New Structural Insights into Phosphorylation-free Mechanism for Full Cyclin-dependent Kinase (CDK)-Cyclin Activity and Substrate Recognition*

Background: Unlike cell cycle CDK-cyclin complexes, the Pho85-Pcl CDK-cyclin superfamily requires no phosphorylation for full kinase activity. Results: Crystal structures of Pho85-Pcl10 solidified a key role of an invariant cyclin aspartate in bypassing CDK phosphorylation and attaining full activity. Conclusion: Identical phosphorylation bypass mechanism governs the entire Pho85-Pcl superfamily. Significance: The emergence of other CDK-cyclin complexes has added new facets beyond those regulating cell cycle. Pho85 is a versatile cyclin-dependent kinase (CDK) found in budding yeast that regulates a myriad of eukaryotic cellular functions in concert with 10 cyclins (called Pcls). Unlike cell cycle CDKs that require phosphorylation of a serine/threonine residue by a CDK-activating kinase (CAK) for full activation, Pho85 requires no phosphorylation despite the presence of an equivalent residue. The Pho85-Pcl10 complex is a key regulator of glycogen metabolism by phosphorylating the substrate Gsy2, the predominant, nutritionally regulated form of glycogen synthase. Here we report the crystal structures of Pho85-Pcl10 and its complex with the ATP analog, ATPγS. The structure solidified the mechanism for bypassing CDK phosphorylation to achieve full catalytic activity. An aspartate residue, invariant in all Pcls, acts as a surrogate for the phosphoryl adduct of the phosphorylated, fully activated CDK2, the prototypic cell cycle CDK, complexed with cyclin A. Unlike the canonical recognition motif, SPX(K/R), of phosphorylation sites of substrates of several cell cycle CDKs, the motif in the Gys2 substrate of Pho85-Pcl10 is SPXX. CDK5, an important signal transducer in neural development and the closest known functional homolog of Pho85, does not require phosphorylation either, and we found that in its crystal structure complexed with p25 cyclin a water/hydroxide molecule remarkably plays a similar role to the phosphoryl or aspartate group. Comparison between Pho85-Pcl10, phosphorylated CDK2-cyclin A, and CDK5-p25 complexes reveals the convergent structural characteristics necessary for full kinase activity and the variations in the substrate recognition mechanism.

cyclin, were originally discovered to be the main driving forces for the sequential events of eukaryotic cell cycle progression through the phosphorylation of substrates (for review, see Ref. 1). Other CDK and cyclin complexes or families have also emerged and were investigated as playing important regulatory/signaling functions in diverse processes, adding new facets beyond classical cell cycle progression regulation.
The latter is exemplified by CDK5; its complex with the p35 cyclin is critical in the development of the central nervous system (for review, see Refs. 2 and 3). It is also typified by Pho85 CDK. In the budding yeast (Saccharomyces cerevisiae), 10 different cyclins, called Pcls (Pho85 cyclins), Pho80, and Clg target Pho85 to different substrates that regulate numerous cellular functions, including autophagy, gene expression, morphogenesis, cell cycle control, stress adaptation, carbon source utilization, amino acid biosynthesis, cell polarity and actin cytoskeleton regulation, and phosphate and glycogen metabolism, among others (for review, see Refs. 4 and 5); listed in Table 1 are the Pcls and substrates). Thus, Pho85 has emerged as an important model of CDKs to investigate different cellular processes in mammalian cells. Like all CDKs, Pho85 and CDK5 are proline-directed serine/threonine kinases that require substrates have a proline residue immediately after the phosphorylatable serine or threonine.
Interestingly, Pho85 is the closest functional homolog of CDK5 (6,7). Moreover, it is noteworthy that, unlike cell cycle CDKs, which require phosphorylation of a Ser/Thr residue on the so-called "activation loop" by CDK-activating kinase for full activation (1), CDK5 and Pho85 require no phosphorylation despite the presence of equivalent serine residues (8 -11). The crystal structures of CDK5 complexed with p25, a truncated form of p35 cyclin, and Pho85 with full-length Pho80 Pcl have been determined (12,13).
The Pho85-Pho80 complex is the best characterized among the Pho85-Pcl superfamily (for a recent review, see Ref. 14); see also Refs. 4 and 5). It epitomizes the CDK-cyclin complex role in sensing and responding to environmental changes as well as regulating metabolism. Specifically, Pho85-Pho80 signals a response to the stress of phosphate starvation through regulating the location and activity of the substrate Pho4, a transcription factor required for expression of phosphate-responsive genes.
