An Atypical Tyrosine Kinase with a Key Role in CDK1 Regulation

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Structure, Vol. 13, 541–550, April, 2005, ©2005 Elsevier Ltd All rights reserved. DOI 10.1016/j.str.2004.12.017 Structure and Inhibition of the Human Cell Cycle Checkpoint Kinase, Wee1A Kinase: An Atypical Tyrosine Kinase with a Key Role in CDK1 Regulation

Christopher J. Squire, James M. Dickson, cyclins and by activating and inhibitory phosphoryla- Ivan Ivanovic, and Edward N. Baker* tion (Dunphy, 1994; Morgan, 1997). The association of School of Biological Sciences and Centre for CDK1 with cyclin B during S phase facilitates the phos- Molecular Biodiscovery phorylation of CDK1 at T14, Y15, and T161 (Atherton- University of Auckland Fessler et al., 1993; Draetta et al., 1988; Meijer et al., Auckland 1001 1991; Parker et al., 1991; Solomon et al., 1992). T161 New Zealand phosphorylation allows stable association with cyclin B that is essential for CDK1 activation and the initiation of mitosis. On the other hand, phosphorylation of T14 Summary and Y15 is required to maintain CDK1 in an inactive state during the preceding S and G2 phases. Phosphorylation is critical to regulation of the eukary- The inhibitory inactivation of CDK1 is mediated by otic cell cycle. Entry to mitosis is triggered by the enzymes of the Wee1 kinase family, which thus play a cyclin-dependent kinase CDK1 (Cdc2), which is inac- critical role in the cell cycle by controlling the entry tivated during the preceding S and G2 phases by into mitosis. In humans and other vertebrates, Y15 is phosphorylation of T14 and Y15. Two homologous ki- phosphorylated by Wee1 kinase (Honda et al., 1992; nases, Wee1, which phosphorylates Y15, and Myt1, McGowan and Russell, 1993; McGowan and Russell, which phosphorylates both T14 and Y15, mediate this 1995; Parker and Piwnica-Worms, 1992; Watanabe et inactivation. We have determined the crystal struc- al., 1995), and another Wee1 family member, the mem- ture of the catalytic domain of human somatic Wee1 brane bound, dual-specificity Myt1 kinase, phosphory- (Wee1A) complexed with an active-site inhibitor at lates both T14 and Y15 (Liu et al., 1997; Mueller et al., 1.8 Å resolution. Although Wee1A is functionally a ty- 1995). Late in G2, Cdc25 phosphatase dephosphory- rosine kinase, in sequence and structure it most lates T14 and Y15, activating CDK1 and thus promoting closely resembles serine/threonine kinases such as mitosis (Dunphy and Kumagai, 1991; Gautier et al., Chk1 and cAMP kinases. The crystal structure shows 1991; Sebastian et al., 1993). that although the catalytic site closely resembles that Wee1 kinase is itself regulated by phosphorylation of other protein kinases, the activation segment con- and degradation (Watanabe et al., 1995). Phosphoryla- tains Wee1-specific features that maintain it in an tion of residues in its N-terminal domain, described be- active conformation and, together with a key substitu- low, generates a phospho-degron (signal for degradation in its glycine-rich loop, help determine its sub- tion) that allows recognition by an F box protein and strate specificity. targets Wee1A for ubiquitination and degradation (Wa- tanabe et al., 2004). Removal of Wee1 thus promotes Introduction entry into mitosis (Ayad et al., 2003; Watanabe et al., 2004). Phosphorylation of the C-terminal tail allows 14- Typical eukaryotic cells cycle through four phases dur- 3-3 peptides to bind to Wee1A, an event that influences ing cell division: G1, S (DNA synthesis), G2, and M (mi- the intracellular distribution of Wee1 and its level of ac- tosis). This process must accurately replicate the par- tivity (Wang et al., 2000; Lee et al., 2001). Recent evi- ent cell’s genetic material in the S phase and distribute dence shows that Wee1 levels are also controlled by identical chromosomal copies evenly between two components of the circadian clock, making it the key daughter cells at M phase. Progress through the cell link between the circadian clock and the cell cycle, and cycle is coordinated by the actions of CDKs (cyclin- hence cellular proliferation (Matsuo et al., 2003). dependent kinases) that are activated by association The role of Wee1 in controlling entry to mitosis also with specific cyclin proteins (for reviews, see Pines, gives it considerable potential as a drug target. Cellular 1993; Morgan, 1997). CDK activity is further regulated DNA damage can initiate the G2 checkpoint, a pause through posttranslational modification of the CDKs, at G2 that prevents the cell from entering mitosis, pre- transcriptional control, the association of cofactors, in- sumably giving time for other cellular mechanisms to tracellular localization, and the balance between syn- make repairs. Experiments in fission yeast indicate that thesis and degradation of cyclins and other regulators the G2 checkpoint functions by maintaining Y15 phos- (Morgan, 1995; Pavletich, 1999; Malumbres and Bar- phorylation on CDK1 until DNA damage is repaired bacid, 2001). Defects in this strict regulation of the cell (O’Connell et al., 1997; Rhind et al., 1997). As a regula- cycle can result in the development of tumors and other tor of CDK1 Y15 phosphorylation, Wee1 is thus integral diseases that involve diverse cellular processes includ- to this G2 checkpoint function, and Wee1 inhibitors ing apoptosis, differentiation, angiogenesis, and inflam- have potential use as G2-abrogating compounds. As mation. one example, studies using seven cancer cell lines In higher eukaryotes, entry to mitosis is triggered by lacking a functional p53 demonstrated that inhibition of the key CDK known as Cdc2 (CDK1). The activity of Wee1 with a nanomolar inhibitor, PD0166285, effec- CDK1 is regulated both by association with B type tively abrogated the G2 checkpoint (Wang et al., 2001). This allowed the p53−/− cells to proceed to premature *Correspondence: [email protected] mitosis and cell death, suggesting that effective Wee1 Structure 542

