Implications for Chk1 Regulation: The 1.7 Å Crystal Structure of Human Cell Cycle Checkpoint Kinase Chk1

Cell, Vol. 100, 681–692, March 17, 2000, Copyright 2000 by Cell Press

The 1.7 A˚ Crystal Structure of Human Cell Cycle Checkpoint Kinase Chk1: Implications for Chk1 Regulation Ping Chen,* Chun Luo,* Yali Deng, Kevin Ryan, James Register, Stephen Margosiak, Anna Tempczyk-Russell, Binh Nguyen, Pamela Myers, Karen Lundgren, Chen-Chen Kan, and Patrick M. O’Connor Agouron Pharmaceuticals, Inc. 3565 General Atomics Court San Diego, California 92121

The checkpoint kinase Chk1 is an important mediator of cell cycle arrest following DNA damage. The 1.7 A˚ resolution crystal structures of the human Chk1 kinase domain and its binary complex with an ATP analog has revealed an identical open kinase conformation. The secondary structure and side chain interactions stabilize the activation loop of Chk1 and enable kinase activity without phosphorylation of the catalytic domain. Molecular modeling of the interaction of a Cdc25C peptide with Chk1 has uncovered several conserved residues that are important for substrate selectivity. In addition, we found that the less conserved C-terminal region negatively impacts Chk1 kinase activity.

et al., 1998). In S. pombe, Chk1 is essential for cell cycle arrest following DNA damage (Walworth et al., 1993; Walworth and Bernards, 1996; Lindsay et al., 1998). A checkpoint role for Chk1 has also been uncovered in Xenopus (Kumagai et al., 1998), and in cells from individuals defective in ATM (Chen et al., 1999). In this latter case, overexpression of Chk1 rescued the G2 checkpoint defect of this cancer-prone phenotype, an outcome consistent with a G2 delay induced by overexpression of Chk1 (Walworth et al., 1993; Chen et al., 1999). In yeast, Chk1 also phosphorylates Wee1 (O’Connell et al., 1997) and Pds1 (Sanchez et al., 1999). In the case of Pds1, which regulates DNA damage–induced arrest in S. cerevisiae, Chk1-dependent phosphorylation of Pds1 correlates with Pds1 resistance to degradation (Sanchez et al., 1999). Taken together, these results point to an important role for the Chk1 in regulating cell cycle arrest following DNA damage. The human Chk1 is a nuclear protein of 476 amino acids (Sanchez et al., 1997). It contains a highly conserved N-terminal kinase domain (residues 1–265), a flexible linker region, and a less conserved C-terminal region with undefined function. The crystal structure of the human Chk1 kinase domain (residues 1–289) described here provides important insights into Chk1 regulation and substrate selectivity.

Introduction

Results

DNA damage activates a series of checkpoints that arrest cell cycle progression for repair of damaged DNA (Hartwell and Weinert, 1989; Weinert, 1998). These checkpoints interfere with the cyclin-dependent kinases (Cdks), which coordinate cell cycle progression (Murray, 1992; Elledge, 1996; O’Connor, 1997). At the G2 checkpoint in most species, DNA damage blocks the removal of Cdc2-inhibitory phosphorylations (Thr14 and Tyr15 of Cdc2) to prevent activation of this mitosis promoting Cdk (Jin et al., 1996; Poon et al., 1996; Rhind et al., 1997). Dephosphorylation and thereby activation of Cdc2 is carried out in human cells by the dual specificity Cdc25C phosphatase (Millar et al., 1991; Lee et al., 1992; Hoffmann et al., 1993). Activation of the Cdc25C is proposed to occur through an autocatalytic feedback loop mechanism involving nuclear Cdc2 (Hoffmann et al., 1993). The nuclear accumulation of Cdc25C is negatively regulated by phosphorylation at Ser216, which enables binding to 14–3–3 proteins (Peng et al., 1997; Sanchez et al., 1997; Dalal et al., 1999). In S. pombe, 14–3–3 proteins shuttle the phosphorylated Cdc25 out of nucleus after DNA damage (Lopez-Girona et al., 1999), and the G2 checkpoint is abrogated by expressing a Cdc25 mutant severely impaired for 14–3–3 binding (Zeng et al., 1998). One of the kinases that phosphorylates Cdc25C on Ser216 is Chk1, which is conserved from yeast to humans (Peng et al., 1997; Sanchez et al., 1997; Matsuoka

The Overall Structure The crystal structures of Chk1KD apoenzyme and its binary complex with an ATP analog, AMP-PNP, have been determined to 1.7 A˚ resolution. Chk1KD contains a C-terminal 6⫻His tag and the first 289 amino acids of human Chk1, consisting of the conserved kinase domain and part of the linker region. Table 1 summarizes our crystallographic analysis. Chk1KD has a canonical kinase two-lobe fold with the ATP-binding cleft residing between the two lobes (Figure 1A). Sequence homology in the Chk1 kinase domain suggests overall structural conservation across species (Figure 2). Residues that are not modeled in the current structure, due to lack of electron density, are not conserved among Chk1 proteins (Figures 1 and 2). There is no conformational change in Chk1KD between the AMP-PNP bound binary complex and the apoenzyme. In both structures, the ATP binding site, catalytic residues, and the activation loop are well ordered. The conformation of Chk1KD is close to active kinase structures. The N- and C-terminal lobes can be separately superimposed on the corresponding lobes of phosphorylase kinase (PhK, PDB code 2PHK) (Lowe et al., 1997), but, the two lobes of Chk1KD adopt a more open conformation compared to PhK (Figure 1B). A nearly identical Chk1KD conformation has been observed in other crystal space groups of Chk1KD with different molecular packing, suggesting that the apoenzyme and the AMP-PNP bound binary complex favor an open conformation.

Summary

* To whom correspondence should be addressed (e-mail: ping.chen@ agouron.com [P. C.], [email protected] [C. L.]).

Cell 682

Table 1. Statistics for the Crystallographic Analysis Crystal

Nat1

Nat2

AMP-PNP

Hg

Au

Internal merging and scaling Resolution (A˚) Reflections measured Unique reflections Completeness (%) Average I/␴ Rsyma

1.7 162418 35032 93.6 (88.3) 29.9 (9.0) 3.2 (18.1)

2.1 46947 19145 95.4 (94.6) 15.47 (4.38) 5.0 (23.3)

1.7 107449 35285 94.1 (91.1) 26.4 (12.5) 3.0 (10.0)

2.4 64881 12821 95.4 (96.4) 27.1 (11.6) 6.0 (13.2)

2.0 125728 22086 97.5 (84.8) 33.5 (14.8) 4.2 (11.8)

15–3.0 0.49 2.27/1.98

15–3.0 0.55 2.39/1.48

SIRSAS analysis Resolution (A˚) Rcullisb Phasing power c (SIR/SAS) Figure of merit (combined)

0.764

Refinement statistics Resolution range (A˚) Reflections usedd (F ⬎ 1␴F) Total nonhydrogen atoms Rcryste (%) Rfreef (%) Rmsd from ideal bond length (A˚) Rmsd from ideal bond angle (⬚) Average B (A˚2; all atoms)

7–1.7 30132 2372 21.6 23.5 0.005 1.30 28.9

7–2.1 15804 2354 20.8 25.0 0.006 1.27 29.7

7–1.7 31794 2460 22.6 24.9 0.010 1.58 23.22

Data for the outermost resolution shell are given in parentheses. a

Rsym ⫽

N

N

⫽

⫽

兺h i 兺1 | I¯(h) ⫺ I(h)i|/ 兺h i 兺1 |(h)i*100, where I(h)i is the ith measurement of reflection h and I¯(h) is the mean value of the N equivalent

reflections. b Rcullis ⫽ ⌺ | | FPH ⫾ FP | ⫺ FH(calc) | / ⌺ |FPH ⫾ FP | for all centric reflections. c Phasing power ⫽ rms (|FH|/E), where |FH| is the heavy-atom structure factor amplitude and E is the residual lack of closure. d Number of reflections used in working set. e Rcryst ⫽ ⌺ ||Fobs| ⫺ ||Fcalc| |/⌺|Fobs|, where summation is over data used in the refinement. f Rfree is the same calculation including the 10% of data excluded from all refinements.