The highlight of the crystal structure of the Pho85-Pho80 complex, the first for a member of the Pho85-Pcl CDK-cyclin complexes, is the revelation of a more direct participation of the cyclin to the activity of the complex than previously observed in other CDK-cyclin complexes (13). This is manifested by the finding of the distinct critical functional roles played by the proximal residues, Asp-136 and Phe-138, located at the bending tip loop between two helices (␣3-␣4 loop) of the core cyclin box motif of Pho80 cyclin. The electrostatic interactions of Asp-136 with residues in the Pho85 active site region, principally the salt links with arginine residues, provided the initial major clue on the mechanism for circumventing phosphorylation of a serine on the activation loop. The Pho80 cyclin box Phe-138 residue, on the other hand, is strongly indicated to be involved in the recognition of the Ile/Leu residue at the ϩ3 position of the consensus sequence motif, SPX(I/L (where S is the phosphorylatable serine at position 0, P is proline at position ϩ1, and X is any residue at position ϩ2), of several phosphorylation sites in the Pho4 substrate (13,15). In sharp contrast, the ϩ3 Lys/Arg residue of the canonical phosphorylation motif of (S/T)PX(K/R) of substrates of CDK2 is recognized by the phosphoryl adduct of the phosphorylated CDK2 (or pCDK2) (16).
Similar to Pho80, Pcl10 or Pcl8 has a major role in regulating metabolism and sensing environmental changes. Pcl10/Pcl8 targets Pho85 to negatively regulate the substrate, Gsy2. Gsy2 is the predominant nutritionally regulated isoform of glycogen synthase in yeast whose activity is inhibited by Pho85-Pcl8/10 mediated phosphorylation, preventing the hyper-accumulation of glycogen (11,17). Here we report the crystal structures of Pho85-Pcl10 and its complex with the ATP analog, ATP␥S. The structures considerably deepen our understanding of activation loop phosphorylation bypass mechanism and its key role in attaining full activity and provide new insights into substrate recognition by the Pho85-Pcl CDK-cyclin superfamily.

EXPERIMENTAL PROCEDURES
Bacterial Co-expression and Purification of Pho85-Pcl10-Plasmids of full-length subunits Pho85 (35 kDa and 305 residues) and Pcl10 (49 kDa and 433 residues) were kindly provided by Dr. Erin K. O'Shea. Pho85 were cloned into pQE60 vector from Qiagen with a His tag at the C terminus. Pcl10 was cloned in pSBET vector, which harbors a helper gene ARG tRNA to improve the expression of eukaryotic genes in Escherichia coli (18). ARG tRNA gene encodes a low-abundance tRNA in E. coli corresponding to codons AGG and AGA (Arg), which is used in much higher frequency in eukaryotic coding sequences. Pcl10 has nine AGG and AGA codons; therefore, pSBET expression vector was expected to improve the expression of Pcl10. Pho85 and Pcl10 plasmids were co-transformed into E. coli BL21 DE3 (Novagen) cells and grown overnight at 37°C on LB plates with 100 g/ml ampicillin and 70 g/ml kanamycin. Freshly transformed clones were inoculated into 6 liters of LB media and allowed to grow at 37°C to A 600 of 0.6 -0.8. Protein expression was then induced by adding isopropyl-␤-D-thiogalactopyranoside to a final concentration of 200 M at 16°C overnight. Cell pellets were harvested by centrifugation at 3000 rpm for 15 min and stored at Ϫ80°C. The cell pellets were resuspended in the lysis buffer consisting of 300 mM NaCl, 50 mM NaH 2 PO 4 , 10% glycerol, 5 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, pH 7.0. The cells were lysed by passing the suspension twice through microfluidizer processor (Microfluidics). After centrifugation at 20,000 rpm for 30 min at 4°C, the clear lysate was loaded onto a self-packed TALON column (Clontech) preequilibrated with the lysis buffer (without phenylmethylsulfonyl fluoride) and then washed with at least 10 column volumes of the same buffer. The bound protein was then eluted by a  solution of 10% glycerol, 300 mM NaCl, 5 mM ␤-mercaptoethanol, 250 mM imidazole, and 50 mM NaH 2 PO 4 , pH 7.0. The eluted protein was concentrated to 2-3 ml by centrifugation using Amicon Ultra-15 centrifugation filter devices (Millipore) at 4000 ϫ g at 4°C. The protein was further purified by gel filtration using a Superdex 200 16/60 (GE Healthcare) column equilibrated with 100 mM NaCl, 5 mM ␤-mercaptoethanol, 5% glycerol, and 10 mM HEPES, pH 7.0. Surprisingly, the size exclusion chromatography revealed two major peaks, indicating two types of Pho85-Pcl10 complexes, one eluting first, which is indicative of a larger mass followed by another somewhat smaller mass. PAGE analysis revealed that the first peak is composed of full-length Pho85 and Pcl10, and the second peak is made up of full-length Pho85 and a truncated Pcl10 of a ϳ20 kDa mass, which likely arose from endogenous proteolytic cleavage. Both complexes appear stable. N-terminal amino acid residue sequencing of the Pcl10 fragment gave a stretch of sequence of SLPHDEEEDQE, which matches exactly the Pcl10 sequence of residues 227-237. This result and the 443 residue composition of the full-length Pcl10 indicate that the ϳ20-kDa fragment represents the C-terminal fragment of residues 227-443. Sequence comparison shows that a large core region of the 207-residue Pcl10 fragment aligns very well with a region of the full-length 293-residue Pho80 cyclin encompassing the fivehelix cyclin box fold previously revealed by its crystal structure in complex with Pho85 (13).
