Mechanism of Activation of the RAF-ERK Signaling Pathway by ...

Cell, Vol. 116, 855–867, March 19, 2004, Copyright 2004 by Cell Press

Mechanism of Activation of the RAF-ERK Signaling Pathway by Oncogenic Mutations of B-RAF Paul T.C. Wan,1,6 Mathew J. Garnett,2,6 S. Mark Roe,1 Sharlene Lee,3 Dan Niculescu-Duvaz,4 Valerie M. Good,1 Cancer Genome Project,5 C. Michael Jones,3 Christopher J. Marshall,2 Caroline J. Springer,4 David Barford,1,* and Richard Marais2,* 1 Section of Structural Biology 2 Cancer Research UK Centre for Cell and Molecular Biology 3 Section of Gene Function and Regulation The Institute of Cancer Research Chester Beatty Laboratories 237 Fulham Road London SW3 6JB United Kingdom 4 Cancer Research UK Centre for Cancer Therapeutics The Institute of Cancer Research 15 Cotswold Road Sutton, SM2 5NG United Kingdom 5 Cancer Genome Project The Wellcome Trust Sanger Institute Wellcome Trust Genome Campus Hinxton United Kingdom

Summary Over 30 mutations of the B-RAF gene associated with human cancers have been identified, the majority of which are located within the kinase domain. Here we show that of 22 B-RAF mutants analyzed, 18 have elevated kinase activity and signal to ERK in vivo. Surprisingly, three mutants have reduced kinase activity towards MEK in vitro but, by activating C-RAF in vivo, signal to ERK in cells. The structures of wild type and oncogenic V599EB-RAF kinase domains in complex with the RAF inhibitor BAY43-9006 show that the activation segment is held in an inactive conformation by association with the P loop. The clustering of most mutations to these two regions suggests that disruption of this interaction converts B-RAF into its active conformation. The high activity mutants signal to ERK by directly phosphorylating MEK, whereas the impaired activity mutants stimulate MEK by activating endogenous C-RAF, possibly via an allosteric or transphosphorylation mechanism. Introduction The RAF-MEK-ERK signal transduction pathway is a conserved RAS-activated protein kinase cascade that regulates cell growth, proliferation, and differentiation in response to growth factors, cytokines, and hormones *Correspondence: [email protected] (D.B.), richard.marais@ icr.ac.uk (R.M.) 6 These authors contributed equally to this work.

(Robinson and Cobb, 1997). The transforming activities of the viral oncogenic forms of RAS and RAF were key to their discovery (Rapp et al., 1983), and RAF was the first effector identified downstream of RAS (Avruch et al., 2001; Malumbres and Barbacid, 2003). RAF activation is initiated by RAS-GTP association with the RAS binding domain (RBD) situated within the N-terminal regulatory region of the kinase. Concomitant conformational changes and recruitment to the cell membrane promote changes in RAF phosphorylation that combine to stimulate its serine/threonine kinase activity, triggering sequential phosphorylation and activation of MEK and ERK (Kolch, 2000; Morrison and Cutler, 1997). The three functional RAF proteins in humans, A-RAF, B-RAF, and C-RAF (also termed c-Raf-1), are dependent on activation segment phosphorylation for activity (Chong et al., 2001; Zhang and Guan, 2000). However, the details of their regulatory mechanisms differ because C-RAF and A-RAF require additional serine and tyrosine phosphorylation within the N region of the kinase domain for full activity (Mason et al., 1999), and B-RAF has a much higher basal kinase activity than either A-RAF or C-RAF. Communication between RAF, MEK, and ERK requires the scaffolding protein KSR1, a RAF homolog that is devoid of kinase activity (Morrison, 2001). Constitutive activation of the RAS-ERK signaling pathway is common to numerous cancers. Approximately 15% of human cancers have activating RAS mutations (Malumbres and Barbacid, 2003), and recently mutations in B-RAF were identified in a large-scale screen for genes mutated in human cancer (Davies et al., 2002). Somatic mutations of B-RAF are associated with 60% of malignant melanomas and occur with moderate to high frequency in colorectal (Rajagopalan et al., 2002), ovarian (Singer et al., 2003), and papillary thyroid carcinomas (Brose et al., 2002; Cohen et al., 2003), implicating activating oncogenic mutations of B-RAF as critical promoters of malignancy. Significantly, B-RAF and RAS mutations are restricted to the same tumor types, usually in a mutually exclusive fashion, suggesting that these genes are on the same oncogenic signaling pathway and that RAS acts to activate B-RAF in these tumors. Sequence analysis of the B-RAF gene associated with human cancers has identified over 30 single site missense mutations, mostly within the kinase domain. The mechanism of oncogenic activation of B-RAF therefore differs fundamentally from that of v-Raf, a retroviral oncogene derived from C-RAF. The constitutive activity and high transforming potential of v-Raf most likely results from loss of the autoinhibitory N-terminal region combined with targeting to the plasma membrane. Most of the mutations of B-RAF are clustered to two regions: the glycine-rich P loop of the N lobe and the activation segment and flanking regions. A Glu for Val substitution at residue 599 in the activation segment, adjacent to the conserved DFG motif, accounts for 90% of B-RAF mutations in human cancers. The V599E mutant of B-RAF possesses the hallmarks of a conventional oncogene. The kinase activity of this mutant protein is greatly ele-

Cell 856

vated, it constitutively stimulates ERK activity in vivo independent of RAS, and it potently transforms NIH3T3 cells. Interestingly, the conserved regulatory phosphorylation sites within the activation segment of B-RAF, Thr598 and Ser601, flank Val599, leading to the suggestion that the Glu substitution at this position functions as a phospho-mimetic (Davies et al., 2002). Analysis of three other oncogenic mutants of B-RAF showed that they stimulate kinase activity in a manner similar to V599E B-RAF (Davies et al., 2002). Intriguingly however, extensive analysis of B-RAF mutations in cancer shows that seven of the mutations involve highly conserved or invariant residues in the catalytic domain (Davies et al., 2002; Yuen et al., 2002) that in other kinases are known to be required for optimal catalytic activity (Johnson et al., 1996; Manning et al., 2002), raising the question of how these mutants promote tumorigenesis. To investigate the mechanisms by which mutant oncogenic forms of B-RAF promote cancer, we have examined a panel of 22 mutants. We show that eighteen mutants activate B-RAF in vitro and stimulate ERK signaling in vivo, conforming to the conventional model of an activating oncogene. However, four mutants have reduced kinase activity in vitro, but surprisingly, three of these can activate wild-type C-RAF and thereby signal to ERK. We present a crystallographic analysis of the WTB-RAF and V599EB-RAF kinase domains in complex with the C-RAF inhibitor BAY43-9006 (Lyons et al., 2001). These structures suggest that many of the residues that are mutated in cancer contribute to stabilization of an inactive conformation of the B-RAF kinase domain. Mutation of these residues destabilizes this inactive conformation, promoting the active state. For most mutants, this stimulates enhanced B-RAF kinase activity toward MEK. However, a few mutants act through a different mechanism because although their activity toward MEK is reduced, they adopt a conformation that activates wild-type C-RAF, which then signals to ERK. Results and Discussion The Majority of Mutant B-RAF Proteins in Human Cancer Have Elevated Kinase Activity To characterize B-RAF mutants, we transiently expressed them as myc-epitope-tagged fusion proteins in COS cells and measured their in vitro kinase activity. Some of the mutants have an increased Km for ATP compared to WTB-RAF, resulting in a relative underestimation of their activity when assayed at subphysiological concentrations of ATP (Supplemental Figure S2 at http://www.cell.com/cgi/content/full/116/6/855/DC1). However, there was little apparent difference in the Km of wild-type and mutant B-RAF for MEK (Supplemental Figure S3 online). Equivalent amounts of each mutant protein were immunoprecipitated (Supplemental Figure S1), and their kinase activity was determined by measuring direct MEK phosphorylation using ATP at a physiological concentration of 5 mM and MEK at 30 ␮M. Oncogenic RAS (G12VRAS) activates wild-type B-RAF (WTB-RAF) by ⵑ95-fold (Figures 1A and 1B, Supplemental Table S1 on Cell website). Seven of the mutants had basal kinase activities that exceeded G12VRAS-stimulated WTB-RAF activity, being ⵑ130 (E585KB-RAF) to 700

