Identification of Substrate Specificity Determinants for the Cell Cycle-regulated NIMA Protein Kinase*

NIMA is a cell cycle-regulated protein kinase required for the GzfM transition in the filamentous fungus Aspergillus nidulans. Previous biochemical characterization of the recombinant enzyme indicated that NIMA is a protein serindthreonine specific kinase with p-casein being the best substrate from the many proteins and peptides tested (Lu, K. P., Osmani, S. A, and Means, A. R. (1993) J. Biol. Chem. 268,8709-8776). However, substrate specificity or physiologically relevant substrates for NIMA remained unknown. In search for a peptide substrate for this enzyme, we screened an assembled library of synthetic peptides that each contained a phosphorylation site for a known protein kinase and found an excellent peptide substrate for NIMA, phospholemman 42-72 (PLM(42-72)). NIMA kinase phosphorylated PLM(42-72) uniquely and stoichiometrically on S e P with a V,, of 1.4 pmollmidmg and apparent K,,, of 20.0 p ~ . These kinetic constants were about 10-fold higher and %fold lower than those for p-casein, respectively. A detailed analysis of substrate specificity determinants using synthetic peptide analogs of PLM(42-72) indicated that Phe-ArgXaa-Serfl'hr represents the optimal pri-mary sequence for NIMA kinase phosphorylation. Re- while These results reveal the unique nature of substrate recognition and should prove valuable in the for biologically relevant substrates. kinase is the product of a cell cycle regulatory gene that was isolated by genetic complementation of a tem-perature-sensitive mutation in the nimA gene of Aspergillus nidulans (1). Cells carrying temperature-sensitive mutations in the nimA gene were specifically arrested in Gz at the restric-tive temperature, but rapidly and synchronously entered mitosis when shifted to the permissive temperature. In contrast, overexpression of the nimA gene product induced premature mitotic arrest (1). These results indicate that NIMA plays a critical role in the progression of cells into mitosis. Examina-*

The NIMA kinase is the product of a cell cycle regulatory gene that was isolated by genetic complementation of a temperature-sensitive mutation in the n i m A gene of Aspergillus nidulans (1). Cells carrying temperature-sensitive mutations in the n i m A gene were specifically arrested in Gz at the restrictive temperature, but rapidly and synchronously entered mitosis when shifted to the permissive temperature. In contrast, overexpression of the n i m A gene product induced premature mitotic arrest (1). These results indicate that NIMA plays a critical role in the progression of cells into mitosis. Examinat Present address: Molecular Biology and Virology Laboratory, The tion of the amino acid sequence of NIMA deduced from the nimA cDNA suggested that the nimAgene product belonged to the family of Serrrhr protein kinases. Antibodies specific for NIMA precipitated a fungal protein, which phosphorylated p-casein in vitro. Using p-casein phosphorylation, the activity of NIMA was shown to fluctuate during the nuclear division cycle, peaking in late Gz and mitosis (2). To determine the biochemical properties of the protein kinase, we expressed NIMA in bacteria (3). Biochemical characterization of the purified recombinant enzyme revealed it to be a unique SerPThrspecific protein kinase, the activity of which was also regulated by SerPThr phosphorylatioddephosphorylation (3). However, nothing was known concerning NIMA substrate specificity or the identity of physiologically relevant substrates. One strategy that can be used to search for relevant substrates of a novel protein kinase is to define the structural determinants that are required for phosphorylation of protein or peptide substrates. Of the many proteins and synthetic peptides utilized as substrates for well characterized SerTIkr-specific protein kinases, p-casein was the best substrate for NIMA (3). However, we showed that under optimal assay conditions, NIMA phosphorylated p-casein on multiple residues and with a relatively low V,,, of 156 nmollmidmg (3). In this study, we have identified an excellent peptide substrate for NIMA, namely phospholemman 42-72 (PLMl(42-7211, by screening an assembled library of 56 synthetic peptides representing the phosphorylation sites in many different proteins. NIMA kinase phosphorylated PLM(42-72) with a 10-fold higher V, , and a lower K, than p-casein. After establishing that the unique residue phosphorylated by NIMA in this peptide (SeF3) was different from those phosphorylated by cyclic AMP-dependent protein kinase (PKA) and protein kinase C (PKC), we synthesized two series of peptide analogs to define the structural determinants required for NIMA kinase specificity. Our results reveal hitherto unknown unique sequence determinants required for phosphorylation of synthetic peptide substrates by the NIMA kinase. These results also demonstrate the utility of using a synthetic peptide screen to identify efficient substrates for new protein kinases and provide essential information to begin the search for the biological substrates for NIMA.