Crystallization and Structure Determination-Both purified complexes of full-length Pho85 with full-length Pcl10 and truncated Pcl10 in 5% glycerol, 100 mM NaCl, 5 mM ␤-mercaptoethanol, and 10 mM HEPES, pH 7.0, were concentrated to 10 mg/ml using Amicon Ultra-15 centrifugation filter devices (Millipore). Crystallization screens for both complexes were carried out at 20°C. Only the Pho85-truncated Pcl10 complex could be crystallized; hence, the complex is hereafter identified for brevity as Pho85-Pcl10. Crystals grew by vapor diffusion in hanging drops by mixing 1:1 ratio volumes of the concentrated protein complex with the reservoir solution of 22-25% polyethylene glycol 5000 monomethyl ether (PEG 5K MME), 0.1 M MES, pH 6.4, at room temperature. Plate shaped crystals were formed within a week and grew to full size in 2 weeks. To obtain bigger crystals, microseeding was performed. Microseeds were generated by breaking the existing crystals with a fine fiber and then dipped into fresh hanging drops of protein and reservoir. Crystals of the Pho85-Pcl10 in complex with ATP␥S were obtained by soaking Pho85-Pcl10 crystals in a solution of 2 mM ATP␥S and 4 mM Mg 2ϩ in 25% PEG 5K MME, 0.1 M MES, pH 6.4.
The crystals were cryo-protected in a solution of 25% PEG 5K MME, 0.1 M MES, pH 6.4, and 20% glycerol before being flashcooled in liquid nitrogen. X-ray diffraction data of crystals of Pho85-Pcl10 and its complex with ATP␥S were collected at Ser-CAT 22ID and 19ID, respectively, at Advanced Photon Source at the Argonne National Laboratory. Native data were processed with HKL2000 and were merged with SCALEPACK (19). Statistics for diffraction data are shown in Table 2.
The Pho85-Pcl10 structure was determined by a molecular replacement technique using the Pho85-Pho80 structure (PDB 2PK9) (13) as the search model in Molrep-auto molecular replacement under CCP4 package (20). A satisfactory molecular replacement solution was obtained for a model in which the side chains of Pho80 were excluded. After multiple rounds of annealing refinement, the model of the complex with the correct sequence for Pcl10 was then refined in Refmac5 under CCP4 package and Phenix (21). The refinements were interspersed with model building and fitting of water molecules in Coot (22). The structure was evaluated using Molprobity, and the statistics for the final refinement are shown in Table 2.
Kinase Assay of Pho85-Pcl10 Using Peptides of the Phosphorylation Site of the Gsy2 Substrate-The kinase activity of Pho85-Pcl10 (Pcl10 residues 227-433) was assayed using the most active peptide KKLMVPGSPRDLRS (called SPRDL) containing the phosphorylatable Ser-654 (bold S residue in the peptide) of the Gys2 as the substrate (11). The assay mixture was made by combining 1 l of Pho85-Pcl10 (0.7 mg/ml), 5 l of peptide (0.01 M), and 14 l of kinase buffer (20 mM MgCl 2 , 40 mM TRIS, pH 7.5). The reaction was initiated by the addition of 5 l 85 mM ATP (0.15 mCi/ml [␥-32 P]ATP). After 30 min of incubation at 30°C, 10 l of the reaction mixture was spotted onto P81 Whatman square paper. After allowing the mixture to be absorbed for a few seconds, the P81 paper was dipped into a beaker containing 1% (v/v) phosphoric acid. The paper was washed 3 times for 15 min with 500 ml of 1% (v/v) phosphoric acid, rinsed briefly with acetone, and dried overnight. Incorporated radioactivity was determined by scintillation counting. Kinase activities are expressed as the amount of radioactive phosphate incorporated into the peptide substrate in counts/ min. Assays were done in triplicate.

RESULTS
Crystal Structures of the Pho85-Pcl10 CDK-Cyclin Complex-As indicated under "Experimental Procedures," only the complex of Pho85 with the Pcl10 fragment (Pho85-Pcl10 and Pho85 Pcl10 for only the Pho85 in the complex) could be crystallized, and its structure was determined by molecular replacement using as the search model the 2.9 Å structure of the Pho85-Pho80 complex composed of intact subunits (13). The structure was refined at 1.95 Å resolution to R-crys and R-free values of 19.4 and 24.1%, respectively, and excellent geometry ( Table 2 and Fig. 1). Because the structure was determined at much higher resolution than the isomorphous 2.6 Å structure with bound ATP analog (described below), it is the basis of a larger part of the results narrative. The structure of Pho85-Pcl1, one of the highest resolution CDK-cyclin structures, is very similar overall to the 2.9 Å Pho85-Pho80 complex structure. The backbone r.m.s.d. between the two complexes is 4.5 Å, a somewhat high value attributable to the difference in the relative orientations between the kinase and cyclin subunits. This is confirmed by better r.m.s.d. values obtained in superpositioning only the Pho85s (1.2 Å) or Pcl10 and Pho80 cyclins (1.9 Å).
