Characterization of Raf-1 activation in mitosis.

We have used site-directed mutagenesis to explore the mechanisms underlying Raf-1 activation in mitosis, and we have excluded most previously characterized activating interactions. Our results indicate that the primary locus of activation lies in the carboxyl-half of the molecule, although the extent of activation can be influenced by the amino-proximal region, particularly by the Raf-1 zinc finger. We also found that Raf-1 is hyperphosphorylated in mitosis at multiple sites within residues 283-302 and that these hyperphosphorylations are not required for activation. In addition, neither Mek1 nor Mek2 are stably activated in coordination with Raf-1 in nocodazole-arrested cells. Overall, the data suggest that the mechanism(s) responsible for activating Raf-1 during mitosis, and the subsequent downstream effects, are distinct from those involved in growth factor stimulation.

Raf-1 is a critical mediator of mitogenic and developmental stimuli to the mitogen-activated protein (MAP) 1 kinase cascade in eukaryotic cells. Activated Raf-1 phosphorylates and activates the mitogen-activated or extracellular signal-regulated kinases (Meks), which in turn phosphorylate and activate the MAP kinases known as extracellular signal-regulated kinases (Erks) (reviewed in Ref. 1). Inactive Raf-1 is maintained in the cytoplasm in a large macromolecular complex with heat-shock proteins (2) and 14-3-3 proteins (reviewed in Ref. 3).
Multiple lines of evidence suggest that Src and/or Src family members can regulate Raf-1 activity by tyrosine phosphorylation, When Raf-1 is coexpressed with activated Src in insect cells or Cos cells, it is phosphorylated at Tyr-340 and Tyr-341 and becomes activated (9,10). In Jurkat cells, CD4-mediated activation of Raf-1 is dependent on the Src family kinase Lck and is associated with enhanced Raf-1 tyrosine phosphorylation (11). Moreover, in Ras-transformed NIH 3T3 cells, plasma membrane-associated activated Raf-1 has increased tyrosine phosphorylation, and Raf-1 kinase activity is lost when Raf-1 immunoprecipitates are incubated with a tyrosine phosphatase (12). Interferon-, oncostatin M-, and growth hormone-induced activation of Raf-1 may also depend on phosphorylation at Tyr-340 and Tyr-341 by the Jak kinases (13)(14)(15).
Raf-1 can also be activated by coexpression with protein kinase C (PKC) family members in insect and mammalian cells (16,17). In addition, direct activation of Raf-1 immunoprecipitates by PKC␣ in vitro has been reported (16). The major PKC phosphorylation site in Raf-1 is Ser-499, and mutation of either Ser-499 or Ser-259 has been reported to inhibit Raf-1 activation by PKC in vivo (16,17), although this has been disputed (18,19). Raf-1 activation by PKC can also be inhibited by mutations that disrupt Ras-Raf-1 interactions (8,19,20). Another Raf-1 activator is CAP kinase, which has been reported to activate Raf-1 by phosphorylation at Thr-268 and Thr-269 (21). These are also the major Raf-1 autophosphorylation sites (6).
In addition to activation at G 0 /G 1 , Raf-1 is hyperphosphorylated at serine and activated in mitosis in a wide variety of cell types by an unknown mechanism (22)(23)(24). The hyperphosphorylation reduces Raf-1 electrophoretic mobility, but we do not know how many sites are phosphorylated or if this is causally connected to activation. In Jurkat cells, the Src family kinase Lck binds Raf-1 in a mitosis-specific manner, and Raf-1 is not activated in mitosis in an Lck-deficient Jurkat cell line (25). This is interesting in light of the well documented activation of Src family kinases in mitosis and the requirement of Src family function for fibroblast entry into mitosis (26 -28). However, it is not known whether direct phosphorylation of Raf-1 by Lck is involved or whether the more ubiquitous Src family members found in fibroblasts participate in Raf-1 mitotic activation in those cells.
Thus, there are many known mechanisms that potentially could be involved in the mitotic hyperphosphorylation and activation of Raf-1. We explored these possibilities by mutagenesis and excluded most known direct interactions. In addition, we identified a locus containing multiple hyperphosphorylation sites and showed that the primary mitotic activating event occurs in the carboxyl-half of Raf-1.
For Mek kinase assays, NIH 3T3 cells were grown in Dulbecco's modified Eagle's medium supplemented with 5% calf serum. Mitotic cells were prepared by treating cells with 0.4 g/ml nocodazole for 7 h and gently rinsing the plates to dislodge mitotic cells that were pelleted, washed, and lysed in a 1% Triton X-100 lysis buffer described in Jelinek et al. (32). Unsynchronous lysates were prepared from untreated duplicate plates. PDGF-stimulated cell lysates were prepared by serumstarving cells for 24 h, adding PDGF (20 ng/ml final) for 5 min, and lysing in RIPA buffer. Serum-starved, unstimulated cells were used as controls.
