Mitogen Regulation of c-Raf- 1 Protein Kinase Activity toward Mitogen-activated Protein Kinase-Kinase*

The c-raf- 1 protooncogene encodes a SerIThr protein kinase. A mitogen-activated protein kinase-kinase (MAPKK) purified from bovine brain is phosphoryl- ated and activated 4-%fold in vitro by c-Raf-1 from mitogen-treated cells. c-Raf-1 protein kinase activity, measured by the phosphorylation of brain MAPKK substrate, is detectably activated within 1 min after addition of platelet-derived growth factor (PDGF) to 3T3 cells, increasing more rapidly than the endogenous NIH 3T3 cell MAPKK activity. c-Raf-1 activation is also induced by insulin, phorbol ester, thrombin, and endothelin. PDGF-, epidermal groth factor-, and in-sulin-stimulated 32P-c-Raf- 1 yield very similar, com- plex tryptic “P-peptide maps, wherein only 2 of 10 32P-peptides appear entirely de novo after growth factor addition. Mitogen-activated protein kinaselextracellular signal-regulated kinase-2 can phosphorylate c-Raf-1 in vitro on 4-6 tryptic 32P-peptides, all of which comigrate with tryptic 32P-peptides derived from c-Raf- 1 labeled in situ. Mitogen-activated protein kinase phosphorylation of c-Raf-1

I The first two authors contributed equally.
$$ Established Investigator of the American Heart Association.
ing murine retroviral oncogene product, v-Raf (1). The ability of v-Raf to activate cell division indicates that the product of the protooncogene, c-ruf-1, participates in the intracellular signal transduction response to mitogens. Numerous studies have demonstrated that mitogen treatment of cells increases c-Raf-1 autophosphorylation in vitro, measured after cell disruption, shifts the c-Raf-1 polypeptide to a slower mobility on SDS-PAGE' and increases c-Raf-1 kinase activity toward a number of model substrates, such as a syntide and histone H1 (2,3). The rate of phosphorylation of the model substrates by c-Raf-1 immunoprecipitates is, however, very low and diminishes progressively on washing. The lack of a reliable quantitative assay for c-Raf-1 protein kinase activity has hampered study of the mechanism by which c-Raf-1 is normally regulated during mitogenic signal transduction, inasmuch as the most reliable assay of c-Raf-1 "activation" remains its transforming potential, rather than its biochemical activity toward a genuine physiologic substrate.
Recently, strong evidence has identified mitogen-activated protein kinase-kinase (MAPKK) as a likely physiologic substrate of c-Raf-1. MAPKK is a novel, dual specificity protein kinase (4-7) which mediates the activation of the MAP kinases (MAPKs) (8,9), Erk-1 (p44 MAP kinase), and Erk-2 (p42 MAP kinase) by catalyzing the phosphorylation of nearly adjacent Tyr and Thr residues in the motif, TEY, in subdomain VI11 of the MAP kinase catalytic domain (10)(11)(12)(13)(14)(15)(16). MAPKK is itself regulated by Ser/Thr phosphorylation, in so far as it can be inactivated with Ser/Thr phosphatase-2A, but not Tyr phosphatase (16,17). MAPKK is constitutively activated in v-ruf-transformed NIH 3T3 cells, suggesting that c-Raf-1 may be situated upstream of MAPKK in the mitogen signal transduction pathways (18). Direct evidence that c-Raf-1 is (at least one of) the immediate upstream activators of MAPKK was provided by the demonstration that a constitutively active, amino-terminally truncated c-Raf-1 could phosphorylate in uitro and reactivate a MAPKK previously deactivated with phosphatase-2A (18). Similar data have subsequently been provided by others (19,20).
In this study, we purified a 48-kDa MAPKK to near homogeneity from bovine brain and established that this enzyme is a substrate for c-Raf-l-catalyzed phosphorylation and ac-The abbreviations used are: PAGE, polyacrylamide gel electrophoresis; MAP, mitogen-activated protein; MAPK, mitogen-activated protein kinase; MAPKK, mitogen-activated protein kinase-kinase; btMAPK, bacterially expressed, recombinant MAPK; MBP, myelin basic protein; Mops, 4-morpholinepropanesulfonic acid; Erk, extracellular signal-regulated kinase; Mes, 4-morpholineethane sulfonic acid; DTT, dithiothreitol; PDGF, platelet-derived growth factor; EGF, epidermal growth factor; PMA, phorbol 12-myristate, 13-acetate, PMSF, phenylmethylsulfonyl fluoride; TPA, 12-O-tetradecanoylphorbol- 13-acetate. tivation. This MAPKK provides the basis for a sensitive assay for c-Raf-1 kinase activity, enabling a more precise characterization of the changes in c-Raf-1 catalytic function on mitogen stimulation in situ, as well as direct inquiry to the role of the previously described MAP kinase-catalyzed phosphorylation of the c-Raf-1 polypeptide (21,22) in the regulation of c-Raf-1 protein kinase.