The Pho85-Pcl10 structure is also similar overall to the fully activated, phosphorylated CDK2-cyclin A complex, with bound ATP analog AMPPNP, Mg 2ϩ , and a peptide substrate (HHASPRK) containing the SPRK canonical substrate phosphorylation motif (16) (the structure identified hereafter for brevity as pCDK2-SPRK) ( Fig. 2A)). Consistent with high amino acid sequence identities (50 -60%), the CDK structures are much more conserved than those of cyclins. Conservation of sequences and structure is the highest for the core cyclin box motif, which is composed of five ␣ helices.
Superpositioning of the all helix cyclin structures of Pcl10, Pho80, cyclin A, and p25 shows a better, although imperfect, overlap of the cyclin box helices (Fig. 2B) as compared with that based on the CDKs superpositioning (Fig. 2C). This reflects the different orientations each cyclin subunit assumes relative to its cognate CDK subunit in the different CDK-cyclin complexes. It should be noted that Pho80 is the only intact cyclin; Pcl10, cyclin A and p25 are truncations, proteolytic products of their respective full-length activators.
The distinctiveness of the different cyclin structures arises mainly from the presence of ␣ helices and loops between them that precede or follow the cyclin boxes, which in most cases differ in number and relative orientations (Fig. 2, B and C). Unique to Pcl10, its cyclin box is preceded by two well separated helices (␣NT1 and ␣NT2) and followed by one helix (␣CT1) connected with ␣5 by a very long loop. In contrast, the cyclin boxes in Pho80 and p25 are preceded by one helix (␣NT1) and followed by two helices (␣CT1 and ␣CT2). The positions of the ␣NT1 helices in Pcl10, Pho80 and p25, are nearly clustered, whereas those of the ␣CT1 helices are far apart. The ␣CT2 helices in Pho80 and p25 are in a similar orientation. Totally different patterns are seen for the additional ␣ helices in cyclin A; not only is there a single ␣NT and ␣CT, but neither helix is close to the corresponding positions in Pcl10, Pho80, and p25 (Fig. 2, B and C). The ␣NT1 of cyclin A is positioned at the interface with CDK2, where it influences formation of the heterodimer interface and active site region. In contrast since ␣NT1 helices of Pcl10, Pho80, and p25 are found clustered at the opposite side, they have no influence on the interface or active site region.
Binding of the ATP Analog, ATP␥S, and the Kinase "Glycinerich" Loop-The structure of Pho85-Pcl10 with bound ATP␥S-Mg 2ϩ was determined by molecular replacement using the Pho85-Pcl10 structure as the template and refined at 2.6 Å resolution ("Experimental Procedures" and Table 2). The difference electron density of the nucleotide is shown in Fig. 3A. The backbone r.m.s.d. between the nucleotide-free and -bound Pho85-Pcl10 structures is 0.65 Å. The binding geometry of ATP␥S-Mg 2ϩ is extremely similar to that of the ATP analog, AMPPNP-Mg 2ϩ , in the pCDK2-SPRK structure, underscoring the highly conserved nature of the ATP binding sites in the two different groups of CDKs. Both ATP␥S-Mg 2ϩ and AMPPNP-Mg 2ϩ are superposable, as are the conserved nonpolar and polar residues contacting the nucleotide analogs (Fig. 3, B and C). The conserved nonpolar residues surrounding the adenine rings are Ile-10, Val-18, Phe-80, Phe-82, and Leu-134 in CDK2, which matches with Leu-13, Val-21, Phe-82, Phe-84, and Leu-140, respectively, in Pho85. The conserved polar residues include the invariant catalytic triad of KED residues, which is considered to direct the position and orientation of the ATP phosphate tail for an in-line attack by the substrate threonine/serine hydroxyl group (1, 16) (further discussed below).
The only pronounced difference between the two ATP binding sites is the conformations of the Gly-rich loop (14 -19 in Pho85 and 11-16 in pCDK2), which forms a lid or roof on the site, provides a phosphate anchor in ATP binding, and deploys conserved tyrosine residues (Tyr-18 in Pho85 and Tyr-15 in CDK2) on the tip of the loop which may or may not be important for activity (Fig. 3B). Interestingly, the loop lid in Pho85 is closer to the ATP analog than in pCDK2, but this apparently has no effect in keeping the identical nucleotide binding mode.