The two lobes of Chk1KD are linked by ␤6⬘ of the hinge region and held together at the lobe interface by an extensive hydrogen-bond network that involves ␤6⬘, the loop linking ␣C and ␤4 of the N-terminal lobe, and strands ␤7 and ␤8 of the C-terminal lobe. Residues from this network also form part of the pocket that interacts with the adenine moiety of AMP-PNP. The conserved Lys69 and Lys145 in this hydrogen-bond network interact with the projected side chain of Glu85 that is conserved at the end of ␤5 in Chk1 proteins. These interactions appear to maintain the relative position of the N-terminal lobe and the Asp-Phe-Gly motif in the C-terminal lobe, allowing a coordinated movement during lobe closure. Mutation of the equivalent Glu to Asp in S. pombe Chk1 led to a temperature-sensitive phenotype (Francesconi et al., 1997). The Chk1KD crystal structure indicates that disruption of the interactions formed by this conserved Glu residue could render Chk1 proteins structurally less stable. Conserved Motifs The IEPDIG motif of Chk1KD (residues 96-Ile-Glu-ProAsp-Ile-Gly-101) limits ␣D to a one-turn helix because Pro98 initiates a tight turn between ␣D and ␣E (Figures 1 and 2). This turn interacts with the C terminus of ␣F through a backbone hydrogen bond between Asp99 and an invariant Gly204. Within this turn, Glu97 forms backbone hydrogen bonds with Ile100 and Gly101. Compared with ␣G of PhK, ␣G of Chk1KD is positioned differently (Figure 1B) and is flanked by two pairs of invariant PW residues (Pro207/Trp208 and Pro230/Trp231) that

form van der Waals contacts and interact with the hydrophobic core of the C-terminal lobe. Interactions of the PW residues may help position the side chain of Leu206 that forms a hydrophobic pocket with side chains from the IEPDIG motif. These motifs could be important for substrate interaction as described below. Helix ␣E contains a conserved motif Ala-Gln-X-PhePhe-X-Gln-Leu (residues 107–114), in which the hydrophobic residues are buried inside the C-terminal lobe. The side chain of Gln108 projects out of the kinase domain and forms hydrogen bonds directly or through a water molecule to backbone atoms of the linker region that follows ␣I. Although Chk1 sequences diverge in this linker region, these backbone interactions with Gln108 could still be conserved, thereby helping to anchor the linker to ␣E. The Activation Loop Phosphorylation of the activation loop serves as a molecular switch for activation of many kinases (Johnson et al., 1996). For Cdk2, phosphorylation of Thr160 is required for its activity (Russo et al., 1996). There is no equivalent Thr in the activation loop of human Chk1 (Figure 2), indicating that Chk1 is regulated through a different mechanism than Cdk2. Indeed, the activation loop of Chk1KD is already folded into an active conformation (Figure 3). The activation loop of Chk1KD appears to be stabilized by its secondary structure and side chain interactions (Figures 1 and 2). The antiparallel ␤ strands of ␤10 and ␤11 are almost perpendicular to the antiparallel ␤

Chk1 Kinase Domain Structure 683

Figure 1. Chk1KD Crystal Structure and Comparison with PhK (A) A ribbon diagram of the binary complex structure of Chk1 with AMP-PNP showing secondary structural elements and loops discussed in the text. The ␣ helices are shown in blue, ␤ strands in cyan, catalytic loop in orange, activation loop in red. AMP-PNP and sulfate ion are shown as ball and stick models. The triphosphate moiety of AMP-PNP is not modeled due to lack of electron density and the ribose ring is in C2⬘-endo pucker conformation. The protein termini are denoted by N and C. The figure was prepared with Molscript (Kraulis, 1991). (B) Stereo diagrams comparing the Ca tracing of the Chk1KD (open) and PhK (closed). The Chk1 (purple) and PhK (blue) structures were superimposed using the C-terminal lobes as a reference. The C␣ tracing of MC-peptide of PhK is colored orange. Glu110 of PhK and Arg(P⫺3) of MC-peptide are shown as ball and stick models. The N-terminal lobes (residues 2–90 of Chk1KD and residues 14–108 of PhK) can be aligned with an rms deviation of 1.1 A˚ between positions of C␣ atoms. The C-terminal lobes (residues 91–276 of Chk1KD and 110–290 of PhK) superimposed with an rms deviation for C␣ atoms of 0.9 A˚, excluding ␣G and connecting loops. The N-terminal lobe of Chk1KD can be rotated ⵑ15⬚ around the hinge region near Glu91, whose side chain almost overlaps with the corresponding side chain of Glu110 in PhK, to the position of the N-terminal lobe of PhK. The figure was prepared with Molscript.

strands of ␤6 and ␤9 due to a twist at Phe155 between ␤9 and ␤10. Phe155 forms backbone hydrogen bonds with Gly125 in ␤6 and Arg162 in ␤11. Between ␤10 and ␤11 is a ␤ turn formed through backbone interaction between Tyr157 and Arg160. These interactions pack the N terminus of the activation loop against the catalytic loop, and position the highly conserved Arg156 and Glu161 whose side chains interact with the invariant His122 of ␣E and His185 that precedes ␣F, respectively. Together, these interactions anchor the N-terminal part of the activation loop to the core of the C-terminal lobe. The center of the activation loop interacts with the remainder of the C-terminal lobe through two backbone hydrogen bonds between Leu164 and Phe184. The C terminus of the activation loop is supported by ␣EF, which is anchored at two positions to the core of the

C-terminal lobe through two ion pairs; one is the invariant kinase ion pair between Glu177 and Arg253, the other is between Lys180 and Glu248. This extra ion pair restrains the movement of ␣EF, and in turn the movement of the C terminus of the activation loop. Interestingly, Lys180 and Glu248 are only conserved among known vertebrate Chk1 proteins. Phosphorylation of the activation loop provides a negative charge to cluster a set of positively charged side chains, thereby stabilizing the active conformations of both the activation and catalytic loops (Johnson et al., 1996). In Chk1KD, a cluster of positive charges are contributed from Lys54, Arg129, Arg162, and Lys166 without phosphorylation in the activation loop. A sulfate ion lies close to the corresponding phosphate position of the phosphothreonine (Thr197) in cyclic AMP–depen-