EXPERIMENTAL PROCEDURES
Synthesis of Peptides-Peptides were synthesized and purified as described previously (4). All peptides were dissolved in distilled water at a concentration of 4 mg/ml and a series of dilutions was made using 0.25 mg/ml bovine serum albumin as a carrier. The precise concentration of each peptide was determined by amino acid analysis, which also established the predicted amino acid composition of all the synthesized pep- Peptide Phosphorylation-NIMA protein kinase was purified from a bacterial expression system as described previously (3). Purified kinase was diluted prior to use in a buffer containing 50 m HEPES, pH 7.5, 0.5 mg/ml bovine serum albumin, and 1 m dithiothreitol. Phosphorylation of synthetic peptides was camed out in a 30-pl reaction mix containing 50 m HEPES, pH 7.5, 0.25 mg/ml bovine serum albumin, 10 rn magnesium acetate, 1 m dithiothreitol, 100 p d or 300 PM [?"-PIATP (300-1,000 countdminlpmol), and 50-100 ng of NIMA. After incubation at 30 "C for 15 min, 25-pl aliquots were removed from the reaction mix and applied to phosphocellulose P81 filters. The filters were washed with 0.5% phosphoric acid, and the radioactivity was counted as described previously (3). Under these conditions, all kinase reactions were linear for at least 30 min. Every peptide phosphorylation assay was repeated three to six times using at least two different prepa-rations of NIMA kinase. Less than 10% standard deviation among experiments using different kinase preparations and less than 5% standard deviation among duplicates within a given experiment were noted. The kinetic data were analyzed as described previously (5). Calmodulindependent protein kinase I1 and IV were assayed as described previously (6). Protein kinase A (PKA), protein kinase C (PKC), cyclin-dependent protein kinase CDC2, and casein kinase I and I1 (Upstate B~otechnology Inc.) were assayed according to the manufacturer's eonditions. Mitogen-activated protein kinase was kindly provided by Dr. Perry Blackshear (Dept. of Biochemistry, Duke University Medical Center) and assayed as described (7).
Analysis of Phosphorylated Residues-For phosphoamino acid and phosphopeptide analyses, peptides were phosphorylated in the presence of 50 p,t [f'z-PIATP. Phosphopeptides were separated from the free radioactive ATP on an AGl-X8 ion-exchange column in the presence of 30% acetic acid as described previously (8). After repeated lyophilization in water, peptides were subjected to partial acid hydrolysis for phosphoamino acid analysis or repeated trypsin digestion for phosphopeptide analysis, followed by two-dimensional separation by thin layer chromatography as described previously (9).
For phosphopeptide sequencing, peptides were phosphorylated in the presence of 500 unlabeled ATP containing a trace amount of [?"-PIATP. The phosphopeptides were separated from unphosphorylated peptides by reversed-phase HPLC using a gradient of 2041% acetonitrile (v/v) in 0.1% trifluoroacetic acid (v/v) with a flow rate of 1 mumin. The purified phosphopeptides were digested overnight at 37 "C in 100 pl of 0.1 M NH4HC03 containing about 2% (w/w) protease V8 (Boehringer Mannheim). Following lyophilization, peptides were again separated by reversed-phase HPLC using 520% acetonitrile (v/v). The single radioactive peak was sequenced by Dr. Richard Cook (Baylor College of Medicine) using an Applied Biosystems Inc. 470A gas-phase sequenator and a 120A phenylthiohydantoin amino acid analyzer.
RESULTS AND DISCUSSION Screening the Assembled Peptide Library-To identify peptide substrates of the NIMA kinase, we used 56 synthetic peptides representing sites identified in a wide range of proteins that are phosphorylated by many different protein kinases (10). As shown in Table I, dramatic differences in the NIMA activity toward these peptides were observed. Most of the peptides were not phosphorylated by NIMA. Only eight peptides were phosphorylated a t a rate above 20% of that observed using p-casein as a substrate. However, a single peptide PLM(42-72) was observed to be a better substrate than p-casein. NIMA phosphorylated this peptide at a rate that was more than 10-fold faster than p-casein. To distinguish whether NIMA phosphorylated Ser62-Ser63 a n d o r S e P , phosphopeptide analysis was undertaken. The phosphorylated synthetic peptide was cleaved exhaustively with trypsin, and the digest was separated by thin layer chromatography according to the protocol of Boyle et al. (9). A single phosphorylated peptide, peptide 1 (Fig. E ) , was detected, indicating that NIMA could phosphorylate either Ser62-Ser63 or S e P , but not both, since trypsin cleaved a t arginines located between these serines. To identity the phosphorylated residue(s), peptide PLM(42-72) was phosphorylated by NIMA in the presence of ATP containing a trace amount of [+'z-P]ATP, the phosphorylated product was purified by reversed-phase HPLC, and digested repeatedly with protease V8 under conditions where the protease cleaves peptides specifically at the COOH-terminal side of Glu residues (Fig. lA). Subsequent separation by HPLC yielded a single radioactive peptide (data not shown). This peptide was subjected to amino acid sequencing. The sequence obtained was GTFRSS*IRRLSTRRR, where * indicates that the detection of SeF3 was selectively reduced, compared to the other residues. These results indicated that Ser63 was most likely the unique residue phosphorylated by NIMA.