Full Activation of Pho85-Pcl without Phosphorylation by Mimicry-In the cell cycle, phosphorylation of a specific threonine or serine residue (e.g. Thr-160 in CDK2) on the CDK activation loop by a CDK-activating kinase is absolutely required for maximal activation of the CDK-cyclin complexes (1). In sharp contrast, it is not obligatory for the complexes of Pho85 with Pcls even though an equivalent residue (Ser-166) is present (10,11). Moreover, neither Pho85-Pcl10 nor Pho85 alone is phosphorylated in vitro or in vivo (11). A comparison between the structure of Pho85-Pcl10 with those of unphosphorylated and phosphorylated CDK2-cyclin A sheds new light into the mechanism by which the Pho85-Pcl10 and, by exten- Only the first of two cyclin boxes of CDK2, which match with only one cyclin box in Pcl10 is shown (see also panels B and C). AMPPNP is the ATP analog and the SPRK is the phosphorylation motif of the HHASPRK substrate bound in the active site between the N-terminal ␤-strand-rich lobe and the C-terminal ␣-helix-rich lobe. For positional reference, the functionally critical ionic lock (Fig. 1) is shown. B, relative orientations of ␣ helices in the cyclins Pcl10 (green), Pho80 (red), cyclin A (gray), and p25 (yellow). They are based on superpositioning of the cyclins as they occur in the structures of the Pho85-Pcl10, Pho85-Pho80 (PDB entry 2PK9) (13), pCDK2-SPRK (16), and CDK5-p25 (PDB entry 1H4L) (12). Like Pcl10, Pho80 and p25 contain only one cyclin box. Protein Data Bank analysis of the coordinates of p25 in the CDK5-p25 structure as well as those performed by the authors indicated that p25 lacks cyclin box ␣4 helix. The ␣1Ј helix of cyclin A is the first helix of the second cyclin box. C, similar to panel B, except that it was obtained by superpositioning of the kinases.
sion, the superfamily of Pho85-Pcl complexes circumvent activation loop phosphorylation.
Whereas the flexibility and accessibility of the region surrounding the activation loop Thr-160 are important for modulation of the CDK2-cyclin A catalytic activity, they play a less significant role in Pho85-Pcl10. In the CDK2-cyclin A structure (23), the salt links in CDK2 between glutamate (Glu-162) and two (Arg-126 and Arg-150) of the set of three close-by conserved arginines keep the short activation loop segment (residues 158 -164), which we named the "substrate recognition segment" (or SRS) for this portion of the activation loop, in a conformation that for the most part sterically hinders substrate binding in the catalytic site while maintaining Thr-160 in an exposed position (Fig. 4, A and B). By pushing against the SRS, Tyr-180 also contributes in restricting the SRS to the inhibitory configuration. After phosphorylation of Thr-160 (16,24), the SRS and adjoining sections undergo conformational change, lifting them out of its inhibitory position (Fig. 4, B and C). The conformational change is further accompanied by the following prominent rearrangements (Fig. 4, B and C): 1) major displacement of Glu-162 to the protein surface by the phosphate group, which in turn forms two salt links with two arginines; 2) flipping of Tyr-180 away from the segment; 3) transformation of Val-164 to a left-handed configuration concomitant with the repositioning of the Val-163 carbonyl group. The phosphoryl adduct further engages in hydrogen bond array with RYT (Arg-126 guanidinium-Tyr-180hydroxyl-Thr(P)-160 main chain carbonyl oxygen) that further stabilizes the SRS conformation (Fig. 4C). The overall net effect is an SRS secured in place and a completely open, accessible, and rigidified active site with a geometry tailored for productive substrate binding.
In contrast to the unphosphorylated CDK2 complexed with cyclin A, the equivalent SRS in Pho85 (residues 164 -170) complexed with Pcl10 is already poised and stabilized in an active conformation, with the functionally important tandem valines, Val-169 -Val-170, exhibiting the identical unusual geometry of the equivalent Val-163-Val-164 in pCDK2 (Fig. 5, A and B). Remarkably, this is achieved largely with the carboxylate side chain of Asp-376 on the ␣3-␣4 loop of the Pcl10 cyclin box almost exactly taking the place of the phosphoryl adduct in pCDK2 (Figs. 1, 5B, 6A, and 7A). Asp-376 carboxylate tethers the SRS in the active configuration through several electrostatic interactions with Pho85 residues, including the following (Figs. 5, A and B, and 6A); via salt links with two of the three conserved arginines (Arg-132 and Arg-156) and a hydrogen bond array with RYS (Arg-132-Tyr-186 -Ser-166), both mimicking those involving the phosphoryl group in pCDK2, and also via a charge-neutral hydrogen bond with the main chain NH of Ser-166. In addition to the hydrogen bonds involving its main chain NH and CO groups, the side chain hydroxyl of Ser-166 is engaged in a neutral-charge hydrogen bond with Glu-168, the equivalent in sequence and position of Glu-162 of pCDK2, and van der Waals contact with Phe-377 on the Pcl10 ␣3-␣4 loop.
Asp-376 side chain of Pcl10 serves as a surrogate of the phosphoryl adduct. The observations that the charge-coupling interactions between the Pcl10 aspartate and Pho85 arginines (collectively called "ionic lock") are also closely matched in the Pho85-Pho80 structure and that the aspartate is invariant in all 10 Pcls are concrete evidence for a common phosphorylation bypass mechanism in the Pho85-Pcl CDK-cyclin superfamily ( Fig. 7A and Table 1). However, as a consequence of crystal packing, the SRS conformation in the structure of Pho85 in complex with Pho80 deviates considerably from that in the complex with Pcl10 (Fig. 7B), indicating the critical importance of the latter productive complex in understanding function and of the nonproductive complex with Pho80 in being mindful of possible deleterious effects of crystal packing in crystal structure studies.