Cell 684

Figure 2. Sequence Alignment of Chk1KD and PhK ClustalW sequence alignment of Chk1 kinase domains of human (hs), mouse (mm), Xenopus (xl), Drosophila (dm), C. elegans (ce), S. cerevisiae (sc), and S. pombe (sp), as well as human PhK. Secondary structural elements of human Chk1, colored as in Figure 1A, are shown above the alignment. Residues not modeled in current structures are indicated by a dotted line. The numbers of amino acids are shown on the right. Invariant residues of Chk1 among these species are in red and human Chk1 residues that are also conserved in other species are in cyan. Conserved motifs discussed in the text are underlined. Numbers within the PhK sequence indicate the number of residues not shown.

dent kinase (cAPK, PDB code 1ATP) (Knighton et al., 1991a). This sulfate ion interacts with side chains of Arg129, Arg162, and Thr153 of Chk1KD. We investigated whether this sulfate ion could mimic phosphate and thereby stabilize the activation loop by solving the Chk1KD structure under sulfate-free conditions. The resultant 2.1 A˚ resolution structure, however, revealed a nearly identical conformation of the activation loop, except for the side chain of the conserved Arg162, which turns toward solvent (Figure 3). The nearly identical structures of the activation loop in Chk1KD with and without sulfate ion, together with their well-defined electron density, underscore the importance of secondary structure and side chain interactions that stabilize the Chk1KD activation loop in its active conformation. An intriguing feature of Chk1KD is the permuted positions of Lys166 and Thr153. Kinases that are activated by phosphorylation of the activation loop have Thr/Tyr in the position corresponding to Lys166 of Chk1KD and Lys/Arg in the position corresponding to Thr153 of Chk1KD. Kinases that are not controlled by phosphorylation of the activation loop such as PhK have other residues in these corresponding positions (Johnson et al., 1996). In Chk1KD, Lys166 points its side chain to the thiol atom of Cys168, and together with the side chain of Arg129, creates a basic environment for a potential thiolate ion that could stabilize the positive charge of Arg129, a highly conserved residue adjacent

to the catalytic residue Asp130. Lys166 also appears to play a role in determining the substrate selectivity as discussed below. The side chain of Thr153 of Chk1KD forms a hydrogen bond with the side chain of Lys54 projected from ␣C (Figure 3). Lys54 is conserved in all but S. pombe Chk1 and lies adjacent to Glu55, which interacts with Lys38 forming an invariant ion pair of active kinases. The interaction between Thr153 and Lys54 may thus play a similar role to the interaction between the phosphate of phosphorylated Thr197 and His87 of cAPK (Knighton et al., 1991a). While Thr153 is conserved in Chk1 proteins (Thr or Ser), its permuted position suggests that it is not a phosphorylation site, since disruption of the secondary structure of the activation loop would have to accompany phosphorylation of Thr153. Active Site and AMP-PNP Binding Residues 18–21 at the apex of the glycine-rich loop between ␤1 and ␤2 of Chk1KD showed poor electron density in the crystal structures. Tyr20 present in human Chk1 corresponds to Tyr15 of Cdc2, phosphorylation of which inhibits Cdc2 activity (Krek and Nigg, 1991; Parker and Piwnica-Worms, 1992). However, mutation of Tyr20 in full-length Chk1 to either Glu or Phe does not significantly affect its kinase activity (data not shown). A comparison of Chk1KD with PhK and cAPK in their ternary complexes revealed similar local conformation

Chk1 Kinase Domain Structure 685

Figure 3. Stereoview of the Activation Loop and Its Relationship to the Catalytic Loop and Helix ␣C Chk1 in purple and PhK in green are superimposed with their C-terminal lobes. The structures displayed correspond to Chk1 residues 50–58 of helix ␣C, residues 129–132 of the catalytic loop, and residues 148–170 of the activation loop. The displayed Chk1KD structure is from the sulfate-free crystal. The conformation of Glu182 (E182) of PhK may represent the conformation of Ser173 of S. pombe Chk1 and Asp189 of S. cerevisiae Chk1 which substitute Lys166 of human Chk1. The figure was prepared with Molscript.

of the catalytic residues (Figure 4). For example, in the N-terminal lobe, the invariant ion pair of active kinases is present in Chk1KD between Lys38 of ␤3 and Glu55 of ␣C. Despite the more open lobe conformation of Chk1KD, the relative side chain positions of Lys38, Glu55, and Asp148 are similar to those of the corresponding residues in cAPK and PhK. This is partly due to the position of the Asp-Phe-Gly motif, which is closer to the N-terminal lobe in Chk1KD than in PhK and cAPK. These residues, together with the glycine-rich loop, would coordinate one of the two magnesium ions and anchor the ␣ and ␤ phosphates of ATP (Johnson et al., 1996). In the C-terminal lobe, the conformation of the catalytic loop (residues 129–135) of Chk1KD is nearly identical to that of PhK, with the side chains of Asp130, Lys132, and Asn135 of Chk1KD nearly superimposable on the corresponding residues Asp149, Lys151, and Asn154 of PhK (Figures 3 and 4). In PhK, Lys151 binds to the ␥-phosphate of AMP-PNP and Asn154 chelates the second magnesium ion that in turn binds to the ␤ and ␥ phosphates. Thr170 of Chk1KD is conserved in other Ser/Thr protein kinases and contributes to the specificity of Ser/Thr versus Tyr kinases (Taylor et al., 1995). The side chain of Thr170 in Chk1KD is within hydrogen-bonding distance to side chains of Asp130 and Lys132. Thr170 may thus aid to position the carbonyl of the catalytic residue Asp130. Mutation of Thr170 in full-length Chk1 to either Ala or Glu abolishes Chk1 kinase activity toward Ser216 of human Cdc25C (data not shown). A difference between Chk1KD and the active ternary complexes of PhK and cAPK is the distance between the side chains of residues important for catalysis, and this appears due to the more open lobe conformation of Chk1KD (Figure 4). Specifically, the residues in the N-terminal lobe and the Asp-Phe-Gly motif (Lys38, Glu55, and Asp148) of Chk1KD are further apart from the

catalytic loop (Asp130, Lys132, and Asn135) in Chk1KD than in PhK or cAPK. Indeed, separation of these residues into two clusters, together with the flexible glycinerich loop in Chk1KD, could be responsible for the lack of electron density of the AMP-PNP triphosphate in the binary complex structure. The adenine and ribose moieties of AMP-PNP are clearly defined in the Chk1KD binary complex structure (Figure 5B), but the conformation of the ribose ring in the present structure would not allow correct orientation of the ATP phosphates for catalysis. In contrast to the C3⬘-endo conformation found in active kinase structures, the ribose ring in the Chk1KD binary structure adopts a C2⬘-endo conformation (Figures 1 and 5B) similar to that seen in the inactive form of Cdk2 (PDB code 1HCK) (De Bondt et al., 1993). However, this is not due to displacement of ␣C in Chk1KD as is the case in Cdk2. This C2⬘-endo conformation of ribose probably results from the more open lobe conformation of Chk1KD, as well as the flexible triphosphate moiety of AMP-PNP that showed no electron density. Taken together, conformational changes in both Chk1KD and the ribose ring of AMP-PNP would be necessary to align the ATP phosphates for catalysis. Structural Basis of Substrate Selectivity The ordered activation loop of Chk1KD and the overall structural resemblance to PhK enabled us to assess the interactions of a substrate peptide with Chk1KD. Residues phosphorylated by Chk1 have been identified in Cdc25 proteins across species (Peng et al., 1997; Sanchez et al., 1997; Kumagai et al., 1998; Zeng et al., 1998). We have modeled the interaction of Chk1KD with the human Cdc25C peptide Leu-Tyr-Arg-Ser-Pro-SerMet-Pro-Glu (residues 211–219) which spans P⫺5 to P⫹3 based on the ternary complex structure of PhK with the MC-peptide (PDB code 2PHK) (Lowe et al.,