(V599DB-RAF) fold more active than basal WTB-RAF (Figure 1A). We refer to these mutants as the high activity group. A further eleven mutants had basal kinase activities some 1.3 (G468EB-RAF) to 64 (L596VB-RAF) fold higher than WT B-RAF (Figure 1B), and because their activities were between basal and G12VRAS-activated WTB-RAF, we refer to these mutants as the intermediate activity group. The activities relative to WTB-RAF we observe are significantly higher than those recently published for four of the mutants (Ikenoue et al., 2003). However, in that study, ATP was used at the subphysiological concentration of ⵑ130 ␮M, which may account for their relatively low activity. Next, we examined the ability of these mutants to activate endogenous ERK in COS cells. Western blotting with an antibody that specifically recognizes the dually phosphorylated and active forms of ERK1 and ERK2 indicated that all eighteen mutants stimulate endogenous ERK in COS cells (Figures 1C and Supplemental Table S1 online). The high activity mutants stimulated ERK phosphorylation to a similar level as that stimulated by G12VRAS, whereas the intermediate activity mutants were less efficient at stimulating ERK phosphorylation (Figure 1C). We used an immunoprecipitation kinase assay to measure ERK activity directly. G12VRAS activated ERK by approximately 4-fold (Figure 1D). G465AB-RAF, G468E B-RAF, N580SB-RAF, and V599EB-RAF, whose in vitro basal kinase activities are between 1.3- and 480-fold higher than WTB-RAF activity (Figures 1A and 1B), activated ERK by 2- to 4.6-fold (Figure 1D). Thus, despite possessing widely differing in vitro kinase activities, individual B-RAF mutants activated ERK to similar levels as G12V RAS. Finally, several high and intermediate activity mutants transformed NIH3T3 cells (Supplemental Table S1 on Cell website). Three B-RAF Cancer Mutants Have Impaired Kinase Activity but Activate ERK in Cells We identified four cancer-associated B-RAF mutants whose basal kinase activities were reduced to between 30% and 80% of the activity of WTB-RAF (Figure 2A). We term these mutants the impaired activity group. Strikingly, however, despite their reduced in vitro kinase activity, three of the mutants still activated endogenous ERK in COS cells, although to lower levels than other mutants (Figures 2B and 2C). One mutant, D593VB-RAF, failed to activate ERK in COS cells and behaved similarly to a mutant (KDB-RAF) in which the catalytic lysine (Lys482) is mutated to methionine (Figures 2B and 2C). The three impaired activity mutants that activate ERK also activate MEK in COS cells (Figure 2D). To further analyze these mutants in vivo, we expressed them in developing Xenopus embryos and stained for active ERK. WTB-RAF did not activate ERK in Xenopus cells, whereas V599EB-RAF and the three impaired activity mutants that activate ERK in COS cells induced strong activation in Xenopus cells (Figure 2E). As in COS cells, D593VB-RAF and KDB-RAF did not activate ERK in Xenopus cells (Figure 2E). Impaired Activity Mutants Stimulate C-RAF In Vivo One explanation for how impaired activity mutants stimulate MEK in vivo is that they activate endogenous C-RAF. Previous studies had demonstrated that C-RAF

Mechanism of Oncogenic Activation of B-RAF 857

Figure 1. Characterization of Activated Mutant B-RAF Proteins from Human Cancer The indicated myc-epitope-tagged wild-type and mutant B-RAF proteins were expressed in COS cells and their kinase activity was measured. Where indicated, G12VRAS (RAS) was included in the transfections. (A) B-RAF kinase activity for high activity mutants. (B) B-RAF kinase activity for intermediate activity mutants. (C) Western blot of myc-tagged B-RAF, total ERK2, and ppERK in COS cells expressing the high and intermediate activity mutants. (D) ERK kinase activity in COS cells expressing B-RAF mutant proteins.

and B-RAF form complexes in mammalian cells (Weber et al., 2001). Consistent with these data, we found that exogenously expressed WTB-RAF formed a complex with endogenous C-RAF (Figure 3A). This complex was not disrupted by G12VRAS, and C-RAF also formed complexes with high (V599EB-RAF), intermediate (G465AB-RAF, G468E B-RAF), and impaired (G465EB-RAF, G465VB-RAF, G595R B-RAF) activity B-RAF mutants (Figure 3A). Thus the complexes are not affected by either B-RAF activity or mutational status. To test whether C-RAF was activated, we used an antibody to immunoprecipitate C-RAF and measured the immunoprecipitates for RAF kinase activity in a coupled kinase cascade assay at 800 ␮M ATP, conditions where intermediate and impaired activity mutants have negligible activity (Supplemental Figure S2 on Cell website), whereas C-RAF activity remains high (Mason et al., 1999). Remarkably, the three impaired activity B-RAF mutants that induce ERK activation in

COS and Xenopus cells induced strong C-RAF activation, as did the intermediate activity mutants (Figure 3B). WT B-RAF, D593VB-RAF, and KDB-RAF, which do not activate ERK in COS cells, do not activate C-RAF (Figure 3B). To test whether the high activity mutant V599EB-RAF can also activate C-RAF, we expressed HA-tagged C-RAF with myc-tagged V599EB-RAF in COS cells, immunoprecipitated C-RAF, and measured its activity. To demonstrate that the activity being measured is due to C-RAF, we show that G465AB-RAF, which activates endogenous C-RAF in COS cells (Figure 3B), activates exogenous WTC-RAF but not KDC-RAF (Figure 3C). When V599E B-RAF was coexpressed with WTC-RAF, substantial RAF kinase activity was observed in the C-RAF immunoprecipitate, but this was reduced by ⵑ80% when KD C-RAF was substituted for WTC-RAF (Figure 3C). We assume that the residual activity represents the contribution from coimmunoprecipitated V599EB-RAF and con-

Cell 858

Figure 2. Characterization of Impaired Activity B-RAF Mutants (A) B-RAF kinase activity. (B) ppERK staining in COS cells extracts. (C) ERK kinase activity in COS cell extracts. (D) MEK phosphorylation in COS cells expressing B-RAF mutants or G12VRAS. (E) ERK activation in Xenopus embryos expressing the indicated B-RAF mutants or uninjected control.