ble 1. The final model comprises 258 amino acid residues; the inhibitor PD0407824 (9-hydroxy-4-phenyl-6H- pyrrolo[3,4-c]carbazole-1,3-dione); 2 Mg2+ ions, which were identified as such by their coordination geometry; and 144 water molecules. All 279 residues of the kinase Figure 1. Domain Structure for Human Somatic Wee1 Kinase domain construct have been modeled except for resi- (Wee1A) dues 436–455 and 570–575, which had no interpretable The N-terminal regulatory domain is shown as an open box with electron density. The polypeptide has good geometry, the known serine phosphorylation sites (S53 and S123) and nuclear with 91.5% of nonglycine residues in the most favored localization signal (NLS) indicated. The kinase domain is indicated regions of the Ramachandran plot, as defined in PRO- by a gray box, with the first and last residues of the construct ex- CHECK (Laskowski et al., 1993), and only one outlier. pressed in this work indicated. The C-terminal serine residue impli- cated in 14-3-3β binding is also indicated (S642). Overall Structure The Wee1A catalytic domain displays a standard two- inhibition might be used in this way to enhance conven- lobed kinase fold (Figure 2A). The mainly β sheet tional cancer therapy. N-terminal lobe is folded around a five-stranded anti- In humans, two Wee1 kinases with different temporal parallel β sheet and has the typical glycine-rich loop, and spatial expression, Wee1A and Wee1B, have been with consensus sequence GxGxxG (residues 306–311 characterized (Nakanishi et al., 2000). The somatic in Wee1A), that is found in other kinases. The mainly Wee1 kinase, Wee1A, is a 646 amino acid protein (Wa- α-helical C-terminal lobe contains a characteristic four- tanabe et al., 1995) comprising three domains: a long helix bundle. The two lobes are joined by a linker pep- N-terminal regulatory domain, a central kinase domain, tide (residues 377–381) that connects strand β5inthe and a short C-terminal regulatory domain (Figure 1). N lobe to helix αD in the C lobe; this has been shown The N-terminal domain contains two phosphorylation to allow relative movement of the two lobes in other sites, at Ser53 and Ser 123, which provide the signal kinases, for example cAMPK (Narayana et al., 1997). that targets Wee1 for destruction. It also contains three The active-site cleft is located between the two lobes potential PEST regions (Watanabe et al., 1995) and a and provides the binding site for the Wee1A inhibitor putative nuclear localization signal (Arg-Arg-Arg-Lys- PD0407824 (Figure 2B), an antagonist of the natural Arg). Phosphorylation of the penultimate residue in the ATP substrate. The active-site region contains a short short C-terminal domain, Ser645, generates the 14- section of extended polypeptide (the catalytic seg- 3-3 binding site. Although it has a strict specificity for ment) that includes the essential catalytic aspartate phosphorylation of Y15 in CDK1, the central kinase do- Asp426, and a large loop, referred to as the activation main of Wee1A is not a typical tyrosine kinase. Se- segment, that provides a substrate binding platform quence searches of the human genome do not place in protein kinases and can undergo conformational Wee1A in any of the identified tyrosine kinase subfami- changes that control the activation state (Johnson et lies (Robinson et al., 2000), and comparisons with the al., 1996). A sequence alignment of Wee1 family mem- full complement of protein kinases place Wee1A and bers from various species suggests structural conser- Myt1 in a separate family (Manning et al., 2002). vation across the whole kinase domain (Figure 3). The Here we present the 1.8 Å crystal structure of the loop 436–455, not located in the structure, has been kinase domain of Wee1A, complexed with a known in- identified from its sequence as a potential PEST region hibitor of full-length Wee1A kinase, PD0407824 (Booth (Watanabe et al., 1995). PEST regions are found in et al., 2003). The structure, which is to our knowledge many short-lived eukaryotic proteins and play a role in the first for any Wee1 family member, reveals a close their degradation (Rogers et al., 1986). The solvated, structural relationship with prototypical serine/threo- disordered nature of this loop is consistent with its nine kinases, coupled with Wee1-specific sequence PEST annotation. changes. These differences maintain Wee1A in an active A search of the current protein structure database conformation and provide a basis for understanding the with SSM (http://www.ebi.ac.uk/msd-srv/ssm/) reveals substrate specificity of Wee1A. The structure also pro- many protein kinases that can be superimposed onto vides a starting point for structure-based drug design the Wee1A kinase domain with root-mean-square dif- of Wee1 inhibitors for cancer therapy. ferences in Cα positions of less than 2.0 Å over 220– 240 residues. The closest structural matches are mostly Results and Discussion Ser/Thr kinases, including the active forms of CDK2 (Jeffrey et al., 1995), Aurora kinase (Nowakowski et al., Structure Determination 2002), and Chk1 kinase (Chen et al., 2000); these cell Sequence and structural alignments of other kinase do- cycle-associated kinases are all close metabolic neigh- mains led to the design of a construct comprising resi- bors to Wee1A. It is notable that the top hits are all dues 291–575 of human Wee1A kinase, which was then functionally Ser/Thr kinases and that not until the four- expressed in Sf9 insect cells. The crystal structure was teenth-highest hit is a Tyr kinase encountered. Although solved by single isomorphous replacement with anom- protein kinases vary widely in amino acid sequence, the alous scattering (SIRAS) and refined at 1.8 Å resolution Wee1A sequence is also more closely related to Ser/ to an R factor of 0.218 (Rfree = 0.237). Statistics for the Thr kinases than to Tyr kinases. This is illustrated by a structure determination and refinement are given in Ta- simple BLAST comparison (Altschul et al., 1997)ofthe Structure of Human Wee1 Kinase 543

Table 1. Crystallographic Statistics

Crystal Native1 Native2 Hg Data collection statisticsa Resolution (Å) 1.81 (1.87–1.81) 2.25 (2.33–2.25) 2.60 (2.69–2.60) Reflections measured 734,175 189,188 750,884 Unique reflections 34,834 19,169 12,493 Completeness (%) 95 (92) 95 (77) 100 (99) Average I/σ 23 (2.6) 25 (4.9) 45 (13) b Rmerge 0.082 (0.362) 0.079 (0.401) 0.081 (0.275) SIRAS statistics (phasing to 3.0 Å resolution) SOLVE FOM (3 × Hg) 0.32 SOLVE z-score 12.07 Resolve FOM 0.63 Refinement statistics Resolution (Å) 47–1.81 Reflections (F > 0.0) 34,800 Total non-H atoms 2213 c Rcryst 0.218 d Rfree 0.237 Residues in most favored regions of the Ramachandran 90.5 plot (%) Rmsd ideal bond length (Å) 0.022 Rmsd ideal bond angles (o) 2.09 a Data for the highest resolution shell is given in parentheses b Rmerge = S|Ii − |/SIi c Rcryst = S||Fobs|-|Fcalc||/S|Fobs|, calculated for the 95% data used in refinement d Rfree = same as for Rcryst but using only the 5% of the data excluded from refinement