Our results indicate that wild-type c-Raf-1 endogenous to NIH 3T3 cells or H4 cells phosphorylates and activates MAPKK in uitro; this activity of c-Raf-1 is substantially increased upon stimulation by a number of agonists in situ, rising more rapidly than the activity of endogenous MAPKK, consistent with the idea that c-Raf-1 lies upstream of MAPKK in situ. Activation of c-Raf-1 in situ is accompanied by multisite (Ser/Thr) Raf-1 phosphorylation. Although MAPK probably contributes to the mitogen-stimulated phosphorylation of c-Raf-1, present evidence indicates that MAPK neither initiates nor down-regulates c-Raf-1 activation.

EXPERIMENTAL PROCEDURES
Preparation of Bovine Brain MAPKK-All procedures were carried out at 4 "C. Three bovine brains (Pel-Freez, Rogers, AR), frozen as l-cm3 pieces at -80 "C, were powdered, frozen, and the powder homogenized in 1 liter of buffer A (10 mM Tris-HCI, pH 8.0, 2 mM EGTA, 1.5 mM DTT, 5% (v/v) glycerol, 1 mM PMSF, 2 p~ pepstatin, 2 p M leupeptin, 10 kallikrein-inhibiting units/ml aprotinin) The lysate was centrifuged at 3,000 X g for 10 min. Often a fat cake was observed at this point, floating on top of the extract. This was removed and the remaining material was centrifuged for 1.5 h at 100,000 X g at 4 "C. The supernatant was filtered through glass wool and mixed with 300 ml of (settled) DEAE-cellulose equilibrated in buffer A. The slurry was stirred gently for 1 h, and the breakthrough material was collected by suction filtration. Mes was added to the breakthrough from a 1 M, pH 6.5, stock, to a final concentration of 15 mM, and the pH of the extract was adjusted to 6.5. The extract was then mixed with 50 ml of (settled) Fast-S Sepharose which was equilibrated with buffer M (20 mM Mes, pH 6.5, 2 mM EGTA, 1.5 mM DTT, 5% (v/v) glycerol, 0.03% (w/v) Brij 35, 1 mM PMSF). This slurry was gently stirred for 1 h, at which time the resin was poured into a column.
The column was washed with 200 ml of buffer M and developed with a 500-ml linear gradient of NaCl (0-250 mM) in buffer M. The flow rate was kept at 1.5 ml/min and 7-ml fractions were collected. Fractions containing maximal MAPKK specific activity were pooled and dialyzed (2 h) into buffer M' (buffer M containing 5 mM MgC12).
The dialysate was then applied to a 5-ml column of DEAE-Cibacron blue 3GA-agarose equilibrated with buffer M'. The column was washed with 30 ml of buffer M' and developed with a 50-ml linear gradient of NaCl (0-250 mM) and ATP (0-2 mM) in buffer M'. The flow rate was kept at 0.3 ml/min and 1-ml fractions were collected. Fractions containing maximal MAPKK specific activity were pooled and dialyzed into buffer M and concentrated by Mono-S chromatography. The Mono-S column was run at 1 ml/min and developed with a 10-ml linear NaCl gradient (0-250 mM) in buffer M. Fractions (0.5 ml) containing MAPKK activity were pooled and snap frozen in liquid Nz as 100-p1 aliquots and stored at -80 "C. Routinely, this procedure gave a preparation with a specific activity of 3500 units/ mg (16, 18) with an overall recovery of 6% (see Table I). Fig. 1A shows a silver-stained gel of Mono-S fractions from a representative MAPKK preparation.
Preparation of MAPKK from PDGF-stimulated NIH 3T3 Cells for Use in Time Course Experiments-NIH 3T3 cells were treated with PDGF (20 ng/ml) for the times indicated in the figure legends. Cell extracts were prepared as described elsewhere (18) except that the buffer contained 50 mM NaC1. MAPKK was separated from MAPKs by passing the extracts through a Mono-Q column equilibrated with extraction buffer. MAPKK flows through the Mono-Q under these conditions, while MAPKs bind (18). MAPKK was then assayed as previously described (18). c-Raf-1 was purified from these cells as described below, in parallel. Phosphotyrosine immunoblotting was performed as described elsewhere (23).
Cells-NIH 3T3 cells and EC4Al cells (an NIH 3T3 cell stably overexpressing c-Raf-1) were cultured to confluence in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. H4 hepatoma cells were cultured as previously described (24). For mitogen stimulation, cells were serum-starved (0.25% fetal calf serum for 3T3 cells, 0% serum for H4 cells) overnight. PDGF was added as indicated, or at 20 ng/ml. EGF was added at 100 ng/ml. PMA was used at 1 p~, endothelin was used at 10 nM, thrombin was added at 100 milliunits/ ml, and insulin was used at 150 milliunits/ml. Unless indicated, agonist incubations were for 10 min at 37 "C. For 3zP cell labeling, cells were switched to phosphate-free medium. Carrier-free (3ZP)orthophosphate (1 mCi/plate, 0.3 mCi/ml) was then added and the cells allowed to incubate for 2 h.