clude that V599EB-RAF activates C-RAF. Mutant B-RAF proteins were extremely efficient at activating C-RAF, inducing 5- to 8-fold more activity than was induced by G12V RAS (Figure 3C). To provide additional evidence that the impaired activity B-RAF mutants stimulate ERK activity in vivo through activation of WTC-RAF, we used an RNA interference approach. A C-RAF-specific siRNA oligonucleotide reduced C-RAF protein by 50%–60% compared to the nonspecific, scrambled control but did not alter the levels of either B-RAF (Figure 3D) or A-RAF protein (C. Wellbrook and R.M., unpublished data). Depleting C-RAF had little effect on ERK activity in COS cells expressing WTB-RAF, the intermediate activity mutant G465A B-RAF, or the high activity mutant V599EB-RAF (Figures 3D and 3E). In contrast, in COS cells expressing the impaired activity mutants, C-RAF depletion significantly suppressed ERK activation (Figure 3E). Similarly, in the human cancer cell line WM266-4, which expresses the high activity mutant V599DB-RAF (Davies et al., 2002),

C-RAF depletion did not affect ERK activity, whereas B-RAF depletion blocked ERK activity (Figure 3F). In H1666 cells, however, which express the impaired activity mutant G465VB-RAF (Davies et al., 2002), both B-RAF and C-RAF are required for ERK activation (Figure 3F). Thus, although C-RAF is not required for ERK activation by either the high or intermediate activity B-RAF mutants, the ability of impaired activity B-RAF mutants to stimulate ERK is dependent on C-RAF. Expression, Purification, and Enzyme Activity of WTB-RAF and V599EB-RAF Kinase Domains To understand how cancer-associated mutations influence the activity and function of the protein, we determined the crystal structures of the WTB-RAF and V599E B-RAF kinase domains. We expressed the kinase domain of human B-RAF equivalent to v-Raf using the baculovirus/insect cell system. B-RAF is a client protein of the Hsp90/p50Cdc37 chaperone (Blagosklonny, 2002; Stancato et al., 1993), and coexpression of B-RAF with


Figure 3. C-RAF Is Required for ERK Activation by the Impaired B-RAF Mutants (A) B-RAF and C-RAF complex analysis in COS cells. Levels of exogenous B-RAF and endogenous C-RAF are shown in the lower two panels and B-RAF in the C-RAF IP in the upper panel. Identical results were obtained using two different anti-C-RAF antibodies. (B) Kinase activity of endogenous C-RAF from COS cells expressing the indicated B-RAF mutant. (C) The RAF activity in C-RAF immunoprecipitates from COS cells expressing WTC-RAF (WT) or KDC-RAF (KD) together with the indicated B-RAF constructs. (D) C-RAF depletion in COS cells. COS cells expressing the B-RAF mutant indicated were treated with C-RAF or scrambled control (SCR) siRNA oligonucleotides as indicated and the levels of mutant B-RAF, endogenous C-RAF, phosphorylated ERK, and total ERK are shown. (E) Endogenous ERK kinase activity in COS cells expressing B-RAF mutants and treated with siRNA. (F) C-RAF depletion in cancer cell lines. WM266-4 and H1666 cells were treated under mock conditions or with the C-RAF, B-RAF, and scrambled control (SCR) siRNA oligonucleotides as indicated, and the levels of endogenous B-RAF, C-RAF, phosphorylated ERK, and total ERK are shown.

human p50Cdc37 significantly increased expression levels (data not shown), most likely because endogenous insect cell Hsp90 stabilizes the human B-RAF protein.

Using this system, B-RAF was isolated as a complex with p50Cdc37 and two insect cell proteins, Hsp90 and 14-3-3. The role of 14-3-3 binding to the C terminus

Cell 860

of C-RAF is controversial. Some studies suggest that binding is necessary for activity (Avruch et al., 2001; Kolch, 2000). However, under some conditions, dissociation of 14-3-3 from C-RAF does not alter its kinase activity toward MEK in vitro (Michaud et al., 1995), and 14-3-3 binding to the C terminus of B-RAF is not required for the ability of B-RAF to phosphorylate MEK in vitro but instead couples the kinase to downstream effector complexes (MacNicol et al., 2000). To isolate a more homogeneous preparation of B-RAF for crystallographic studies, we expressed a C-terminal truncated form of the kinase that removed the 14-3-3 binding site. This isolated B-RAF kinase domain (termed WT⌬B-RAF) was separated from Hsp90/p50Cdc37 during purification and was then dephosphorylated. To analyze the consequence of a Glu substitution for Val599, we also expressed the kinase domain of V599EB-RAF (V599E⌬B-RAF). The V599E⌬B-RAF kinase domain was ⵑ500-fold more active than WT⌬B-RAF (data not shown). Thus the Glu substitution for Val599 exerts its stimulatory effect on kinase activity within the context of the kinase domain of B-RAF, and this suggests that B-RAF activation is not simply a release from the autoinhibitory restraints conferred by the N-terminal region. Moreover, because we dephosphorylated the purified B-RAF kinase domain, the high activity of V599E⌬B-RAF indicates that activation segment phosphorylation is not required for the kinase activity of the mutant protein. Overall Structure and B-RAF BAY43-9006 Interactions The crystals that we obtained with the WT⌬B-RAF and V599E ⌬B-RAF kinase domains alone were not suitable for analysis. However, in the presence of the RAF inhibitor BAY43-9006, suitable crystals were obtained with both preparations. The crystals in the presence of BAY439006 grew as thin needles with a maximum thickness of 15–20 ␮m and diffracted to resolutions of 2.9 A˚ and 3.4 A˚ for WT⌬B-RAF and V599E⌬B-RAF, respectively. Details of the structure determination are described in Experimental Procedures and Supplemental Table S2 online. Wild-type and V599E⌬B-RAF kinase domains adopt essentially identical conformations in the presence of the inhibitor (rmsd between equivalent C␣-atoms of 0.64 A˚, Supplemental Figure S4). The protein is well ordered, except for the N terminus (Gln432–Ser446), and a region of the activation segment, residues Lys600–Gln611. The kinase domain of ⌬B-RAF adopts the bilobal architecture characteristic of other members of the protein kinase family (Figure 4). In both wild-type and mutant ⌬B-RAF, the BAY43-9006 molecule is well resolved in the electron density maps, with the inhibitor spanning the length of the interfacial cleft, buried deeply between the N and C lobes (Figures 4 and 5A). Numerous interactions are formed between the protein and inhibitor, and residues that contact BAY43-9006 are conserved in C-RAF, consistent with much of the structure-activity relationships of C-RAF inhibition by BAY43-9006 and its derivatives (Lowinger et al., 2002). The distal pyridyl ring of the inhibitor occupies the ATP adenine binding pocket, interacting with three aromatic residues: Trp530 of the hinge region, Phe582 at the end of the catalytic loop, and Phe594 of the DFG motif, which also contacts