Wee1A catalytic domain sequence with consensus se- is often an arginine in kinases that are activated by quences for Ser/Thr kinases and Tyr kinases; expecta- phosphorylation of residues in the activation segment, tion values of 6e-40 and 3e-22, respectively, show that referred to as RD kinases (Johnson et al., 1996). In Wee1A shares features of both but is more Ser/Thr ki- Wee1A, however, the preceding residue is nonpolar nase-like. A dendrogram of 491 eukaryotic protein ki- Met425, which forms part of a hydrophobic cluster with nase domain sequences (Manning et al., 2002) likewise Val487 and Leu495 that helps to anchor the activation recognizes Wee1 as a separate family, closest to sev- segment by interaction with Val475. eral Ser/Thr kinase families. Taken together, the sequence Superposition of the Wee1A structure onto the inhibi- and structural comparisons, and the Myt1-Wee1 famil- tor or ATP bound structures of cAMPK and Chk1 ki- ial relationship, suggest that the Wee1-family proteins nases (Zheng et al., 1993; Chen et al., 2000) shows a may have evolved from a Ser/Thr kinase, a few key mu- close overlay of the catalytically important residues tations converting Wee1 to a Tyr kinase that acts on (Figure 4A), including one of the two Mg2+ ions found CDK1. in the Wee1A structure. This Mg2+ is bound to Asn431 from the catalytic segment and Asp463 from the start Active Site of the activation segment, together with three water The well-defined active-site cleft is bounded by the molecules, and has a coordination environment virtu- five-stranded β sheet and glycine-rich loop from the N ally identical to that of the catalytically important Mg2+ lobe, which curve over it from above; the β5-αD con- ion in the cAMPK/ATP structure. The sixth coordination necting loop (the hinge region), which defines the end site is vacant in both structures. of the cleft; and two parts of the C lobe, helix αD and One intriguing variation exists in the catalytic seg- the catalytic segment, which make up the other wall ments. In cAMPK and Chk1, the carboxyl group of the of the cleft. The floor is formed by the N lobe helix αC catalytic aspartate (Asp166 in cAMPK, Asp130 in Chk1) and the region between β8 and β9 leading into the acti- is hydrogen bonded to a threonine residue (Thr170 in vation segment. Many invariant or conservatively sub- Chk1) that contributes to the specificity of Ser/Thr ki- stituted residues are found in the active site, particu- nases versus Tyr kinases (Taylor et al., 1995). This helps larly in the catalytic segment. position the aspartate side chain for its role in catalysis. The catalytic segment, residues 422–433, extends In Wee1A, however, this threonine is substituted by from strand β6 to the beginning of strand β7 and pres- Asp479 (a Wee1-specific substitution), precluding a ents the essential catalytic aspartate, Asp426, into the similar interaction with the catalytic Asp426. Instead, active-site cleft. Its sequence, which is highly con- Asp426 is held in an appropriate geometry by hydrogen served (Figure 3), contains the active-site motif HxD bonds with Asp479 NH and a network of water mole- and closely matches the protein kinase consensus se- cules, and by coordination to the Mg2+ ion bound to quence IVHxDLKPxNIx. The residue preceding Asp426 Asn431. A water molecule directly in front of Asp426 Structure 544

Figure 2. X-Ray Structure of the Wee1A/PD0407824 Complex (A) Ribbon diagram showing the canonical two-lobe protein kinase fold with inhibitor PD0407824 (ball-and-stick model) binding in the active- site cleft between the N and C lobes, hydrogen bonding to the hinge region. Two Mg2+ ions are shown as yellow spheres. β strands are colored blue, α helices green, the catalytic (CS) and activation (AS) segments red. Figure prepared with Ribbons (Carson, 1997). (B) Electron density for the PD0407824 inhibitor bound to Wee1A, contoured at a level of 3 σ from an Fo − Fc omit map. Hydrogen bonds linking the inhibitor to the hinge region (residues 376–379) and to the two well-ordered water molecules, W614 and W724, are shown with broken lines. Drawn with Ribbons (Carson, 1997). (C) Stereo Cα plot for the Wee1A catalytic domain structure. Every tenth residue is indicated with a small green sphere. mimics a substrate tyrosine hydroxyl. A similarly placed Consistent with this, the inhibitor PD0407824 (Figure water molecule is found in cAMPK, where it is sug- 2B) is bound deeply in the active-site cleft, largely bur- gested to function as a nucleophile, initiating hydrolysis ied within the protein, in a mostly hydrophobic envi- (Madhusudan et al., 2002). ronment. Inhibitor binding is marked by highly favorable hy- ATP and Inhibitor Binding drogen bonding with surrounding protein groups, in Crystals of Wee1A were only obtained when the protein which the maleimido ring plays a key role. It is located was complexed with high-affinity inhibitors, suggesting internally, with its two oxygen atoms and intervening that inhibitor binding may have induced some confor- NH hydrogen bonded to main chain and side chain mational change and/or stabilization of the structure. atoms of residues 376–379 of the β5-αD loop, which Structure of Human Wee1 Kinase 545

Figure 3. Multiple Sequence Alignment of Wee1 Kinase Family Members The sequence numbering is that of human Wee1A kinase, with the Wee1A secondary structure shown above the alignment. Invariant residues (white on red background) and conservatively substituted residues (red) in the Wee1 family are indicated. Unique Wee1 residues are marked with an asterisk and are defined as those residues conserved in Wee1 but not found in protein kinase, serine/threonine kinase, or tyrosine kinase consensus sequences from the BLAST database (Altschul et al., 1997). The catalytic and activation segments are indicated with a red bar. Sequences shown are for Wee1A kinase of human (hu), mouse (mm), Carassius auratus (ca), Drosophila (dm), and Arabidopsis (at); Wee1B kinase of Xenopus (Weeb_xl); Mik1 kinase of S. pombe (Mik1_sp); and Myt1 kinase of human (Myt1_hu). Figure prepared using ESPript (Gouet et al., 1999).