Cell 686

Figure 4. Catalytic Site of Chk1KD and PhK (A) Cross section of the catalytic site of Chk1KD with AMP-PNP. Protein C␣ ribbon representations are shown in purple for Chk1KD. The side chains of the catalytic site residues are shown as ball and stick models and are color-coded by atom type: carbon, green; nitrogen, blue; oxygen, red. The distances (A˚) along the dotted lines between the catalytic site residues are shown. In addition, Lys38N⑀ is 10 A˚ away from Asp130O␦2, compared with 8.2 A˚ in PhK. Asp148O␦1 is 6 A˚ away from Asp130O␦2, compared with 3.8 A˚ in PhK. In Chk1KD, one water molecule is located between Asp148 and Asp130 and is hydrogen-bonded to Asp130O␦2 and Asn135O␦1. The figure was prepared with INSIGHT II (INSIGHT II User Guide, 1997, Molecular Simulations Inc., San Diego, CA). (B) The corresponding section of PhK.

1997). The extended backbone conformation and antiparallel alignment with respect to the activation loop appear to be general features for peptides that interact with kinases at the substrate binding site, as has been observed in the insulin receptor tyrosine kinase ternary complex structure (PDB code 1IR3) (Hubbard, 1997), and in cAPK ternary complex structure (PDB code 1ATP) (Knighton et al., 1991b). Similar to the MC-peptide bound to PhK, the Cdc25C peptide was modeled in an extended conformation from the P⫺5 to P⫹3 positions (Figure 6). In this conformation, the peptide fits into a shallow groove on the surface of the C-terminal lobe of Chk1KD. The O␥ atom of Ser(P), the presumed nucleophile in the phosphate transfer reaction, is placed close to an ordered water molecule in the Chk1KD structure. This water molecule forms hydrogen bonds with both the Asp130O␦2 and Lys132N⑀. Superposition of Chk1KD and PhK shows that this water molecule would be 3.4 A˚ from the ␥-phosphorus atom of the AMP-PNP in PhK. This water molecule may thus approximate the location of the O␥ atom of Ser(P) during catalysis. The side chains at P⫺5 and P⫹1 of the peptide could establish hydrophobic interactions with Chk1KD and anchor the peptide at its two ends. The hydrophobic side

chain of Leu(P⫺5) fits into a hydrophobic pocket formed by Phe93, Ile96, Pro98, and Leu206. All of the residues except Leu206 are invariant in Chk1 proteins, and Ile96 and Pro98 are in the conserved IEPDIG motif. The hydrophobic side chain of Met(P⫹1) projects into another hydrophobic pocket formed by residues of Leu171, Val174, Leu178, Leu179, and Met167 of Chk1KD. Although these residues diverge in Chk1 proteins, the hydrophobicity of this pocket is conserved (Figure 2). The Arg(P⫺3) side chain of the modeled Cdc25C peptide points toward Glu91 of Chk1KD similar to Arg(P⫺3) in the MC-peptide of the PhK ternary complex structure (Lowe et al., 1997). Arg is a preferred residue at the P⫺3 position of 14–3–3 binding sites (Yaffe et al., 1997). An extended side chain of Arg(P⫺3) in the Cdc25C peptide would place the guanidinium group 3 A˚ away from the carboxyl of Glu91. In both ternary complex structures of cAPK and PhK, the guanidinium of Arg(P⫺3) forms a somewhat closer ion-pair (2.5 A˚) with the carboxyl of the corresponding Glu residues (Knighton et al., 1991a, 1991b; Lowe et al., 1997). A more favorable ionic interaction between Arg(P⫺3) and Glu91 would be established in Chk1KD after lobe closure. The side chain of Ser(P⫺2) does not appear to interact with Chk1KD, rather it makes a hydrogen bond to the

Chk1 Kinase Domain Structure 687

Figure 5. Stereoview of Representative Electron Density Map (A) Stereoview of a representative portion of the experimental density at 1.5 ␴ calculated to 3.0 A˚ with the use of phases after solvent flattening. Superposed on the density is the final refined model. (B) Difference Fourier map calculated with native model-derived phases and coefficients |Fo(amppnp)| ⫺ |Fo(native)| to the diffraction of 1.7 A˚ and contoured at 2.5 ␴. The triphosphate moiety of AMP-PNP is disordered and is omitted from the model. No magnesium ions are observed. N6 of adenine forms hydrogen bond with the main chain carbonyl of Glu85; N1 of adenine forms hydrogen bond with amide of Cys87. Not shown in this figure is the interaction of N7 of adenine with the side chain of Ser147 of Chk1KD via a water molecule. The figure was prepared with Xfit (McRee, 1992).

backbone carbonyl oxygen of Pro(P⫺1) (Figure 6). In comparison, Gln(P⫺2) of the MC-peptide interacts with Ser188 of PhK, providing an anchor point for the MCpeptide. This interaction is not available to Chk1KD since it has an invariant Pro172 in the corresponding position of Ser188 in PhK. Pro172 of Chk1KD may thus contribute to the selectivity for a Ser or Thr residue at P⫺2 position of the peptide. The P⫹2 position requires a residue with a small side chain or one that can participate in an abrupt turn in order to avoid steric clash with Lys166 in Chk1KD. The side chain of Lys166 appears to be fixed since its welldefined positions are identical in the Chk1KD structures with or without sulfate. Lys166 is conserved among vertebrate Chk1 proteins. Correspondingly, Pro is found at the P⫹2 position in vertebrate Cdc25B and Cdc25C proteins. Indeed, Pro at P⫹2 position is critical to create a consensus 14–3–3 binding site once the Ser(P) is phosphorylated (Yaffe et al., 1997). Phosphorylation of human Cdc25C by human Chk1 is very selective such that the Ser(P⫺2) is not phosphorylated (Peng et al., 1997; Sanchez et al., 1997). Indeed, phosphorylation of Ser(P⫺2) alone or together with Ser(P) diminished the ability of a Raf peptide to bind