Assays-MAPKK was assayed as described elsewhere (18). Phosphatase 2A treatment of MAPKK was as previously described (18). e-Raf-1 kinase was assayed using 48-kDa bovine MAPKK as a substrate. c-Raf-1 was immunopurified from NIH 3T3 cells or EC4Al cells as before (18); dose response and time course experiments, however, employed a more dilute extract (two plates/ml) than that described (18) to prepare c-Raf-1 immunoprecipitates. c-Raf-1-cata-Iyzed phosphorylation of MAPKK was assayed in a volume of 76 gl containing 20 pl of (settled) c-Raf-1 beads, 20 ~1 of MAPKK preparation (-0.25-1 unit of MAPKK activity), and 16 p1 of Mg-ATP mix (50 mM MgCIz, 500 p~ [y3'P]ATP (4000-10,000 cpm/pmol)). The mixture was incubated for 30 min at 30 "C at which time the beads containing c-Raf-1 were removed by centrifugation. SDS-EDTA was added to the supernatant, and the proteins were resolved by SDS-PAGE. The 48-kDa MAPKK band was excised, and radioactivity was measured by liquid scintillation counting. All c-Raf-1 kinase assays were corrected for MAPKK autophosphorylation and for any contaminating MAPKK contributed by the c-Raf-1 immunoprecipitate, as described elsewhere (18). One unit of c-Raf-1 activity will transfer 1 pmol of POdmin from ATP to MAPKK. Where shown, c-Raf-1 and MAPKK assays were carried out in triplicate and are expressed as means 2 S.D.
Phosphorylation of c-Raf-1 by MAPK employed c-Raf-1 immunoprecipitated from control or PDGF-treated EC4A1 cells and treated with MAPKs (-0.33 unit) purified from insulin-stimulated H4 hepatoma cells (25) or vehicle plus [ Y -~~P ] A T P (100 pM, 11,000 cpm/ pmol) and MgClz (10 mM) for 30 min at 30 "C. The immunoprecipitates were washed twice with Raf-1 lysis buffer, twice with LiCl wash (181, and once with buffer A' (20 mM Mops, pH 7.2, 2 mM EGTA, 1 mM DTT, 0.1% Triton X-100,lO mM MgClZ). Any remaining MAPK was inactivated by incubating the immunoprecipitates with recombinant rat brain protein-tyrosine phosphatase-I (300 units/ml; see Ref. 26) for 20 min at 30 "C. The tyrosine phosphatase was washed away as above, except that the buffer A' contained 2 mM NaaV04. Samples were then assayed for c-Raf-1 activity as described above.
Tryptic Phosphopepetide Mapping and Phosphoamino Acid Analysis-MAPKK or c-Raf-1 were excised from dried stained gels and extracted by exhaustive incubation of the macerated, rehydrated gel piece in 1.5% SDS, 20 mM DTT at 55 "C. Proteins were precipitated with chloroform/methanol using phosphorylase b as a carrier. Digestion with 0.2 mg/ml L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin was for 16 h at 37 "C. Two-dimensional tryptic phosphopeptide mapping (thin layer electrophoresis, pH l.S/TLC) was as previously described (27). Phosphoamino acid analysis was performed as described elsewhere (27,28).

Purification of Bovine Brain MAPKK and Activation by C-
Raf-1-To study mitogen activation of c-Raf-1 and the role of c-Raf-1 in the activation of MAPKK by extracellular agonists, an abundant source of MAPKK was necessary for use as an in vitro substrate for assay of c-Raf-1 protein kinase activity. Inasmuch as MAPKs are highly abundant in brain (29), we examined soluble extracts of bovine brain for erkspecific activator/MAPKK activity. A spontaneously active MAPKK was readily detected in the flow-through fractions of a DEAE-cellulose column and was purified -200-fold (Table I)  Treatment of the purified brain MAPKK with phosphatase-2A failed to decrease activity, in contrast to the extensive phosphatase-2A inactivation observed previously using TABLE I Purification of MAPKK from bovine brain MAPKK was purified from cytosolic extracts of bovine brain as descrihed in "Experimental Procedures." Activity is listed in units which are defined in Refs. 16 and 18. Activity in the crude extract was not detectahle, and purification is calibrated from the DEAE steD. Step

67-
MAPKK purified from mitogen-treated cells (18). This suggests that the bovine brain MAPKK as isolated has kinase activity independent of its prior phosphorylation (or contains phosphatase-2A-resistant sites). Nevertheless, as observed previously with phosphatase-2A-inactivated MAPKK, the bovine brain MAPKK Mono-S fractions are activated by incubation with c-Raf-1 (Fig. 1B). Thus, c-Raf-1 immunoprecipitated from serum-starved NIH 3T3 cells increases MAPKK activity approximately 2-4-fold in an ATP-dependent reaction, whereas c-Raf-1 immunoprecipitated from PDGF-stimulated cells produces a much larger activation of brain bovine MAPKK, ranging from 5-to 20-fold. Concomitant with the ability of c-ruf-1 to increase MAPKK activity in vitro, c-Raf-1 catalyzes the selective phosphorylation of the coeluting 48-kDa band in the MAPKK preparation, further supporting the identification of the 48-kDa polypeptide as the MAPKK (Fig.  lB, inset). Activation of MAPKK by c-Raf-1 is reflected both by an enhanced ability of MAPKK to catalyze '*P incorporation from [-p3'P]ATP into the btMAPK polypeptide (Fig. 2B), as well as by the greater ability of MAPKK to activate btMAP kinase catalytic activity toward MBP ( Fig. 2A); the greater -fold increase in btMAPK phosphorylation (Fig. 2B) than in MAPK-catalyzed MBP phosphorylation probably reflects a MAPKK-induced increase in MAPK autophosphorylation. Phosphatase-2A treatment of the c-Raf-1-activated MAPKK reverses completely both measures of MAPKK activity, back to their basal level. A second incubation of the phosphatase-2A-treated MAPKK with c-Raf-1 and ATP can restore completely both aspects of MAPKK catalytic function (Fig. 2, A  and B ) .