forms the hinge region at the back of the cleft, and to Attempts to cocrystallize Wee1A with staurosporine, the first of two very well-defined water molecules, or with ATP or ATP analogs, have been unsuccessful. W614 and W724 (Figure 2B). Both waters are tetrahe- This may be because of their lower affinity, but it may drally coordinated, completing a network of interac- also reflect a change in the closure of the glycine-rich tions between the inhibitor and the protein structure. loop over the PD0407824 inhibitor, which could affect W614 sits over the inhibitor phenyl ring, which occupies crystal packing. Differences of this type, accompanying a pocket bounded by hydrophobic side chain atoms ligand binding, have been seen for cAMPK (Narayana of Val313, Lys328, and Ile374. Further stabilization of et al., 1997). inhibitor binding is provided by the packing of side Modeling ATP binding (Figure 4B) shows that the ad- chains against the planar carbazole ring system: the enine moiety binds in the hydrophobic pocket formed conserved Wee1 residue Phe433 stacks against one by Ile305, Val313, Ala326, and Phe433, occupied by the face of the ring system; Ile305 and Val313, which imme- carbazole system in the inhibitor bound structure. The diately precede and follow the glycine-rich loop, and modeling places the ATP adenine more shallowly in the Ala326 from strand β3 together present a hydrophobic active site than PD0407824, however, with fewer con- surface to the other face. The glycine-rich loop itself, tacts to the protein backbone, hydrogen bonding only residues 306–311, makes no close contacts with to Glu377 O. This is entirely consistent with their rela- PD0407824 and has higher B factors than the rest of tive binding affinities, with ATP binding at least three the active-site region. The most exposed part of the orders of magnitude more weakly than the PD0407824 inhibitor is the edge of the carbazole moiety, including inhibitor (A.J. Kraker, personal communication). Two ly- its pyrrole nitrogen. sine residues, Lys328 and Lys428 (equivalent to Lys72 Comparisons with other protein kinase inhibitor com- and Lys168 in cAMPK), are positioned such that Lys328 plexes (reviewed in Noble et al., 2004) show that can hydrogen bond to the α- and β-phosphates, and PD0407824 has particular features that enhance bind- Lys428 to the γ-phosphate and the catalytic Asp426, ing in Wee1A. Thus, although the core structure of the like their analogs in cAMPK (Zheng et al., 1993). Lys428 protein kinase inhibitor staurosporine is similar to that is not found in other Tyr kinases but is conserved of PD0407824, its IC50 for inhibition of Wee1A is more among Ser/Thr kinases. than 100× higher (26 ␮M compared with 97 nM for PD0407824). By consideration of the mode of staurosporine binding found for Chk1 (Zhao et al., 2002), we Conformation of the Activation Segment attribute the stronger Wee1A inhibition by PD0407824 The activation segment is a key functional region of to (1) the additional hydrogen bonds made by the sec- protein kinases that extends from the back of the active ond maleimido oxygen, and (2) improved hydrophobic site, where a conserved Asp residue (Asp463 in Wee1A) interactions and reduced steric repulsion resulting from binds the essential Mg2+ ion, and forms an irregular ex- its rotatable phenyl substituent and lack of the tetrahy- tended loop that finishes with helix αEF in the C-ter- dropyran ring of staurosporine (which would clash minal domain (Figure 2A). In many kinases this loop can with Phe433). undergo conformational changes, dependent on the Structure 546

Figure 4. Stereo Views of the Catalytic Sites of Wee1A and cAMP Kinase (A) Overlay of the Wee1A/PD0407824 and cAMP kinase/ATP structures. The Wee1A Cα ribbon is shaded dark gray and the cAMP ribbon is outlined in black (side chains are shaded similarly). Hydrogen bonds are indicated with broken lines. The experimentally located magnesium ions found in the two structures are shown as larger spheres. After superposition, the catalytic sites have a root- mean-square difference in Cα atomic positions of 0.76 Å for the residues shown. (B) Wee1A/ATP model. The ATP molecule, the tip of the glycine-rich loop, the side chain of Lys328, and a second magnesium (Mg2) were modeled on the basis of the cAMP kinase/ATP structure. The ATP base hydrogen bonds to Glu377 O at the back of the active site while the phosphates coordinate the two magnesium cations, Lys328, and Lys428. Both figures drawn with MolScript (Kraulis, 1991).

phosphorylation state of certain residues (Johnson et gen bonds between Thr468 and Ser472, and between al., 1996). Arg469 and the C terminus of helix αC in the N lobe, In Wee1A the activation segment is 25 residues in further stabilize this portion of the activation segment. length, residues 462–486. It is well ordered throughout Hydrophobic interactions involving side chains at each its whole length, with a conformation that corresponds end of the activation segment (Leu464 and Val467 at its very closely to that of the activation segments in Chk1 N-terminal end and Leu483 at its C-terminal end) fur- (Chen et al., 2000), Aurora kinase (Nowakowski et al., ther anchor it to the C-terminal lobe, as does an invari- 2002), and phosphorylase kinase (Owen et al., 1995), ant kinase ion pair between Glu486 and Arg557. A num- none of which appear to require phosphorylation for ac- ber of well-ordered water molecules form bridges tivation. Structural superpositions show that 21 Cα between the activation segment and residues in both atoms of the Wee1A activation loop can be superim- the N-terminal and C-terminal lobes. posed on the corresponding atoms of Chk1, Aurora ki- Many of the side chains that stabilize the activation- nase, and phosphorylase kinase with rms differences segment conformation are contributed by residues that of 1.56, 1.55, and 1.62 Å, respectively. The activation are invariant or highly conserved in all Wee1 family loop does not occlude any part of the active site such members and differ from the sequence patterns seen that it could prevent either ATP or substrate peptide in other protein kinases (Figure 3). Of the 25 residues, binding. We conclude, therefore, that it is also in an 17 are strongly conserved in the Wee1 family and 7 are active conformation in our Wee1A structure. unique to it. This suggests that the conformation seen Like that in Chk1, the Wee1A activation segment is here is common to all Wee1 family members and has stabilized in its active state by secondary structure and particular functional importance. Of particular note is side chain interactions. In the N-terminal half of the ac- Asp479, one of the Wee1-specific residues, which tivation segment, a short β strand, β9 (468–470), pairs forms a doubly hydrogen-bonded salt bridge with with β6 in the C-terminal domain, and the tip of the loop Arg481, holding it in a position in which it can play a key has consecutive β turns, 469–472 and 472–475, that are role in substrate recognition and binding (see Substrate buttressed against the C lobe by hydrophobic inter- Binding and Specificity). actions through Ile470 and Val475. Side chain hydro- Our structure suggests that regulation of Wee1A is Structure of Human Wee1 Kinase 547

