Identification of the 14.3.3 zeta domains important for self-association and Raf binding.

The 14.3.3 ζ protein is a ubiquitous and abundant arachidonate-selective acyltransferase and putative phospholipase A2, which self-assembles into dimers and binds to c-Raf-1 and other polypeptides in vitro and in intact cells. The 14.3.3 polypeptides endogenous to Sf9 cells associate in situ with both active and inactive recombinant Raf and copurify at a fairly reproducible molar ratio that is probably ≥1. Purified baculoviral recombinant Raf, despite its preassociated 14.3.3 polypeptide, binds additional recombinant 14.3.3 ζ polypeptide in vitro, in a saturable and specific reaction, forming a complex that is resistant to 1 M LiCl. A two-hybrid analysis indicates that 14.3.3 ζ binds primarily to Raf noncatalytic sequences distinct from those that bind Ras-GTP, and in vitro 14.3.3 ζ binds to Raf without inhibiting the Ras-Raf association or Raf-catalyzed MEK phosphorylation. Deletion analysis of 14.3.3 ζ (1-245) indicates that the 14.3.3 domain responsible for binding to Raf extends over the carboxyl-terminal 100 amino acids, whereas 14.3.3 dimerization is mediated by amino-terminal sequences. As with Ras, the 14.3.3 ζ polypeptide does not activate purified Raf directly in vitro. Moreover, expression of recombinant 14.3.3 ζ in COS cells beyond the substantial level of endogenous 14.3.3 protein does not alter endogenous Raf kinase, as judged by the activity of a cotransfected Erk-1 reporter. Coexpression of recombinant 14.3.3 with recombinant Myc-tagged Raf in COS cells does increase substantially the Myc-Raf kinase activity achieved during transient expression, which is attributable primarily to an increased level of Myc-Raf polypeptide, without alteration of Myc-Raf specific activity or the activation that occurs in response to epidermal growth factor or 12-O-tetradecanoylphorbol-13-acetate. Nevertheless, evidence that 14.3.3 actively participates in Raf activation in situ is provided by the finding that although full-length 14.3.3 ζ binds active Raf in situ, truncated versions of 14.3.3, some of which bind Raf polypeptide in situ nearly as well as full-length 14.3.3 ζ, are recovered in association only with inactive Raf polypeptides. Thus, 14.3.3 polypeptides bind tightly to one or more sites on c-Raf. Overexpression of 14.3.3 ζ enhances the expression of recombinant Raf, perhaps by stabilizing the Raf polypeptide. In addition, Raf polypeptides bound to truncated 14.3.3 polypeptides are unable to undergo activation in situ, indicating that 14.3.3 participates in the process of Raf activation by mechanisms that remain to be elucidated.

An important insight into the initial step in Raf activation was the discovery that Raf binds directly to the GTP-bound form of Ras (1)(2)(3). This Raf-Ras-GTP interaction does not result directly in Raf activation, inasmuch as addition of Ras GTP to inactive, baculoviral recombinant Raf in vitro does not alter Raf kinase activity. Presumably, Ras GTP functions in situ to translocate Raf to the surface membrane so as to enable its activation by other processes. Support for this model is provided by the demonstration that fusion of plasma membrane targeting (CAAX) sequences onto the Raf carboxyl terminus is transforming and bypasses the need for Ras in Raf activation; a large increase in the activity of membrane-associated Raf is observed in growth factor-deprived cells, and EGF stimulates Raf CAAX activity a further 10-fold in a Ras-independent reaction (4,5).
The inability of Ras to directly activate Raf, together with the finding that mitogen activation of Raf becomes Ras independent if Raf is targeted directly to the plasma membrane, implies that physiologic activation of (Ras bound) Raf requires Raf interaction with other plasma membrane components, e.g. lipids, polypeptides, or both. Ghosh et al. (6) reported that the Raf amino-terminal noncatalytic sequences bound to liposomes in a phosphatidylserine-dependent reaction that is independent of Ca 2ϩ and diacylglycerol. In this report, we describe the binding of c Raf-1 in vitro and in situ to the 14.3.3 polypeptide, an arachidonate-selective acyl transferase and putative phospholipase A2 (7). We define the Raf domain employed for the binding of 14.3.3 in situ and the 14.3.3 domains necessary for self-association and Raf binding; we find that while carboxyl-terminal fragments of 14.3.3 bind Raf in situ nearly as well as full-length 14.3.3, only the latter is found in association with catalytically active Raf polypeptides in situ.