Since phospholemman was previously shown to be phosphorylated by both PKA and PKC (111, we examined whether peptide PLM(42-72) was also a substrate for these protein kinases and if so, which residue(s) was phosphorylated. PKA and PKC each phosphorylated the peptide exclusively on serine predicted proteolytic products generated by digestion with either trypsin (solid arrows) or V8 protease (open arrows) which cleaves the peptide specifically at the COOH-terminal side of Arg or Glu residues, respectively. B, phosphoamino acid analysis. PLM(42-72) was phosphorylated by the different protein kinases as indicated and phosphorylated peptides were isolated and subjected to acid hydrolysis, followed by two-dimensional separation on thin layer chromatography plates. S, phosphoserine; T, The phosphorylated peptides generated by the enzymes shown at the phosphothreonine; Y , phosphotyrosine. C, phosphopeptide analysis. top of each panel were exhaustively digested with trypsin and then subjected to two-dimensional separation on thin layer chromatography plates separately (left panel ) or in combination (right panel ). The ratios of phosphopeptide radioactivity loaded on the thin layer chromatography plates containing the combined samples were 3:l for N I W K A , N I W K C , and PKAPKC. residues, as did NIMA (Fig. 1B). Phosphopeptide analysis indicated that a single tryptic peptide, peptide 2, was phosphorylated by PKA and that the mobility of this peptide was different from that phosphorylated by NIMA, as determined by mixing experiments (Fig. E ) . These results indicate that PKA predominantly phosphorylated Ser68, which contains the appropriate consensus sequence RRXS for phosphorylation by PKA. Two tryptic peptides (phosphopeptides 1 and 2) were phosphorylated by PKC a t a ratio of approximately 4:l ( Fig.  1 0 . In phosphopeptide-mixing experiments, these two peptides comigrated with peptides 1 and 2 which were independently phosphorylated by NIMA and PKA, respectively. Given the specificity requirements of PKC, it seems reasonable that it phosphorylates SeP3 (peptide 1) and S e P (peptide 2) where peptide 2 contains the most favored site which corresponds to tively. PKA appears to suppress PKC phosphorylation of SeF3 (peptide l ) , but this has not been quantified (Fig. 1C). A more detailed comparison between PKC and PKA phosphorylation preferences is beyond the scope of this study. Effect of Peptide Length on Kinetics of Phosphorylation-Since the NIMA phosphorylation site in PLM(42-72) was different from the major phosphorylation sites for PKA and PKC, we wished to define the determinants for substrate recognition by the NIMA kinase. Our strategy was to first determine the minimal peptide that could still be phosphorylated by NIMA with reasonable K,,, and V,,, values. Once this was determined, we could then introduce individual amino acid replacements with confidence that interpretation of results would be independent of variation of peptide length. We first synthesized a series of peptides with different numbers of residues deleted from the N H 2 and/or COOH terminus of the parent peptide and examined them as substrates for NIMA (Table 11). The removal of residues from the NH2 terminus up to Pro53 did not affect the kinetics of peptide phosphorylation. Although omitting Asp54, Glu, Glu, Glu57 slightly increased the apparent K , (2-fold), it did not alter the V,,,. Further deletion of GlP8 and ThP9 substantial19 reduced the V,,, for peptide phosphorylation and also increased the apparent K,, suggesting that the glycine and threonine residues contribute to some extent to the affinity of the peptide for the kinase. As expected, negligible phosphorylation was observed with the peptide in which the phosphorylation site, SeF3, was deleted (PLM-65-72). Deletions up to S e P residue from the COOH-terminal side had only a small effect on the kinetics, i.e. all peptides were phosphorylated at rates comparable to those observed with the peptide, PLM(58-72). These results indicated that the COOH-terminal residues up to S e P do not play a major role in substrate recognition by the NIMA kinase.

