Phosphorylation of Raf-1 by p21-activated Kinase 1 and Src Regulates Raf-1 Autoinhibition*

Exposure of cells to mitogens or growth factors stimulates Raf-1 activity through a complex mechanism that involves binding to active Ras, phosphorylation on multiple residues, and protein-protein interactions. Recently it was shown that the amino terminus of Raf-1 contains an autoregulatory domain that can inhibit its activity in Xenopus oocytes. In the present work we show that expression of the Raf-1 autoinhibitory domain blocks extracellular signal-regulated kinase 2 activation by the Raf-1 catalytic domain in mammalian cells. We also show that phosphorylation of Raf-1 on serine 338 by PAK1 and tyrosines 340 and 341 by Src relieves autoinhibition and that this occurs through a specific decrease in the binding of the Raf-1 regulatory domain to its catalytic domain. In addition, we demonstrate that phosphorylation of threonine 491 and serine 494, two phosphorylation sites in the catalytic domain that are required for Raf-1 activation, is unlikely to regulate autoinhibition. These results demonstrate that the autoinhibitory domain of Raf-1 is functional in mammalian cells and that its interaction with the Raf-1 catalytic domain is regulated by phosphorylation of serine 338 and tyrosines 340 and 341.

Activation of the extracellular signal-regulated kinases (ERK) 1 1 and 2 is required for cellular proliferation in many cell types and is also a requisite event in neoplastic transformation (1)(2)(3)(4). The activities of ERKs 1 and 2 are regulated through the activation of a sequentially acting protein kinase cascade, known generically as a mitogen-activated protein kinase module. At the top of this cascade is the serine/threonine protein kinase Raf-1. Once activated, Raf-1 phosphorylates and activates the mitogen and extracellular signal-regulated kinase kinases 1 and 2 (MEKs 1 and 2), which subsequently phosphorylate and activate ERKs 1 and 2. ERKs then phosphorylate a large number of nuclear and cytoplasmic substrates that ultimately regulate the ability of a cell to grow and divide (2).
The Raf-1 activation mechanism is a complex process that involves binding to the small GTP binding protein Ras, phosphorylation on multiple residues, altered protein-protein inter-actions, and perhaps direct interaction with lipids (5)(6)(7). Binding to active Ras relocalizes Raf-1 from the cytosol to the plasma membrane. This interaction is initially mediated by a conserved domain in the Raf-1 amino terminus known as the Ras binding domain (RBD). Binding of Ras to the RBD then promotes contact with an adjacent domain known as the cysteine-rich domain (CRD), and binding of Ras to both of these domains is required for full activation of Raf-1 (5). Interaction between Ras and Raf-1 also stimulates the release of 14 -3-3 from its amino-terminal binding site, which is centered on serine 259, and the subsequent dephosphorylation of this site (8 -10). This is thought to contribute to the separation of the catalytic domain from the amino-terminal regulatory domain.
Activation of Raf-1 by Ras is also accompanied by phosphorylation on multiple residues, including serines 338 and 339 and tyrosines 340 and 341. Phosphorylation of these sites is essential for Raf-1 activation by extracellular ligands such as epidermal growth factor, phorbol esters, and integrin binding (11)(12)(13). Kinases that may catalyze the phosphorylation of these sites in the cell are the p21-activated kinases (PAKs) 1-3, which phosphorylate serines 338 and 339 (13,14), and the Src family of tyrosine kinases, which phosphorylate tyrosines 340 and 341 (15,16). In addition, Raf-1 activation may require phosphorylation on two conserved sites within the activation loop of its kinase domain (threonine 491 and serine 494) (17). Kinases that phosphorylate these residues have not been identified. Despite the clear importance of phosphorylation of Raf-1 on each of these sites, it is not yet understood how these events contribute to Raf-1 activation.
