How Tyrosine 15 Phosphorylation Inhibits the Activity of Cyclin-dependent Kinase 2-Cyclin A*

Inhibition of cyclin-dependent kinase 1 (CDK1) activity by Tyr-15 phosphorylation directly regulates entry into mitosis and is an important element in the control of the unperturbed cell cycle. Active site phosphorylation of other members of the CDK family that regulate cell cycle progression instates checkpoints that are fundamental to eukaryotic cell cycle regulation. Kinetic and crystallographic analyses of CDK2-cyclin A complexes reveal that this inhibitory mechanism operates through steric blockade of peptide substrate binding and through the creation of an environment that favors a non-productive conformation of the terminal group of ATP. By contrast, tyrosine phosphorylation of CDK2 alters neither its Km for ATP nor its significant intrinsic ATPase activity. Tyr-15-phosphorylated CDK2 retains trace protein phosphorylation activity that should be considered in quantitative and qualitative cell cycle models.

The cyclin-dependent protein kinase (CDK) 4 family constitutes a highly conserved but functionally diverse family that is essential for many cellular regulatory processes, notably in the regulation of transcription and cell cycle progression (reviewed in Refs. 1 and 2). For full activity, CDKs absolutely require association with an activating cyclin molecule, and they may also require phosphorylation of a conserved threonine residue that lies within the activation loop (Thr-160 in CDK2). Beyond these effects, CDKs are regulated by multiple mechanisms that include inhibitory phosphorylation of conserved residues within the active site (3) and binding of additional regulatory molecules (4).
Early studies in Schizosaccharomyces pombe demonstrated that Cdc2 (CDK1) Tyr-15 phosphorylation directly regulates entry into mitosis and is an important element in the control of the unperturbed cell cycle (5)(6)(7). Subsequent studies showed that in higher vertebrates the adjacent threonine residue (Thr- 14) is also a site of inhibitory phosphorylation, (8,9) and that phosphorylation at either or both of these sites abrogates CDK1 kinase activity (5,9,10). In human cells, the mechanism of inhibition of CDK1-cyclin complexes by phosphorylation is conserved and plays a major role in controlling passage through the G 2 /M transition point (reviewed in Ref. 11). Saccharomyces cerevisiae appears to be the exception in that phosphorylation of Tyr-19 is detectable (the residue equivalent to CDK1 Tyr-15) but its dephosphorylation is not a requirement for entry into mitosis, suggesting that other mechanisms must regulate the transition (12,13).
Phosphorylation of CDK1 on Tyr-15 is catalyzed by members of the Wee1/Mik1 kinase family (14 -17) (reviewed in Ref. 3). The related kinase Myt1 phosphorylates CDK1 Tyr-15 and Thr-14 (18 -20). In S. pombe and S. cerevisiae there is a single Cdc25 dual specificity phosphatase that regulates tyrosine dephosphorylation of Cdc2 and Cdc28 (21,22) (reviewed in Refs. 1,3). In human cells there are three Cdc25 isoforms (Cdc25 A, B, and C). Initial characterization suggested that they had discrete and non-overlapping roles in regulating CDK activation. However, this model has had to be re-evaluated as an increasing body of evidence supports a model in which the three isoforms cooperate to regulate the major cell cycle transitions (reviewed in Ref. 23).
Other members of the human CDK family that regulate the cell cycle (CDKs 2, 4, and 6) all contain a tyrosine residue equivalent to CDK1 Tyr-15, and phosphoforms of these CDKs, phosphorylated on this tyrosine residue, have been detected in cultured cell lines (24,25). Whether inhibitory phosphorylation is as crucial for their activities in the unperturbed cell cycle as it is for the activity of CDK1 is still unresolved. CDK2 is phosphorylated on Thr-14 and Tyr-15 during S and G 2 phases and is dephosphorylated by Cdc25 (24, 26 -28). Indirect evidence suggests that this phosphorylation might regulate CDK2 activity in this context: modulation of Cdc25A activity either by ectopic overexpression (28) or by immunodepletion (29,30) results, respectively, in an accelerated G 1 /S transition or in a G 1 /S block. As CDK2 is the major cell cycle regulator during these cell cycle transitions, it has been proposed that these effects are as a result of Cdc25A activity against CDK2. Indeed Cdc25A overexpression resulted in an increase in CDK2-cyclin A and cyclin E activity concomitant with CDK2 Tyr-15 dephosphorylation (30). However, recent work showing that CDK1 can substitute for CDK2 to promote S-phase, as revealed by experiments in CDK2 Ϫ/Ϫ p27 Ϫ/Ϫ cells (31), would suggest that further work is required to establish whether CDK2 and/or CDK1 is the essential Cdc25A target at G 1 /S.