Figure 5. Model for Peptide Binding to Wee1A (A) The peptide Gly-Glu-Gly-Thr-Tyr-Gly-Val- Val (residues 11–18 of human CDK1) is shown as a stick model, color coded by atom type. Putative hydrogen bonds are shown as broken lines. The peptide passes between Arg518 and Arg481, partly hidden below it, and a hydrogen bond to T14 helps position Y15 for phosphorylation by Wee1A. Glu309 would limit the approach of substrate for phosphorylation at T14. Figure drawn with SYBYL (Blanco, 1991), with solvent-accessible sur- faces calculated with a probe radius of 1.4 Å and colored by electrostatic potential (nega- tive potential blue, positive potential red, and neutral regions green, as indicated by the color-bar legend). (B) Specific difference between Wee1 kinases and other tyrosine kinases in the peptide binding region, shown by comparison of Wee1A (green) with the insulin receptor tyrosine kinase IRK3 (mauve). In typical tyrosine kinases, the arginine residue equivalent to Arg481 hydrogen bonds to a nearby loop, causing a surface bulge. In Wee1A, however, Arg481 is sequestered by a Wee1-specific salt bridge with Asp479 that instead creates a shallow groove, assisted by an invariant leucine. Drawn with Ribbons (Carson, 1997). achieved not by conformational changes in the activa- side chain is positioned in the active site for phosphoryl tion segment but through regions outside the catalytic transfer, whereas the preceding T14 side chain is ori- domain. This is consistent with previous experiments ented outwards and can hydrogen bond to Arg518. suggesting that regulation of Wee1 kinases is achieved From the Wee1A structure, two Wee1-specific features through its N- and C-terminal regulatory domains, both seem likely to play important roles in substrate specific- by phosphorylation (Watanabe et al., 1995) and by 14- ity. First, Arg481 in the activation loop is the site of a 3-3 binding (Lee et al., 2001; Wang et al., 2000). Recent key structural deviation from other tyrosine kinases, in results with both Xenopus and human Wee1 proteins which an equivalent arginine forms a surface bulge as- demonstrate that phosphorylation of serine residues in sociated with a conserved PxR (or PxK) sequence motif their N-terminal domains induces recognition by F box- in the activation segment (Figure 5B). In Wee1A, in con- containing proteins of the SCF E3 ubiquitin ligase com- trast, Arg481 is sequestered by a salt-bridge interaction plex and targets Wee1 for destruction (Ayad et al., with the Wee1-specific Asp479. This removes the bulge 2003; Watanabe et al., 2004). This temporary removal and results in the surface groove in which the substrate of Wee1 leads to timely entry into mitosis. peptide is docked in Figure 5A. Second, for phosphorylation of T14 the peptide would have to approach more Substrate Binding and Specificity closely the glycine-rich loop of Wee1A. The presence The structure of the Wee1A kinase catalytic domain of the Glu309, another Wee1-specific feature, at the tip provides a basis for understanding the specificity of of this loop may inhibit this closer approach in the same this enzyme, which is so critical to the G2 checkpoint way that phosphorylation of Thr14 or Tyr15 in CDK1 and proper entry into mitosis in the cell cycle. The natu- is believed to inhibit substrate binding through steric ral substrate is CDK1, which is specifically phosphory- hindrance (Endicott et al., 1999). lated on Y15, located in its glycine-rich loop (sequence GEGTYG, residues 11–16). Attempts to cocrystallize Conclusion Wee1A or soak Wee1A crystals with 10- or 20-residue Wee1 kinase plays a key role in cell cycle regulation by peptides containing this target sequence were unsuc- inactivating the cell cycle kinase Cdc2 (CDK1) through cessful, presumably because binding depends also on phosphorylation of Y15 on CDK1. This inactivation is parts of the CDK1 structure outside this region. We important in preventing entry into mitosis until DNA therefore docked an octapeptide with an appropriate damage is repaired, giving Wee1 an important function sequence onto our Wee1A structure, as described in in the G2 checkpoint. Our crystal structure of the hu- Experimental Procedures. In doing so, we assumed man Wee1A catalytic domain shows that Wee1 is an that the substrate peptide has sufficient flexibility to atypical tyrosine kinase in that it groups most closely conform to the contours of the Wee1A surface. The with Ser/Thr rather than Tyr kinases. Its specificity presence of three glycines and the known flexibility of for CDK1 Y15 phosphorylation appears to arise from this glycine-rich region in other protein kinases, includ- Wee1-specific features in its activation loop, coupled ing the very closely related CDK2, support this as- with the presence of Glu309 in the glycine-rich loop of sumption. Wee1A, which may disfavor phosphorylation of the pre- The need to position Y15 for phosphorylation, with ceding CDK1 T14 residue by steric hindrance. its hydroxyl replacing a water molecule from the Wee1A The importance of substitutions in the glycine-rich crystal structure, imposes some constraints on where loop in controlling activity is a recurring theme in pro- the peptide must bind. In this proposed binding mode tein kinases and arises because of the need for this (Figure 5A), the peptide occupies a shallow groove on flexible loop to fold over ATP for stable binding. In the Wee1A surface, passing between the side chains CDK1 and CDK2, phosphorylation of the glycine-rich of two arginine residues, Arg481 and Arg518, under loop residues T14 and Y15 (equivalent to Glu309 and Glu309 from the Wee1A glycine-rich loop, and docking Phe310 in Wee1A) causes inactivation by disfavoring V17 of the peptide into a hydrophobic pocket. The Y15 protein-substrate binding and/or productive ATP bind- Structure 548

ing (Endicott et al., 1999; Bartova et al., 2004). By anal- above. Data were collected to 2.6 Å resolution. All data were pro- ogy, we propose that the Wee1-specific presence of cessed with the Denzo/HKL package (Otwinowski and Minor, 1997). Glu309 in the glycine-rich loop influences substrate specificity in a similar way. Structure Determination and Refinement The Wee1A crystal structure, in complex with a nano- The structure was determined by SIRAS (single isomorphous re- molar inhibitor, provides a starting point for structure- placement with anomalous scattering) using the HgCl2 derivative based drug design of inhibitors that might be targeted data and the rotating anode native data. Three mercury atoms were at Wee1A itself or at other Wee1 family members. Inhi- located and refined with respect to their positions, thermal param- bition of human Wee1A has been shown to induce pre- eters, and occupancy using SOLVE (Terwilliger and Berendzen, mature entry into mitosis, before DNA damage has 1999). The SOLVE phases were input to RESOLVE (Terwilliger, 1999; Terwilliger, 2000) for density modification of the resulting map. An been repaired at the G2 checkpoint (Wang et al., 2001), initial Wee1 model was built automatically into the RESOLVE map demonstrating the potential use of such inhibitors as using MAID (Levitt, 2001). This resulted in a backbone trace for enhancers of conventional cancer therapy. Our struc- approximately 90% of the secondary structure elements. Side ture defines features of Wee1A sequence and surface chains were manually built into the experimental electron density topology that could be targeted by inhibitor molecules using O (Jones et al., 1991). Further refinement used the 1.8 Å syn- in this and other Wee1 family members, and the binding chrotron data and iterative cycles of CNS simulated annealing, mode of the PD0407824 molecule provides a guide to conjugate gradient minimization, and B factor refinement (Brünger et al., 1998). Throughout refinement, 2Fo − Fc maps (including ex- positions that might be appropriate for medicinal perimental phases) and the SIRAS map were used for model build- chemistry or combinatorial chemistry substitutions ing. CNS density-modified maps (solvent flipping) and composite (provisional US patent application filed May 29, 2003, omit maps further aided the manual building process. All maps US Serial No. 60/473887). showed clear density for the inhibitor, allowing manual docking and positional refinement. Finally, 144 water molecules and 2 Mg2+ ions (identified by octahedral coordination geometry) were located in Experimental Procedures Fo − Fc maps and were refined. The final refinement cycles used TLS and CGMAT refinement in REFMAC (Murshudov et al., 1997) Construct Design, Expression, and Purification and eased restraints. The construct used for the structure analysis, comprising residues