MATERIALS AND METHODS
cDNAs encoding murine 14.3.3 were isolated from a murine T cell DNA library by two-hybrid expression cloning according to Durfee et al. (8), using the c-Raf sequences 1-25/305-648 as bait (see "Results"). cDNAs encoding rat Erk-1, human MEK-1, and human C-Ha-Ras were expressed in Escherichia coli as GST 1 fusion proteins using the p-GEX kg vector (9) and purified by glutathione-agarose affinity chromatogra-* This work was supported by an award from the American Cancer Society. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by a National Research Service Award from NCI, National Institutes of Health.
phy. The free Erk-1 and 14.3.3 polypeptides were obtained after thrombin cleavage. Recombinant Raf polypeptide containing a hexahistidine tag at the carboxyl terminus was expressed in Sf9 cells using a recombinant baculovirus and purified by nickel chelate affinity chromatography. Active baculoviral Raf kinase was obtained by co-infection with baculoviruses encoding v-Ras and v-Src (10).
The transacylation activity of the recombinant 14.3.3 was measured according to Zupan et al. (7). The Raf kinase assay was performed as previously described (11,12).
The binding in vitro of various polypeptides to GST or GST fusion proteins immobilized on glutathione-Sepharose was carried out at 30°C for 30 min in buffer A containing 25 mM Tris-Cl, pH 7.5, 50 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol; polypeptide concentrations are described in the figure legends. The beads were washed in excess binding buffer three times; the retained polypeptides were eluted directly into SDS-containing buffer, separated by SDS-PAGE, and analyzed by protein staining, immunoblot, or autoradiography as described.
The association of polypeptides in situ was assessed during transient expression in COS M7 cells and transfected by the DEAE-dextran method. The cDNA sequences encoding Raf were inserted into two mammalian expression vectors; Myc-Raf contains a 33-amino acid epitope from human c-Myc, known to be reactive with the monoclonal antibody 9B7.3 (13), appended to the Raf amino terminus, and inserted into pMT2. Raf was also expressed as a GST fusion protein using the vector pEBG, which encodes glutathione S-transferase driven by the EF1␣ promoter/enhancer. The cDNA encoding 14.3.3 was introduced unmodified into the vector CMV5, into the pEB vector (lacking the glutathione S-transferase sequences) with a 9-amino acid epitope from the influenza hemagglutinin (HA epitope, Ref. 14) added to its carboxyl terminus, and into pEBG for expression in situ as a GST fusion. Deletion mutation of the 14.3.3 was made by polymerase chain reaction from the 5Ј-and 3Ј-ends of the cDNA. The polymerase chain reaction products were subcloned into the pEBG vector, and the structures were verified by DNA sequence analysis.
All transfections utilized a total of 20 g of DNA; 48 h after transfection, extracts were prepared by homogenization in a buffer containing 25 mM Tris-Cl, pH 7.5, 1 mM EDTA, 2 mM EGTA, 1 mM dithiothreitol, 25 mM ␤-glycerophosphate, 1 mM sodium vanadate, 1% Triton X-100, and proteinase inhibitors. Immunoprecipitations were carried out for 1 h at 4°C using the monoclonal antibody 12CA5 for the HA epitope or the anti-Myc monoclonal antibody 9B7; immune complexes were harvested with protein G-Sepharose. GST-Raf and GST-14.3.3 fusions were recovered using glutathione-Sepharose beads. Immunoblots of Raf were carried out using 9B7.3 for Myc-Raf or a polyclonal antipeptide antibody raised to the carboxyl-terminal 12 amino acids of human c-Raf-1. Immunoblot of 14.3.3 was carried out using a polyclonal antibody raised to cleaved, purified recombinant murine 14.3.3 .