the central phenyl ring of the inhibitor. At the opposite end of the inhibitor, the lipophilic trifluoromethyl phenyl ring inserts into a hydrophobic pocket formed between the ␣C and ␣E helices and N-terminal regions of the DFG motif and catalytic loop. The aliphatic side chains of Lys482, Leu513, and Thr528 contact the central phenyl ring of the inhibitor. In addition to the van der Waals interactions that dominate B-RAF-inhibitor contacts, polar interactions also contribute to binding. Importantly, the urea group of the inhibitor forms two hydrogen bonds with the protein, one via its amide nitrogen atom to the carboxylate side chain of the catalytic Glu500 residue and the second via the carbonyl moiety to the main chain nitrogen of Asp593 of the DFG motif. An equivalent constellation of hydrogen bonds contributes to the interactions between c-Abl and p38 MAP kinase and their respective inhibitors STI-571 and BIRB-796 (Nagar et al., 2002; Pargellis et al., 2002; Schindler et al., 2000). Finally, the ring nitrogen atom of the pyridyl moiety, which enhances affinity some 5-fold compared with a carbon atom (Lowinger et al., 2002), accepts a hydrogen bond from the main chain nitrogen of Cys531 of the interdomain hinge region, whereas the methyl amide side group contacts the main chain carbonyl of Cys531. The structure of ⌬B-RAF is markedly reminiscent of the inactive conformation of the c-Abl tyrosine kinase domain associated with the inhibitor STI-571 (Gleevec) (Figure 5B) (Nagar et al., 2002; Schindler et al., 2000), and it also shares significant similarities with the p38 MAP kinase-BIRB 796 complex (Pargellis et al., 2002). There is considerable overlap between the B-RAF and c-Abl inhibitors, and each molecule interacts with their cognate kinases in essentially the same way. Specifically, in the two proteins, the relative dispositions of the N and C lobes, ␣C helix, and DFG motif are virtually identical. The DFG motif Phe residues in both ⌬B-RAF and c-Abl (Phe594 in ⌬B-RAF) are shifted by some 8 A˚ relative to their counterparts in active protein kinases. BAY43-9006 contributes to the inactive conformation of the DFG motif both by interacting with the phenyl ring of Phe594 and because the trifluoromethyl phenyl moiety of the inhibitor inserts into the site that Phe594 would presumably occupy in the active state. Significantly, the structure of the c-Abl STI-571 complex indicates that the STI-571 inhibitor promotes the inactive conformation of the DFG motif of c-Abl by a similar mechanism. In the active conformation of protein kinases, the activation segment adjacent to the DFG motif forms a ␤ sheet interaction with the ␤6 strand. However, in the inactive ⌬B-RAF and c-Abl structures, similar to the unphosphorylated insulin receptor kinase (Hubbard et al., 1994), the flipped conformation of the DFG motif orients this region of the activation segment toward the P loop of the N lobe. Consequently, in ⌬B-RAF, residues Gly595–Val599 of the activation segment form an array of hydrophobic interactions with the P loop residues Gly463–Val470 (Figure 6A). These interactions, which comprise the side chains of Leu596 and Val599 of the activation segment contacting Gly465, Phe467, and Val470 of the P loop, function to establish an inactive conformer of the ⌬B-RAF kinase domain, linking the two regions of the kinase in which the cancer-associated mutations occur. Apart from the inactive conformation


Figure 4. Structure Domain

of

WT

⌬B-RAF

Kinase

(A) Ribbons diagram of WT⌬B-RAF kinase domain in complex with BAY43-9006. The positions of Asp593 and Phe594 of the DFG motif, Asn580 of the catalytic loop, and Glu585 are shown, DFG and APE motif in yellow, rest of activation segment and the N region are in red. N lobe is in magenta, C lobe in marine, and P loop in orange. Residues 600–611 of the activation loop are disordered (dashed lines). (B) Schematic of B-RAF primary structure, showing functional domains and position of 32 observed cancer-associated mutants of B-RAF. The amino acid substitutions are color coded according to their activity class. Figures were made with PyMOL (http://www. pymol.org).

of the DFG motif/activation segment, which is not aligned for ATP and peptide substrate recognition and is partially disordered, all other key residues of ⌬B-RAF required for the phosphotransfer reaction are aligned as in active kinases. Specifically, the P loop, which coordinates the phosphate and adenine moieties of ATP, the catalytic Lys482 and Glu500 of the N lobe and the C lobe catalytic loop (Asn580, Arg574, and Asp575) are all correctly aligned. Therefore only a change in position of the DFG motif/activation segment is required for conversion to the active state. Structural Basis for Activating Oncogenic Mutations We have classified cancer-associated mutants of B-RAF into high, intermediate, or impaired activity groups according to their ability to stimulate in vitro B-RAF kinase activity (Figures 1, 2, and 4B, Supplemental Table S1 on Cell website). Given the phenotypic differences between the various oncogenic mutants, it is striking that nearly all mutants (⬎80%) are located within either the P loop or the activation segment within or adjacent to the DFG motif (Figure 4B). An apparent paradox of some mutants is that depending on the amino acid substitution, the mutant protein has either an elevated or reduced kinase activity (Figure 4B). Significantly, the high and intermedi-

ate activating mutations comprise substitutions of amino acids that are compatible with catalysis. Such mutations either occur at nonconserved positions or replace conserved residues with amino acids observed in other protein kinases (Supplemental Table S1 online). In contrast, the impaired activity mutants correspond to amino acid substitutions of highly conserved/invariant residues that are required for an optimal protein kinase catalytic reaction. As discussed below, a unifying feature of oncogenic mutations located in the P loop and DFG motif is the prediction that the inactive conformation of the DFG motif/activation segment will be destabilized because the hydrophobic cluster linking the DFG motif to the P loop is disrupted (Figure 6A). To explain the molecular mechanism of B-RAF oncogenic mutations, we discuss the three classes of mutants in turn, and the structural consequences of all mutations are described fully in Supplemental Table S3 online. Here we illustrate the principles underlying oncogenic B-RAF activation by reference to specific examples. Beginning with the high activity mutants, we note that all the Val599 mutants belong to this group, as does the neighboring mutant K600E and the P loop mutant G468A (Figure 1A). Activating mutants of B-RAF presumably mimic the conformational changes promoted by activation segment phosphorylation, and importantly the V599E B-RAF mutation appears to obviate the requirement

Cell 862

Figure 5. Interactions between BAY43-9006 and ⌬B-RAF and Comparison with c-Abl Inhibitor Complexes (A) Stereoview of BAY43-9006 bound to the interdomain cleft of ⌬B-RAF. The 2Fo-Fc electron density omit map was calculated from phases derived from a simulated annealing refinement of the structure omitting the inhibitor. The map is contoured at 1␴. For clarity, the activation segment is not shown. (B) c-Abl-STI-571 and (C) c-Abl-PD173855. The color scheme for ⌬B-RAF is the same as Figure 4, c-Abl is cyan. The side chain of the DFG motif Phe residue is shown for both kinases. In the STI-571 complex, the DFG motif overlaps that of ⌬B-RAF.

for activation segment phosphorylation. The structure of the wild-type ⌬B-RAF kinase domain indicates that the aliphatic side chain of Val599 interacts with the phenyl ring of Phe467 in the P loop (Figure 6A). Replacing the medium sized hydrophobic Val side chain with a larger and charged residue as found in human cancer (Glu, Asp, Lys, or Arg) would be expected to destabilize the interactions that maintain the DFG motif in an inactive conformation, so flipping the activation segment into the active position. Consistent with this notion is the discovery that mutation of Phe467 (F467CB-RAF) also occurs in cancer. In the V599E⌬B-RAF structure, the DFG motif adopts the same inactive conformation seen in

the wild-type protein, although the side chain of Glu599 is disordered and does not contact the Phe467 side chain. The presence of BAY43-9006 presumably restrains the DFG motif in the inactive conformation in V599E ⌬B-RAF. The diverse amino acid substitutions of Val599 capable of activating B-RAF suggest that the mutations all function by disrupting the inactive conformation. However, it is possible that acidic amino acids at position 599 could also interact with Lys506 of the ␣C helix to provide additional contacts that promote the active state (Figure 6B). In the active state of the AGC kinase PKB/Akt, Glu299 (equivalent to 599 of B-RAF) contacts