291–575 of human somatic Wee1A kinase (Wee1A291–575), was cho- sen on the basis of molecular-modeling experiments. Superpo- Molecular Modeling sition of seven tyrosine kinase structures downloaded from the ATP was modeled into the Wee1A active site using the programs HOMSTRAD website (Mizuguchi et al., 1998) led us to choose ap- SYBYL (Blanco, 1991), GOLD (Jones et al., 1995), and MODELLER propriate start and end points that would not compromise the in- (Marti-Renom et al., 2000). The glycine-rich loop of the Wee1A tegrity of the catalytic domain. Human U937 cells (T cell lymphoma) were used as a source of structure was modified to conform to the more closed conforma- mRNA for human Wee1A cDNA production. Amplicons using tion of the cAMPK/ATP/peptide structure (Zheng et al., 1993) using 291–575 2+ the above primers were generated from U937 cDNA via the poly- MODELLER. Side chains in the active site and an additional Mg merase chain reaction using Platinum Pfx DNA Polymerase (Invitro- ion were also positioned as in the cAMPK structure. The gen). The amplicons were purified using a High Pure PCR kit PD0407824 inhibitor and water molecules were removed, charges (Roche), restricted with NcoI and EcoRI, and ligated into the NcoI/ were assigned to the remaining atoms using SYBYL (Gasteiger EcoRI sites of the pFastBacHTb vector. The resultant recombinant method), and an ATP molecule was then docked into the apo struc- plasmids produced an in-frame fusion of the human Wee1A se- ture using GOLD. GOLD assigned atom types by default, and a quence with the vector His-tag and rTEV cleavage sequences distance restraint between Glu377 O and ATP N6 (min 2.4 Å, max 3.2 Å, spring constant 5.0) was included in the GOLD docking pro- downstream of the polyhedrin promoter. Human Wee1A291–575 was purified from Sf9 insect cells by 6×His affinity chromatography cess to position the adenine moiety of ATP in the back of the using Talon resin (CLONTECH), rTEV cleavage, dialysis, affinity Wee1A active-site cleft. GOLD defined the active site by a cavity chromatography, and, finally, size-exclusion chromatography on a search within 10 Å of Glu377. Multiple GOLD solutions consistently Superdex 200 HR 10/30 column (Pharmacia). placed the ATP molecule in the same site, with the ATP phosphates bound to the active-site magnesium ions. The top solution was then subjected to minimization in SYBYL using the Powell method Crystallization and Data Collection (default parameters). A 50 µl aliquot of Wee1A protein (concentration w7 mg/ml) 291–575 An octapeptide GEGTYGVV, comprising the target sequence of was mixed with 2.5 µl of PD0407824 (9-hydroxy-4-phenyl-6H-pyr- CDK1, was manually docked into the Wee1A/ATP model. The rolo[3,4-c]carbazole-1,3-dione) at a 4 mM concentration in dimethyl central tyrosine, corresponding to CDK1 Y15, was placed beside sulfoxide. Crystals were grown in hanging drops by mixing equal the catalytic Asp426 with its hydroxyl replacing a bound water volumes of the protein complex and the reservoir solution, com- molecule and its location also guided by that observed in a peptide prising 100 mM HEPES (pH 7.5), 4.3 M NaCl. Crystals formed in 2–3 days, growing to approximately 0.5 mm in their longest dimen- complex of the insulin receptor tyrosine kinase (Hubbard, 1997). sion within 3 weeks. The crystals were dipped briefly in a solution The remainder of the peptide was placed to complement the containing 85% (v/v) reservoir and 15% (v/v) glycerol, and flash- Wee1A surface in a conformation similar to that found for peptide frozen in liquid nitrogen for data collection. Data to 1.8 Å resolution binding to the closely related cAMPK (Zheng et al., 1993). The re- were collected at −170°C at the undulator beam line 17-ID of APS sulting Wee1A-peptide model was then energy minimized using using a 165 mm MarResearch CCD detector with Pt/Pd-coated CNS conjugate gradient minimization with the X-ray term excluded.

ULE mirrors. The crystals belong to space group P41212 with unit cell dimensions a=b=69.78, c=157.02 Å. An additional 2.25 Å resolution native data set was collected at −160°C on a Rigaku rotating anode generator equipped with Supper-like total reflection Acknowledgments mirrors and a Mar345 image-plate detector. A mercury derivative crystal was produced by supplementing a crystallization drop with We thank Erli Zhang for collecting data at the IMCA-CAT station at

1 mM HgCl2 and soaking a crystal for 10 min. The derivative crystal APS; John Booth, Al Kraker, Dan Ortwine, and Charles Hasemann was soaked in cryoprotectant without the addition of HgCl2 and for many useful discussions; and Pfizer Global Research, Ann Ar- flash-frozen for data collection on the rotating anode instrument bor, Michigan, for support of the research. Structure of Human Wee1 Kinase 549