RESULTS
Seeking proteins other than Ras and related small GTPases that interact with Raf, we utilized the Raf sequences corresponding to 1-25/305-648 in a two-hybrid screening (8). This truncated Raf polypeptide, also known as BXB-Raf (15), is a transforming protein that encodes a dispensable region of the Raf amino terminus (residues 1-25) fused to the Raf carboxylterminal 344 amino acids, which includes the entire catalytic domain (residues 335-627) flanked by short segments of amino-terminal (residues 305-334) and carboxyl-terminal (residues 628 -648) noncatalytic sequences. 20 His ϩ and LacZ ϩ cDNAs were recovered in a screen of 2.5 ϫ 10 6 transformants. Five of these cDNAs gave much more rapid complementation of LacZ ϩ activity than the remaining cDNAs; the former all encoded MEK1 or MEK2. Among the remaining 15 cDNAs were 5 that encoded polypeptide sequences 100% identical to the rat isoform of the 14.3.3 polypeptide, and that differed from a human platelet PLA 2 polypeptide by a single conservative substitution (7). Inasmuch as 14.3.3 proteins have been reported to copurify or associate with a relatively large number of proteins (16), and 14.3.3 has been identified as a protein cofactor for a number of enzymes in vitro, including the ADP-ribosylation of Ras and other small GTPases by exoenzyme S (17), we exam-ined several proteins other than Raf for their ability to associate with 14.3.3 in the two-hybrid system. No interaction of 14.3.3 with p70 S6 kinase and amino-terminal regulatory domain of protein kinase C (residues 1-245) or c-Ha-Ras (1-185) was detected (data not shown). The relative selectivity of the 14.3.3-Raf association led us to undertake a further characterization of this interaction.
The two-hybrid method was employed to identify the region on Raf that interacts with 14.3.3 in comparison to MEK1 and Ras, proteins known to interact with Raf in a physiologic context. The Ras binding site has previously been localized to Raf residues 50 -150 (3,18), whereas neither MEK nor 14.3.3 interacts with Raf 1-257. MEK, a known Raf substrate, interacts strongly with the BXB-Raf and holo-Raf (1-648) but not at all with Raf 1-332 (Table I). By contrast, 14.3.3 interacts weakly with BXB-Raf and holo-Raf (1-648) but associates strongly with the Raf 1-332 (Table I). Thus, 14.3.3 binds in situ most avidly to a segment of Raf between amino acids 257 and 332, a noncatalytic region distinct from those that bind to Ras or MEK.
Binding of Raf to 14.3.3 in Vitro-The ability of recombinant Raf to interact directly with recombinant 14.3.3 was investigated. The 14.3.3 polypeptide was expressed in E. coli as a glutathione S-transferase fusion protein (9), purified by GSH affinity chromatography, and employed with and without thrombin cleavage (Fig. 1). The functional integrity of the recombinant 14. Binding of Raf to 14.3.3 in Situ-The interaction of recombinant 14.3.3 with Raf in situ was examined further by cotransfection (Fig. 4). An epitope-tagged recombinant 14.3.3 polypeptide was created by fusion of a 9-amino acid epitope from the influenza hemagglutinin to either the amino terminus or carboxyl terminus, enabling the selective immunoprecipitation of the tagged 14.3.3 polypeptide with anti-HA monoclonal antibody 12CA5 (14). After transfection of either tagged version of 14.3.3 , the anti-HA immunoprecipitates contain, in addition to a 30-kDa band of HA-tagged recombinant 14.3.3 , a 28-kDa polypeptide that is not reactive with the HA antibody but is reactive with the anti-14.3.3 antiserum (data not shown). Thus, the recombinant HA-tagged 14.3.3 forms complexes (presumably dimers) with endogenous 14.3.3 polypeptides, and free amino or carboxyl termini are not crucial for dimerization.