Previously it was shown (18) that the amino terminus of Raf-1 contains an autoinhibitory domain that can block the function of the Raf-1 catalytic domain in Xenopus oocytes. This supports a model in which the catalytic activity of inactive Raf-1 is inhibited by interaction with the autoinhibitory domain. The mechanism whereby the catalytic domain is released from the autoinhibitory domain has not been determined. In this report we show that the autoinhibitory domain of Raf-1 can block the ability of a separately expressed Raf-1 catalytic domain to stimulate ERK2 activity in mammalian cells. We also show that this domain minimally consists of the first 147 amino acids of Raf-1 and encompasses portions of both the RBD and the CRD. We also demonstrate that phosphorylation of Raf-1 on serine 338 and tyrosines 340 and 341 relieves this autoinhibitory effect and that this occurs through a reduction in the affinity of the amino terminus for the phosphorylated catalytic domain. Furthermore, we demonstrate that putative phosphorylation sites within the kinase loop (threonine 491 and serine 494) are not likely to regulate autoinhibition. These data demonstrate that phosphorylation of Raf-1 on serines 338 and 339 and tyrosines 340 and 341 contributes to Raf-1 activation by blocking the ability of the autoinhibitory domain to regulate Raf-1 catalytic activity.
Transfection, Immunoprecipitation, and in Vitro Kinase Assays-HEK 293 cells were maintained in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum and 10 g/ml penicillin/streptomycin (all cell culture reagents from Invitrogen). Cells in 60-mm dishes were transfected by calcium phosphate coprecipitation (19). For all assays when shown, 2 g of HA-tagged ERK2 and 20 ng of Raf BXB plasmids were transfected. The amounts transfected for other constructs were as noted in the corresponding figure legends. Twenty hours after transfection, the culture medium was replaced with Dulbecco's modified Eagle's medium without serum, and the cells were incubated for a further 24 h. Cells were lysed in 0.5 ml of Triton lysis buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 80 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, 1 mM EDTA, 0.5% Triton X-100, 10 g/ml aprotinin, 10 g/ml pepstatin A, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride). Insoluble proteins were pelleted by centrifugation at 16,000 ϫ g for 10 min at 4°C, and the soluble supernatant was stored at Ϫ80°C. Immunoprecipitation-kinase assays measuring hemagglutinin epitope-tagged ERK2 activity were performed as previously described (19). Briefly, HA-ERK2 was immunoprecipitated from soluble cell lysates using 2 g of mouse anti-HA antibody (Santa Cruz Biotechnology) and 40 l of a 50% slurry of protein A-Sepharose (Amersham Biosciences). Immunoprecipitates were washed three times with 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, and once with 20 mM Tris-HCl, pH 8.0. The kinase activity of immunoprecipitated HA-ERK2 was assayed using 10 g of myelin basic protein (MBP) (Sigma) as a substrate. Incubations were for 30 min at 30°C in kinase buffer (20 mM Tris, pH 8.0, 10 mM MgCl 2 , 1 mM dithiothreitol, 100 M ATP, 2 Ci of [␥ 32 P]ATP (ICN Biomedicals)). Phosphorylated MBP was resolved by 15% SDS-PAGE. After drying, the gel was exposed to autoradiography film, and the phosphorylation of MBP was quantitated by scintillation counting of the excised MBP.
RafCAT Binding Assays-RafCAT protein (1 g) was phosphorylated or not by incubation in kinase buffer (20 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 100 M ATP) for 30 min at 30°C with or without recombinant GST⅐PAK1 (22) or recombinant Src (Upstate Biotechnology Inc.). It was previously determined that after 30 min the phosphorylation of RafCAT by either kinase had reached a maximum. After phosphorylation, onetenth of the phosphorylated RafCAT (100 ng) was mixed with 10 g of GST⅐Raf-1 1-330 bound to glutathione-agarose, and phosphate-buff-ered saline was added to a final volume of 300 l. Proteins were mixed for 1 h at 4°C. Glutathione-agarose-protein complexes were then pelleted by centrifugation, and the supernatant was saved to test for the presence of unbound RafCAT protein. The pellets were washed three times with 1 ml of ice-cold phosphate-buffered saline. Pelleted complexes were solubilized with an equal volume of 2ϫ Laemmli sample buffer. Bound and unbound proteins were loaded on 12% SDS-PAGE and transferred to nitrocellulose. Western blots were then performed for RafCAT (rabbit anti-Raf-1, C-12, Santa Cruz Biotechnology), phospho-serine 338 (rat anti-Raf-1 phospho-serine 338, Upstate Biotechnology Inc.), and phospho-tyrosine (mouse anti-phospho-tyrosine 4G10, Upstate Biotechnology Inc.). Western blots were quantitated by densitometry.