14–3–3 protein (Muslin et al., 1996). This selectivity is important for Cdc25 regulation since phosphorylation of Ser216 of human Cdc25C confers binding to 14–3–3 proteins, which in turn sequester Cdc25C into the cytoplasm (Peng et al., 1997). The structural basis for such selectivity is suggested by our model: In Chk1KD, the two conserved hydrophobic pockets, together with residues Glu91 and Lys166, are critical for substrate selectivity. In the substrate peptide, hydrophobic side chains at the P⫺5 and P⫹1 positions, Arg at P⫺3 position, and the steric restriction at the P⫹2 position specify the interaction with Chk1KD. This model not only is consistent with all known Chk1 phosphorylation sites, but also could explain the differences between human and yeast substrates, as discussed below. Kinase Activity The Chk1KD structure suggests that phosphorylation of the kinase domain is not required for its activity. Indeed, Chk1KD (residues 1–289) and the conserved kinase domain (residues 1–265) of human Chk1 were over 20fold more active than the recombinant full-length Chk1 protein toward various substrates including Ser216 of Cdc25C (Figure 7, comparing lanes 1–3, and data not

Cell 688

Figure 6. Molecular Surface of the Chk1 with Modeled Cdc25C Peptide The molecular surface of Chk1KD is colored as follows: basic side chains in blue, acidic side chains in red, and nonpolar side chains in violet. Substrate peptide (Leu-Tyr-Arg-Ser-Pro-Ser-Met-Pro-Glu, residues 211–219 of human Cdc25C) is shown as stick model and color-coded by atom type: carbon, green; nitrogen, blue; oxygen, red; sulfur, yellow. The figure was prepared with INSIGHT II.

shown). These results suggested that the C-terminal region of Chk1 negatively impacts Chk1 kinase activity. Phosphorylation of Chk1 induced by DNA damage required Chk1 kinase activity in S. pombe, suggesting a potential role of autophosphorylation in regulating Chk1 activity (Walworth and Bernards, 1996). The recombinant full-length Chk1 was readily autophosphorylated and there were multiple phosphorylation sites (Figure 7A). The linker region (residues 261–340 of human Chk1) can be phosphorylated by Chk1 in vitro (Kaneko et al., 1999). Indeed, autophosphorylation was observed for Chk1KD, which contains part of the linker region (residues 266–289), but not for the shorter conserved kinase domain (residues 1–265) (data not shown). Interestingly, autophosphorylation did not change the kinase activity of the recombinant full-length Chk1 (Figure 7, lanes 3–8). These data further support the notion that Chk1 kinase activity is not regulated by phosphorylation of its kinase domain. Discussion The crystal structures of human Chk1KD apoenzyme and its AMP-PNP bound binary complex revealed that AMPPNP binding does not cause conformational changes in Chk1KD. Compared to structures of catalytically active kinases (Knighton et al., 1991a; Johnson et al., 1996; Russo et al., 1996; Lowe et al., 1997), Chk1KD adopts an open lobe conformation with slight misalignment of residues critical for catalysis. Lobe closure could be achieved through a simple rotation of the N-terminal

lobe of Chk1KD by ⵑ15⬚ around the hinge region near Glu91 with respect to its C-terminal lobe (Figure 1B). This rotation would bring the N-terminal lobe of Chk1KD into the corresponding position of the N-terminal lobe of PhK and align the catalytic residues. Our model for interaction of Chk1KD with a substrate peptide suggests a structural basis for the substrate selectivity of human Chk1. In addition, this model may help to explain different substrate preferences between human and yeast Chk1 proteins. The amino acid sequences of the activation loops in yeast Chk1 proteins suggest that the loops are more flexible in the yeast proteins compared to human Chk1. Specifically, yeast Chk1 proteins lack the corresponding ion-pair between Lys180 and Glu248 that helps to anchor ␣EF in human Chk1. The residue equivalent to Lys166 of Chk1KD is substituted by Ser173 in S. pombe Chk1 and Asp189 in S. cerevisiae Chk1. These changes could alter substrate selectivity, since substitution of Lys166 with Ser or Asp would provide more room for a larger side chain at the P⫹2 position of the substrate peptide. Indeed, Phe and Thr residues are present at the P⫹2 position in S. pombe Cdc25 phosphorylation sites. Our model also allows prediction of potential Chk1 substrates. Human Chk1 has been shown to phosphorylate all human Cdc25 isoforms (Sanchez et al., 1997). Although the phosphorylation sites have not been mapped, there are several Ser residues in Cdc25A and Cdc25B lying in sequences that match the consensus derived from our model. In addition, S. pombe Chk1 phosphorylates the S. pombe Wee1 protein (O’Connell

Chk1 Kinase Domain Structure 689

Figure 7. Chk1KD Is More Active Than the Full-Length Chk1 and Autophosphorylation of Full-Length Chk1 Does Not Change Its Activity Kinase assays were performed using GST-Cdc25C(200–256) as substrate at 30⬚C for 5 min in the presence of ␥-32P-ATP (Sanchez et al., 1997), with 0.05 ␮M (1.7 ␮g/ml) of conserved kinase domain (residues 1–265), 0.05 ␮M of Chk1KD (residues 1–289), or 0.5 ␮M of recombinant full-length Chk1. Autophosphorylation of the recombinant full-length Chk1 was carried out by incubating the full-length protein at 36 ␮M (2 mg/ml) with ATP at 30⬚C for the indicated period of time. (A) Coomassie blue staining of 18 pmol of each protein resolved by SDS-PAGE with phosphorylated Chk1 bands marked on the right. (B) A typical phosphorimage of GST-Cdc25C(200–256) phosphorylated by Chk1. Chk1KD was 20-fold more active than the full-length Chk1 whereas the kinase activities of the full-length Chk1 after autophosphorylation were within 1- to 1.5-fold of the control.

et al., 1997; Boddy et al., 1998), in which Ser104 in the sequence Val-Pro-Arg-Arg-Pro-Ser-Leu-Phe-Asp is a consensus Chk1 phosphorylation site. Interestingly, this potential phosphorylation site lies in the highly divergent regulatory region, N-terminal to its kinase domain, and is not conserved in human Wee1 (Boddy et al., 1998). In the S. cerevisiae anaphase inhibitor Pds1, another substrate of Chk1 (Sanchez et al., 1999), there are several consensus Chk1 phosphorylation sites. One of them, Ser98, in the sequence 93-Asn-Asn-Arg-Ser-LysSer-Phe-Ile-Phe-101, lies next to the destruction box (residues 85-Arg-Leu-Pro-Leu-Ala-Ala-Lys-Asp-Asn-93) of Pds1. Identification of the actual phosphorylation site(s) in the aforementioned proteins would test our model of substrate selectivity for Chk1KD. The Chk1KD structure has revealed that the activation loop is stabilized in the absence of phosphorylation. In addition, Tyr20 in the glycine-rich loop of Chk1 did not appear to play a role in regulating Chk1 kinase activity. Taken together, our results suggested that Chk1 kinase activity is not regulated by phosphorylation of its kinase domain. Chk1 is phosphorylated after DNA damage and this phosphorylation has been correlated with DNA damage–induced G2 arrest (Walworth and Bernards, 1996; Sanchez et al., 1997; Kumagai et al., 1998; Martinho et al., 1998; Sanchez et al., 1999). However, we have not observed any significant impact of autophosphorylation on the kinase activity of recombinant Chk1 protein in