Two tagged versions of recombinant Raf-1 were constructed by introduction at the Raf amino terminus of a Myc epitope (13) or by fusion to GST. After cotransfection of HA-tagged 14.3.3 with either version of Raf into COS cells, anti-HA 14.3.3 immunoprecipitates (Fig. 4, lanes 3, 4, 7, 8) contain substantial immunoreactive Raf-1, ranging from 30 to 100% of the amount of Raf polypeptide recovered from an identical aliquot of the COS cell extract by anti-Myc immunoprecipitation (Fig. 4, lanes 5 and 6) or by binding to GSH-Sepharose ( Fig. 4 lanes 1  and 2).  (Fig. 5B, lower panel). These results indicate that although a free NH 2 terminus is not critical for 14.    (Fig. 5B, lower panel, lane 8).
As to the effects of truncation on the ability of GST-14.3.3 to bind cotransfected Myc-Raf, the 14.3.3 amino-terminal 139 residues are largely dispensable; optimal Myc-Raf recovery is observed with GST-14.3.3 (139 -245) (Fig. 5C, middle panel, lanes  2, 7), and considerable Myc-Raf is recovered with GST-14.3.3 (179 -245) (Fig. 5C, middle panel, lane 8), whereas no binding of Myc-Raf occurs to GST-14.3.3 (1-140) (Fig. 5C, middle panel,  lane 4). Each of these GST-14.3.3 isolates was also assayed for the presence of active Raf kinase, estimated by its ability to phosphorylate and activate GST-MEK (Fig. 5C, bottom panel). Notably, only the GST-14.  (Fig. 6, upper panel, lanes 1 and 2). In addition, after washing the immobilized GST-Ras-Raf complex, 14.3.3 is seen to have been retained by the GST-Ras, but only if Raf is present (Fig. 6, lower panel, compare lanes 1 and 2). These data demonstrate that 14.3.3 does not bind to Ras directly nor displace Raf from Ras, but rather it is capable of binding Raf in vitro so as to allow the formation of a ternary complex with Ras and Raf.
Addition  (Fig. 7A, bottom panel, compare lane 1 to   FIG. 6. Raf binds 14.3.3  lane 2) and in response to EGF (Fig. 7A, bottom panel, compare lane 5 to lane 6); EGF itself gave 2-3-fold activation of Myc-Raf (Fig. 7A, bottom panel, compare lane 1 to lane 5). The ability of GST-14.3.3 to increase Myc-Raf activity is not due to Raf activation, as occurs with EGF or TPA, but appears to be attributable to an increased Myc-Raf polypeptide abundance, as seen by immunoblot of the whole cell extract (Fig. 7A, 2nd panel  from top, compare lanes 1 to 2, 5 to 6, 9 to 10) and in the Myc immunoprecipitates (Fig. 7A, 3rd panel from top). Such an increase in Myc-Raf expression and recovery of activity was observed repeatedly on cotransfection with GST-14.3.3  and was present but less pronounced with the truncated 14.3.3 GST fusion proteins that were capable of binding Myc-Raf in proportion to their somewhat lesser expression than GST-14.3.3 (1-245) (Fig. 7A, top panel and 2nd panel from top, compare  lanes 2 to 3, 6 to 7, 10 to 11).