In parallel to the activation of MAPKK catalytic function, c-Raf-1 catalyzes the phosphorylation of the bovine brain MAPKK (Fig. 2C). The stoichiometry of this phosphorylation is uncertain, primarily because the amount of MAPKK polypeptide is too low to be quantitated accurately; nevertheless, several MAPKK tryptic 32P-peptides are visualized on peptide mapping, and phosphoamino acid analysis of c-Raf-1-phosphorylated "'P-MAPKK shows both 32P-Ser and 32P-Thr phosphate, as before (18). Phosphorylation of MAPKK by c-Raf-1 is also accompanied by a slight slowing of the MAPKK polypeptide on SDS-PAGE, a phenomenon seen with many polypeptides whose activity is regulated by phosphorylation, presumably reflecting a phosphorylation-induced alteration in the SDS binding and/or conformation of the substrate. Treatment of c-Raf-1-activated MAPKK with phosphatase-2A reverses the slowed mobility concomitant with deactivating fully the c-Raf-1-induced increase in MAPKK activity, but results in the hydrolysis of only -50% of the "*P incorporated into MAPKK in the presence of c-Raf-1. A second incubation with c-Raf-1 and [y-"PIATP restores the 32P content of MAPKK to the level achieved by the first cycle of c-Raf-1-catalyzed phosphorylation, concomitant with a slowing of mobility in SDS-PAGE (Fig. 2C) and complete reactivation of the MAPKK activity (Fig. 2 A ) .
The phosphorylated MAPKKs shown in Fig. 2C were subjected to complete tryptic digestion and two-dimensional phosphopeptide mapping; a composite diagram of MAPKK tryptic phosphopeptides is seen in Fig. 3E. Basal autophosphorylation of MAPKK is insignificant (Fig. 2C) and was not analyzed. Phosphorylation of bovine brain MAPKK with c-Raf-1 results in the phosphorylation of seven 32P-peptides (Fig. 3, A and E ) ; spots 1-5, and spots X and Y; spots 1, X, and Y are quantitatively dominant. Treatment of MAPKK with phosphatase-2A is accompanied by the complete disappearance of "P-peptides X and Y with little or no change in the relative 32P content of spots 1-5. Thus, phosphorylation of one or both of the MAPKK 32P-peptides X and Y is necessary for c-Raf-1-induced MAPKK activation. Spot 1, which contains about 50% of the 32P incorporated into MAPKK during the initial phosphorylation by c-Raf-1, ac- counts for the bulk of the residual 32P-MAPKK after phosphatase-2A deactivation of MAPKK seen in Fig. 2C, lane 3.
Reactivation and rephosphorylation of MAPKK catalyzed by a second Raf-1 treatment (see Fig. 2, A-C) results in the selective reincorporation of 32P into spots X and Y (Fig. 3C) with a return in overall MAPKK 32P content to the level achieved in the initial c-Raf-1-catalyzed reaction (Fig. 2C). A mixture of samples from Fig. 3, B and C (Fig. 3 0 ) is identical to Fig. 3A, confirming spots X and Y as the dominant MAPKK 32P-peptides selectively dephosphorylated by phosphatase-2A concomitant with inactivation of MAPKK. Thus, mitogenstimulated c-Raf-1 catalyzes the phosphorylation of MAPKK at multiple sites that are both resistant and sensitive to phosphatase-2A; occupancy of one or both of the phosphatasesensitive-"P-peptides ( X and/or Y ) is essential for the persistence of MAPKK activation, whereas phosphorylation of the phosphatase-2A-resistant sites (especially spot 1 ) is not per se sufficient to maintain MAPKK activation.