The closest functional and structural homolog of Pho85, the neuron-specific CDK5, does not require phosphorylation either (8,9) despite the presence in its SRS (residues 157-163) of a serine residue (Ser-159) at a position equivalent to the CDK2 Thr-160 and Pho85 Ser-166. The SRS of the CDK5 complexed with p25 (12) also adopts a similar overall conformation to those in pCDK2 and Pho85-Pcl10 (Figs. 5C and 6A). Surprisingly, we observed a buried water or preferably hydroxide molecule taking the place of the phosphate in pCDK2 or the Asp-376 in Pcl10 (Figs. 5C and 6A). Much like the phosphoryl group in pCDK2 and Asp-376 in Pcl10, the hydroxide is tightly held in place by salt links with two of the three conserved CDK5 arginine residues (Arg-125 and Arg-149). It further engages in two hydrogen bonds with the carbonyl oxygens of Gly-238 and Asn-239 on the loop after the ␣3 helix in p25. The hydroxide is also linked to an RY(T/S) hydrogen bond array (Arg-125-Tyr-179 -Ser-159). The side chain atoms of CDK5 Ser-159 are in van der Waals contacts with the Asn-239 and Ile-241 side chains of p25. It is noteworthy that, whereas several interactions are made by the Pho85 Ser-166 and CDK5 Ser-159, none are formed by the counterpart Thr-160 in the unphosphorylated CDK2 (Fig. 4, A  and B).
Although the water/hydroxide molecule is accounted for in the deposited coordinates for the CDK5-p25 complex structure (PDB entry 1H4L) (12) to our knowledge, it is described here for the first time, notably within the context of the activation loop segment productive conformation. Because of its association with two guanidinium side chains, we favor a hydroxide anion with its lone proton making bifurcated hydrogen bonds to both peptide carbonyl oxygens of Gly-238 and Ile-241 of p25 (Fig.  5C). If it were a water molecule, its two protons could ideally form bidentate hydrogen bonds with the carbonyl oxygens (i.e. each carbonyl oxygen accepting a proton).
An overlap of the SRSs and conserved/invariant residues depicted in Fig. 6 shows the overall conservation of the features for full activity of pCDK2, Pho85, and CDK5 kinases in association with cognate cyclins. Notably, the phosphoryl adduct of pCDK2, Asp-376 of Pcl10, and the hydroxide in the CDK5-p25 are in very similar locations and serving as a major conduits for very similar interactions (Fig. 6A). The combined configuration of the tandem VV residues of the SRSs imposed by the left-handed conformation of the second valine (pCDK2 Val-164, Pho85 Val-170, and CDK5 Val-163) perfectly overlap (Fig. 6B). . The x-ray structure of the pCDK2-SPRK has revealed the molecular basis for the functional requirements of the positions of the three residues in the motif (16). The position 0 threonine hydroxyl side chain is in a precise location to attack the ␥-phosphate of ATP. The proline in the ϩ1 position is docked in a shallow pocket created by the phosphorylation-induced transformation of Val-164 to a lefthanded configuration coupled with the repositioning of the Val-163 carbonyl (Figs. 4B and 5B). The foregoing two features of substrate motifs recognition are most likely preserved in CDK5 and Pho85 in complexes with their cognate cyclins (portrayed in Fig. 8). Docking of the ϩ1 P is a key feature in facilitating the proper positioning of the serine/threonine nucleophile. The ϩ3 position lysine is recognized principally by the salt link with the phosphoryl adduct of Thr-160 (Figs. 4B and 5B), highlighting an additional critical role of the phosphoryl group. However, this role absolutely cannot be duplicated in CDK5-p35 complex and the Pho85-Pcl group.
Studies of substrate recognition by other CDK-cyclin complexes have come to rely on the pCDK2-SPRK structure as a model for substrate binding, given the high structural homology of the CDKs (e.g. see Figs. 2A and 3), including the active site region (e.g. see Fig. 6), followed by assessing the modeling results through binding studies of modified peptide substrates and/or active site residues by site-directed mutagenesis. This approach provided a credible explanation for the recognition by CDK5-p25 of the position ϩ3 residue of the SPXK substrate motif (12). Although the positively charged ϩ3 residue is identical to the cell cycle CDK substrates, the mechanism for its recognition by CDK5-p25 differs totally as the kinase requires no activation loop phosphorylation. Instead, the carboxylate side chain of Glu-240 on the loop that follows the p25 cyclin box ␣3 helix is expected to form a salt link with the ϩ3 Lys/Arg side chain (Fig. 8A). Using the same approach, we derived a similar mechanism for the ϩ3 Leu/Ile recognition of the SPX(L/I) motif of several phosphorylation sites in the Pho4 substrate of the of Pho85-Pho80 (13) ( Table 1). The Phe-138 deployed by the ␣3-␣4 loop is strongly implicated in the recognition of the large aliphatic residues at the ϩ3 position (Fig. 8A). Being engaged in the ionic lock, the close by Asp-136 fulfills another role by keeping the Phe-138 in place (Fig. 8A).