vitro. Autophosphorylation in vitro may not reflect physiological events induced by DNA damage and phosphorylation of Chk1 by another kinase may be required for its activation. Alternatively, phosphorylation-dependent interaction of Chk1 with another protein could contribute to Chk1 regulation. Chk1 has been shown to interact with two yeast checkpoint proteins, Cut5 and Crb2 (Saka et al., 1997), and Crb2 may regulate Chk1 phosphorylation (Esashi and Yanagida, 1999). The physiological significance of the finding that Chk1KD is more active than the full-length Chk1 requires further investigation of the cellular Chk1 activity. The C-terminal region may regulate Chk1 kinase activity directly. This region contains several motifs conserved within Chk1 proteins but shows no sequence homology to other proteins. Although this region is not included in Chk1KD, the crystal structure suggested potential mechanisms of regulation. The C-terminal region could control the access of substrates to the Chk1 kinase domain by interacting with the substrate binding site at the front of the kinase domain. Such an autoinhibitory mechanism has been observed for several kinases including titin (PDB code 1TKI) (Mayans et al., 1998). Like Chk1KD, titin is not regulated through phosphorylation of its activation loop. The titin structure showed that its C-terminal region extends into the ATP binding site and blocks substrate peptide binding. However, the short C-terminal linker visible in Chk1KD structure follows a different path. In addition, the conserved turn formed by the IEPDIG motif between helices ␣D and ␣E of Chk1KD blocks access to the front of the kinase domain. Alternatively, the C-terminal region of Chk1 could inhibit Chk1 kinase activity through interaction with the back of the kinase domain to maintain the open lobe conformation, as observed in the current crystal structures. Regulation of kinase activity by restraining lobe movement has been proposed for Src family kinases (Xu et al., 1997). The structure of Chk1KD also has practical implications for anticancer drug discovery since compounds that abrogate G2 arrest following DNA damage have been shown to improve the killing of tumor cells (Hartwell and Kastan, 1994; O’Connor and Fan, 1996). This is envisaged to occur through limiting the time available for repair of damaged DNA. Interestingly, p53-deficient cancer cells appear to be more susceptible to G2 checkpoint abrogation, suggesting such a strategy might prove selective to certain tumors (Fan et al., 1995; O’Connor and Fan, 1996). UCN-01, which is under clinical evaluation, is also a potent Chk1 kinase inhibitor (Sarkaria et al., 1999; Graves et al., 2000). Our work provides a structural base upon which selective inhibitors of Chk1 can now be designed. Key residues in the catalytic domain to target for kinase inhibition include Glu85 and Cys87. Further exploration of selectivity could be obtained through targeting residues that are special to Chk1, such as Asn59 and Leu84. Our model for interaction of Chk1KD with its substrate peptide also provides a structural basis for the design of selective peptidomimetic Chk1 inhibitors. Indeed, accumulation of Cdc25C peptides into cultured cells abrogates checkpoint control (Suganuma et al., 1999). In conclusion, the Chk1KD structure reveals structural

Cell 690

features important for kinase activity, substrate selectivity, and potential modes of Chk1 kinase regulation. Specifically, secondary structure and side chain interactions stabilize the activation loop of Chk1. We propose that Chk1 becomes fully active upon substrate binding, and that its activity is not regulated by phosphorylation within the kinase domain. Our results further suggest a potential regulatory role of the less conserved C-terminal region of Chk1. In addition, our model of Chk1 substrate interaction suggests a consensus sequence for phosphorylation by Chk1 that could aid identification of additional Chk1 substrates through which Chk1 affects cell cycle progression. Finally, the Chk1KD crystal structure should prove useful in the design of selective inhibitors of Chk1 for pharmacological and therapeutic investigations of checkpoint regulation. Experimental Procedures Protein Expression and Purification Human Chk1 cDNA was amplified by PCR using Vent plymerase (New England Biolabs) from human thymus and testis cDNA (Clontech). Large scale insect cell fermentation and purification will be described elsewhere (C. L., unpublished data). Chk1KD contains the conserved kinase domain (residues 1–265), part of the linker region (residues 266–289), and 6 His residues at the C-terminus. Chk1KD was purified from insect cells by sequential chromatography using Q-FastFlow (Pharmacia), Ni-NTA-Agarose (Qiagen), ATPSepharose (Xu et al., 1997), and HiPrep Sephacryl S-200 (Pharmacia) columns. Purified protein was concentrated to 7–7.5 mg/ml and stored at ⫺80⬚C before use. Crystallization and Data Collection Chk1KD crystals were obtained using a hanging-drop vapor-diffusion method. The Nat1 crystal was obtained by mixing equal volumes of protein solution and reservoir solution of 13% PEG 8000 (w/v), 0.15 M (NH4)2SO4, 0.1 M NaCacodylate (pH 6.8), and 2% glycerol. The Nat2 crystal was crystallized using reservoir solution of 12% PEG8000 (w/v), 15% iso-propanol, and 0.1 M HEPES (pH 7.5). Both conditions produced the exact same form of crystals belonging to the space group P21 with unit cell dimensions a ⫽ 45.2 A˚, b ⫽ 65.7 A˚, c ⫽ 58.1 A˚, ␤ ⫽ 93.9⬚. The crystals contain one molecule per asymmetric unit and 53% solvent by volume. The crystals of binary complex with AMP-PNP were obtained by cocrystallization first under Nat1 condition in the presence of 1.25 mM AMP-PNP and 2.5 mM MgCl2, then soaking in mother liquor with 5 mM MgCl2 and 20 mM AMP-PNP for two days. The cocrystals also have the same space group and cell dimensions as the native crystals. Crystals were introduced into a cryoprotectant solution containing reservoir solution and 20% glycerol. For AMP-PNP cocrystals, additional 10 mM MgCl2 and AMP-PNP were included in the cryoprotectant solution. Crystals were then flash frozen in a stream of nitrogen gas at ⫺170⬚C. All data were collected at ⫺170⬚C using a copper K␣ radiation produced by a Rigaku rotating anode Ru 200 X-ray generator equipped with Osmic confocal mirrors and a Mar345 image-plate detector. All data were processed with the Denzo/HKL package (Otwinowski, 1993). Structure Determination and Refinement Initial apoenzyme structure determination using Nat1 crystal data was carried out by molecular replacement (MR) using a modified Cdk2 structure (omitted loop regions) (Russo et al., 1996) as a search model. AmoRe (Navaza, 1994) rotation and translation functions revealed a solution using Nat1 data from 10 to 4 A˚. The MR model was refined by simulated annealing using XPLOR (Bru¨nger, 1992). However, 2Fo-Fc and Fo-Fc electron density maps were poorly defined at the loop regions, which were omitted from the initial model. Additional phase information was obtained from multiple isomorphous replacement with two heavy metal derivatives: 0.5 mM HgCl2 (soaked for 15 hr) and 5 mM KAu(CN)2 (soaked for 17 hr). Five Hg