In contrast, the importance of CDK inhibition by phosphorylation in establishing cell cycle arrest in response to checkpoint activation is well documented. Active site phosphorylation of CDK1 (32,33), CDK2 (34 -36), and CDK4 (37) has been observed as a response to DNA damage (reviewed in Refs. 38 -40). Exposure to transforming growth factor ␤ results in a cell cycle arrest that is correlated with elevated levels of tyrosine-phosphorylated CDKs 4 and 6 (25) and that may be mediated by down-regulation of Cdc25A.
Genetic and biochemical studies to dissect the effect of tyrosine and threonine phosphorylation on CDK activity have focused on CDK1 because of its established role in controlling entry into M-phase. In mouse cells, expression of a CDK1 (Y15E) mutant produces a dominant negative phenotype (9). Similarly chicken mutants (CDK1(T14E), CDK1(Y15E), CDK1(T14D, Y15F), and CDK1(T14A, Y15D)) were unable to rescue fission yeast cells lacking wild-type CDK1 function and caused a cdc phenotype upon overexpression (7). These results suggest that negative charge at either position 14 or 15 is sufficient to inhibit kinase activity.
Early biochemical studies suggested that Tyr-15-phosphoylated CDK1 is inactive (15). Molecular models of CDK1 have suggested that a phosphate group located on Tyr-15 would present a steric and charge impediment to peptide substrate binding (41,42). Biochemical characterization of suicide inhibitor labeling of Tyr-15-phosphorylated CDK1 has further suggested that phosphorylation of Tyr-15 might not inhibit ATP binding (43), an hypothesis supported by molecular modeling studies (44). Despite extensive x-ray crystallographic analysis of members of the CDK family, an atomic structure for a CDK in this phosphorylated state has not been described. In this report we present both the structure and the kinetic properties of CDK2-cyclin A phosphorylated on Tyr-15 and Thr-160. These results are discussed with respect to the molecular mechanism of CDK inhibitory phosphorylation and its physiological consequences.

EXPERIMENTAL PROCEDURES
Purification of Phosphorylated CDK2-Cyclin A Complexes-T160pCDK2-cyclin A was purified as described (45). Y15pT160pCDK2-cyclin A was purified by one of two methods. In the first method (Method 1), a modified pGEX-6P-1 vector (Amersham Biosciences) that independently expresses human CDK2 and the human Wee1 kinase catalytic domain (46) as glutathione S-transferase fusion proteins (pGEX-6P-CDK2-Wee1) was transformed into Escherichia coli strain B834(DE3)/LysS. Notably, only GST-CDK2 contains a cleavage site for PreScission protease. GST-CDK2 and GST-Wee1 fusion proteins were purified together by affinity chromatography, and then GST-CDK2 was phosphorylated in vitro on Thr-160 by GST-CIV1 as described (47). Following a buffer exchange step, the mixture of GST-Y15pT160pCDK2, GST-CIV1, and GST-Wee1 was re-applied to a glutathione-Sepharose 4B column. Lysate from a culture of E. coli B834(DE3)/ pLysS cells expressing a truncated version of human cyclin A (residues 174 -432) (45) was passed over the column. The binary Y15pT160pCDK2-cyclin A complex, GST-Wee1, and GST-CIV1 were eluted with glutathione; the mixture was incubated with PreScission protease and then subjected to gel filtration (Superdex 75 HR26/60; Amersham Biosciences) followed by a glutathione-Sepharose 4B column. The binary complex was recovered in the flow-through, and any remaining GST-CIV1 or GST-Wee1 was retained on the final column.