Kinase Assays and Western Blots-Raf-1 kinase assays were carried out exactly as described previously (22) using lysates adjusted to contain equal Flag-Raf-1 levels with two changes: 1) following clearing by centrifugation at 14,000 rpm at 4°C for 20 min, the lysates were precleared by rotation at 4°C for 1 h with protein G-Sepharose beads (Amersham Pharmacia Biotech, Uppsala, Sweden) (5 l of a 50% v/v slurry per 100 l of lysate), and 2) 6 g of anti-FLAG M2 monoclonal antibody (Eastman Kodak Co.) was used for each immunoprecipitation. Reaction products were resolved by 10% SDS-PAGE and transferred to Immobilon (Millipore, Bedford, MA). The membranes were then exposed to a PhosphorImager plate, and the resulting image was captured using a Storm 840 PhosphorImager (Molecular Dynamics, Sunnyvale, CA). Appropriate membrane segments were probed using a Flag polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) (used in Fig. 5B only) or Raf-1 or Mek1 monoclonal antibodies (Transduction Laboratories, Lexington, KY, Antibodies R19120 and M17020 respectively), as indicated.
Mek activity was assayed as described by Jelinek et al. (32), using GST-kinase-defective ERK (KD-Erk) as a substrate (GST-KD-Erk bacterial expression plasmid was a gift from Alessandro Alessandrini), except that Mek1 and Mek2 monoclonal antibodies were used for immunoprecipitation and Western blot analysis (Transduction Laboratories M17020 and M24520, respectively). Reaction products and separate aliquots of the Mek immunoprecipitates (used to measure Mek concentrations) were separately resolved on 8% SDS-PAGE gels.
Western blots in Fig. 6 were visualized using goat anti-mouse antibodies coupled to horseradish peroxidase (Sigma) and the Renaissance chemiluminescence reagent (NEN Life Science Products). In all other cases antibodies were detected with electrochemical fluorescence reagents (Amersham Pharmacia Biotech) using a Storm 840 PhosphorImager. Images were analyzed and processed using ImageQuant 1.1 (Molecular Dynamics) and Adobe Photoshop 4.0 (Adobe Systems, San Jose, CA).

RESULTS
We and others (22)(23)(24) have previously shown that Raf-1 undergoes mitosis-specific hyperphosphorylation and that Raf-1 immunoprecipitates from mitotic cells exhibit enhanced autokinase and Mek kinase activities. Since we planned to test the effects of Raf-1 mutations using amino-terminal Flag epitope-tagged Raf-1 protein (Flag-Raf) transiently expressed in Cos cells, we began by testing wild-type Raf-1 in this system. As anticipated, both the ability of Flag-Raf to phosphorylate KD-Mek1 and to autophosphorylate were increased in mitosis (Fig. 1, lanes 1 and 2, panels B and C). The average increase in mitotic Mek kinase activity relative to that seen for Raf-1 from unsynchronized cells in 16 experiments of this type was 11.5 Ϭ ϫ 1.1, similar to the 10-fold increase we have reported for stably overexpressed untagged wild-type Raf-1 in NIH 3T3 cells (22). In addition, as previously observed with untagged Raf-1, multiple electrophoretically retarded isoforms were present in mitotic cells (panel A).
To confirm that the phosphorylation of Mek1 was catalyzed by Raf-1 and not by a contaminating kinase activity bound to Raf-1 such as Cdc2 (which phosphorylates Mek on Thr-286 and Thr-292), we probed the kinase assay reaction products with a Mek1 polyclonal antibody that is specific for Mek1 phosphorylated on Ser-218, one of the Raf-1-specific phosphorylation sites (panel D). Raf-1 specific phosphorylation of Mek1 was not detectable in unsynchronized cells (compare lanes 1 and 3). However, antibody binding was increased ϳ3-fold for the mitotic sample (lane 2), indicating that Raf-1 itself was responsible for increased phosphorylation.
We examined the mechanisms underlying Raf-1 activation in mitosis using an array of Raf-1 mutants known to interfere with phosphorylation of and/or signaling to Raf-1 by Src, Ras, 14-3-3 proteins, protein kinase C (PKC), subcellular relocalization, Raf-1 autophosphorylation, and ceramide-activated protein kinase (CAP kinase) ( Table I). Amino-terminal Flag epitope-tagged wild-type (WT) and mutant Raf-1 proteins were transiently expressed and tested in unsynchronized and mitotic Cos cells, and their abilities to phosphorylate KD Mek in an immune complex kinase assay and phosphorylation-induced changes in Raf-1 electrophoretic mobility were compared.