Received: July 14, 2004 sity maps and the location of errors in these models. Acta Crystal- Revised: December 21, 2004 logr. A47, 110–119. Accepted: December 22, 2004 Jones, G., Willet, P., and Glen, R.C. (1995). Molecular recognition Published: April 12, 2005 of a recptor site using a genetic algorithim with a description of solvation. J. Mol. Biol. 245, 43–53. References Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both de- tailed and schematic plots of protein structures. J. Appl. Crystal- Altschul, S.F., Madden, T.L., Schäffer, A.A., Zhang, J., Zhang, Z., logr. 24, 946–950. Miller, W., and Lipman, D.J. (1997). Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Laskowski, R.A., MacArthur, M.W., Moss, D.S., and Thornton, J.M. Acids Res. 25, 3389–3402. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291. Atherton-Fessler, S., Parker, L.L., Geahlen, R.L., and Piwnica- Worms, H. (1993). Mechanisms of p34cdc2 regulation. Mol. Cell. Lee, J., Kumagai, A., and Dunphy, W.G. (2001). Positive regulation Biol. 13, 1675–1685. of Wee1 by Chk1 and 14-3-3 proteins. Mol. Biol. Cell 12, 551–563. Ayad, N.G., Rankin, S., Murakami, M., Jebanathirajah, J., Gygi, S., Levitt, D.G. (2001). A new software routine that automates the fit- and Kirschner, M. (2003). Tome-1, a trigger of mitotic entry, is de- ting of protein X-ray crystallographic electron density maps. Acta graded during G1 via the APC. Cell 113, 101–113. Crystallogr. D57, 1013–1019. Bartova, I., Otyepka, M., Kriz, Z., and Koca, J. (2004). Activation Liu, F., Stanton, J.J., Wu, Z., and Piwnica-Worms, H. (1997). The and inhibition of cyclin-dependent kinase-2 by phosphorylation; a human myt1 kinase preferentially phosphorylates cdc2 on threo- molecular dynamics study reveals the functional importance of the nine 14 and localizes to the endoplasmic reticulum and Golgi com- glycine-rich loop. Protein Sci. 13, 1449–1457. plex. Mol. Cell. Biol. 17, 571–583. Blanco, M. (1991). Molecular silverware.1. General solutions to ex- Madhusudan, P.A., Xuong, N.-H., and Taylor, S.S. (2002). Crystal cluded volume constrained problems. J. Comput. Chem. 12, 237– structure of a transition state mimic of the catalytic subunit of 247. cAMP-dependent protein kinase. Nat. Struct. Biol. 9, 273–277. Booth, R.J., Denny, W.A., Dobrusin, E.M., Kraker, A.J., Mitchell, Malumbres, M., and Barbacid, M. (2001). To cycle or not to cycle: L.H., Smaill, J.B., Thompson, A.M., Lee, H.H., McCarthy, F.O.J., and a critical decision in cancer. Nat. Rev. Cancer 1, 222–231. Palmer, B.D. April 2003. Inhibitors of Checkpoint Kinases (Wee1 Manning, G., Whyte, D.B., Martinez, R., Hunter, T., and Sudarsa- and Chk1). PCT International Patent WO 03/091255. nam, S. (2002). The protein kinase complement of the human ge- Brünger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., nome. Science 298, 1912–1934. Grosse-Kunstleve, R.W., Jiang, J.-S., Kuszewski, J., Nilges, M., Marti-Renom, M.A., Stuart, A., Fiser, A., Sánchez, R., Melo, F., and Pannu, N.S., et al. (1998). Crystallography & NMR system: a new Sali, A. (2000). Comparative protein structure modeling of genes software suite for macromolecular structure determination. Acta and genomes. Annu. Rev. Biophys. Biomol. Struct. 29, 291–325. Crystallogr. D54, 905–921. Matsuo, T., Yamaguchi, S., Mitsui, S., Emi, A., Shimoda, F., and Carson, M. (1997). Ribbons. Methods Enzymol. 277, 493–505. Okamura, H. (2003). Control mechanism of the circadian clock for Chen, P., Luo, C., Deng, Y., Ryan, K., Register, J., Margosiak, S., timing of cell division in vivo. Science 302, 255–259. Tempczyk-Russell, A., Nguyen, B., Myers, P., Lundgren, K., et al. McGowan, C.H., and Russell, P. (1993). Human Wee1 kinase inhibits (2000). The 1.7 Å crystal structure of human cell cycle checkpoint cell division by phosphorylating p34cdc2 exclusively on Tyr 15. kinase Chk1: implications for Chk1 regulation. Cell 100, 681–692. EMBO J. 12, 75–85. Draetta, G., Piwnica-Worms, H., Morrison, D., Druker, B., Roberts, McGowan, C.H., and Russell, P. (1995). Cell cycle regulation of hu- T., and Beach, D. (1988). Human cdc2 protein kinase is a major cell- man WEE1. EMBO J. 14, 2166–2175. cycle regulated tyrosine kinase substrate. Nature 336, 738–744. Meijer, L., Azzi, L., and Wang, J.Y.J. (1991). Cyclin B targets Dunphy, W.G. (1994). The decision to enter mitosis. Trends Cell p34cdc2 for tyrosine phosphorylation. EMBO J. 10, 1545–1554. Biol. 4, 202–207. Mizuguchi, K., Deane, C.M., Blundell, T.L., and Overington, J.P. Dunphy, W.G., and Kumagai, A. (1991). The cdc25 protein contains (1998). HOMSTRAD: a database of protein structure alignments for an intrinsic phosphatase activity. Cell 67, 189–196. homologous families. Protein Sci. 7, 2469–2471. Endicott, J.A., Noble, M.E.M., and Tucker, J.A. (1999). Cyclin- Morgan, D.O. (1995). Principles of CDK regulation. Nature 374, dependent kinases: inhibition and substrate recognition. Curr. 131–134. Opin. Struct. Biol. 9, 738–744. Morgan, D.O. (1997). Cyclin-dependent kinases: engines, clocks, Gautier, J., Solomon, M.J., Booher, R.N., Bazan, J.F., and Kirschner, and microprocessors. Annu. Rev. Cell Dev. Biol. 13, 261–291. M.W. (1991). Cdc25 is a specific tyrosine phosphatase that directly Mueller, P.R., Coleman, T.R., Kumagai, A., and Dunphy, W.G. (1995). activates p34cdc2. Cell 67, 197–211. Myt1: a membrane-associated inhibitory kinase that phosphory- Gouet, P., Courcelle, E., Stuart, D.I., and Metoz, F. (1999). ESPript: lates Cdc2 on both threonine 14 and tyrosine 15. Science 270, analysis of multiple sequence alignments in PostScript. Bioinfor- 86–90. matics 15, 305–308. Murshudov, G.N., Vagin, A.A., and Dodson, E.J. (1997). Refinement Honda, R., Ohba, Y., and Yasuda, H. (1992). The cell cycle regulator, of macromolecular structures by the maximum-likelihood method. human p50wee1, is a tyrosine kinase and not a serine/tyrosine ki- Acta Crystallogr. D53, 240–255. nase. Biochem. Biophys. Res. Commun. 186, 1333–1338. Nakanishi, M., Ando, H., Watanabe, N., Kitamura, K., Ito, K., Oka- Hubbard, S.R. (1997). Crystal structure of the activated insulin re- yama, H., Miyamoto, T., Agui, T., and Sasaki, M. (2000). Identifica- ceptor tyrosine kinase in complex with peptide substrate and ATP tion and characterisation of human Wee1B, a new member of the analog. EMBO J. 16, 5573–5581. Wee1 family of Cdk-inhibitory kinases. Genes Cells 5, 839–847. Jeffrey, P.D., Russo, A.A., Polyak, K., Gibbs, E., Hurwitz, J., Mas- Narayana, N., Cox, S., Xuong, N.-H., Ten Eyck, L.F., and Taylor, S.S. sague, J., and Pavletich, N.P. (1995). Mechanism of CDK activation (1997). A binary complex of the catalytic subunit of cAMP-depen- revealed by the structure of a cyclinA-CDK2 complex. Nature 376, dent protein kinase and adenosine further defines conformational 313–320. flexibility. Structure 5, 921–935. Johnson, L.N., Noble, M.E.M., and Owen, D.J. (1996). Active and Noble, M.E.M., Endicott, J.A., and Johnson, L.N. (2004). Protein ki- inactive protein kinases: structural basis for regulation. Cell 85, nase inhibitors: insights into drug design from structure. Science 149–158. 303, 1800–1805. Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1991). Im- Nowakowski, J., Cronin, C.N., McRee, D.E., Knuth, M.W., Nelson, proved methods for the building of protein models in electron den- C.G., Pavletich, N.P., Rogers, J., Sang, B.-C., Scheibe, D.N., Swan- Structure 550

