated by a Ca 2ϩ -dependent, most likely S100B-mediated autophosphorylation on Ser-281 and by phosphorylation by an as yet unidentified Ca 2ϩ -dependent upstream kinase on Thr-444.
Thr-74 Is Required for the Association of HA-NDR1 with Ca 2ϩ /S100B-We have demonstrated previously that NDR1 forms functional complexes with S100B in vivo and that this interaction depends on an intact N-terminal domain of NDR1. Now, we asked whether this association depends on phosphorylation status of NDR1 (12). Therefore, we transfected COS-1 cells with S100B and NDR1 wild-type or alanine-mutant expression plasmids and monitored the NDR1-S100B interaction by coimmunoprecipitation of NDR1 with S100B. As shown in Fig. 5, we found that the formation of NDR1⅐S100B complexes was independent of NDR1 kinase activity or the phosphorylation status of the regulatory residues Ser-281 and Thr-444, and thus the NDR1-S100B interaction appears to be constitutive. However, mutation of Thr-74 (the minor in vitro autophosphorylation site) markedly impaired interaction between these two proteins, and therefore this residue seems to be essential for NDR1 to undergo the interaction with Ca 2ϩ /S100B. In the context of the results gained with the recombinant GST-NDR1 (Fig. 2) and the fact that Thr-74 is a part of a putative, S100binding ␣-helix formed by amino acids 65-81, it would not be surprising if Thr-74 was one of the crucial residues forming the contact interface between NDR1 and S100B. We are currently addressing these questions by generating a phosphoepitopespecific antibody against this residue, and the results will be presented elsewhere. DISCUSSION We found that NDR1 autophosphorylates on three residues in vitro. The first of them, Thr-74, is located in the N-terminal S100B binding domain of NDR1. The second, major site Ser-281, which is well conserved among all AGC group kinases, constitutes an essential part of the activation segment in subdomain VIII of the kinase catalytic domain, immediately after nuclear localization signal and kinase domain insert of NDR1. The third site, Thr-444, also conserved in the AGC superfamily, is located outside of the kinase catalytic domain in a region enriched with hydrophobic amino acid residues (therefore "hydrophobic motif"). Mutational analysis showed, however, that the residue Ser-281 alone was responsible for the major part of phosphate incorporation into the autophosphorylated NDR1, whereas Thr-74 and Thr-444 could merely account for a minor percentage of the incorporated phosphate. Moreover, mutation of Ser-281 led to most dramatic decrease of NDR1 autophos-phorylation-induced kinase activity, thereby confirming this residue as the major NDR1 autophosphorylation site.
The addition of Ca 2ϩ /S100B led to a 2-fold increase of NDR1 autophosphorylation (on both Ser-281 and Thr-444) accompanied by about a 5-fold increased kinase activity. As expected, mutation of Ser-281 again led to an almost total decline of both autophosphorylation and kinase activity. Nevertheless, we also observed a marked, about 5-fold reduction of NDR1 kinase activity for the T444A mutant. These facts suggest that both residues are necessary for the full-active NDR1 and point to a synergy between these two residues in activation of NDR1. The outstanding importance of these residues is not surprising because similar observations were made also for other AGC group kinases (for review, see Refs. 21 and 22). Based upon structural analysis, the PKA activation segment residue Thr-197 was found in its phosphorylated form to align the catalytic site of that enzyme, thereby generating an active kinase conformation (23). On the other hand, the recently resolved structure of PKB␤ confirmed the crucial role of the hydrophobic motif with its residue Ser-474 which, upon phosphorylation, undergoes a series of interactions with ␣B and ␣C helices of the catalytic domain, thereby promoting the disorder to order transition of this part of molecule with concomitant restructuring of the activation segment and reconfiguration of the kinase bilobal structure (24,47).