Src Family Kinases-To explore the possible role of Src family kinases in the mitotic activation of Raf-1, we tested Flag-Raf(Y340F/Y341F) and Flag-Raf(Y340D/Y341D) which contain, respectively, loss-of-function (Tyr 3 Phe) and gain-offunction (Tyr 3 Asp) mutations at Tyr-340 and Tyr-341, the major Src phosphorylation sites (9). The activities of these mutants in unsynchronized cells have previously been shown to be decreased and increased, respectively (9), and both of these mutants are known to be poorly stimulated when coexpressed with activated Src in Cos cells (residual activation is believed to result from a modest effect of Src on Ras activation levels (10)). Flag-Raf(Y340F/Y341F) exhibited reduced kinase activity in unsynchronized cells as anticipated (Fig. 2, lane 3; Table II). Furthermore, it was activated to a significantly reduced extent in mitosis (lanes 3 and 4; Table II). In contrast, Flag-Raf(Y340D/Y341D) had elevated activity in unsynchro-nized cells and was activated to an extent similar to Flag-Raf(WT) in mitosis (lanes 5 and 6; Table II). The electrophoretic mobility of both mutants was retarded during mitosis, with Flag-Raf(Y340D/Y341D) more retarded than Flag-Raf(WT) (upper panel). We conclude that negative charges at these sites (either caused by mutagenesis or by Src-mediated phosphorylation) act synergistically with the mitotic activator(s) of Raf-1, but that changes in phosphorylation of these sites are not required for mitotic activation of Raf-1. This is consistent with our observations 2 and other observations (33) that Flag-Raf(WT) which has been activated by coexpression with activated Src in unsynchronized human embryonic kidney 293T cells or Cos cells is further activated during mitosis.
Ras-We tested a Raf-1 point mutant Flag-Raf (R89L) that is unable to stably associate with Ras (29), and we found that it behaved similarly to Flag-Raf(WT) both in unsynchronized and mitotic cells (Fig. 3A, compare lanes 1 and 2 to lanes 5 and 6; Table II).
The absence of significant Mek kinase activity in control immunoprecipitates of kinase-defective Flag-Raf(K375M) (Fig.  3B, lanes 1 and 2) provides further evidence that the Raf-1 immunoprecipitates are not contaminated with another Mek kinase activity (above).
These data suggest that changes in 14-3-3 binding are not central to Raf-1 activation in mitosis. However, based on the 2 A. D. Laird and D. Shalloway, unpublished observations.

FIG. 1. Raf-1 is active in mitosis in Cos cells.
Cos cells were transiently transfected with a plasmid encoding amino-terminal FLAG wild-type (WT) Raf-1. Lysates were then prepared from cells arrested in mitosis using nocodazole (M), or from untreated, unsynchronized cells (U). Flag-Raf-1 proteins were immunoprecipitated using the anti-FLAG M2 monoclonal antibody and captured on protein G-Sepharose beads. Immune complex Raf-1 kinase assays used [␥-32 P]ATP and His-tagged kinase-defective Mek1 (KD-Mek) (82) as a substrate. Reaction products were resolved by 10% SDS-PAGE and transferred to Immobilon. 32 P-Labeled products were detected by autoradiography, and Raf-1 and Mek1 proteins were subsequently detected by Western blot analysis. Panel A, Western blot using a Raf-1 monoclonal antibody. Panel B, autoradiograph of autophosphorylated Raf-1. Panel C, autoradiograph of phosphorylated KD-Mek. Panel D, detection of Mek1 using a phospho-specific polyclonal antibody that binds Mek1 phosphorylated at Ser-218/222 with a much greater affinity (Ͼ1000-fold) than inactive Mek1. Panel E, Western blot using a non-phospho-specific Mek1 monoclonal antibody. Anti-Flag antibody was included in experimental (ϩ) or omitted in control (Ϫ) immunoprecipitations as indicated. Molecular mass standards are indicated in kDa. Raf-1 is underloaded ϳ50% in lane 1 relative to lane 2 in this experiment.
C165S/C168S mutant results, Raf-1 zinc finger integrity appears to be important for Raf-1 mitotic activation and hyperphosphorylation, suggesting that interaction of an unknown regulator(s) with the zinc finger may play a role in these phenomena. Alternatively, Flag-Raf(C165S/C168S) may be crippled in a nonspecific manner, as suggested by its inability to respond efficiently to EGF stimulation in Cos cells (34). We were unable to test the involvement of the carboxyl-proximal region 14-3-3-binding site using similar approaches since its mutation eliminates Raf-1 kinase activity (6,8).