c-Raf-1 Kinase, Assayed by Phosphorylation of MAPKK in Vitro, Is Agonist-stimulated-Based on the results shown in Figs. 1-3, the c-Raf-1-catalyzed phosphorylation of the 48-kDa brain MAPKK serves as a direct assay of c-Raf-1 kinase activity and was employed to characterize further the mitogen activation of c-Raf-1 kinase in situ. PDGF activation of c-Raf-1 was examined in NIH 3T3 cells, as well as a line of NIH 3T3 cells (EC4A1) that stably overexpress recombinant human c-Raf-1. Serum-starved cells were harvested 15 min after addition of PDGF or carrier; c-Raf-1 kinase activity in immunoprecipitates was increased 9-16-fold after PDGF with half-maximal activation observed at approximately 0.3-0.5 ng/ml PDGF (Figs. 4 and 6). The rate of activation of c-Raf-1 after addition of a supramaximal dose of PDGF to serumstarved NIH 3T3 cells was compared to the rate of activation of endogenous MAPKK (Fig. 5A) and PDGF-stimulated tyrosine phosphorylation (Fig. 5B). A doubling in c-Raf-1 kinase is evident as early as 1 min after PDGF addition, rising to a peak at 3 min, approximately 15-fold over basal levels. Endogenous 3T3 cell MAPKK activity also increases 10-15fold in response to PDGF, but more slowly; MAPKK is not yet half-maximal at 3 min and peaks at 10 min (Fig. 5A). Both c-Raf-1 and MAPKK activities are sustained for at least 30 min, whereas PDGF-stimulated Tyr phosphorylation (Fig.  5B), already maximal at 20 s after mitogen addition, declines rapidly after 3 min. In addition to PDGF, a variety of agonists whose receptors act through different signal transduction pathways, also activate c-Raf-1 in the NIH 3T3, EC4Al and H4 hepatoma cells (Fig. 6). Thrombin and endothelin, agents that act through sepentine receptors that contain seven transmembrane segments and are coupled through heterotrimeric G proteins, each activate c-Raf-1, as does the active phorbol ester, PMA. In H4 hepatoma cells, insulin activates c-Raf-1. Each of these agonist has been shown previously to activate MAP kinase activity.
Effect of Mitogens on Site-specific Phosphorylation of c-Raf-1 in Situ-Numerous studies have shown that the c-Raf-1 Addition of PDGF to 3T3 cells stimulates overall 3'P incorporation into c-Raf-1 and shifts a portion of the c-Raf-1 polypeptides to a slower mobility on SDS-PAGE (30) (see Fig. 8, inset). The ability of immunoprecipitated c-Raf-1 to catalyze autophosphorylation in vitro is also increased after PDGF treatment, and the c-Raf-1 polypeptides that undergo the most extensive autophosphorylation are primarily those that exhibit the most retarded mobility on SDS-PAGE (Fig.  8, inset). These findings suggest that these slowly migrating c-Raf-1 polypeptides are the most active c-Raf-1 kinase molecules. An analogous situation is seen with the S6 kinases (31,32).
The effect of PDGF on the site-specific phosphorylation of c-Raf-1 32P-labeled in situ was evaluated by two-dimensional peptide mapping of tryptic digests prepared from 3'P-c-Raf-1 (diagram of c-Raf-1 32P tryptic peptides is shown in Fig. 7A). The 3'P-c-Raf-1 isolated from serum-starved NIH 3T3 cells exhibits four major (spots [1][2][3][4] and four minor (spots [5][6][7][8] 3'P-peptides (Fig. 6B, top left). PDGF induced a 3-fold increase in overall 3'P incorporation into immunoprecipitated c-Raf-1 (basal, 548 cpm 32P; PDGF, 1518 cpm 32P); however, when equal counts/min 32P of 3'P-~-Raf-l tryptic digests from control and PDGF-treated cells were compared by peptide mapping, the relative 3' P content of spots 1-8 is seen to be unaltered by PDGF, and two additional minor 3'P-peptides appear de nouo, designated A and B (Fig. 7B, top right). A generally similar result is seen with c-Raf-1 obtained from 32P-labeled EC4A1 cells, before and after EGF treatment (Fig.  7C). Tryptic digests of overexpressed recombinant human 32P-~-Raf-l appear almost identical to those seen with endogenous murine c-Raf-1 (compare upper panets in Fig. 6, B and C), except that spot I is not detectable in the recombinant human c-Raf-1, whereas spot 7 is visualized as a major 32Ppeptide in murine c-Raf-1; thus, spots I and 7 may be structurally related. Although absolute recovery of 32P-~-Raf-1 from EC4A1 cells is greater due to c-Raf-1 overexpression, overall stimulation of 32P incorporation into c-Raf-1 by EGF is less (basal, 4000 cpm 32P; EGF, 6900 cprn 3'P); peptide maps of tryptic digests show no significant change in relative 32P content in spots 2-8 after EGF, with the de novo appearance of spots A and B (B > > A ) . Insulin stimulation of H4 cells also increases 3' P incorporation into the endogenous c-Raf-1 polypeptide (3'P-Ser only (Fig. 7E); control, 3600 cpm; insulin, 7100 cpm) and retards the mobility of the c-Raf-1 polypeptide on SDS-PAGE (Fig. 7 0 , left). Tryptic 3'P-peptide maps of endogenous rat c-Raf-1 exhibit a pattern strikingly similar to that seen with the murine and recombinant human c-Raf-1, including the presence of spots A and B (Fig. 70,