Notably, we also observed a strongly compromised capabil-ityoftheNDR1T74Amutanttobesignificantlyautophosphorylated and activated by Ca 2ϩ /S100B. As mentioned above, this residue is a part of the S100B binding domain of NDR1 (amino acids 62-86), and it is broadly conserved among protein kinases of the NDR group from different origin (Fig. 6A). However, we cannot unambiguously distinguish between the possibility that solely a presence of a threonine or serine residue at this position is necessary for activation of NDR1 by Ca 2ϩ /S100B and the possibility that their phosphorylation is required as well. Nevertheless, the N-terminal domain of NDR1, which is highly conserved in its S100B binding region within the whole NDR subgroup of AGC kinases, seems to exert an autoinhibitory effect on the kinase catalytic domain which can be relieved by binding S100B or other potential interacting proteins. Analogous intramolecular domain-domain interactions are well known for a number of AGC and other kinases, such as PKB (pleckstrin homology domain), PKCs (C1 domain), PRK (HR1 domain), or calmodulin (CaM)dependent kinases (AID domain). Intriguingly, the mechanism of NDR1 regulation by Ca 2ϩ /S100B is reminiscent of the regulatory features of CaM kinase II, another Ca 2ϩ -controlled kinase, which is known to undergo an intramolecular (intersubunit) autophosphorylation on two threonine residues within its autoinhibitory and a nearby CaM binding domain upon binding of CaM (25). We have demonstrated previously that NDR1 also possesses the capacity to interact with CaM, although to a lesser extent than S100B, but CaM failed to activate NDR1 (12). Notably, the S100B relative S100A1 and, in part, S100B itself, are known to activate through a direct interaction the invertebrate giant sarcomeric kinase twitchin and its vertebrate counterpart titin; however, the exact molecular mechanism of how this activation occurs has not been elucidated so far (26,27).
As we reported earlier, the Ser-281 and Thr-444 residues become phosphorylated also in vivo (13). However, it was not known how these two regulatory sites are targeted in living cells. Here, we report that the Ser-281 site is an autophosphorylation residue in vivo because the kinase-inactive K118A mutant did not get phosphorylated on this position, i.e. the kinase activity of NDR1 is indispensable for the phosphoryla- FIG. 5. Thr-74 is required for association of HA-NDR1 with Ca 2؉ /S100B in COS-1 cells. COS-1 cells were transfected with HA-NDR1 wild-type, kinase-dead (K118A), and alanine mutants of phosphorylation residues Thr-74, Ser-281, or Thr-444, and S100B or corresponding empty vectors as indicated. 48 h later, nondetergent nuclear and cytoplasmic cell lysates were prepared, pooled, and analyzed for expression of HA-NDR1 variants (top panel). 1 mg of protein extracts was immunoprecipitated further with anti-S100B-Sepharose and analyzed for S100B expression (middle panel) and association of NDR1 with anti-S100B immunoprecipitates (bottom panel) as described under "Experimental Procedures." tion of Ser-281. Nevertheless, we cannot entirely exclude the possibility that this mutation in the substrate binding site leads, for instance, to conformational changes in the NDR1 molecule which do not permit phosphorylation action by a putative upstream kinase. In fact, in most other AGC kinases including PKB, p70 S6K , serum and glucocorticoid-regulated kinase (SGK), p90 RSK , and PKC isoforms, the activation segment residue (mostly threonine) has been reported to be a target of PDK1 (for review, see Refs. 28 and 29). The pivotal role of PDK1 in phosphorylation of AGC kinases is supported by high conservation of activation segment residues within the AGC family (Fig. 6A) and by the fact that a number of AGC kinases are known to undergo a direct interaction with PDK1. However, more recent studies employing PDK1 Ϫ/Ϫ embryonic stem cells confirmed that there are several exemptions to this general concept. Some AGC kinases such as PKA, PKC␦, AMPactivated protein kinase (AMPK), MSK1, and PRK2 become efficiently phosphorylated at their activation segment residues also in the absence of PDK1 (30,31), and at least for the two former kinases, it has been proven that this may occur through autophosphorylation (32,33). Our earlier observation on NDR1 also corroborated that this kinase is not modified by PDK1 either, inasmuch as cotransfection of PDK1 with NDR1 did not significantly change the NDR1 activity in vivo, and immunoprecipitated PDK1 failed to transfer phosphate to the GST-NDR1 K118A mutant in vitro (13). Most likely, this can be explained by the fact that the NDR1 sequence around Ser-281 displays apparent variations in Pϩ1 and ϩ2 positions to the highly conserved PDK1 consensus target site (Ser/Thr-Phe-Cys-Gly-Thr-Xaa-Asp/Glu-Tyr-Xaa-Ala-Pro-Glu, where Ser/ Thr is the phosphoacceptor site, and Xaa stands for a hydrophobic residue; Fig. 6A). Taken together, our findings suggest that NDR1, in contrast to most AGC group kinases, is not targeted by PDK1 on its activation segment residue Ser-281 but instead, becomes efficiently autophosphorylated at this residue both in vivo and in vitro.