Protein Kinase C-The ability of protein kinase C (PKC) to phosphorylate and activate Raf-1 in vitro is abolished by mutation of Ser-499 (16). To test whether phosphorylation at this site might be important for mitotic activation, we tested Flag-Raf(S499A). However, we found that it was activated to normal levels and exhibited a normal electrophoretic mobility shift in mitosis (Fig. 4, lanes 11 and 12; Table II).
The demonstration that the R89L and S259A mutants are still activated in mitosis provides further evidence that PKC is not the primary activator of Raf-1 in mitosis since 1) Raf (R89L) has recently been shown to have reduced responsiveness to PKC-mediated stimulation (8,20), and 2) Ser-259 is required for efficient activation of Raf-1 by coexpressed PKC (17) (although this has been disputed (18)).
Subcellular Localization-Ras activates Raf by altering its subcellular localization and recruiting it to the plasma membrane where it may participate in further activation steps (reviewed in Ref. 3). It is possible that Ras-independent Raf-1 activation in mitosis could also act by altering subcellular localization. We tested this by constitutively localizing Raf-1 to the plasma membrane by adding a CAAX box motif to its carboxyl terminus. Raf-1-CAAX has a greatly increased basal kinase activity which can be additionally stimulated by EGF (37,38). We found that Flag-Raf-1-CAAX was activated in mitosis and had a mitosis-specific electrophoretic mobility shift (Fig. 3B, lanes 5 and 6; Table II), suggesting that constitutive localization to the plasma membrane is neither sufficient for nor prevents the mitosis-specific phenomena.
CAP Kinase and Autophosphorylation-Cross-linking of Raf-1 proteins results in their activation (39,40), suggesting that Raf-1 might become activated by trans-autophosphorylation. The major sites of Raf-1 autophosphorylation have been mapped to Thr-268 and Thr-269 (6). Moreover, these residues are reportedly targets of the ceramide-activated protein kinase (CAP kinase) (21). We tested Flag-Raf(T268A/T269A) and found it to be activated in mitosis to an extent similar to Flag-Raf(WT) and to undergo a similar mitosis-specific electrophoretic mobility shift (Fig. 4, compare lanes 5 and 6 with lanes 3 and 4; Table II).
Phosphorylation at Ser-338 and Ser-339 -Ser-338 has been recently identified as a site whose phosphorylation is required for Ras-, CAAX box-, and Src-dependent activation of Raf-1 (41). We tested Flag-Raf(S338A/S339A) and found that it exhibits a lower basal activity than Flag-Raf(WT) (Fig. 4, compare lanes 7 and 3; Table II) as previously reported (19) and is activated to a reduced extent (lanes 7 and 8; Table II). Flag-Raf(S338A/S339A) undergoes mitosis-specific changes in electrophoretic mobility similar to that seen for Flag-Raf(WT). This phenotype is similar to that seen for Flag-Raf(Y340F/Y341F) (see above and lanes 9 and 10; Table II), suggesting that there is a general requirement for phosphorylation at either Ser-338 and Ser-339 or Tyr-340 and Tyr-341 to facilitate complete activation of Raf-1 in mitosis. In both cases, the fact that some mitotic stimulation is observed implies that it is not initiated  by phosphorylation at these sites.
Mapping of Minimal Raf-1 Sequences Required for Activation in Mitosis-We next constructed a series of truncation mutants to map the minimal sequence elements required for Raf-1 activation in mitosis (Fig. 5A). These mutants were Flag epitope-tagged at their carboxyl terminus (the eight carboxylterminal residues were replaced by the Flag epitope). Proteins were expressed in unsynchronized and mitotic Cos cells, and immune-complex kinase assays were performed to measure their activities (Fig. 5B, middle panel). The Raf-1 proteins in the kinase reactions were visualized by Western blotting (upper panel). The carboxyl-tagged wild-type Raf-1 (Raf(WT)-Flag) was activated only ϳ5-fold in mitosis (lanes 3 and 4), as compared with ϳ10-fold for Flag-Raf(WT), suggesting that the addition of the Flag epitope and/or removal of the eight carboxyl-terminal residues might inhibit Raf-1 activation in mitosis. Activation of the truncation mutants was compared with Raf(WT)-Flag.
To define more precisely the required region, we constructed and tested two internal deletion mutants lacking residues between 282 and 294 or between 282 and 303 (Fig. 5A). These constructs, like the point mutants, contained an amino-terminal Flag epitope. We found that Flag-Raf(⌬282-294) and Flag-Raf(⌬282-303) (Fig. 5C, lanes 11 and 12 and lanes 13 and 14; Table II) were also normally activated in mitosis. Combined with the truncation mutant data this implies that Raf-1 is activated during mitosis by events acting between residues 303 and 640, possibly also involving residues 275-282.