In the second method (Method 2), pGEX-6P-CDK2-Wee1 and a pACYC-derived vector to express GST-CIV1 were cotransformed into E. coli strain B834(DE3). This expression system produced CDK2 phosphorylated on Tyr-15 and Thr-160. The Y15pT160pCDK2-cyclin A complex was subsequently prepared by a modified method to that described for preparation of T160pCDK2-cyclin A (45). Briefly, Y15pT160pCDK2cyclin A was purified by sequential affinity (glutathione-Sepharose 4B), size-exclusion, and then affinity chromatography steps as described (45). The sample was then enriched for Y15pT160pCDK2 by exploiting a GST-Cdc25A affinity column. The Cdc25A construct used (GST-Cdc25A C430S (336 -523) ) encodes glutathione S-transferase fused to Cdc25A residues 336 to 523 and is a substrate-trapping mutant in which the catalytic cysteine (Cys-430) is mutated to serine. The construct was expressed in E. coli strain B834(DE3), and an S100 lysate was prepared and loaded onto a glutathione-Sepharose 4B column pre-equilibrated with HEPES-buffered saline. 10 mg of purified Y15pT160pCDK2-cyclin A was then applied to the GST-Cdc25A C430S (336 -523) affinity column and washed to baseline with HEPES-buffered saline. Bound Y15pT160pCDK2-cyclin A was eluted with a gradient of dihydrogen ammonium phosphate (0 -0.5 M). The eluted complex was buffer exchanged into HEPES-buffered saline and glycerol added to 50%. The inclusion of this additional affinity step in the purification procedure consistently enriched for Tyr-15-phosphorylated CDK2-cyclin A. Y15pT160pCDK2-cyclin A prepared by Method 2 was used for the ATPase assays and kinetic analysis.
Analysis of Phosphorylated CDK2-The extinction coefficient (⑀ 280 ) of T160pCDK2-cyclin A was determined by amino acid analysis. Subsequently, protein concentrations were determined from measured A 280 values and an ⑀ 280 ϭ 67420 M Ϫ1 cm Ϫ1 . The extent of Tyr-15 phosphorylation was determined by N-terminal sequencing (described in supplemental information).
Estimating the Extent of Phosphorylation of Tyr-15-The amount of amino acid liberated at each cycle of N-terminal sequencing was recorded for amino acid Gly-3 and for amino acids Gly-13 through Gly-19 in each of three separate sequencing runs of T160pCDK2-cyclin A (data presented in supplemental information for material prepared by Method 2, Table  1). Data for the different runs were placed on a common scale by dividing the amount of amino acid liberated at each cycle of a given run by the mean amount of amino acid liberated for cycles corresponding to Gly-3, Gly-13 through Thr-14, and Gly-16 through Tyr-19 of that run. The data were further normal-ized for cycle-to-cycle effects (principally expected to result from the relative efficiency of hydrolysis and derivatization of different amino acids) by dividing the scaled datum corresponding to a given cycle of a given run by the mean of the scaled data observed for that particular cycle in all runs. The variance of the resulting distribution of values, which should have a mean value of 1.0 for homogenous samples, is an estimate of the accuracy of the N-terminal sequencing protocol in determining the occupancy of the expected amino acid at any given sequence position. In this case, an accuracy of Ϯ 0.11 was observed.
A further four runs of N-terminal sequencing were carried out on the doubly phosphorylated sample, of which two yielded sequencing data over an amino acid range comparable with that sequenced in the singly phosphorylated samples. Run-to-run scaling was carried out as above, but the cycle-to-cycle scaling was performed using the scale factors determined for the singly phosphorylated sample. Correction by this method of the observed yields of unphosphorylated tyrosine at position 15 of the doubly phosphorylated samples yielded occupancies of 0.042 and 0.055, i.e. a mean of 0.049, with an S.E. of the mean of 0.077.
ATPase Assay-The hydrolysis of ATP was monitored using a coupled spectrophotometric assay using material prepared by Method 2 as described previously (Refs. 48, 49 and supplemental information).