son, R.V., et al. (2002). Structures of the cancer-related Aurora-A, Zheng, J., Knighton, D.R., Ten Eyck, L.F., Karlsson, R., Xuong, FAK, and EphA2 protein kinases from nanovolume crystallography. N.-h., Taylor, S.S., and Sowadski, J.M. (1993). Crystal structure of Structure 10, 1659–1667. the catalytic subunit of cAMP-dependent protein kinase com- O’Connell, M.J., Raleigh, J.M., Verkade, H.D., and Nurse, P. (1997). plexed with MgATP and peptide inhbitor. Biochemistry 32, 2154– Chk1 is a wee1 kinase in the G2 DNA damage checkpoint inhibiting 2161. cdc2 by Y15 phosphorylation. EMBO J. 16, 545–554. Accession Numbers Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffrac- tion data collected in oscillation mode. Methods Enzymol. 276, The coordinates and structure factor amplitudes for the Wee1A ki- 307–326. nase domain structure have been deposited in the Protein Data Owen, D.J., Noble, M.E.M., Garman, E.F., Papageorgiou, A.C., and Bank with the entry code 1X8B. Johnson, L.N. (1995). Two structures of the catalytic domain of phosphorylase kinase: an active protein kinase complexed with substrate analogue and product. Structure 3, 467–482. Parker, L.L., and Piwnica-Worms, H. (1992). Inactivation of the p34cdc2-cyclin B complex by the human wee1 tyrosine kinase. Science 257, 1955–1957. Parker, L.L., Atherton-Fessler, S., Lee, M.S., Ogg, S., Falk, F.L., Swenson, K.I., and Piwnica-Worms, H. (1991). Cyclin promotes the tyrosine phosphorylation of p34cdc2 in a wee1+ dependent man- ner. EMBO J. 10, 1255–1263. Pavletich, N.P. (1999). Mechanisms of cyclin-dependent kinase regulation: structures of Cdks, their cyclin activators, and Cip and INK4 inhibitors. J. Mol. Biol. 287, 821–828. Pines, J. (1993). Cyclins and cyclin-dependent kinases: take your partners. Trends Biochem. Sci. 18, 195–197. Rhind, N., Furnari, B., and Russell, P. (1997). Cdc2 tyrosine phosphorylation is required for the DNA damage checkpoint in fission yeast. Genes Dev. 11, 504–511. Robinson, D.R., Wu, Y.-M., and Lin, S.-F. (2000). The protein tyrosine kinase family of the human genome. Oncogene 19, 5548–5557. Rogers, S., Wells, R., and Rechsteiner, M. (1986). Amino acid sequences common to rapidly degraded proteins: the PEST hypothe- sis. Science 234, 364–368. Sebastian, B., Kakizuka, A., and Hunter, T. (1993). Cdc25M2 activation of cyclin-dependent kinases by dephosphorylation of threonine 14 and tyrosine 15. Proc. Natl. Acad. Sci. USA 90, 3521–3524. Solomon, M.J., Lee, T., and Kirschner, M.W. (1992). The role of phosphorylation in p34cdc2 activation: identification of an activating kinase. Mol. Biol. Cell 3, 13–27. Taylor, S.S., Radzio-Andzelm, E., and Hunter, T. (1995). How do protein kinases discriminate between serine/threonine and tyrosine? Structural insights from the insulin receptor protein-tyrosine kinase. FASEB J. 9, 1255–1266. Terwilliger, T.C. (1999). Reciprocal-space solvent flattening. Acta Crystallogr. D55, 1863–1871. Terwilliger, T.C. (2000). Maximum likelihood density modification. Acta Crystallogr. D56, 965–972. Terwilliger, T.C., and Berendzen, J. (1999). Automated MAD and MIR structure solution. Acta Crystallogr. D55, 849–861. Wang, Y., Jacobs, C., Hook, K.E., Duan, H., Booher, R.N., and Sun, Y. (2000). Binding of 14-3-3beta to the carboxyl terminus of Wee1 increases Wee1 stability, kinase activity, and G2-M cell population. Cell Growth Differ. 11, 211–219. Wang, Y., Li, J., Booher, R.N., Kraker, A., Lawrence, T., Leopold, W.R., and Sun, Y. (2001). Radiosensitization of p53 mutant cells by PD0166285, a novel G2 checkpoint abrogator. Cancer Res. 61, 8211–8217. Watanabe, N., Broome, M., and Hunter, T. (1995). Regulation of the human WEE1Hu CDK tyrosine 15-kinase during the cell cycle. EMBO J. 14, 1878–1891. Watanabe, N., Arai, H., Nishihara, Y., Taniguchi, M., Watanabe, N., Hunter, T., and Osada, H. (2004). M-phase kinases induce phospho- dependent ubiquitination of somatic Wee1 by SCFβ-TrCP. Proc. Natl. Acad. Sci. USA 101, 4419–4424. Zhao, B., Bower, M.J., McDevitt, P.J., Zhao, H., Davis, S.T., Johan- son, K.O., Green, S.M., Concha, N.O., and Zhou, B.-B.S. (2002). Structural basis for Chk1 inhibition by UCN-01. J. Biol. Chem. 27, 46609–46615.