Mitotic Hyperphosphorylations in the 267-302 Region-Several isoforms of Raf(267-640)-Flag (Fig. 5B, lanes 5 and 6) having differing mitosis-specific electrophoretic mobility retardations were observed in Western blots of the kinase reactions (top panel) or of whole cell lysates (bottom panel). This was evident, but to a lesser extent, in Raf(275-640)-Flag (lanes 7 and 8); Raf(292-640)-Flag (lanes 9 and 10) migrated as a single species in both unsynchronized and mitotic cells and had a slight mobility retardation in mitosis (lanes 9 and 10). Mitotic mobility changes were not evident in Raf(302-640)-Flag ( lanes  11 and 12). This suggests that there are several mitosis-specific phosphorylations located between residues 267 and 302. Flag-Raf(⌬282-294) from both unsynchronized and mitotic cells exhibited multiple electrophoretic isoforms and an evident mitosis-specific mobility retardation (Fig. 5C, lanes 11 and 12). Flag-Raf(⌬282-303) exhibited two isoforms in both unsynchronized and mitotic cells and had only a very small mitotic mobility retardation (lanes 13 and 14). Combined with the truncation data, this suggests that at least some mitosis-specific phosphorylation sites reside between residues 283 and 301 (see sequence in Fig. 5A).
Six of the nine potential sites were examined by constructing and testing Flag-Raf(S294A/S296A/S301A), Flag-Raf(S287A/ S289A), and Flag-Raf(S289A/S291A). The triple mutation had no effect on electrophoretic mobility, but both double mutants exhibited reduced electrophoretic mobility retardation in mitosis compared with Flag-Raf(WT) (Fig. 5C, compare lanes 5-10  with lanes 3 and 4). This indicates that phosphorylation at Ser-294, Ser-296, and Ser-301 is not responsible for mitosisspecific changes in Raf-1 electrophoretic mobility and that either Ser-289 or both Ser-287 and Ser-291 are mitosis-specific phosphorylation sites. In agreement with the deletion mutations, these mutations did not significantly affect the mitotic increase in Raf kinase activity (Fig. 5C, lanes 5-10; Table II).
Mek1 and Mek2 in Mitosis-Genetic and biochemical evidence suggests that the major targets of Raf-1 are the mitogenactivated or extracellular signal-regulated kinases (Mek1 and Mek2); these in turn phosphorylate and activate the extracellular signal-regulated kinases (Erks) (1). We and others (42)(43)(44)(45) have shown that the Erks are not activated in mitosis in Chinese hamster ovary, HeLa, and Swiss 3T3 cells, and Ros- somando et al. (46) have reported that Mek1 is not activated in mitosis in HeLa cells. Mek2 is the only known Raf-1 target other than Mek1 (1,47). It differs from Mek1 in that it lacks two negative-regulatory Cdc2 phosphorylation sites (46,47).
We examined Mek1 and Mek2 activities in mitotic and unsynchronized cells using an immune complex kinase assay with kinase-defective Erk as the substrate. Platelet-derived growth factor (PDGF) treatment was used as a positive control in these experiments. In NIH 3T3 pLJRaf(A) cells, which overexpress wild-type Raf-1 ϳ3-fold (22), we observed that PDGF treatment resulted in a high level of Mek1 activation (Fig. 6A, lane 1). However, in agreement with Rossomando et al. (46), Mek1 was not activated and may even have been inhibited in mitotic (as compared with unsynchronized) cells (lower panel, lanes 2 and   3). In addition, we found that in NIH 3T3 cells Mek2 behaves similarly to Mek1, its activity was not increased in mitosis (Fig. 6B). DISCUSSION We and others (22)(23)(24)48) have previously shown that Raf-1 is hyperphosphorylated and activated during mitosis. Many of these experiments involved the use of nocodazole, and it has recently been shown that treatment with nocodazole or colchicine, another microtubule-destabilizing drug, can regulate Src family kinases and the Erk cascade in a non-cell cycle-dependent manner (49,50). We believe that such as inhibitor-based mechanism is not responsible for the observed activation of Raf-1 because of the following observations. 1) We have shown that Raf-1 activation can be detected in mitotic cells isolated by shake-off alone (with no drug treatment) (22). 2) Raf-1 hyperphosphorylation and activation is also detectable in cells arrested in mitosis using taxol, which acts by a different mechanism (48). 3) Only a small fraction of Raf-1 exhibits an electrophoretic mobility shift in lysates isolated from the nonmitotic cells in the nocodazole-treated adherent monolayer which remains after the rounded mitotic cells are rinsed off. 2 Cdc2 is capable of phosphorylating Mek1 (46) as well as Raf-1 2 and is present at high levels in mitotic lysates. But it is unlikely for the following reasons that contaminating Cdc2 in the anti-Raf-1 immune complexes was responsible for the observed increases in Mek-directed kinase activity. 1) Significant Mek1 kinase activity was not present in control immunoprecipitates where the Flag antibody was omitted (e.g. Fig. 1, lane  3 and 4) or in immunoprecipitates of a kinase-defective Raf-1 mutant (Flag-Raf(K375M). 