Phosphorylation of Peptides in Which Individual Residue(s)
Are Changed to Alanine-% identify the amino-terminal specificity determinants in substrates for NIMA kinase, we investigated the role of individual residues in the minimal peptide substrate PLM(58-72). A series of PLM(58-72) analogs was synthesized in which a single or multiple residues were substituted with alanines. As mentioned above, this approach allowed the identification of the functionally important amino acids independent of the potentially confounding effects of peptide length variation. The kinetic constants for phosphorylation of these peptides by NIMA are shown in Table 111. Substitution of Ser63 with Ala gave a peptide that was not phosphorylated by NIMA, while substitution of SeP8 and ThP9 or Ser6', Ser68, and ThP9 with alanines had little effect on the overall kinetics of peptide phosphorylation by NIMA. These results confirmed that NIMAphosphorylated a single residue, SeP3. WhenAre5, Are6, or Arg70 were replaced singly or collectively with Ala, the kinetic constants were only slightly altered. However, replacing Are1 with Ala increased the apparent K , about 6-fold and also reduced V,,, resulting in a much poorer peptide substrate and indicated that the Arg at position -2 from the phosphorylated Ser(P-2) may be important for substrate recognition. The most striking finding of these studies was that the substitution of PheG0 with Ala resulted in a near complete inhibition of peptide phosphorylation by NIMA. This indicates that a Phe at P-3 is absolutely required for substrate phosphorylation and may represent a distinguishing feature of NIMA substrate recognition when compared to other known protein kinases.
In support of this unique requirement for NIMA substrates, we compared phosphorylation of two peptide substrates of NIMA as potential substrates for several other Ser/Thr protein kinases. As shown in Table IV  and PKC can both phosphorylate SeP3, there is a distinctly different requirement for the amino acid at P-3 for these protein kinases. To our knowledge, NIMA is the first protein kinase to be characterized that requires a Phe at P-3 of the phosphorylation site. PKA has a requirement for a n aromatic residue at the P-11 position for high affinity binding of the protein kinase I peptide (121, and Colbran et al. (13) have shown that Phe at the P+4 position can act as a negative determinant for PKA. We have not attempted to model the NIMA structure (11, but inspection of the NIMA catalytic domain sequence in the light of the recognition requirements and three-dimensional structure of PKA (14, 15) reveals some similarities and some important differences between these two enzymes. Both share Glu residues analogous to GIu'~' and GluZ3' of the PKA that are important for recognition of an Arg at the P-2 position relative to the serine that is phosphorylated (PO). On the other hand G1ulZ7 which is required for P-3 Arg recognition and GluZo3 which is required for P-6 Arg recognition in PKA are replaced in NIMA by Asp and Phe, respectively. These differences favor the view that P-3 and P-6 basic residue determinants are not important for NIMA. The aromatic pocket in the PKA structure contains w35 and Phe239 which are required for high affinity binding of protein kinase inhibitor at the P-11 position. As these residues are replaced by Glu and Asn in NIMA, respectively, we surmise that NIMA substrates do not have a similar requirement to PKA for this aromatic pocket. However, a second hydrophobic pocket in the PKA structure that is utilized for recognition of the P+l position, namely Leulg8, Prozo2, and Leuzo5 is reasonably conserved in NIMA by Tyr, Pro, and Met, respectively. The differences in the substrate recognition residues between the two enzymes are consistent with the observation that substrates for PKA, such as Kemptide are poor substrates for NIMA. The likely recognition site for the Phe at P-3 is not known although inspection of the NIMA sequence reveals that in the sequence corresponding to the region in the PKA structure between the p strand 9 and the F-a! helix, SHD-" FASTYVGTPFYMSPEIC, 3 aromatic residues are present that may provideda recognition pocket for PheGo in the PLM(42-72) peptide.
Whereas the NIMA protein kinase is critical for mitotic progression in A. nidulans (1, 2, 16, 17), it is not known whether homologs of this enzyme exist in other organisms. Even in the fungal system the physiological relevant substrates for NIMA remain to be identified. The pair of peptides developed and characterized in this report (Table IV) should be useful reagents in both searches. On the one hand enzyme activities can be identified that recognize PLM(5&72)A62,68,69 but will not phosphorylate PLM(58-72)A60,6s.69. On the other hand, one can now search for potential substrates that contain a Phe at P-3 of the phosphorylated residue. In such instances it is also possible that PLM(58-72)A62,6s,69 in which the SerG3 is also converted to a nonphosphorylatable residue might serve as a competitive inhibitor of NIMA-like protein kinases.