The fact that the position ϩ3 residues of the Pho85-Pcl superfamily substrates vary considerably (Table 1) poses challenges in further understanding the mechanism for substrate recognition. This is compounded by the observation that, based on the sequences of the six-residue ␣3-␣4 loops across the Pcls and the assumption of similar loop conformations (e.g. Fig. 7A), the residues in the ␣3-␣4 loop matching that of Pho80 Phe-138 are not fully complementary to the position ϩ3 residues of sub- strates of the other Pho85-Pcl complexes (Table 1). Possibly the most severe case is the complete incompatibility between the ϩ3 aspartate of the Pho85-Pcl10 Gys2 substrate primary phosphorylation site motif ( 654 SPRD) and the valine residue on the ␣3-␣4 loop of Pcl10 (Val-378, underlined V in Table 1). To assess the significance of this observation, we obtained a model of SPRD bound to the Pho85-Pcl10-ATP␥S structure by superpositioning with pCDK2-SPRK (Fig. 8B) as was done for the motifs of the substrates of CDK5 and Pho85 described above. The model indicates that the Val-378 side chain is ϳ7 Å from the ϩ3 aspartate, implying no contact between the two residues as would be expected. The histidine residue that follows Val-378 is not positioned to interact with the ϩ3 aspartate either (Fig. 8B).
These observations raise the question of whether the ϩ3 aspartate plays a role in substrate recognition. To answer the question, we investigated the phosphorylation incorporation by the full-length Pho85 CDK complexed with the fragment of Pcl10 cyclin to the peptide KKLMVPGSPRDLRS, which encompassed the phosphorylatable Ser-654 of the Gys2 substrate (11) (referred to as SPRDL), and its variants KKLMVPG-SPRALRS (SPRAL) and KKLMVPGSPRDARS (SPRDA) in which the residues in the ϩ3 and ϩ4 positions, respectively, were replaced by alanine ("Experimental Procedures"). Unexpectedly, the SPRAL substrate was ϳ2-fold more active than the normal peptide substrate (Fig. 8C). For comparison, identical replacement of the ϩ3 residue arginine of CDK5-p25 peptide substrate and leucine of the Pho85-Pho80 peptide substrate caused 2.4-and 9.5-fold reductions in activity, respectively (12,13). Although an explanation for the enhanced activity toward substrate with the ϩ3 aspartate to alanine mutation is not straightforward, it could be attributed to relieving the unfavorable proximity (ϳ3.5 Å) of the ϩ3 carboxylate to the carbonyl oxygen of Asp-376 (Fig. 8B). The SPRDA substrate was tried to test the supposition that the leucine in the ϩ4 position could contribute to substrate binding (11), but the test proved negative (Fig. 8C).   (13), whereas that of the Pho80-Pcl10 complex contains only one heterodimer. The packing of two Pho85-Pho80 heterodimers in the asymmetric unit results in the formation of two hydrogen-bonding interactions that distort the segment to a non-productive conformation; 1) bifurcated hydrogen bonds between Glu-168 carboxylate in Pho85 (magenta) in the A heterodimer and the backbone NH group of C152 in Pho80 (red) in the B heterodimer and 2) hydrogen bond between Ser-167 in Pho85 (magenta) in the A heterodimer and Lys-146 in Pho80 (red) in the B heterodimer. An identical set of hydrogen-bonding interactions occurs between Pho85 in the B heterodimer and Pho80 in the A heterodimer. These interactions distort the conformation of the activation loop segment as compared with the productive segment in Pho85-Pcl10. No distortion is seen for the equivalent activation loop segment in the Pho85-Pcl10 complex structure primarily, as it has one heterodimer per asymmetric unit.

DISCUSSION
The overall conservation of the SRS conformations as well as the surrounding invariant/similar residues in the three different CDK-cyclin complexes (pCDK2-cyclin A, Pho85-Pcl10, and CDK5-p25) (Fig. 6), attest to the essential nature of the ensemble in attaining full kinase activity. Surprisingly, conservation is achieved via different means; that is, phosphorylation, ionic lock formation, and hydroxide-mediated electrostatic interactions.
The structure determination of Pho85-Pcl10 combined with that of Pho85-Pho80 establishes the mechanism for circumventing activation loop phosphorylation for the Pho85-Pcl superfamily. It further sheds new light into activation and substrate recognition. The ionic lock between Pcl10 Asp-376 and a cluster of Pho85 arginines combined with proximal interaction networks, accomplishes two key functions; it confers the fully active conformation of the activation loop, especially the SRS, and circumvents phosphorylation. The combination of interactions could further inhibit potential activation loop phosphorylation by preventing Pho85 Glu-168 from assuming a position equivalent to its counterpart Glu-162 in the unphosphorylated CDK2 (Fig. 4, A and B) and by keeping Ser-166 shielded (Figs. 5,  A and B, and 6A). Interestingly, the same features may be true for Glu-161 and Ser-159 in the CDK5-p25, the counterpart of Pho85 Glu-168 and Ser-166, respectively (Figs. 5C and 6A).