Figure 6. Position of Oncogenic Mutants of B-RAF and Mechanism of Activation by Thr598 Phosphorylation (A) The molecular surface generated by the P loop (residues 463–468) and the DFG motif/ activation segment (residues 593–599) are colored green and yellow, respectively, demonstrating the hydrophobic cluster created by the association of the P loop with the DFG motif, a conformation stabilized by the hydrogen bond between the amide side chain of Asn580 and the main chain amide of Phe594. (B) A ribbon representation of the structure of the activation segment of PKB is superimposed onto ⌬B-RAF, together with the side chains of Asp293 (Asp593), Lys298 (Thr598), Glu299 (Val599), Ile301 (Ser601), phosphoThr309, and the catalytic Asp275 (Asp575) and Arg274 (B-RAF residues in parenthesis). In B-RAF, Glu599 and Glu600 (equivalent to Glu299 and Glu300 of PKB) could contact Lys569 of the ␣E helix. Any substitution at Val599 will be tolerated. The position of Glu600 is modeled from Gly300 of PKB. Structure of the activation segment is taken from PKB (Yang et al., 2002). Phosphorylation of Thr598 of B-RAF would position a phosphate group in the equivalent position as the phosphate of Thr309 of PKB, allowing interaction with Arg575 of the B-RAF catalytic loop RD motif. The equivalent of B-RAF Val599 is a Glu in PKB, suggesting that in the V599EB-RAF mutant, Glu599 will be solvent exposed and likely to interact with Lys506 of the ␣C helix.

Arg202 (equivalent to Lys506 of B-RAF) (Yang et al., 2002). In support of this notion, Glu and Asp substitutions at position 599 of B-RAF are 2- to 4-fold more active than the Arg and Lys substitutions (Figure 1A). Although all four substitutions at position 599 possess similar activity in vitro and in vivo, the Glu substitution accounts for over 95% of the mutations at this site. It is likely that a Glu mutation occurs at high frequency because it only requires a single base substitution, rather than the two necessary for conversion to the other amino acids. Intriguingly, a Glu for Lys substitution at the neighboring position 600 is also a high activity mutant (Figure 1A). In the wild-type ⌬B-RAF kinase domain structure, Lys600 is disordered, suggesting that it does not participate in stabilization of the inactive conformation. However, in the active conformation of ⌬B-RAF, a Glu at residue 600 may be positioned to interact with Lys506 and/or Lys569 of the ␣C and ␣E helices, respectively (Figure 6B). The only high activity mutant lying outside of the P loop and DFG motif is E585KB-RAF. Glu585 lies on the opposite surface of the kinase domain from the DFG motif (Figure 4A), and it is possible that its mutation disrupts potential interdomain interactions within full-

length B-RAF that relieves N-terminal domain autoinhibition. Intermediate activating mutants are most likely explained by their ability to disrupt the inactive conformation of the DFG motif; however, the overall kinase activity is reduced relative to the strongly activating mutants due to suboptimal B-RAF catalytic efficiency. This is best illustrated by comparing the properties of the strongly activating G468AB-RAF mutant with the intermediate activity mutant G468EB-RAF, different substitutions of the same amino acid which differ 200-fold in their activities (Figures 1A and 1B, Supplemental Table S1 online). Gly468 of the P loop is wedged against Leu596 of the activation segment and substitutions of Gly468 would therefore perturb P loop-DFG contacts, destabilizing the inactivated state (Figure 6A). The difference in activity between G468AB-RAF and G468EB-RAF may be explained by the observation that in all known protein kinases, only Gly, Ala, or Ser occur at this position (the third Gly residue of the P loop) (Manning et al., 2002) and that an Ala for Gly substitution at the equivalent position in PKA has little influence on PKA activity (Grant et al., 1998). We propose that both the Ala and Glu substitutions of Gly468 of B-RAF share the property of destabilizing the

Cell 864

inactive conformation, but whereas the Ala is permissive for high activity, the bulky Glu interferes with P loop/ ATP interactions reducing its potential to elevate kinase activity. In another example, while replacing Gly463 (the 1st Gly of the P loop) with Val would destabilize the inactive conformation of the DFG motif, it may also interfere with the optimal interactions between the P loop and phosphate groups of ATP. A survey of all human kinase sequences indicates that Gly is observed in 94% of kinases, and only Ala and Ser are also tolerated at this position (Manning et al., 2002). Asn580 is a highly conserved metal coordinating residue of the catalytic loop of the kinase domain. Surprisingly the N580SB-RAF mutant is 6-fold more active than WTB-RAF. Although Asn580 is located outside of the P loop and DFG/activation segment, its replacement by Ser is likely to disrupt the inactive conformation of the DFG motif. In the inactive ⌬B-RAF structure, the amide side chain of Asn580 accepts a hydrogen bond from the main chain amide group of Phe594 of the DFG motif (Figure 6A), indicating that the Asn580 and Phe594 interactions contribute to the inactive conformation and explaining how mutation of either of these residues disrupts the inactive DFG conformation. The properties of other intermediate mutants are listed in Supplemental Table S3 on the Cell website. The third class of B-RAF oncogenic mutants is the most surprising because these comprise nonconservative amino acid substitutions of catalytic site residues that are required for optimal kinase activity. Although these mutations result in the expected reduction of in vitro B-RAF kinase activity, the activity of ERK is stimulated in vivo. We propose that the oncogenic potential of these mutants results from a disruption of the inactive DFG conformation. Experiments described above suggest that this conformational change is transmitted to elevate ERK activity via a cooperative activation of wildtype C-RAF (and possibly B-RAF; M.J.G. and R.M., unpublished data) although the mechanism of cooperativity is unclear (discussed below). The second Gly of the P loop (Gly465 in B-RAF) is invariant in all kinases (Manning et al., 2002), suggesting that any substitution of this residue cannot be readily tolerated. Mutations to the corresponding residue in PKA are detrimental to kinase activity (Grant et al., 1998). However, by analogy to mutants of the 1st and 3rd Gly residues of the P loop that generate activated B-RAF, we predict that these mutations destabilize the inactive conformation of the DFG motif (Figure 6A). Similarly, within the DFG motif itself, the introduction of a bulky charged residue at Gly595 in the G595RB-RAF mutant is not compatible with the DFG motif in the inactive conformation. Our analysis provides a satisfactory explanation for the prevalence of B-RAF mutations within the P loop and DFG motif/activation segment. Such mutations disrupt the inactive conformation of the DFG motif, converting the activation segment into the active state, thereby mimicking its phosphorylation. However, one cancer-associated mutant D593VB-RAF behaves differently from the three classes of mutants we have characterized. Significantly, this mutant is almost completely inactive in vitro and is unable to activate either ERK or C-RAF in vivo. Consistent with these findings, it is not