Because the kinase-inactive K118A as well as the S281A mutant became phosphorylated at Thr-444 virtually to the same extent as wild-type NDR1 in vivo, we surmise that this residue is targeted by an upstream kinase in OA-treated COS-1 cells. However, we cannot rule out the possibility that Thr-444 autophosphorylation also might contribute to the overall phosphorylation rate at this residue because the vigorous treatment by OA is known to lead to a stoichiometric phosphorylation of Thr-444 (13) and may, perhaps, favor an upstream kinase at the expense of autophosphorylation (thereby being sufficient to achieve full phosphorylation also for the K118A and S281A mutants). Hence, it remains to be elucidated whether NDR1 becomes exclusively targeted by the putative upstream kinase also if stimulated by comparably mild, as yet unknown physiological stimuli. Remarkably, Thr-444 is flanked by hydrophobic amino acids, which in this region are also conserved among all AGC group kinases (consensus Phe-Xaa-Xaa-Phe-Ser/Thr-Phe/Tyr; except NDR1 has a tyrosine instead of phenylalanine at the PϪ1 residue, and PKA is truncated at this position). There is an ongoing dispute in the field concerning the mechanism of hydrophobic motif phosphorylation of several AGC protein kinases. We have reported recently that the corresponding residue in PKB␣ (Ser-473) becomes phosphorylated independently of PKB␣ activity status (34); and furthermore, we have characterized a constitutively active, staurosporine-insensitive Ser-473 kinase activity enriched in buoyant, detergent-insoluble plasma membrane rafts (35). Because this activity also depends on phosphoinositide 3-kinase, it is sometimes referred to as PDK2. Although few candidates such as integrin-linked kinase or mitogen-activated protein kinase-activated protein kinase-2 were originally reported to be capable of in vitro phosphorylation of PKB␣ at Ser-473 (36,37), these proteins turned out later to be physiologically irrelevant hydrophobic motif kinases (38,39). Thus, despite considerable efforts of several laboratories, Ser-473 kinase resisted identification and cloning so far. Moreover, because the sequence similarity between PKB␣ and NDR1 is low in this region (except the three conserved hydrophobic residues) and because NDR1 is FIG. 6. Structure of NDR1 and model of its regulation. A, NDR1 domain structure is shown including the N-terminal regulatory domain containing S100 binding region and residue Thr-74; the catalytic domain spliced by a 30-amino acid insert (encompassing a nonconsensus NLS sequence) and containing the activation segment residue Ser-281; and a C-terminal part comprising the hydrophobic motif with Thr-444. Below, the conservation of S100B binding region is shown among several NDR-related kinases from different origin as well as conservation of parts of activation segment and hydrophobic motif residues within the AGC family. The identical residues are boxed in black, and similar residues are in gray. The phosphorylation sites are in bold. B, model of NDR1 regulation. NDR1 becomes autophosphorylated on the three indicated residues, Thr-74, Ser-281, and Thr-444 in vitro. However, only Ser-281 is an autophosphorylation site in vivo. In contrast, Thr-444 is targeted mainly by an as yet unidentified upstream kinase. Nevertheless, both phosphorylation events are Ca 2ϩ -dependent and essential for NDR1 activity in vivo. Although Thr-74 is also a minor autophosphorylation