2) Following incubation with mitotic Raf-1 immunoprecipitates, the Mek1 substrate exhibited enhanced immunoreactivity with a Mek1 antibody prepared against a Mek1 phosphopeptide containing phosphoserines at residues 218 and 222, which are Raf-1 (and not Cdc2) phosphorylation sites (Fig. 1). 3) Phosphoamino acid analysis of Mek-1 eluted from an SDS-PAGE gel following a Raf kinase assay using immunoprecipitates from mitotic NIH 3T3 cells stably expressing Flag-Raf(WT) showed that Mek1 was phosphorylated on serine, but not threonine, 2 consistent with phosphorylation by Raf-1 (Ser-218 and Ser-222) but not Cdc2 (Thr-286 and Thr-292). 4) The RIPA buffer used in lysate preparation contains 0.1% SDS in addition to Nonidet P-40 and sodium deoxycholate, which minimizes coprecipitation of contaminating kinases (51).
Circumstantial evidence suggests that either Src, Ras, 14-3-3, PKC, or CAP kinase could be responsible for the activation of Raf-1 that occurs in mitosis. However, we have shown that this is not so, at least by known direct mechanisms, in each case.
Src kinase activity is stimulated in mitosis, and Src family activity is required for cell cycle progression of fibroblasts through late G 2 (reviewed in Refs. 26, 27, and 52). Moreover, Raf-1 is activated in mitosis in regular Jurkat lymphocytes but FIG. 6. Activities of Mek1 and Mek2 in mitosis. A, lysates were prepared from pLJRaf(A) cells that had been serum-starved followed by 5 min, 20 ng/ml PDGF treatment (ϩGF), arrested in mitosis by nocodazole-treatment (M), or unsynchronized (U). Immune complex kinase assays were performed with anti-Mek1 immunoprecipitates from 1 mg of total cell protein using [␥-32 P]ATP and GST kinase-defective Erk (KD-ERK) as a substrate. Anti-Mek1 antibody was omitted from control immunoprecipitates. Upper panel, aliquots of the immunoprecipitates were Western-blotted with Mek1 monoclonal antibody. Lower panel, reaction products were analyzed by 8% SDS-PAGE and autoradiography. B, lysates were prepared from NIH 3T3 cells that had been serumstarved for 24 h (ϪGF), serum-starved followed by 5 min, 20 ng/ml PDGF treatment (ϩGF), arrested in mitosis by nocodazole treatment (M), or unsynchronized (U). Immune complex kinase assays were performed as described above with anti-Mek2 immmunoprecipitates from 700 g of total cell protein. Anti-Mek2 antibody was omitted from control immunoprecipitates. Upper panel, aliquots of the immunoprecipitates were Western blotted with Mek2 monoclonal antibody. Lower panel, reaction products were analyzed by 8% SDS-PAGE and autoradiography. Molecular mass standards are indicated in kDa. not in a Lck-deficient Jurkat cell line (25). A direct mechanism that could be involved is mitosis-specific Src phosphorylation of Raf-1 Tyr-340 and Tyr-341; Src phosphorylates these sites when Src and Raf are coexpressed in insect and mammalian cells, and this can activate Raf-1 in a Ras-independent (insect cells) or Ras-dependent (mammalian cells) manner (9,10,12). Consistent with this, Raf(Y340F/Y341F) or Raf(Y340D/Y341D) mutants have reduced or increased basal kinase activities, respectively (9). We (and recently Ziogas et al. (33)) found that Raf(Y340F/Y341F) exhibits reduced activation in mitosis, suggesting that phosphorylation at this site is needed for full mitotic activation. However, our finding that Raf-1 is normally activated in mitosis when negative charges are constitutively placed at these sites (Raf(Y340D/Y341D) shows that mitotic Src-induced changes in tyrosine phosphorylation are not responsible for the mitotic activation. Thus, the evidence that Lck is at least indirectly involved in Raf-1 activation in lymphocytes is in contrast to our evidence that Src is not directly involved in Raf-1 activation in Cos cells. It would be interesting to test whether Raf(Y340F/Y341F) and Raf(Y340D/Y341D) mutants were activated in mitosis in Jurkat cells. Genetic and biochemical evidence indicates that Ras acts directly upstream from Raf-1 in transmitting proliferative and developmental signals to the MAP kinase cascade (reviewed in Ref. 1). A triple Saccharomyces cerevisiae mutant that lacks Ras function undergoes cell cycle arrest late in mitosis; this phenotype can be rescued by exogenous mammalian Ras (53). Furthermore, overexpression of Ras in mammalian cells induces a rapid increase in abnormal mitoses (54,55). However, the demonstration that Raf(R89L), a mutant which cannot stably interact with Ras (29), is normally activated in mitosis, indicates that Ras binding is not required for this (similar data was recently published by others (33)). Interaction of the recently identified kinase suppressor of Ras (KSR) with Raf-1 is Ras-dependent (56,57) and thus is presumably blocked by the R89L mutation. Therefore it too is unlikely to participate in mitotic activation of Raf-1. Since the R89L mutation significantly suppresses Raf response to EGF (8,20), we conclude that the mechanism which activates Raf during mitosis is distinct from that which activates it following the addition of growth factors.