Assay of the Kinetics of Phosphorylation of Histone H1 by T160pCDK2-Cyclin A and Y15pT160pCDK2-Cyclin A-
T160pCDK2-cyclin A (5.2 nM, prepared by Method 2) was incubated with histone H1 substrate at varying concentrations in the range 6.25-100 M. The kinase reaction was started by the addition of [␥-32 P]ATP/Mg 2ϩ (600 cpm/pmol) in a final reaction volume of 10 or 20 l containing 1 mM ATP, 10 mM MgCl 2 , 200 mM NaCl, 0.01% monothioglycerol, 100 mM Tris, pH 7.4. Triplicate reactions were allowed to proceed at 30°C for 10 min before being terminated by the addition of 50% acetic acid. Phosphorylated histone was separated from free ATP by p81 paper cation exchange chromatography. Phosphate incorporation and hence the rates of reaction were determined by Cerenkov counting. The kinetic parameters from the data were determined by non-linear regression analysis. Y15pT160pCDK2-cyclin A was analyzed similarly except that a higher final enzyme concentration (1.6 M) and a higher substrate concentration range (up to 200 M) were used, reflecting the lower activity and higher K m of this sample. Analysis of the variation of initial rate with concentration of ATP for Y15pT160pCDK2-cyclin A suggested trace contamination of the sample with the more highly active T160pCDK2-cyclin A, particularly apparent at lower concentrations of substrate (Fig.  4). Regression analysis for this sample therefore used data from the higher concentration regime (50 -200 M). Attempts to model the contribution of T160pCDK2-cyclin A to the kinetic data recorded from the Y15pT160pCDK2-cyclin A sample indicated a contamination of Ͻ1%, a value within the estimated accuracy of the amount indicated by N-terminal sequencing.
Crystallization of Phosphorylated CDK2-Cyclin A Binary Complexes-Y15pT160pCDK2-cyclin A complex prepared by Method 1 in 40 mM HEPES, pH 7.4, 200 mM NaCl to 10 mg/ml and 5Ј-adenylylimido-diphosphate (AMPPNP) (Sigma) was added to an aliquot of the protein solution to a final concentra-tion of 1 mM. Crystals were grown at 4°C in 100 mM HEPES, pH 7.0, 5 mM dithiothreitol, 0.7 M KCl, and 1.20 M (NH 4 ) 2 SO 4. Crystals of T160pCDK2-cyclin A-AMPPNP were grown under very similar conditions as previously described (45). A native Y15pT160pCDK2-cyclin A-AMPPNP dataset was collected from a crystal that, although soaked in a substrate peptide solution (sequence PKTPKKAKKL), had no bound peptide. After failing to bind peptide to these crystals despite repeated attempts, protein cross-linking was used to allow soaking at higher peptide concentrations. The crystals (T160pCDK2-cyclin A-AMPPNP and Y15pT160pCDK2-cyclin A-AMPPNP) were first cross-linked for 30 -60 min at 4°C by gentle vapor diffusion of glutaraldehyde added to the well solution (0.1%) and then soaked for 15 min in mother liquor supplemented with 20 mM substrate peptide (HHASPRK, previously shown to bind to T160pCDK2-cyclin A) (45). For all data collection, crystals were briefly cryoprotected in 8 M sodium formate and flash frozen in liquid nitrogen. Data were integrated, scaled, and merged using the programs MOSFLM and SCALA. Statistics for the datasets are presented in Table 1.
Structure Determination and Refinement-The structure of Y15pT160pCDK2-cyclin A-AMPPNP was solved by molecular replacement using the program Molrep via the CCP4 GUI (50). The coordinates of a T160pCDK2-cyclin A complex (Protein Data Bank entry 1JST) (51) were used as the search model. Several cycles of rigid body refinement were initially pursued using REFMAC5. Subsequent refinement was then pursued using REFMAC in conjunction with manual rebuilding using the programs O (52) and Coot (53). A detailed description of all steps in the structure determination and refinement procedure is provided in the supplemental information. Statistics for the final models are presented in Table 1.