obvious from the ⌬B-RAF structure how replacing Asp593 would promote the active conformation of the kinase. Implications for Regulation of B-RAF by Activation Segment and N Region Phosphorylation Thr598 is the major activation segment phosphorylation site whereas Ser601 is a relatively minor one (Zhang and Guan, 2000). Substitution of Thr598 with alanine blocks B-RAF activity (Zhang and Guan, 2000), whereas substitution with isoleucine as occurs in cancer strongly activates B-RAF (Figure 1B). Presumably, the bulky side chain of isoleucine disrupts the inactive conformation of the activation segment, whereas the small Ala side chain does not (Figure 6A). In a similar fashion, we propose that phosphorylation of Thr598 would also destabilize the inactive conformation. In addition, by comparing B-RAF to other protein kinases that are regulated by activation segment phosphorylation, we suggest that pThr598 provides additional contacts to the conserved Arg574 residue adjacent to the catalytic Asp575 of the catalytic loop, stabilizing the active form (Johnson et al., 1996) (Figure 6B). The structure of ⌬B-RAF also provides insight into the role of the regulatory N region. B-RAF is constitutively phosphorylated on the N region Ser445 within the motif Ser445-Ser-Asp-Asp448, but the equivalent motif in C-RAF, Ser338-Ser-Tyr-Tyr341, is subject to regulatory phosphorylation on both Ser338 and Tyr341 (Mason et al., 1999). In C-RAF, replacing the Tyr residues with Asp mimics the effects of Tyr341 phosphorylation and substituting the Asp residues in B-RAF suppresses basal and RASstimulated kinase activity (Fabian et al., 1993; Mason et al., 1999). In our ⌬B-RAF structure, Asp447 contacts Arg505 of the ␣C helix, suggesting that this interaction stabilizes the active conformation, providing a molecular explanation for why this Asp is important for the basal and RAS-stimulated kinase activity of B-RAF. It is less clear what role Ser445 plays in activating B-RAF. However, in our V599E⌬B-RAF protein, Ser445 is unphosphorylated and yet the kinase domain is highly active and (excluding the activation segment) resembles an activated kinase. These data suggest that phosphorylation of Ser445 (and Ser338 in C-RAF) may relieve constraints imposed by the N terminus. Concluding Remarks The discovery that oncogenic activating mutations of B-RAF occur in two thirds of malignant melanomas and at high frequency in other human cancers confirms the significance of the RAS-RAF-MEK-ERK pathway in promoting tumorigenesis. A significantly high frequency of mutants cluster to the conserved P loop and DFG motif of the kinase, and we show that a unifying feature of these mutants is to destabilize the inactive ⌬B-RAF structure, thereby promoting an active conformation. Mutations of Val599 occur in 92% of oncogenic forms of B-RAF, and V599EB-RAF represents 90% of oncogenic B-RAF mutants (Supplemental Table S1 online). Interestingly, oncogenic activating mutants of c-Kit and Met (a Val for Asp substitution; Glover et al., 1995; Kitayama et al., 1995) correspond to Val599 of B-RAF, raising the possibility that a related structural mechanism may underlie oncogenic activation of these tyrosine kinases.


Figure 7. Activating and Impaired Activity B-RAF Mutants Stimulate ERK Signaling through Distinct Mechanisms (A) The activated mutants can directly phosphorylate and activate MEK and although they also activate C-RAF, their ability to signal to ERK does not require this pathway. (B) The impaired activity B-RAF mutants cannot stimulate efficient activation of MEK, but can stimulate C-RAF activity, which then activates MEK. Their ability to signal to ERK is therefore dependent on C-RAF protein.

We propose that other activating mutants also destabilize the inactive conformation of the DFG motif, although it is noteworthy that these mutants constitute on average 0.3% of B-RAF mutations (Supplemental Table S1 online). Oncogenic mutants of B-RAF also have analogies with STI-571-resistant mutants of BCR-ABL (Azam et al., 2003). Two such mutations, situated within the activation segment (equivalent to Leu596 and Lys600 of B-RAF), are likely to destabilize the inactive conformation of the DFG/activation segment of the kinase, disrupting the binding site for the drug. Our findings that BAY43-9006 interacts with an inactive conformation of B-RAF would suggest that activating oncogenic mutants of B-RAF would be less sensitive to the inhibitor than WTB-RAF, with consequent implications for potential BAY43-9006-induced resistant mutations of the kinase. Consistent with this prediction, kinetic studies from BAYER pharmaceuticals have recently shown that V599EB-RAF is 2-fold less sensitive to BAY43-9006 than WTB-RAF (IC50s of 38 and 22 nM, respectively; Brian Schwartz, personal communication). We have also characterized a class of oncogenic mutants that reduce kinase activity relative to WTB-RAF (Figure 2). These mutants have weak oncogenic potential as judged by their inability to transform NIH3T3 cells (Supplemental Table S1 online), which is correlated to the relatively low level of ERK activation. We present data that impaired activity mutants of B-RAF trigger the activity of wild-type C-RAF and B-RAF, and our siRNA study demonstrates that the ability of these B-RAF mutants to stimulate ERK is dependent on WTC-RAF. Crucially, these mutations are predicted to induce an active kinase conformation. It is this property of the impaired activity mutants that may explain their ability to activate ERK via C-RAF even though their kinase activity toward MEK is severely impaired. However, D593VB-RAF and the kinase dead mutant K482MB-RAF, which we predict would not promote the active state, do not induce either ERK or C-RAF activity. We propose that whereas the high activity B-RAF mutants primarily stimulate ERK via direct MEK phosphorylation, the impaired activity mutants, which are defective in direct MEK phosphorylation, induce ERK by triggering C-RAF activity (Figure 7).

There are various possibilities for how the impaired activity mutants of B-RAF generate elevated wild-type C-RAF and B-RAF activity. First, the presence of the active conformation of B-RAF sequesters negative inhibitors of RAF kinases (for example, protein phosphatases or the RAF inhibitory protein RKIP; Kolch, 2000). Alternatively, the active conformation may recruit positive effectors of the pathway that act on wild-type RAF kinases. Finally, the active conformation of B-RAF may promote the active conformation of C-RAF via an allosteric mechanism or directly phosphorylate the activation segment of wild-type RAF kinases. The possibility that an autocrine feedback loop activates C-RAF is not consistent with our findings that the MEK inhibitor UO126 has no effect on the ability of B-RAF mutants to stimulate C-RAF, whose activity is also independent of RAS (M.J.G. and R.M., unpublished data). It is unlikely that the lack of activity of these mutants is due to their altered binding to Hsp90 and 14-3-3 because a recent study has shown that D593VB-RAF and G595RB-RAF bind these proteins with similar affinity as WTB-RAF (Ikenoue et al., 2003). An interesting possibility is that communication between impaired activity mutants of B-RAF and endogenous wild-type RAF kinases could be mediated via scaffolding molecules such as KSR1, and it has been demonstrated previously that C-RAF forms oligomers with itself (Luo et al., 1996) and with B-RAF (Weber et al., 2001), a finding confirmed in our studies. Furthermore, induced dimerization of C-RAF stimulates ERK activity (Farrar et al., 1996; Luo et al., 1996). The finding that the active conformation of B-RAF is sufficient to stimulate endogenous wild-type RAF and ERK represents a novel paradigm for the action of an oncogene. Experimental Procedures Mammalian Cell Expression Details of the expression constructs for G12VRAS, C-RAF, and B-RAF, COS cell transfection conditions, NIH3T3 cell transformation assays, Western blotting, protein expression measurements, and the RAFcoupled kinase assay have all been described (Marais et al., 1997, 1998; Mason et al., 1999). Details of the MEK phosphorylation assay are provided in the Supplemental Data online. B-RAF mutants were generated using PCR mutagenesis and vector sequence verified