residue in vitro, it is currently not clear whether Thr-74 becomes (auto)phosphorylated also in vivo. Nonetheless, this residue appears to be a crucial component of the S100B binding domain located in N terminus of NDR1 (for details, see "Discussion"). not sensitive to inhibitors of phosphoinositide 3-kinase, 4 it is rather unlikely that Ser-473 kinase will be the same protein as the hydrophobic motif kinase phosphorylating Thr-444 in NDR1. For similar reasons, it is also improbable that NDR1 could be targeted by atypical PKCs such as PKC/ (which curiously possess a phosphate-mimicking, negative-charge glutamate residue in place of Ser/Thr at their hydrophobic motif, making them independent of an upstream kinase), as was postulated for conventional and novel PKCs such as PKC␣, PKC␦, or PKC⑀, and p70 S6K (40 -42). Finally, although precedents for the autophosphorylation at hydrophobic site also exist (e.g. PKC␤II, Ref. 43), as mentioned above, we have unambiguously ruled out this alternative for NDR1.
We observed that both autophosphorylation of NDR1 on Ser-281 as well as phosphorylation by an upstream kinase on Thr-444 are Ca 2ϩ -dependent processes. In this context, it should be noted that OA, which induces phosphorylation on both sites, was reported, in addition, or because of its inhibitory effect on protein phosphatase 2A, also to increase the intracellular Ca 2ϩ concentration. This occurs by at least two mechanisms: activation of L-type Ca 2ϩ voltage-dependent channels (44) or release of intracellular Ca 2ϩ stores (45). This fact implies that OA not only stimulates NDR1 directly through relieving the protein phosphatase 2A-mediated inhibition of phosphorylation of NDR1, but it also may activate NDR1 indirectly, through some Ca 2ϩ -dependent signaling pathways. Combining the in vitro and in vivo data we gathered in this work, the Ca 2ϩ dependence of autophosphorylation of NDR1 on Ser-281 can be explained readily by requirement of NDR1 for Ca 2ϩ /S100B to autophosphorylate efficiently. Indeed, the T74A mutant deficient in binding to S100B also displayed a strongly diminished autophosphorylation rate on Ser-281 in vivo. On the other hand, the Ca 2ϩ dependence of NDR1 phosphorylation on Thr-444 appears to be mediated by a Ca 2ϩ -dependent kinase. However, because the T74A mutant showed a compromised ability to become phosphorylated on Thr-444 as well, it might also be possible that the N-terminal domain of NDR1, together with Ca 2ϩ /S100B, can in part be responsible for the Ca 2ϩ dependence of Thr-444 phosphorylation, e.g. through targeting an upstream kinase to NDR1.
In this report, we defined NDR1 as a new member of a broad group of Ca 2ϩ -regulated protein kinases which currently comprises proteins as different as CaM and CaM kinases, twitchin/ titin giant kinases, conventional PKCs, and to some extent also mitogen-activated protein kinases, PKA, PKB, and PKG (46). We characterized the molecular mechanism for activation of NDR1 by Ca 2ϩ consisting of Ca 2ϩ /S100B-induced autophosphorylation of NDR1 on the activation segment residue Ser-281 and phosphorylation of hydrophobic motif site Thr-444 by a Ca 2ϩ -dependent upstream kinase. The exact nature of the activatory Ca 2ϩ signals as well as the identity of upstream kinase and downstream targets of NDR1 still awaits elucidation but will eventually provide insights into the physiological function of this enzyme and the corresponding signaling pathway.