right). M A P Kinase Phosphorylation of c-Raf-1 in Vitro: Site Specificity and Effects on c-Raf-1 R i m e Activity-A Ser/Thr-rich
segment of the c-Raf-1 amino-terminal regulatory domain between residues 289 and 315 contains numerous potential MAPK phosphorylation sites (Ser/Thr Pro) (33-35). As previously observed (21, 22), p42/44 MAPKs phosphorylate c-Raf-1 in vitro (Fig. 7, B and C, bottom left).' The p42 MAPK appears to phosphorylate c-Raf-1 immunoprecipitated from serum-starved or PDGF-stimulated cells to a nearly equal extent, suggesting that activation of c-Raf-1 by PDGF does not lead to occupancy of a substantial fraction of the MAP kinase sites on c-Raf-1 (Fig. 8, inset). The sites on c-Raf-1 phosphorylated by MAPKs are amino-terminal to residue 302, inasmuch as neither v-Raf nor the amino-terminal truncated BXB-Raf (in frame deletion of residues 26-302) (36) are phosphorylated by p42 MAP kinase (not shown). The extent (i.e. stoichiometry), however, to which basal or PDGF-stimulated c-Raf-1 is phosphorylated by MAP kinase in vitro is unknown, because a reliable estimate of the amount of c-Raf-1 polypeptide in these reactions is not available. The maximal extent of MAP kinase-catalyzed phosphorylation of c-Raf-1 from serum-starved cells achieved in vitro, however, does not lead to slowing of the mobility of a portion of c-Raf polypeptides on SDS-PAGE, whereas obvious slowing of a substantial fraction of c-Raf-1 molecules is observed after PDGF (Fig. 8, inset) and insulin activation (Fig. 7 0 ) in situ. Examination of the sites phosphorylated by p42 MAP kinase on c-Raf-1, immunoprecipitated from serum-starved cells, by two-dimensional tryptic peptide maps indicates that MAPK catalyzes the phosphorylation of c-Raf-1 segments that comigrate with spots 2-8 seen in digests of 32P-c-Raf-1 phosphorylated in situ. Of the major "P-peptides seen in digests of c-Raf-1 labeled in situ (spots 1-4), MAPK appears to generate major "P-peptides corresponding in mobility to spots 2 and 4; in addition, 32P-peptide comigrating with spots 6, 7, and 8 are also generated by MAPK phosphorylation of c-Raf-1 in uitro, although these 32P-peptides are quantitatively minor spots after PDGF stimulation in situ. MAPK does not generate 3'Ppeptides that comigrate with spots A and B. Thus, MAPK phosphorylates c-Raf-1 on multiple sites, all of which appear to comigrate with c-Raf-1 32P-peptides labeled in situ. MAPK, however, phosphorylates only a subset of the c-Raf-1 sites labeled in situ; in particular, MAP kinase does not phosphorylate the sites on c-Raf-1 that appear entirely de nouo (i.e. spots A and B ) after PDGF, EGF, or insulin stimulation in situ.
The phosphorylation of c-Raf-1 catalyzed by MAP kinase in vitro does not alter the c-Raf-1 protein kinase activity, neither increasing it from the basal state nor inhibiting it after PDGF activation in situ (Fig. 8). Taken together with the failure of MAP kinase phosphorylation to slow c-Raf-1 mobility on SDS-PAGE (Fig. 8, inset) or alter c-Raf-1 autophosphorylation in vitro (not shown), it appears that phosphorylation of c-Raf-1 by MAP kinase, as it occurs in uitro, despite the overlapping site specificity with PDGF-stimulated c-Raf-1 phosphorylation in situ, is not capable of modifying c-Raf-1 activity.

DISCUSSION
The goal of this study was %fold: to characterize further the ability of c-Raf-1 to phosphorylate and activate MAPKK in vitro, to determine if this activity was part of the normal program of cellular signal transduction, and to evaluate the role of MAP kinase in the mitogen stimulation of c-Raf-1 phosphorylation and activation.
Bovine brain provides a convenient and abundant source of MAPKK. Although the enzyme cannot be assayed in crude extracts, a vigorous ATP-dependent MAPKK is evident in the breakthrough fraction of a DEAE column, and no such activity is observed in the fractions eluted from DEAE, behavior consistent with that reported for MAPKK characterized previously (10)(11)(12)(13)(14)(15)(16). Inasmuch as approximately 80% of total protein binds to the DEAE column, and assuming complete recovery of the MAPKK in the breakthrough, an overall purification of -1,000-fold gives MAPKK of approximately 30% purity, suggesting that this enzyme may constitute 0.03% of bovine brain soluble proteins (Table I). The identification of the 48-kDa polypeptide as the MAPKK is based on its coelution with the enzyme activity, and its selective phosphorylation by c-Raf-1 concomitant with the activation of MAPKK activity. This size of the bovine brain MAPKK polypeptide is close to the values reported previously for the various reported preparations of MAPKK (10-16), as well as the polypeptide encoded by the MAPKK cDNAs recently described (4-7).