Unlike the first two residues ((S/T)P) of the substrate phosphorylation motifs for the proline-directed CDKs, with the key ϩ1 Pro residue nestled in a shallow pocket, there is no unified mechanism for the substrate recognition of the residues at the ϩ3 position. Even though a basic residue occupies the ϩ3 position of the pCDK2 and CDK5 substrate phosphorylation motif (SPXK), it is recognized with equal high specificity by two different groups, the phosphoryl adduct and p25 glutamate, respectively (Fig. 8A). The structure determination of Pho85-Pho80 uncovered a mechanism similar to CDK5; a phenylalanine (Phe-138) on the ␣3-␣4 loop of the Pho80 cyclin box is strongly implicated in the recognition of the ϩ3 Ile/Leu of the substrate Pho4 consensus sequence (SPX(I/L); Fig. 8, A and B) (13). The same mechanism presumably applies to the Rim15 kinase substrate of Pho85-Pho80, which has a motif identical to that of Pho4 (Table 1). Thus, for CDK5-p25 and Pho85-Pho80 the recognition requirements for the ϩ3 residues are strongly dictated by the cognate cyclins.
The paramount question is whether the highly favorable recognition mechanism between Pho80 Phe-138 and ϩ3 Ile/Leu is replicated for the remaining nine Pho85-Pcl/Clg complexes. The data indicate that the Pho80 Phe-138 counterparts in the other complexes vary, but not in a highly complementary manner to the different substrate ϩ3 residues ( Table 1). The severest is between the valine on the Pcl10 ␣3-␣4 loop and the aspartate at position ϩ3 in the Gys2 substrate of Pho85-Pcl10 (Table 1). Our modeling and phosphorylation studies of peptide substrates containing the Gys2 phosphorylation sequence SPRDL and variants replacing the ϩ3 Asp or ϩ4 Leu with alanine show that neither residue position contributes to substrate recognition (Fig. 8, B and C).
Another severe case of incompatibility is between the Phe-138 of Pho80 and the ϩ3 Lys/Arg in the three Sic1 substrate phosphorylation sites (Table 1), which replicates the ϩ3 Lys phosphorylation motifs of pCDK2 and CDK5 substrates (e.g. see Figs. 5B and 8A). Additionally, Pho85 in combination with two other Pcls (Clg1 and Pcl1) also phosphorylates three sites of the Sic1 substrate, but the regulatory subunits contain threo- FIGURE 8. Dispensability of the aspartate in the ؉3 position of the phosphorylation sequence motif of the Pho85-Pcl10 substrate Gsy2. A, potential residues on the ␣3-␣4 loop of cyclins for recognition of the ϩ3 position residues Lys (yellow) in the CDK5-p25 substrate and Leu (red) of several sites in the Pho4 substrate of Pho85-Pho80 (13) (described in text). Overlap is based on the kinase structures. The phosphorylation peptide substrate motif minus the X residue in the ϩ2 position is from the overlap with the pCDK2-SPRK structure. B, models of the bound peptide phosphorylation motifs of 654 SPRD in the Gsy2 substrate and SPXL in the Pho4 substrate of Pho85-Pho80 obtained as in panel A. The ϩ3 Lys of the SPRK was changed to either Asp or Leu to conform to the ϩ3 residue of Gsy2 or Pho4, respectively. The residue in the ϩ2 (X) position is arginine in both substrates of pCDK2 and Pho85-Pcl10. The carboxylate side chain of the ϩ3 aspartate of Gys2 peptide motif is ϳ7 Å to the side chain of Val-378 (black dashed circle) and ϳ3.5 Å to the backbone carbonyl oxygen of Asp-376 of Pcl10 (magenta dashed circle). The Leu (ϩ3) side chain of Pho4 is ϳ3.5 Å to the Phe-138 side chain of Pho80 and ϳ4 Å to the backbone carbonyl oxygen of Asp-136 of Pho80, the equivalent of Asp-376 of Pcl10. C, kinase activities of Pho85-Pcl10 toward three different synthetic peptides as substrates.
nine and serine residues, respectively, on the ␣3-␣4 loop (Table  1). Although this would be less deleterious than the Pho80 Phe-138, it is still not ideal for recognizing the ϩ3K. Pho85-Pcl1/2 complexes pose a different recognition problem. The two substrates of these complexes have a diverse number of phosphorylation site ϩ3 residues, including Ser, Leu, Pro, and Ala, but the Pho80 Phe-138 counterpart residue is serine in both Pcls (Table  1). There are also incompatibilities of varying degrees between the complexes of Pho85 with Pcl5, Pcl6/7, and Pcl9 and the ϩ3 residues of the respective substrates ( Table 1).
The principal conclusion that can be drawn from the structural and functional studies reported and discussed here are as follows. The ionic lock enables Pcl10 more direct involvement in shaping the active site geometry in Pho85. Remarkably, three different negatively charged groups (phosphate in pCDK2-cyclin, aspartate in Pcls, and hydroxide in CDK5-p25) serve as a lynchpin for conferring or maintaining similar molecular geometry for full kinase activity. Unlike the requirement of the ϩ3 position residues of phosphorylation motifs of substrates of pCDK2-cyclin A, CDK5-p25, and Pho85-Pho80, the same residue position in the Gsy2 substrate of Pho85-Pcl10 makes no contribution to substrate recognition. Little or no contribution is also expected from the substrates of the eight remaining members of the Pho85-Pcl superfamily.