RESULTS
The Structure of CDK2-Cyclin A Phosphorylated on Tyr-15 and Thr-160-To provide a structural understanding of the role of Tyr-15 phosphorylation in regulating CDK activity we have determined the structure of a CDK2-cyclin A complex phosphorylated on Tyr-15 and Thr-160 (Y15pT160pCDK2-cyclin A) ( Table 1) and characterized its catalytic activity (Table 2 and Fig. 4) A Cdc25A affinity column was found to be the most effective method to enrich for Y15pT160pCDK2-cyclin A, and exploiting this method preparations were conservatively estimated to contain 95.1% Ϯ 7.8% CDK2 phosphorylated on Tyr-15 (supplemental Table S1). In our early CDK2/Wee1 coexpression studies we detected low levels of Thr-160 phosphorylation by SDS-PAGE gel shift assay and subsequent dephosphorylation by kinase-associated phosphate (KAP) (results not shown). Using more sensitive protein sequencing techniques we did not detect any phosphorylation on Thr-14 (supplemental Table S1). Taken together, these results suggest that the isolated Wee1 catalytic domain retains a degree of selectivity for Tyr-15 but that in this in vitro system Thr-160, and not Thr-14, is additionally susceptible to phosphorylation.
The initial mF o Ϫ F c maps calculated for the Y15pT160pCDK2-cyclin A complex co-crystallized with AMP-PNP included strong electron density for both phosphate groups attached to Tyr-15 and Thr-160 (Fig. 1). The Thr-160 phosphate group reorganizes the CDK2 activation segment to create the peptide substrate-binding site and, through interactions with the guanidinium groups of Arg-50 in the C-helix and Arg-126 and Arg-150 at the start of the activation loop (residues 145-172 between the conserved DFG and APE motifs), indirectly contributes to a tighter CDK2-cyclin A interface (51).
In contrast, phosphorylation of Tyr-15 does not cause any major structural changes in the enzyme (Fig. 2). The phosphorylated Tyr-15 side chain adopts a "swung-out" conformation so that the phosphate group is exposed to solvent and is coordinated through a network of waters to Ser-46 and Thr-47. This is reminiscent of the Tyr-15 side chain conformation observed in the T160pCDK2-cyclin A-AMPPNP-peptide complex (45) but is in contrast to its conformation in a T160pCDK2-cyclin A-ATP complex (Protein Data Bank code 1JST) (51). In this structure the hydroxylated aromatic ring points into the active site and makes a hydrogen bond with the side chain carboxylate group of Glu-51 that coordinates the ATP ␥-phosphate moiety (Fig. 2).
The asymmetric unit of the Y15pT160pCDK2-cyclin A crystals contains two CDK2-cyclin A dimers, with constituent CDK2-cyclin A chains designated A/B and C/D. The electron density for the phosphorylated tyrosine side chain is stronger for one CDK2 molecule (chain A) than the other (chain C), reflecting the tight packing of chain A against a symmetry-related molecule. The presence of the phosphate group on the tyrosine hydroxyl results in a slight opening of the active site where F o,h and F c,h are the observed and calculated structure factor amplitudes for reflection h, k, l. R free is equivalent to R conv but is calculated using a 5% disjoint set of reflections excluded from the least squares refinement stages. cleft but does not perturb the activation loop or the conformation of residues Lys-33, Glu-51, Asp-127, Lys-129, Asn-132, and Asp-145 that are involved in the phospho-transfer reaction. There are also no major conformational changes in the CDK2 residues involved in substrate binding. Biochemical (43) and molecular modeling studies (44) have proposed that Tyr-15 phosphorylation does not inhibit ATP binding. These conclusions were confirmed by the initial mF o Ϫ F c maps contoured over the active sites of both CDK2 molecules in the asymmetric unit that showed unambiguous electron den-sity for the non-hydrolyzable ATP analogue AMPPNP (Fig. 3). The interactions between the AMPPNP adenine ring and ribose moiety and CDK2 residues are conserved with other CDK2-cyclin A-ATP complexes. In a T160pCDK2-cyclin A-AMPPNP-substrate peptide structure (Protein Data Bank code 1QMZ) (45), the ␣-phosphate moiety is coordinated by Asn-132 and hydrogen bonds to Lys-33, while the ␤-phosphate is projected toward the roof of the glycine-rich loop and interacts with the carbonyl group of Glu-12. These interactions contribute to positioning the ␥-phosphate group to point out of the ATP binding pocket in a position compatible with in-line transfer to the substrate serine hydroxyl. A Mg 2ϩ ion coordinates the three phosphate groups, thereby neutralizing some of the negative charge and stabilizing the AMPPNP in the polarized environment.