Ser-338 is a Ras-responsive phosphorylation site, and the S338A/S339A mutation blocks activation of Raf-1 by coexpressed Ras and inhibits Raf-1 activation by coexpressed Src in Cos cells (19,41). Thus, the finding that Raf(S338A/S339A) is activated in mitosis is consistent with the observations that neither Src nor Ras are required for activation. The phenotype of this mutant is similar to that of Raf(Y340F/Y341F), which may be related to recent evidence suggesting that phosphorylation in the region between residues 338 and 341 participates in lifting repression of Raf-1 kinase activity (58).
Protein kinase C (PKC) plays a critical role in cell cycle progression at G 2 /M (65), and some PKC family members may directly activate Raf-1 by phosphorylation at Ser-499 and Ser-259 (16,17) (although this has recently been disputed (18,19)). PKC-induced activation appears to be Ras-dependent since it is inhibited by the R89L mutation which, as discussed above, blocks the Ras-Raf interaction (8,20). Thus, our demonstration that neither the S499A, S259A, nor R89L mutations blocked mitotic Raf-1 activation indicates that PKC phosphorylation at these sites is not directly involved. Ras-mediated activation of Raf-1 involves its recruitment of Raf-1 to the plasma membrane, and constitutive relocalization by addition of a CAAX sequence to the Raf-1 carboxyl terminus constitutively activates Raf-1 in unsynchronized cells (37,38). However, since we showed that the kinase activity of Raf-(CAAX) is further increased during mitosis, we can conclude that changes in its localization are not sufficient for its mitotic activation.
It has been reported that CAP kinase can activate Raf-1 by phosphorylating the Raf-1 autophosphorylation sites at Thr-268 and Thr-269 (21,66). In addition, the activation of Raf-1 chimeras by artificially induced dimerization (39, 40) might possibly be explained by enhanced trans-autophosphorylation at these sites. However, the possibility that these mechanisms are required for mitotic activation is excluded by the observation that the Raf(T268A/T269A) double mutant was normally activated in mitosis.
Although it did not completely block mitotic activation, the C165S/C168S mutation caused a significant (ϳ3-fold) decrease in activation. This mutation destroys the structural integrity of the Raf-1 zinc finger, suggesting that zinc finger domain-mediated interactions may participate in (although not initiate) the activation.
Other potential candidates to activate Raf-1 in mitosis include the kinases that target the MPM-2 phosphoepitope for phosphorylation-dependent proline isomerization (67,68) and Cdc2. However, we found that mitotic Raf-1 does not crossreact with an MPM-2 antibody capable of recognizing multiple mitotic phosphoproteins. 2 Moreover, Raf-1 does not contain a recently identified consensus sequence found in other proteins targeted by MPM-2 kinases (68). Raf-1 contains 11 Ser/Thr-Pro residues so testing the potential role of Cdc2 by site-directed mutagenesis was difficult. Instead, we used a more general truncation and deletion approach to identify the minimum Raf-1 mutant that can be activated in mitosis. We found that the carboxyl-half of Raf-1 (residues 275-640) contains the sequences essential for Raf-1 mitotic activation. Furthermore, an otherwise full-length Raf-1 mutant lacking residues 283-302 is normally activated. This indicates that the primary locus of Raf-1 activation lies between residues 303 and 640 (and/or between residues 275 and 282). This reduces the number of potential Cdc2 phosphorylation sites that could be required for mitotic activation to three as follows: Thr-310, Thr-383, and Ser-567. Experiments currently underway to test the importance of these residues will allow us to determine whether Cdc2 plays a primary role in Raf-1 activation in mitosis.