The basal activity of the bovine brain MAPKK is resistant to Ser/Thr phosphatase; by contrast, the activity associated with the isolates of MAPKK described previously has exhib-ited over 90% deactivation on treatment in vitro with comparable amounts of Ser/Thr phosphatase (18). The phosphatase resistance of our preparation of bovine brain MAPKK may be attributable to any of several factors. First, the data in Table I, as well as numerous other studies indicate that MAPKs and, by implication, MAPKK, are highly abundant in brain (29), a nonproliferating tissue, where they are likely activated by neuronal signal transduction cascades possibly including those initiated at receptors coupled, via heterotrimeric G proteins, to protein kinase C. MAPKK, in a basal state, present at 0.03% of the total soluble protein, might very well be detectable using the assay conditions described herein. Alternatively, this MAPKK may be' a novel isoform with different regulatory properties. Bovine brain MAPKK may be phosphorylated at phosphatase-2A-resistant sites resulting in partial activation. Full activation may be achieved only after subsequent phosphorylation by c-Raf-1. Our data (Fig.  3) indicate that c-Raf-1 itself can phosphorylate MAPKK at phosphatase-2A-resistant sites. Thus, bovine brain MAPKK may be subject to complex regulation. There is good evidence that MAPKKs with radically different modes of regulation exist. Jaiswal et al. (37) recently reported the identification of a MAPKK present in PC12 cells which is regulated by Ser/ Thr and Tyr phosphorylation. Finally, our preparation of MAPKK may contain several copurifying isoforms, some of which are already active having been activated in situ solely by mechanisms independent of phosphorylation at PP2Asensitive sites. The remainder may be completely inactive requiring c-Raf-1 phosphorylation for activation.
Still, our preparation of bovine brain MAPKK is readily  ( p 4 2 / 4 4 ) . c-Raf-1 was immunoprecipitated and subjected to two-dimensional tryptic phosphopeptide mapping as described under "Experimental Procedures." For in vitro phosphorylations, human c-Raf-1 was immunoprecipitated from unlabeled, unstimulated EC4A1 cells. C, same as B except that EC4A1 cells were used. D, phosphorylation of c-Raf-1 in 32P-labeled H4 cells in response to insulin (left) and tryptic phosphopeptide map of H4 cell c-Raf-1 from insulin-stimulated cells (right). Cells were labeled with ["PI orthophosphate and treated with PDGF (NIH 3T3 cells), EGF (EC4A1 cells), or insulin (H4 hepatoma cells) as described under "Experimental Procedures." E, phosphoamino acid analysis of c-Raf-1 from insulin-stimulated cells. of MAPKK activated by phosphorylation in vitro with c-Raf-1, as observed previously with MAPKK from mitogen-treated cells (18) and skeletal muscle (20); bovine brain MAPKK then can be deactivated by phosphatase-2A and activated a second time by c-Raf-1. Bovine brain MAPKK is phosphorylated on multiple sites by c-Raf 1 (Fig. 3). Phosphorylation at the sites encompassed on 32P-peptides X and/or Y are necessary for MAPKK activation, whereas the role of the phosphatase-2Aresistant phosphorylation sites on spot 1, the other major site(s) of c-Raf-1-catalyzed 32P incorporation, is not yet known.
Measurement of 32P incorporation into brain MAPKK provides a sensitive assay of c-Raf-1 kinase activity and enables the demonstration of a far greater stimulation of c-Raf-1 kinase activity by PDGF treatment in situ than that indicated by previous work which employed model peptide or protein substrates (2, 3). In response to PDGF, clear-cut activation of c-Raf-1 kinase is evident at 1 min, preceding detectable activation of the endogenous MAPKK. This temporal pattern is consistent with an in vivo role for c-Raf-1 as a physiologic activator of MAPKK. Whether A-Raf, B-Raf, or Ser/Thr protein kinases other than ruf isoforms can serve as MAPKK activators is not yet known. While other MAPKK isoforms with different regulatory properties exist (37), our data strongly support the contention that c-Raf-1 is a major MAPKK activator in several signal transduction pathways. Whether the Tyr-phosphorylated MAPKK in PC-12 cells is in part regulated by c-Raf-1 remains to be determined. We have observed3 that expression in NIH 3T3 cells of dominant negative forms of c-ruf can prevent the activation of cotransfected Erk-1 engendered by activated ras, v-src, et^.,^ suggesting that c-Raf-1 is a necessary immediate upstream activator of MAPKK in response to those agents, at least in NIH 3T3 cells. It has been suggested that agonists acting through heterotrimeric G proteins may recruit a MAPKK activator other than c-Raf-1 (38,39); the present work shows that thrombin and endothelin do activate c-Raf-1 kinase (Fig. 6).