Cell 866

using automated dideoxy sequencing. The human cancer cell lines WM266-4 and H1666 were propagated in DMEM or RPMI1040 (Gibco-BRL Life Technologies), respectively, supplemented with 10% FCS. To analyze complex formation between B-RAF and endogenous C-RAF, C-RAF was immunoprecitated using 5 ␮g mouse monoclonal anti-C-RAF antibody (Transductions labs) coupled to Protein G sepharose (Amersham). Samples were tumbled for 2 hr at 4⬚C, washed 3⫻ with 400 ␮l NP40, and resolved on a 7% SDS gel for Western blotting. Ten percent of each extract was run as a control for protein expression. SiRNA Transfections 2.5 ⫻ 105 cells were treated with 200–400 nM siRNA (Qiagen) and 10 ␮l Oligofectamine (Invitrogen) for 6 hr according to the manufacturers protocol. Transfected cells were supplemented with 1 ml media containing 1% (COS cells) or 10% FCS (WM266-4 and H1666) and protein extracts prepared after 48 hr. COS cells were transfected with siRNA 24 hr after transfection with the protein expression constructs. The following sequences were used: B-RAF: AAG UGG CAU GGU GAU GUG GCA; C-RAF: AAU AGU UCA GCA GUU UGG CUA; Scrambled: AAA CCG UCG AUU UCA CCC GGG. Xenopus Embryo Manipulation and Microinjection Xenopus embryos were manipulated for foreign protein expression as previously described (Jones et al., 1996) and fixed at stage 8 (Nieuwkoop and Faber, 1967). Whole-mount antibody staining was performed essentially as previously described (Dent et al., 1989). Protein Expression and Purification The kinase domain of human B-RAF (residues 432–725 ⌬B-RAF) and human p50Cdc37 were cloned into the dual expression vector pFastBac Dual (Life Technologies, Inc). The encoded ⌬B-RAF contained an N-terminal purification tag (MDRGSH6GS). Briefly, WT ⌬B-RAF was purified using TALON metal affinity, cation exchange, and size exclusion chromatography. The protein was concentrated to 2 mg/ml and the inhibitor BAY43-9006 was added to a 4-fold molar excess for crystallization. V599E⌬B-RAF was expressed and purified as for WT⌬B-RAF with minor modifications (see Supplemental Data for full details). Crystallization Crystals of the protein:inhibitor complex were grown by the microbatch method. Crystals of the B-RAF:BAY43-9006 complex grew in 100 mM Tris-HCl (pH 8.5) and 8% (w/v) PEG 8 000 at 20⬚C. Crystals of the V599EB-RAF:BAY43-9006 complex grew in 50 mM KH2PO4 and 20% (w/v) PEG 8 000 at 20⬚C. Data Collection and Structure Determination Data were collected using the micro-focus beam line (ID13) ESRF, Grenoble from a single WTB-RAF crystal to the diffraction limit of 2.9 A˚. The slightly smaller V599EB-RAF crystals diffracted to 3.5 A˚ on ID14-EH4 from a single crystal of dimensions 15 ␮m by 15 ␮m by 0.2 mm in a single sweep. Crystallographic data were processed and scaled using MOSFLM and SCALA (CCP4, 1991). The WTB-RAF structure was solved by molecular replacement using BEAST (CCP4, 1991) with p38 MAP kinase, LCK, and Ephb2 (PDB ID codes: 1WFC, 3LCK, 1JPA, respectively) as a multiple search model. Refinement was assisted by the use of 2-fold NCS restraints using CNS (Brunger et al., 1998) and manually rebuilt using O (Jones et al., 1991) (Supplemental Table S1 online). The cell dimensions of V599EB-RAF differ significantly from the wild-type crystal, and therefore its structure was solved by molecular replacement using the refined wild-type B-RAF model (Supplemental Table S1). Synthesis of BAY43-9006 The synthesis is a modification of published routes (Bankston et al., 2002; Reidl et al., 2000) (Supplemental Data on Cell website).

by Cancer Research-UK, Institute of Cancer Research, and Wellcome Trust. M.J.G. was funded by an Astra-Zeneca studentship. Received: August 21, 2003 Revised: January 14, 2004 Accepted: February 2, 2004 Published: March 18, 2004 References Avruch, J., Khokhlatchev, A., Kyriakis, J.M., Luo, Z., Tzivion, G., Vavvas, D., and Zhang, X.F. (2001). Ras activation of the Raf kinase: tyrosine kinase recruitment of the MAP kinase cascade. Recent Prog. Horm. Res. 56, 127–155. Azam, M., Latek, R.R., and Daley, G.Q. (2003). Mechanisms of autoinhibition and STI-571/imatinib resistance revealed by mutagenesis of BCR-ABL. Cell 112, 831–843. Bankston, D., Dumas, J., Natero, R., Reidl, B., Monahan, M.K., and Sibley, R. (2002). A scaleable Synthesis of BAY 43–9006: A potent raf kinase inhibitor for the treatment of cancer. Organic Process Res. & Dev. 6, 777–781. Blagosklonny, M.V. (2002). Hsp-90-associated oncoproteins: multiple targets of geldanamycin and its analogs. Leukemia 16, 455–462. Brose, M.S., Volpe, P., Feldman, M., Kumar, M., Rishi, I., Gerrero, R., Einhorn, E., Herlyn, M., Minna, J., Nicholson, A., et al. (2002). BRAF and RAS mutations in human lung cancer and melanoma. Cancer Res. 62, 6997–7000. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: A new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921. CCP4 (Collaborative Computational Project Number 4) (1991). The CCP4 Suite: programs for protein crystallography. Acta Cryst D 50, 760–763. Chong, H., Lee, J., and Guan, K.L. (2001). Positive and negative regulation of Raf kinase activity and function by phosphorylation. EMBO J. 20, 3716–3727. Cohen, Y., Goldenberg-Cohen, N., Parrella, P., Chowers, I., Merbs, S.L., Pe’er, J., and Sidransky, D. (2003). Lack of BRAF mutation in primary uveal melanoma. Invest. Ophthalmol. Vis. Sci. 44, 2876– 2878. Davies, H., Bignell, G.R., Cox, C., Stephens, P., Edkins, S., Clegg, S., Teague, J., Woffendin, H., Garnett, M.J., Bottomley, W., et al. (2002). Mutations of the BRAF gene in human cancer. Nature 417, 949–954. Dent, J.A., Polson, A.G., and Klymkowsky, M.W. (1989). A wholemount immunocytochemical analysis of the expression of the intermediate filament protein vimentin in Xenopus. Development 105, 61–74. Fabian, J.R., Daar, I.O., and Morrison, D.K. (1993). Critical tyrosine residues regulate the enzymatic and biological activity of Raf-1 kinase. Mol. Cell. Biol. 13, 7170–7179. Farrar, M.A., Alberol-Ila, and Perlmutter, R.M. (1996). Activation of the Raf-1 kinase cascade by coumermycin-induced dimerization. Nature. 383, 178–181. Glover, H.R., Baker, D.A., Celetti, A., and Dibb, N.J. (1995). Election of activating mutations of c-fms in FDC-P1 cells. Oncogene 11, 1347–1356. Grant, B.D., Hemmer, W., Tsigelny, I., Adams, J.A., and Taylor, S.S. (1998). Kinetic analyses of mutations in the glycine-rich loop of cAMP-dependent protein kinase. Biochemistry 37, 7708–7715.

Acknowledgments

Hubbard, S.R., Wei, L., Ellis, L., and Hendrickson, W.A. (1994). Crystal structure of the tyrosine kinase domain of the human insulin receptor. Nature 372, 746–754.

We are grateful to Harry King for DNA sequencing, staff at ESRF for help with data collection, Vivienne Thompson for tissue culture work, and Laurence Pearl for discussions. The work has been funded

Ikenoue, T., Hikiba, Y., Kanai, F., Tanaka, Y., Imamura, J., Imamura, T., Ohta, M., Ijichi, H., Tateishi, K., Kawakami, T., et al. (2003). Functional analysis of mutations within the kinase activation segment of B-Raf in human colorectal tumors. Cancer Res. 63, 8132–8137.