The carboxyl-proximal localization of the primary activating event is consistent with previous observations that v-Mil, an avian Gag-Raf-1 fusion protein which lacks residues equivalent to residues 1-266 of Raf-1, is activated in mitosis (23). Other precedents for regulation of the carboxyl-half of Raf-1 include the following: 1) the activation of Raf-BXB (which lacks amino acids 26 -302) in Jurkat cells by stimulation of the CD3 receptor or by phorbol ester or okadaic acid treatment (69), and 2) the responsiveness of 22W Raf (a truncation mutant lacking amino acids 1-305) to stimulation by coexpressed activated Src (but not Ras) (70).
Raf(302-640) differs from WT Raf-1 in that during mitosis it does not exhibit electrophoretic isoforms with retarded mobility. However, the longer truncation mutant Raf(267-640) did exhibit such isoforms. The mitosis-specific retardations have been previously shown to be reversed by dephosphorylation (22,24), so we conclude that at least some mobility-retarding phosphorylations lie downstream of residue 267. Since truncation mutants of different lengths (e.g. Raf(275-640) and Ras(292-640)) display isoforms with different mitosis-specific mobility retardations, we conclude that there are multiple mitosis-specific phosphorylation sites. The fact that deletion of residues 283-302 eliminates the mobility retardations suggests that at least two of the serines in this region are mitotic phosphorylation sites. Data from two overlapping double point mutants suggests that Ser-289 is one of these. This may be related to the observation that Raf-1 undergoes multiple phosphorylation events in 32 P-labeled NIH 3T3 cells within residues 283-309 (71). Although we have shown here that phosphorylation within residues 283-302 is not required for mitotic activation, it may be functionally important in other ways. It has been shown that truncation of the region between residues 273 and 302 is associated with a major gain in Raf-1 transforming activity (72).
The fact that the mitosis-specific phosphorylations are completely inhibited by the C165S/C168S mutation within the Raf-1 zinc finger domain and in the kinase-defective K375M mutant shows that non-local effects can be important. Therefore, the site-mapping analysis above must be treated with some caution. However, the simplest explanation of these nonlocal effects is that the mobility-retarding mitotic phosphorylations are secondary effects (occurring after the primary carboxyl region activating event) that require an intact Raf-1 zinc finger and Raf-1 kinase activity. This might be related to the finding that the Raf-1 electrophoretic mobility retardations induced by extracellular stimuli are blocked by inhibitors of its downstream target Mek (73,74).
Overall, the data suggest that the mechanism(s) responsible for mitotic activation of Raf-1 is different from that responsible for activation after growth factor stimulation. In addition, it appears that the downstream effects may differ. Consistent with published findings in HeLa cells (46), we found that Mek1 was not activated in NIH 3T3 cells arrested in mitosis with nocodazole. Moreover, we found that Mek2, the other Mek known to be activated by Raf-1 following growth factor stimulation (47), is not activated in nocodazole-arrested mitotic cells. Similarly, we and others (42)(43)(44) have failed to detect Erk activation in lysates from nocodazole-arrested cells.
However, it is possible that there are highly compartmentalized or transient Raf-1-induced changes in Mek and Erk activity that were not detected using nocodazole-arrested whole cell lysates. Indeed, immunofluorescence microscopy with phosphospecific antibodies detects activated Mek within the nucleus in early prophase and localization of active Erk at the kinetochores at prophase and prometaphase (75,76). In addition, it has recently been proposed that Mek1 and a Golgi-associated Erk direct the dissolution of the Golgi apparatus during the G 2 /M transition of rat kidney cells (77), although this has been disputed (78). These phenomena could reflect Raf-1 activation.
In addition, Raf-1 could act on non-Mek, non-Erk mitotic targets such as a non-Erk MAP kinase-like activity that we 2 and others (42,43) have seen in mitotic cells but not in unsynchronized or growth factor-stimulated cells. Integrins are another potentially interesting family of direct and/or indirect targets. Raf-1 was recently isolated in a screen for integrin suppressors and was subsequently shown to suppress integrin function in an Erk-dependent manner (79), suggesting the pos-sibility that Raf-induced down-regulation could promote the changes in cell shape that are seen at G 2 /M. Another potential target is the microtubule catastrophe factor Oncoprotein 18/ Stathmin which undergoes mitotic hyperphosphorylation that is critical for normal mitotic progression (80). Two of these hyperphosphorylation sites can be phosphorylated by Erks (as well as Cdc2) in vitro, and these sites are rapidly phosphorylated following activation of an inducible Raf-1 construct in NIH 3T3 cells (81).