Characterization of the Raf-1 Autoinhibitory Domain in
Mammalian Cells-Cutler et al. (18) reported that expression of the Raf-1 amino terminus (amino acids 1-330) in Xenopus oocytes blocked both germinal vesicle breakdown and mitogenactivated protein kinase activation stimulated by coexpression of the Raf-1 catalytic domain. To test whether the Raf-1 amino terminus could function as an autoinhibitory domain in mammalian cells, we tested its ability to block ERK2 activation stimulated by expression of the Raf-1 catalytic domain. HEK 293 cells were cotransfected with hemagglutinin epitopetagged ERK2 (HA-ERK2) and a constitutively active Raf-1 catalytic domain construct (Raf BXB, amino acids 1-25 and 304 -648) (19), either with or without increasing amounts of the Raf-1 amino terminus (1-330). After starving the cells for 24 h, the cells were lysed, and the kinase activity of immunoprecipitated HA-ERK2 was assayed using MBP as a substrate (19). As shown in Fig. 1, expression of Raf BXB potently stimulated ERK2 activity in these cells (compare lanes 1 and 2). However, ERK2 activation by Raf BXB was effectively blocked by coexpression of increasing amounts of Raf-1 1-330 (lanes [3][4][5]. Thus, the amino terminus of Raf-1 can inhibit the activity of the Raf-1 catalytic domain in mammalian cells. Because the first 330 amino acids of Raf-1 contain several conserved domains, we next examined which regions of the Raf-1 amino terminus were required for autoinhibition. Conserved sequences present in the amino terminus include the CR1 domain, which contains the RBD, (amino acids 51-131), the CRD, (amino acids 139 -186), and the CR2 domain (amino acids 255-268) ( Fig. 2A) (5). Thus, we tested different aminoterminal constructs for their ability to inhibit HA-ERK2 activation stimulated by coexpression of Raf BXB. In these experiments, similar amounts of each amino-terminal protein were expressed, as judged by Western blot using an antibody that recognizes an N-terminal FLAG epitope present in each protein (data not shown). As shown in Fig. 2B, coexpression of Raf-1 1-330 blocked ERK2 activation by Raf BXB in a dose-dependent manner. Deletions within the carboxyl-terminal end of the regulatory domain did not affect autoinhibition, because Raf-1 1-256, 1-186, and 1-147 were as effective as 1-330 at inhibit- ing ERK2 activation. However, expression of the isolated RBD of Raf-1 (55-131) did not block ERK2 activation, demonstrating that the RBD alone cannot mediate autoinhibition. Similarly, the isolated CRD (amino acids 139 -186) was also relatively inefficient at blocking ERK2 activation by Raf BXB. Interestingly, mutation of cysteines 165 and 168 to serine, which disrupts the zinc finger and blocks the function of the autoinhibitory domain in Xenopus oocytes (18), only slightly affected the inhibition of ERK2 activation by the Raf-1 autoinhibitory domain in these cells (Fig. 2B, Raf-1 1-330 CC/SS). We interpret these data to indicate that the autoinhibitory domain is minimally encompassed by residues 1-147 but also includes sequences within the CRD amino-terminal to residue 147. Furthermore, an intact zinc finger appears not to be required for autoinhibition, because expression of the CC/SS mutant still blocked ERK2 activation. Because mutation of these cysteines also blocks the ability of the CRD to bind Ras, 14-3-3, and phosphatidylserine (5,6,23,24), it appears that interaction with these molecules is not required for autoinhibition in mammalian cells.
Autoinhibition of Raf-1 Activity Is Blocked by Coexpression of Constitutively Active PAK1 or Constitutively Active Src-Raf-1 is phosphorylated on serines 338 and 339 and tyrosines 340 and 341 in response to extracellular stimuli (15,16,25). Among these residues, serine 338 and tyrosine 341 appear to be the major phospho-acceptor sites because their phosphorylation is critical for Raf-1 activation (12,25). However, the role that phosphorylation of these sites plays in the Raf-1 activation mechanism is unknown.