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

logical activity of v-raf, a unique oncogene transduced by a retrovirus. Proc. Natl. Acad. Sci. USA 80, 4218–4222.

Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110–119.

Reidl, B., Dumas, J., Khire, U., Lowinger, T.B., Scott, W.J., Smith, R.A., Wood, J.E., Monahan, M.K., Natero, R., Renick, J., and Sibley, R. (2000). Preparation of w-carboxy aryl substituted diphenyl ureas as p38 kinase inhibitors. In PCT Int. Appl. WO 200041698.

Jones, C.M., Dale, L., Hogan, B.L., Wright, C.V., and Smith, J.C. (1996). Bone morphogenetic protein-4 (BMP-4) acts during gastrula stages to cause ventralization of Xenopus embryos. Development 122, 1545–1554.

Robinson, M.J., and Cobb, M.H. (1997). Mitogen-activated protein kinase pathways. Curr. Opin. Cell Biol. 9, 180–186.

Kitayama, H., Kanakura, Y., Furitsu, T., Tsujimura, T., Oritani, K., Ikeda, H., Sugahara, H., Mitsui, H., Kanayama, Y., Kitamura, Y., et al. (1995). Constitutively activating mutations of c-kit receptor tyrosine kinase confer factor-independent growth and tumorigenicity of factor-dependent hematopoietic cell lines. Blood 85, 790–798. Kolch, W. (2000). Meaningful relationships: the regulation of the Ras/ Raf/MEK/ERK pathway by protein interactions. Biochem. J. 351, 289–305. Lowinger, T.B., Riedl, B., Dumas, J., and Smith, R.A. (2002). Design and discovery of small molecules targeting raf-1 kinase. Curr. Pharm. Des. 8, 2269–2278. Luo, Z., Tzivion, G., Belshaw, P.J., Vavvas, D., Marshall, M., and Avruch, J. (1996). Oligomerization activates c-Raf-1 through a Rasdependent mechanism. Nature 383, 181–185.

Schindler, T., Bornmann, W., Pellicena, P., Miller, W.T., Clarkson, B., and Kuriyan, J. (2000). Structural mechanism for STI-571 inhibition of abelson tyrosine kinase. Science 289, 1938–1942. Singer, G., Oldt, R., 3rd, Cohen, Y., Wang, B.G., Sidransky, D., Kurman, R.J., and Shih, I. (2003). Mutations in BRAF and KRAS characterize the development of low-grade ovarian serous carcinoma. J. Natl. Cancer Inst. 95, 484–486. Stancato, L.F., Chow, Y.H., Hutchison, K.A., Perdew, G.H., Jove, R., and Pratt, W.B. (1993). Raf exists in a native heterocomplex with hsp90 and p50 that can be reconstituted in a cell-free system. J. Biol. Chem. 268, 21711–21716. Weber, C.K., Slupsky, J.R., Kalmes, H.A., and Rapp, U.R. (2001). Active Ras induces heterodimerization of cRaf and BRaf. Cancer Res. 61, 3595–3598.

Lyons, J.F., Wilhelm, S., Hibner, B., and Bollag, G. (2001). Discovery of a novel Raf kinase inhibitor. Endocr. Relat. Cancer 8, 219–225.

Yang, J., Cron, P., Good, V.M., Thompson, V., Hemmings, B.A., and Barford, D. (2002). Crystal structure of an activated Akt/protein kinase B ternary complex with GSK3-peptide and AMP-PNP. Nat. Struct. Biol. 9, 940–944.

MacNicol, M.C., Muslin, A.J., and MacNicol, A.M. (2000). Disruption of the 14–3-3 binding site within the B-Raf kinase domain uncouples catalytic activity from PC12 cell differentiation. J. Biol. Chem. 275, 3803–3809.

Yuen, S.T., Davies, H., Chan, T.L., Ho, J.W., Bignell, G.R., Cox, C., Stephens, P., Edkins, S., Tsui, W.W., Chan, A.S., et al. (2002). Similarity of the phenotypic patterns associated with BRAF and KRAS mutations in colorectal neoplasia. Cancer Res. 62, 6451–6455.

Malumbres, M., and Barbacid, M. (2003). RAS oncogenes: the first 30 years. Nat. Rev. Cancer 3, 459–465.

Zhang, B.H., and Guan, K.L. (2000). Activation of B-Raf kinase requires phosphorylation of the conserved residues Thr598 and Ser601. EMBO J. 19, 5429–5439.

Manning, G., Whyte, D.B., Martinez, R., Hunter, T., and Sudarsanam, S. (2002). The protein kinase complement of the human genome. Science 298, 1912–1934. Marais, R., Light, Y., Paterson, H.F., Mason, C.S., and Marshall, C.J. (1997). Differential regulation of Raf-1, A-Raf and B-Raf by oncogenic Ras and tyrosine kinases. J. Biol. Chem. 272, 4378–4383. Marais, R., Light, Y., Mason, C., Paterson, H., Olson, M., and Marshall, C.J. (1998). Requirement of Ras-GTP-Raf complexes for activation of Raf-1 by protein kinase C. Science 280, 109–112. Mason, C.S., Springer, C., Cooper, R.G., Superti-Furga, G., Marshall, C.J., and Marais, R. (1999). Serine and tyrosine phosphorylations cooperate in Raf-1, but not B-Raf activation. EMBO J. 18, 2137– 2148. Michaud, N.R., Fabian, J.R., Mathes, K.D., and Morrison, D.K. (1995). 14–3-3 is not essential for Raf-1 function: identification of Raf-1 proteins that are biologically activated in a 14–3-3- and Ras-independent manner. Mol. Cell. Biol. 15, 3390–3397. Morrison, D.K. (2001). KSR: a MAPK scaffold of the Ras pathway? J. Cell Sci. 114, 1609–1612. Morrison, D.K., and Cutler, R.E.J. (1997). The complexity of Raf-1 regulation. Curr. Opin. Cell Biol. 9, 174–179. Nagar, B., Bornmann, W.G., Pellicena, P., Schindler, T., Veach, D.R., Miller, W.T., Clarkson, B., and Kuriyan, J. (2002). Crystal structures of the kinase domain of c-Abl in complex with the small molecule inhibitors PD173955 and imatinib (STI-571). Cancer Res. 62, 4236– 4243. Nieuwkoop, P., and Faber, J. (1967). Normal Table of Xenopus laevis (Amsterdam: Daudin). Pargellis, C., Tong, L., Churchill, L., Cirillo, P.F., Gilmore, T., Graham, A.G., Grob, P.M., Hickey, E.R., Moss, N., Pav, S., and Regan, J. (2002). Inhibition of p38 MAP kinase by utilizing a novel allosteric binding site. Nat. Struct. Biol. 9, 268–272. Rajagopalan, H., Bardelli, A., Lengauer, C., Kinzler, K.W., Vogelstein, B., and Velculescu, V.E. (2002). Tumorigenesis: RAF/RAS oncogenes and mismatch-repair status. Nature 418, 934. Rapp, U.R., Goldsborough, M.D., Mark, G.E., Bonner, T.I., Groffen, J., Reynolds, F.H., and Stephenson, J.R. (1983). Structure and bio-

Accession Numbers Data have been deposited in the Protein Data Bank with the following ID codes: wild-type B-RAF, 1uwh (coordinate), rluwhsf (structure factors); V599EB-RAF, 1uwj (coordinates), rluwjsf (structure factors).