However, in neither complex A/B or C/D does the ␥-phosphate group of AMPPNP bound to Y15pT160pCDK2-cyclin A adopt a conformation compatible with an in-line phosphotransfer mechanism (Fig. 3). In the first copy, the ␣-phosphate forms a hydrogen bond with Asp-145, so that it moves away from Asn-132 and Lys-33 and is brought closer to the base of the active site. As a result, the ␤-phosphate moiety is no longer in a position to interact with Glu-12. Instead, the ␥-phosphate interacts with the glycine-rich loop and forms a hydrogen bond with the hydroxyl group of the side chain of Thr-14 and an additional weaker interaction with its backbone amide moiety. In the second copy, the ␥-phosphate, rather than contacting the glycine-rich loop, folds back to form a salt bridge with Lys-33. Overall, the observed active sites are less compact, and it is predicted that the essential salt bridge between Glu-51 and Lys-33 is weakened. In the first molecule, one Mg 2ϩ ion coordinates the three phosphate groups and stabilizes the negative charges in the active site. However, in the second copy of CDK2, there is no electron density to support the presence of an Mg 2ϩ ion and the negative charges of the phosphate groups appear to be compensated by the amino group of Lys-33.
Thus, phosphorylation on Tyr-15 does not bring about inhibition by inducing a unique and unproductive conformation for bound ATP. Instead, it creates an environment in which the ␥-phosphate moiety can become engaged in a range of stable and non-productive networks of interactions with CDK2 residues lining the catalytic site. Such unproductive ATP binding modes appear to characterize all but the catalytically poised T160pCDK2-cyclin A-ATP-peptide complex (45).
Kinetic Characterization of the Y15pT160pCDK2-Cyclin A-To assess whether ATP binding to Y15pT160pCDK2-cyclin A is associated with ATP turnover we measured the ATPase activities of Y15pT160pCDK2-cyclin A and T160pCDK2-cyclin A ( Table 2). T160pCDK2-cyclin A has significant ATPase activity (k cat 0.071 Ϯ 0.009 s Ϫ1 ), and this activity was not significantly   (45). The CDK2 structures are drawn in ribbon representation and colored gold, pink, and cyan, respectively, with selected side chains rendered in ball and stick mode. The structure of cyclin A in the Y15pT160pCDK2-cyclin A-AMPPNP structure is rendered in green. ATP adopts distinct conformations in each structure, and only in 1QMZ (cyan) does it bind so that the phospho-transfer reaction can proceed. Carbon, oxygen, and phosphorus atoms of AMPPNP bound to Y15pT160pCDK2-cyclin A are colored green, red, and purple, respectively. Substrate peptide bound to T160pCDK2-cyclin A (Protein Data Bank code 1QMZ) is drawn in cyan to highlight the incompatibility of the position of the Tyr-15 ␥-phosphate moiety with peptide substrate binding. P0 designates the phospho-accepting residue.
affected by Tyr-15 phosphorylation (k cat 0.096 Ϯ 0.0073 s Ϫ1 , Table 2). These results are in agreement with a previous study that also showed that non-phosphorylated CDK2-cyclin A has ATPase activity comparable with the phosphorylated complex (k cat ϭ 0.2 and 0.17 s Ϫ1 for T160pCDK2-cyclin A and CDK2cyclin A, respectively) (54). Taken together, these results suggest that CDK2 has intrinsic ATPase activity that is not significantly affected by either Tyr-15 or Thr-160 phosphorylation.
It has been proposed that Thr-160 phosphorylation increases CDK2 activity by stabilizing the peptide phosphoacceptor group, leading to a large increase in the rate of phosphoryl group transfer (54). CDK2 inhibition by Tyr-15 phosphorylation could be envisaged to inhibit CDK2 by decreasing the affinity of the enzyme for its substrate or by inducing a disposition of the phospho-accepting residue within the bound substrate incompatible with phospho-transfer.