It has previously been shown that mutation of tyrosine 340 to aspartic acid, which is thought to mimic the phosphorylation of this site, attenuates the autoinhibitory activity of the Raf-1 amino terminus in Xenopus oocytes (18). Thus, we examined whether phosphorylation of serine 338 and tyrosine 341 blocks autoinhibition by the Raf-1 amino terminus in mammalian cells. HEK 293 cells were transfected with Raf BXB, with or without Raf-1 1-330, and the activation of epitope-tagged ERK2 was measured. To stimulate the phosphorylation of serine 338 and tyrosine 341, the cells were also transfected with increasing amounts of constitutively active PAK1 or constitutively active Src, respectively. As shown in Fig. 3A, expression of constitutively active forms of either PAK1 (PAK1*, lanes 5-7) or Src (Src*, lanes 8 -10) blocked the autoinhibitory effect of Raf-1 1-330 on ERK2 activation in a dose-dependent manner. This was accompanied by phosphorylation of Raf-1 on serine 338 (PAK1 transfectants) or tyrosines 340 and 341 (Src transfectants), as determined by Western blotting with antibodies specific for phospho-serine 338 and phospho-tyrosine (data not shown). The results of three independent experiments are quantified in Fig. 3B. Thus, these data indicate that phosphorylation of serine 338 by PAK1, and tyrosines 340 and 341 by Src, block the ability of the autoinhibitory domain to function.
To determine whether expression of active PAK1 or active Src relieved autoinhibition through a mechanism other than phosphorylation of Raf BXB, we examined whether substitution of the PAK or Src phosphorylation sites with non-phosphorylatable residues (S338A or Y340/341F, respectively) affected the ability of active PAK1 or Src to block autoinhibition. As shown in Fig. 4A, expression of active PAK1 or active Src effectively relieved autoinhibition of wild type Raf BXB by the Raf-1 amino terminus (lanes 3-6). However, expression of active PAK1 only marginally rescued ERK2 activation in cells expressing Raf BXB S338A (S/A), whereas coexpression of active Src still blocked autoinhibition (lanes 7-10). The modest rescue of autoinhibition by active PAK1 in cells expressing Raf BXB S338A may be due to a low level of phosphorylation of serine 339 or an increased efficiency of coupling between Raf and MEK1 caused by the phosphorylation of MEK1 by PAK1 (22,26,27). On the other hand, expression of active Src did not relieve autoinhibition in cells expressing Raf BXB Y340/341F, whereas active PAK1 was fully able to block autoinhibition (lanes 11-14). Fig. 4B represents the mean of four independent experiments. Thus, these results indicate that expression of active PAK1 or active Src blocks Raf-1 autoinhibition through the direct phosphorylation of serine 338 and tyrosines 340/341, respectively.
Phosphorylation of Residues within the Raf-1 Kinase Activation Loop Is Not Required to Block Autoinhibition-Based on homology between B-Raf and Raf-1, Chong et al. (17) identified threonine 491 and serine 494 as potential phosphorylation sites that are required for Raf-1 activation. These sites are contained within the activation loop of the Raf-1 kinase domain, and phosphorylation of the corresponding residues in B-Raf is required for its activation (28). We therefore examined whether these sites are involved in the Raf-1 autoinhibition mechanism.
We first substituted these sites with alanine (TS/AA), which precludes their phosphorylation in the cell. As shown in Fig.  5A, this largely blocked the ability of Raf BXB to stimulate ERK2 activity (lane 7). This was most likely due to a reduction in Raf BXB kinase activity, because these sites were shown to be required for the kinase activity of full-length Raf-1 (17). However, some ERK2 activation was still apparent, and this was effectively blocked by coexpression of Raf-1 1-330 (lane 8).
In addition, autoinhibition of this Raf BXB mutant was counteracted by coexpression of active PAK1 or active Src (lanes 9 and 10). Thus, mutation of threonine 491 and serine 494 to nonphosphorylatable residues does not affect the autoinhibition mechanism.