Interestingly, PDGF activation of c-Raf-1 and MAPKK is sustained for at least 30 min, whereas PDGF-stimulated Tyr phosphorylation declines rapidly after 3 min. This indicates that PDGF-induced Tyr phosphorylation, although indispensable for activation of c-Raf-1, is not necessary to sustain the activation of c-Raf-1. The biochemical mechanisms which account for the initiation and maintenance of c-Raf-1 activation after the addition of PDGF are not yet known. In addition to PDGF, many other growth factors, including EGF, colony-stimulating factor-1, and insulin, as well as cytokines, activation of antigen receptors, active phorbol esters, all act to increase c-Raf-1 Ser/Thr phosphorylation (2,3). Previous studies have demonstrated that phosphatase-1 specifically dephosphorylates c-Raf-1 from insulin-treated cells concomitant with a deactivation of c-Raf-1 kinase (40), indicating that Ser/Thr phosphorylation of c-Raf-1 is necessary to maintain the activated state. Nevertheless, the role of c-Raf-1 phosphorylation in initiating the active state of the c-Raf-1 protein kinase is uncertain, as is the identity of the protein kinase(s) that catalyze c-Raf-1 (Ser/Thr) phosphorylation in situ. c-Raf-1 catalyzes an autophosphorylation in vitro, and a contribution of this reaction to PDGF activation in situ has been suggested. The data of Izumi et ul. (30) indicate, however, that the pattern of site-specific c-Raf-1 phosphorylation seen after PDGF stimulation in situ is the same for recombinant wild-type c-Raf-1 as for a c-ruf-1 mutant whose ATP binding site has been mutationally inactivated. This suggests strongly U. R. Rapp, unpublished results. J. Troppmair and U. R. Rapp, manuscript in preparation. that c-Raf-1 autophosphorylation in situ contributes negligibly to overall c-Raf-1 phosphorylation, and that insulin/ growth factor-stimulated c-Raf-l phosphorylation is mediated almost entirely by other mitogen-activated Ser/Thr kinases. It is noteworthy that the tryptic 32P-phosphopeptide maps of c-Raf-1 isolated after PDGF, insulin (30), and EGF stimulation are very similar (Fig. 7 ) , suggesting that these three agonists, acting in two different cell types (H4 hepatoma and 3T3 cells), activate a common set of c-Raf-1 kinases.
Evidence for an important role for kinase C in the c-Raf-1 activation is available in some systems, such as T and B cells, where down-regulation of protein kinase C with TPA prevents activation of c-Raf-1 through the antigen receptor (2,3).
Baculoviral-encoded c-Raf-1 expressed in Sf9 cells can be activated by coinfection with baculoviral-encoded protein kinase Ca plus treatment with TPA (but not either alone). A mutant c-Raf-1 (Ser-499 + Ala) cannot be activated by coinfection with baculoviral protein kinase Ca plus TPA, but can still be activated if coinfected with baculoviral c-rus and lck in a triple infection. Kinase Ca can directly phosphorylate and activate purified c-Raf-1 in vitro; Raf-1 (Ser-449 + Ala), although phosphorylated is not activated. Thus under some circumstances, kinase C a may act directly on c-Raf-1; this is not, however, the mechanism underlying activation by insulin or PDGF in situ. Several studies have shown that downregulation of 3T3 cells with TPA does not prevent activation of endogenous c-Raf-1 by PDGF (2,3); moreover, the mutant c-ruf-1 (Ser-499 + Ala), although refractory to activation by kinase C-mediated phosphorylation in vitro is activated by serum, when expressed in NIH 3T3 cells. 5 With regard to the role of MAPK, the present data indicate that MAPK is capable of phosphorylating c-Raf-1 in vitro at multiple sites, all of which are located on 32P-peptides that comigrate with c-Raf-1 32P-peptides labeled in situ. This is consistent with earlier reports (21, 22) and provides significant evidence that MAPK, or a protein kinase with similar specificity, participates in the insulin, PDGF and EGF-stimulated (Ser/Thr) phosphorylation of c-Raf-1 that occurs in situ. Nevertheless, MAPK phosphorylation of c-Raf-1 does not phosphorylate spots A and B , the new 32P-peptides that appear only after mitogen activation in situ; does not generate a population of c-Raf-1 molecules with slowed mobility on SDS-PAGE, as is seen on activation in situ with PDGF; and neither increases nor decreases c-Raf-1 kinase activity toward MAPKK. Thus phosphorylation of c-Raf-1 in vitro catalyzed by MAPK alone does not reproduce the activation of c-Raf-1 generated by PDGF in situ.
Based on the peptide maps shown in Fig. 7 , it appears likely that at least one Ser/Thr kinase other than MAPK participates in the phosphorylation of c-Raf-1 (i.e. a spot A / B kinase); conceivably, c-Raf-1 phosphorylation by this second enzyme may mediate the initial activation of c-Raf-1 protein kinase. Alternatively, the initial activation of c-Raf-1 may be due to a noncovalent protein-protein, or ligand-protein interaction between the amino-terminal (CR-1) regulatory domain of c-Raf-1 and a mitogen-generated upstream activator, which alters c-Raf-1 conformation, and enables it to undergo multisite phosphorylation. By this scenario, the growth factor-stimulated phosphorylation of c-Raf-1 does not initiate c-Raf-1 activation, but may contribute to the persistent activation of c-Raf-1 observed after PDGF-stimulated, PDGF receptor Tyr phosphorylation has waned. Although the very similar pattern of c-Raf-1 (Ser/Thr) phosphorylation seen after insulin, EGF, or PDGF treatment argues for a common mode of c-Raf-1 activation by different extracellular W. Kolch and U. R. Rapp, manuscript in preparation. stimuli, the existence of multiple alternative mechanisms of c-Raf-1 activation remains a viable possibility. Such a scenario would reflect the central role of c-Raf-1 in growth control and the diverse upstream elements which must recruit c-Raf-1 to successfully initiate mitogenesis.