sites and five Au sites were identified by difference Fourier synthesis using phases generated from MR partial model and were consistent with both isomorphous and anomalous difference Patterson maps. The positional and thermal parameters and relative occupancies for the heavy atom sites were refined using SIR data at 3 A˚ and anomalous data at 3.5 A˚ by program PHASES (Furey and Swaminathan, 1990). Sixteen cycles of solvent flattening were then carried out using phases calculated from refined Hg and Au positions. The resultant electron density maps showed good backbone density and well-defined side chains for most parts of the protein (Figure 5A). Model building utilized the program FRODO (Jones, 1978). The missing loop regions were incorporated into the model using both MIR maps and model phased 2Fo-Fc maps. Further refinement using XPLOR (Bru¨nger, 1992) and then CNS (Bru¨nger et al., 1998) were continued with both conjugate gradient minimization, simulated annealing, then followed by manual rebuilding. Refinement of Nat2 structure was carried out using the refined Nat1 model but omitting residues 153–170 as well as SO42⫺ anion. Fo-Fc maps showed well-defined densities for the omitted region and its conformation is the same as that in Nat1. Refinement of the binary complex with AMP-PNP proceeded with refining the position of the apoenzyme model (Nat1) as a rigid body against the complex data using the CNS program. Fo-Fc maps with ␴A (Read, 1986) weighting showed clear density for the adenine and ribose components of AMP-PNP (Figure 5B). The conformation of residues forming the binding pocket was checked in simulated annealing omit-maps before including the adenine and ribose components of AMP-PNP. The current apoenzyme model (Nat1) includes all atoms for residues 2 to 44, 48 to 276, and 181 ordered solvent molecules and one SO42⫺ anion. Residues 45 to 47 and 277 to 289 plus the 6⫻His tag have not been modeled into the current apoenzyme structure. The refined Nat2 structure contains the same number of residues and 168 solvent molecules but the SO42⫺ anion is not present. The refined AMP-PNP complex contains the same number of residues as the apoenzyme structures with 251 ordered solvent molecules and one SO42⫺ anion. Acknowledgments We thank D. Matthews, Z. Hostomska, R. Love, and D. Knighton for their critical comments on the manuscript; M. Gehring for the expression vector and advice in protein purification; F. Maldonado for technical support; B. Burke and colleagues of the project team for valuable discussions. Received November 30, 1999; revised February 23, 2000. References Boddy, M.N., Furnari, B., Mondesert, O., and Russell, P. (1998). Replication checkpoint enforced by kinases Cds1 and Chk1. Science 280, 909–912. Bru¨nger, A.T. (1992). XPLOR v3.1, A System for X-Ray Crystallography and NMR (New Haven, CT: Yale University Press). Bru¨nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kuntsleve, R.W., Jiang, J.-S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography and NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D54, 905–921. Chen, P., Gatei, M., O’Connell, M.J., Khanna, K.K., Bugg, S.J., Hogg, A., Scott, S.P., Hobson, K., and Lavin, M.F. (1999). Chk1 complements the G2/M checkpoint defect and radiosensitivity of ataxiatelangiectasia cells. Oncogene 18, 249–256. Dalal, S.N., Schweitzer, C.M., Gan, J., and DeCaprio, J.A. (1999). Cytoplasmic localization of human cdc25C during interphase requires an intact 14–3–3 binding site. Mol. Cell. Biol. 19, 4465–4479. De Bondt, H.L., Rosenblatt, J., Jancarik, J., Jones, H.D., Morgan, D.O., and Kim, S.-H. (1993). Crystal structure of cyclin-dependent kinase 2. Nature 363, 595–602. Elledge, S.J. (1996). Cell cycle checkpoints: preventing an identity crisis. Science 274, 1664–1672.

Chk1 Kinase Domain Structure 691

Esashi, F., and Yanagida, M. (1999). Cdc2 phosphorylation of Crb2 is required for reestablishing cell cycle progression after the damage checkpoint. Mol. Cell 4, 167–174. Fan, S., Smith, M.L., Rivet, D.J., Duba, D., Zhan, Q., Kohn, K.W., Fornace, A.J., Jr., and O’Connor, P.M. (1995). Disruption of p53 function sensitizes breast cancer MCF-7 cells to cisplatin and pentoxifylline. Cancer Res. 55, 1649–1654. Francesconi, S., Grenon, M., Bouvier, D., and Baldacci, G. (1997). p56chk1 protein kinase is required for the DNA replication checkpoint at 37⬚C in fission yeast. EMBO J. 16, 1332–1341. Furey, W., and Swaminathan, S. (1990). “PHASES”—a program package for the processing and analysis of diffraction data from macromolecules. PA33, American Crystallographic Association Meeting Summaries, Series 2 18, 73. Graves, P.R., Yu, L., Schwarz, J.K., Gales, J., Sausville, E.A., O’Connor, P.M., and Piwnica-Worms, H. (2000). The Chk1 protein kinase and the Cdc25C regulatory pathway are targets of the anticancer agent UCN-01. J. Biol. Chem., in press. Hartwell, L.H., and Kastan, M.B. (1994). Cell cycle control and cancer. Science 266, 1821–1828. Hartwell, L.H., and Weinert, T.A. (1989). Checkpoint: controls that ensure the order of cell cycle events. Science 246, 629–634. Hoffmann, I., Clarke, P.R., Marcote, M.J., Karsenti, E., and Draetta, G. (1993). Phosphorylation and activation of human cdc25-C by cdc2-cyclin B and its involvement in the self-amplification of MPF at mitosis. EMBO J. 12, 53–63. Hubbard, S.R. (1997). Crystal structure of the activated insulin receptor tyrosine kinase in complex with peptide substrate and ATP analog. EMBO J. 16, 5572–5581. Jin, P., Gu, Y., and Morgan, D.O. (1996). Role of inhibitory CDC2 phosphorylation in radiation-induced G2 arrest in human cells. J. Cell Biol. 134, 963–970.

Nuclear localization of Cdc25 is regulated by DNA damage and a 14–3–3 protein. Nature 397, 172–175. Lowe, E.D., Noble, M.E. M., Skamnaki, V.T., Oikonomakos, N.G., Owen, D.J., and Johnson, L.N. (1997). The crystal structure of a phosphorylase kinase peptide substrate complex: kinase substrate recognition. EMBO J. 16, 6646–6658. Martinho, R.G., Lindsay, H.D., Flaggs, G., DeMaggio, A.J., Hoekstra, M.F., Carr, A.M., and Bentley, N.J. (1998). Analysis of Rad3 and Chk1 protein kinases defines different checkpoint responses. EMBO J. 17, 7239–7249. Matsuoka, S., Huang, M., and Elledge, S.J. (1998). Linkage of ATM to cell cycle regulation by the Chk2 protein kinase. Science 282, 1893–1897. Mayans, O., van der Ven, P.F.M., Wilm, M., Mues, A., Young, P., Fu¨rst, D.O., Wilmanns, M., and Gautel, M. (1998). Structural basis for activation of the titin kinase domain during myofibrillogenesis. Nature 395, 863–869. McRee, D.E. (1992). A visual protein crystallographic software system for X11/Xview. J. Mol. Graph. 10, 44–46. Millar, J.B.A., Blevitt, J., Gerace, L., Sadhu, K., Featherstone, C., and Russell, P. (1991). p55CDC25 is a nuclear protein required for the initiation of mitosis in human cells. Proc. Natl. Acad. Sci. USA 88, 10500–10504. Murray, A.W. (1992). Creative blocks: cell-cycle checkpoints and feedback controls. Nature 359, 599–604. Muslin, A.J., Tanner, J.W., Allen, P.M., and Shaw, A.S. (1996). Interaction of 14–3–3 with signaling proteins is mediated by the recognition of phosphoserine. Cell 84, 889–897. Navaza, J. (1994). AmoRe: an automated package for molecular replacement. Acta Crystallogr. A50, 157–163. O’Connell, M.J., Raleigh, J.M., Verkade, H.M., and Nurse, P. (1997). Chk1 is a wee1 kinase in the G2 DNA damage checkpoint inhibiting cdc2 by Y15 phosphorylation. EMBO J. 16, 545–554.