To distinguish between these two possibilities we first attempted to soak a short peptide substrate (HHASPRK) into matched crystals of Y15pT160pCDK2-cyclin A and T160pCDK2-cyclin A. Crystals of each complex were selected that were of comparable size, had been grown in similar crystallization conditions, diffracted to an equivalent resolution, and had the same cell dimensions. We observed unambiguous electron density for peptide substrate bound to T160pCDK2cyclin A, and the final model agrees with the reported complex structure (45) (results not shown). However, despite repeated attempts at high peptide concentrations (20 mM) using native and cross-linked crystals, we were unable to detect electron density for peptide bound in the active site of Y15pT160pCDK2-cyclin A. A superposition of the Y15pT160pCDK2-cyclin A-AMPPNP and T160pCDK2cyclin A-AMPPNP complexes shows that the Tyr-15 phosphate group is located close (Ͻ3 Å) to the backbone carbonyl group of the substrate residue at position p ϩ 3. In this position it presents an impediment to binding through both steric and charge effects, which is sufficient to prevent peptide binding even at high peptide substrate concentrations (Fig. 2).
We next determined the kinase activity of Y15pT160pCDK2-cyclin A and of T160pCDK2-cyclin A toward histone H1 ( Table 2). The low K m contribution of trace amounts of T160pCDK2-cyclin A in the preparation of Y15pT160pCDK2-cyclin A was clearly distinguishable in plots of activity against substrate concentration (Fig. 4), thus allowing unambiguous determination of kinetic parameters for the doubly phosphorylated species. In agreement with previous reports, T160pCDK2-cyclin A has substantial histone H1 kinase activity yielding K m and k cat values of 28 Ϯ 1.9 M and 5.2 Ϯ 0.4 s Ϫ1 , respectively. However, the kinase activity of Y15pT160pCDK2-cyclin A toward histone H1 was substantially reduced. The K m (78 Ϯ 16 M) and k cat (0.08 Ϯ 0.01 s Ϫ1 ) values for Y15pT160pCDK2-cyclin A, determined using histone H1 as a substrate, characterize a kinase that has measurable but reduced activity (k cat /K m reduced 190-fold) compared with the singly phosphorylated enzyme. These values compare with an increase of ϳ100,000-fold in k cat /K m (peptide) that accompanies phosphorylation of Thr-160 (54).

DISCUSSION
These experiments demonstrate that Tyr-15 phosphorylation significantly reduces CDK2 activity, primarily by reducing its affinity for its peptide/protein substrate but that the tyrosine-phosphorylated species retains trace activity. Sequence conservation suggests that this mechanism will be shared by many CDKs, although an exception may be provided by CDK5 where it has been reported that phosphorylation of the equivalent tyrosine residue brings about activation (55). From an inspection of the Y15pT160pCDK2-cyclin A-AMPPNP structure we predict that phosphorylation of Thr-14 would further block the peptide-binding site and may, moreover, prevent the binding or appropriate configuration of ATP.
There are several reports that suggest that dual phosphorylation of Tyr-15 and Thr-14 exerts a greater suppressive effect  on CDK1 activity than phosphorylation at either site alone (18,20,56). Although in contrast to the above studies, one group reported that they were unable to measure any histone kinase activity for partially purified recombinant Y15pCDK1-cyclin B despite using a 100-fold excess of substrate (43). Residual activity associated with singly phosphorylated CDK1 may be important in initiating an autoamplification loop whereby phosphorylation of Wee1/Myt1 (57, 58) and Cdc25 (59) by CDK1 results in reduced threonine/tyrosine kinase activity and elevated phosphatase activity, respectively. This, and similar processes (60), may contribute to the rapid activation of CDK1 at the G 2 /M transition and of CDK2 at the G 1 /S transition (reviewed in Ref. 11).
Whereas phosphorylation of Thr-160 leads to conformational change that embeds the phosphate group in a network of stable interactions required for structural integrity, the Tyr-15 phosphate moiety remains solvent exposed and is readily accessible to both Wee1/Myt1 kinases and Cdc25 phosphatases. Thus, increased CDK tyrosine phosphorylation provides the cell with a rapid and potentially reversible mechanism for regulating CDK activity in response to checkpoint activation. We suggest that the determination of structural and kinetic details of this key inhibited state may contribute to emerging quantitative models of cell cycle behavior and to our understanding of the molecular mechanisms that underlie them.