To mimic the phosphorylation of these sites, we replaced them with acidic residues (TS/ED) (17). Expression of this Raf BXB mutant stimulated ERK2 activity to a greater degree than the TS/AA mutant (compare lanes 7 and 11). However, this Raf BXB mutant was still subject to autoinhibition by the Raf-1 amino terminus (lane 12). This suggests that phosphorylation of these sites is not required for relief of autoinhibition. Furthermore, autoinhibition of Raf BXB TS/ED was effectively blocked by coexpression of active PAK1 or active Src (lanes 13  and 14). Fig. 5B represents the mean of seven independent experiments. Thus, these results demonstrate that phosphorylation of threonine 491 and serine 494 is not required to block autoinhibition and indicate that phosphorylation of these sites is required for other steps within the Raf-1 activation mechanism.
Phosphorylation of Raf-1 by PAK1 or Src Decreases the Affinity of the Autoinhibitory Domain for the Catalytic Domain-Using purified proteins, we next examined whether phosphorylation of the Raf-1 catalytic domain by PAK1 or Src affected the ability of the autoinhibitory domain to bind to the catalytic domain. For these experiments, Raf-1 1-330 was expressed in bacteria as a GST fusion protein, and the Raf-1 catalytic domain (RafCAT, residues 304 -648) was expressed as an hexahistidine-tagged fusion protein in insect cells (it is insoluble in bacteria). To measure the binding of these proteins, an excess of GST⅐Raf-1 1-330 was incubated with RafCAT, and the binding was allowed to proceed to equilibrium. GST⅐Raf-1 1-330 was then precipitated using glutathione-agarose, and the amount of RafCAT bound was examined by Western blotting using an antibody specific for the RafCAT protein. The amount of RafCAT protein still present in the supernatant was also measured. Under these conditions, ϳ10% of the RafCAT protein did not bind to Raf-1 1-330 (Fig. 6). Furthermore, coprecipitation of RafCAT with the Raf-1 amino terminus required the presence of Raf-1 1-330 because RafCAT did not coprecipitate with glutathione beads bound to glutathione S-transferase alone (data not shown). However, when RafCAT was phosphorylated with Src prior to incubation with Raf-1 1-330, the amount of free RafCAT increased to ϳ30%. The reduction in binding was even more apparent if one measured only the tyrosine-phosphorylated RafCAT population (40% unbound). Similar results were observed when RafCAT was phosphorylated by recombinant PAK1 prior to incubation with the Raf-1 amino terminus. Thus, these data indicate that phosphorylation of Raf-1 on serine 338 or tyrosines 340/341 reduces the affinity of the autoinhibitory domain for the catalytic domain. DISCUSSION Raf-1 activation by growth factors is a complex process that entails recruitment to the plasma membrane by active Ras and the subsequent phosphorylation of Raf-1 on multiple sites. The role of phosphorylation within this mechanism has remained unclear. Previous work has shown that Raf-1 contains an autoinhibitory domain within the first 330 amino acids of its amino terminus (18). This was demonstrated by measuring the ability of the Raf-1 amino terminus to inhibit meiotic maturation in Xenopus oocytes stimulated by expression of the Raf-1 catalytic domain. It was also shown that regulation of autoinhibition did not correlate with changes in 14 -3-3 binding to the Raf-1 amino terminus and that it depended on the integrity of the CRD. In the present study we have shown that the Raf-1 autoinhibitory domain can block the ability of the Raf-1 catalytic domain to stimulate ERK2 activity in mammalian cells. Furthermore, this domain minimally consists of the first 147 amino acids of Raf-1, although sequences within the CRD carboxyl-terminal to residue 147 may contribute to autoregulation. We also show that phosphorylation of Raf-1 on serine 338 and tyrosines 340 and 341 blocks autoinhibition and that this is due to a reduction in the affinity of the autoinhibitory domain for the phosphorylated, catalytic domain. Thus, these results indicate that phosphorylation of Raf-1 on these sites is required for activation because this modification blocks autoinhibition.
Through extensive deletion analysis we have more precisely defined the region encompassing the autoinhibitory domain. This domain includes sequences within the RBD (amino acids 51-131) and the CRD (amino acids 139 -186) but does not include the CR2 domain of Raf-1 (amino acids 255-268). A coincidence of autoinhibitory and small G protein binding domains exists in other kinases. For example, PAK1 contains an autoinhibitory domain that partially coincides with its Rac/ Cdc42 binding domain (20,29). In addition, the interaction of the PAK1 autoinhibitory domain with the catalytic domain is regulated by phosphorylation (30,31).