Johnson, L.N., Noble, M.E.M., and Owen, D.J. (1996). Active and inactive protein kinases: structural basis for regulation. Cell 85, 149–158.

O’Connor, P.M. (1997). Mammalian G1 and G2 phase checkpoints. Cancer Surv. 29, 151–182.

Jones, T.A. (1978). A graphics model building and refinement system for macromolecules. J. Appl. Crystallogr. 11, 268–272.

O’Connor, P.M., and Fan, S. (1996). DNA damage checkpoints: implications for cancer therapy. Prog. Cell Cycle Res. 2, 165–173.

Kaneko, Y., Watanabe, N., Morisaki, H., Akita, H., Fujimoto, A., Tominaga, K., Terasawa, M., Tachibana, A., Ikeda, K., and Nakanishi, M. (1999). Cell cycle-dependent and ATM-independent expression of human Chk1 kinase. Oncogene 18, 3673–3681.

Otwinowski, Z. (1993). DENZO. In Data Collection and Processing of Proceedings of the CCP4 Study Weekend, L. Sawyer, N. Isaacs and S. Bailey, eds. (Warrington, UK: SERC Daresbury Laboratory), pp. 56–62.

Knighton, D.R., Zheng, J., Ten Eyck, L.F., Ashford, V.A., Xuong, N.-H., Taylor, S.S., and Sowadski, J.M. (1991a). Crystal structure of the catalytic subunit of cyclic adenosine monophosphate-dependent protein kinase. Science 253, 407–414.

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.

Knighton, D.R., Zheng, J., Ten Eyck, L.F., Xuong, N.-H., Taylor, S.S., and Sowadski, J.M. (1991b). Structure of a peptide inhibitor bound to the catalytic subunit of cyclic adenosine monophosphate-dependent protein kinase. Science 253, 414–420.

Peng, C.-Y., Graves, P.R., Thoma, R.S., Wu, Z., Shaw, A.S., and Piwnica-Worms, H. (1997). Mitotic and G2 checkpoint control: Regulation of 14–3–3 protein binding by phosphorylation of Cdc25C on serine-216. Science 277, 1501–1505.

Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallogr. 24, 946–950.

Poon, R.Y.C., Jiang, W., Toyoshima, H., and Hunter, T. (1996). Cyclindependent kinases are inactivated by a combination of p21 and Thr-14/Tyr-15 phosphorylation after UV-induced DNA damage. J. Biol. Chem. 271, 13283–13291.

Krek, W., and Nigg, E.A. (1991). Mutations of p34cdc2 phosphorylation sites induce premature mitotic events in HeLa cells: evidence for a double block to p34cdc2 kinase activation in vertebrates. EMBO J. 10, 3331–3341.

Read, R.J. (1986). Improved Fourier coefficients for maps using phases from partial structures with errors. Acta Crystallogr. A42, 140–149.

Kumagai, A., Guo, Z., Emami, K.H., Wang, S.X., and Dunphy, W.G. (1998). The Xenopus Chk1 protein kinase mediates a caffeine-sensitive pathway of checkpoint control in cell-free extracts. J. Cell Biol. 142, 1559–1569. Lee, M.S., Ogg, S., Xu, M., Parker, L.L., Donoghue, D.J., Maller, J.L., and Piwnica-Worms, H. (1992). cdc25⫹ encodes a protein phosphatase that dephosphorylates p34cdc2. Mol. Biol. Cell 3, 73–84. Lindsay, H.D., Griffiths, D.J. F., Edwards, R.J., Christensen, P.U., Murray, J.M., Osman, F., Walworth, N., and Carr, A.M. (1998). S-phase-specific activation of Cds1 kinase defines a subpathway of the checkpoint response in Schizosaccharomyces pombe. Genes Dev. 12, 382–395. Lopez-Girona, A., Furnari, B., Mondesert, O., and Russell, P. (1999).

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. Russo, A.A., Jeffrey, P.D., and Pavletich, N.P. (1996). Structural basis of cyclin-dependent kinase activation by phosphorylation. Nat. Struct. Biol. 3, 696–700. Saka, Y., Esashi, F., Matsusaka, T., Mochida, S., and Yanagida, M. (1997). Damage and replication checkpoint control in fission yeast is ensured by interactions of Crb2, a protein with BRCT motif, with Cut5 and Chk1. Genes Dev. 11, 3387–3400. Sanchez, Y., Wong, C., Thoma, R.S., Richman, R., Wu, Z., PiwnicaWorms, H., and Elledge, S.J. (1997). Conservation of the Chk1 checkpoint pathway in mammals: linkage of DNA damage to Cdk regulation through Cdc25. Science 277, 1497–1501.

Cell 692

Sanchez, Y., Bachant, J., Wang, H., Hu, F., Liu, D., Tetzlaff, M., and Elledge, S.J. (1999). Control of the DNA damage checkpoint by Chk1 and Rad53 protein kinases through distinct mechanisms. Science 286, 1166–1171. Sarkaria, J.N., Busby, E.C., Tibbetts, R.S., Roos, P., Taya, Y., Karnitz, L.M., and Abraham, R.T. (1999). Inhibition of ATM and ATR kinase activity by the radiosensitizing agent, caffeine. Cancer Res. 59, 4375–4382. Suganuma, M., Kawabe, T., Hori, H., Funabiki, T., and Okamoto, T. (1999). Sensitization of cancer cells to DNA damage-induced cell death by specific cell cycle G2 checkpoint abrogation. Cancer Res. 59, 5887–5891. 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. Walworth, N.C., and Bernards, R. (1996). rad-dependent response of the chk1-encoded protein kinase at the DNA damage checkpoint. Science 271, 353–356. Walworth, N., Davey, S., and Beach, D. (1993). Fission yeast chk1 protein kinase links the rad checkpoint pathway to cdc2. Nature 363, 368–371. Weinert, T. (1998). DNA damage and checkpoint pathways: molecular anatomy and interactions with repair. Cell 94, 555–558. Xu, W., Harrison, S.C., and Eck, M.J. (1997). Three-dimensional structure of the tyrosine kinase c-Src. Nature 385, 595–602. Yaffe, M.B., Rittinger, K., Volinia, S., Caron, P.R., Aitken, A., Leffers, H., Gamblin, S.J., Smerdon, S.J., and Cantley, L.C. (1997). The structural basis for 14–3–3:phosphopeptide binding specificity. Cell 91, 961–971. Zeng, Y., Forbes, K.C., Wu, Z., Moreno, S., Piwnica-Worms, H., and Enoch, T. (1998). Replication checkpoint requires phosphorylation of the phosphatase Cdc25 by Cds1 or Chk1. Nature 395, 507–510.