Interestingly, in contrast to the results of Cutler et al. (18), we found that an intact zinc finger within the CRD domain was not required for autoinhibition. This is based on our definition of the minimal autoinhibitory domain (residues 1-147), which does not include the zinc finger region, and the finding that disruption of the zinc finger by mutation of two key cysteine residues to serine (C165/186S) only slightly affected autoregulation by the Raf-1 amino terminus (Fig. 2B). One possible explanation for these conflicting results is the divergent systems used to assay autoinhibition (maturation of Xenopus oocytes versus ERK2 activation in HEK 293 cells). In oocyte maturation, for example, a requirement for the CRD in autoinhibition may reflect a role in other aspects of Raf-1 signaling that are not directly related to mitogen-activated protein kinase activation.
Previous work also suggested that phosphorylation of Raf-1 on tyrosine 340 may block autoregulation. This was proposed because expression of RafCAT containing a phosphorylation mimic at Y340 (Y340D) precluded inhibition of meiotic maturation by the Raf-1 regulatory domain (18). In addition, it was recently published that phosphorylation of residues between serine 338 and tyrosine 341 was necessary for high affinity interaction between Raf-1 and MEK1 (32). Our results are Hexahistidine-tagged Raf-1 catalytic domain (RafCAT) was incubated at 4°C with GST⅐Raf-1 1-330 prebound to glutathione-agarose beads. Prior to incubation with the Raf-1 amino terminus, RafCAT was phosphorylated or not by incubation in kinase buffer with or without recombinant PAK1 or recombinant Src. After binding had reached equilibrium, GST⅐Raf-1 1-330 was precipitated by centrifugation. RafCAT associated with the GST⅐Raf-1 1-330, as well as that still in solution (unbound), was measured by Western blot with an antibody specific for the Raf-1 catalytic domain. Western blots were quantified by densitometry. Shown is the average of three independent experiments. Errors are the S.E. of the mean. consistent with the idea that phosphorylation of serine 338 and/or tyrosine 340 is required to block autoinhibition. We cannot, however, preclude a role for the phosphorylation of these sites in regulating the affinity of Raf-1 for MEK1. In fact, it is possible that phosphorylation of serine 338 and tyrosine 340 serves a dual role in stimulating the activity of Raf-1 toward MEK1, namely to block autoinhibition and to increase interaction between Raf-1 and MEK1.
We have also found that phosphorylation of threonine 491 and serine 494 is not likely to be involved in the regulation of autoinhibition, because substitution of these sites with acidic residues, which mimics their phosphorylation, does not affect autoinhibition. These sites are located within the activation loop of the Raf-1 kinase domain, and in many kinases phosphorylation of sites within the activation loop directly affects catalytic activity (33,34). Thus, given our data, we would predict that phosphorylation of threonine 491 and serine 494 is necessary for an increase in Raf-1 catalytic activity rather than a relief of autoinhibition.
In conclusion, our data detail a role for phosphorylation of Raf-1 on serine 338 and tyrosines 340 and 341 in the Raf-1 activation mechanism. Specifically, phosphorylation of these sites prevents interaction between the autoinhibitory and catalytic domains. Given this data, we can refine a common model for growth factor-mediated Raf-1 activation (5,17). In this model, inactive Raf-1 is recruited to the plasma membrane by GTP-bound Ras. Ras first binds to the RBD of Raf-1 and then interacts with the CRD. Binding of Ras to these domains causes 14 -3-3 to release from its amino-terminal binding site (serine 259), thereby allowing Raf-1 to further unfold. Dephosphorylation of serine 259 may also occur at this time. Raf-1 is then phosphorylated on serine 338 and tyrosines 340 and 341 by kinases whose activities are also regulated by Ras. Phosphorylation of these sites locks Raf-1 in an open conformation by preventing interaction between the catalytic and autoinhibitory domains. However, full activation of Raf-1 would not occur until it is phosphorylated on threonine 491 and serine 494 by one or more as yet unidentified kinases. Once Raf-1 is phosphorylated on all four sites it is fully active and able to phosphorylate its downstream target MEK. Refinement of this model remains an area for further study.