Based on the finding that GST-14.3.3 (139-245) bound Myc-Raf strongly but did not associate in situ with active Raf (Fig. 5C

DISCUSSION
The 14.3.3 class of 29 -33-kDa polypeptides has been found to copurify with a broad array of proteins and has been repeatedly rediscovered as activators (e.g. of tyrosine hydroxylase), inhibitors (e.g. of protein kinase C), or cofactors (exoS) for a number of enzymes in vitro (16). Several recent reports (19,20) have described an interaction of Raf with 14.3.3 polypeptides in vitro and in intact yeast similar to that described here. All studies of the Raf 14.3.3 interaction including the present report have employed one or both as recombinant polypeptides, and it is possible that the interaction observed in vitro, or even in situ examining the overexpressed recombinant polypeptides,   (21) found that microinjection or overexpression of 14.3.3 in Xenopus oocytes led to an increase in the activity of endogenous or recombinant Raf-1. Li et al. (22) reported that transient expression of 14.3.3 in NIH 3T3 cells had little effect on the activity of cotransfected reporters known to be responsive to active Raf-1; however, the ability of transiently expressed Raf-1 or BXB Raf-1 to activate these reporters was substantially augmented by cotransfection with 14.3.3. We find that although overexpression of 14.3.3 in cultured mammalian cells (COS or 293) has no effect on endogenous c-Raf-1 abundance or activity, measured directly or by the activity of a cotransfected Erk-1 reporter (not shown), coexpression of 14.3.3 with recombinant Myc-Raf results in greater Raf kinase activity (Fig. 7), much as found by Li et al. (22) Significantly however, we find that this increase in Raf kinase activity is largely or entirely attributable to a 14.3.3-induced increase in the expression and abundance of recombinant Raf rather than an increase in Raf-1 specific activity. It seems probable that the ability of recombinant mammalian or endogenous yeast 14.3.3 to enhance the activity of mammalian Raf-1 when the latter is expressed in the heterologous milieu of S. cerevisae or Xenopus oocytes may be attributable, in part or in whole, to an ability of 14.3.3 to enhance Raf-1 polypeptide abundance, perhaps e.g. by stabilizing the recombinant Raf-1 polypeptide.
Several observations nevertheless suggest that 14.3.3 may participate more directly in Raf-1 activation. Irie et al. (20) observed that addition of a maltose binding protein-14.3.3 fusion protein directly to mammalian c-Raf-1, immunoprecipitated from yeast extracts, gave a 3-4-fold increase in MEK kinase activity. Takai and colleagues (23) have purified a MEK activator (REKS) from Xenopus oocyte cytosol that is activated in vitro approximately 2-fold by direct addition of GTP Ras and about 1.3-fold by 14.3.3 protein; addition of 14.3.3 in the presence of maximal GTP Ras gives a further 2-fold activation in REKS activity. Li et al. (22) observed that although recombinant 14.3.3 and Ras added directly to inactive baculoviral Raf do not alter Raf kinase activity, the further addition of a crude cellular extract results in a Ras-14.3.3-dependent, 2-3-fold activation of Raf-1 kinase. The biochemical mechanisms that underlie in vitro "activation" of Raf by 14.3.3 are not known, and the extent to which they reflect the ability of 14.3.3 to "stabilize" the Raf polypeptide in vitro, comparable to the effects that underlie the enhanced expression seen on cotransfection, is also not known. The present results, however, provide one persuasive piece of evidence that 14.3.3 participates in Raf activation beyond its ability to bind to and "stabilize" the Raf polypeptide. Carboxyl-terminal fragments of 14.3.3, which bind Raf nearly as well in situ as full-length 14.3.3 and which also provide some enhanced Raf expression in situ, are nevertheless recovered from cells in association only with catalytically inactive Raf polypeptides, whereas full-length 14.3.3 is recovered with catalytically active Raf kinase. This result suggests that the Raf-binding 14.3.3 fragments have lost a function critical to the activation of Raf. This function does not appear to be their ability to dimerize, inasmuch as the Raf polypeptides associated with GST-14. 3.3 (1-180), which dimerizes normally (Fig. 5B, lane 5), are not catalytically active (Fig.   5C, lane 5). We suggest that full-length 14.3.3 contributes to Raf activation either by recruiting an as yet unidentified polypeptide, by providing an intrinsic catalytic function, or both. The present data show that in addition to its ability to bind Raf concomitantly with Ras, the recombinant 14.3.3 polypeptide binds phospholipid and cleaves the sn-2 acyl bond. Whether the acyl transferase function of 14.3.3 is contributory to its role in the regulation of Raf function is not known. Thus, 14.3.3 polypeptides participate in the regulation of Raf activity; however, their specific biochemical function in Raf activation remains to be elucidated.
Subsequent to the completion of these studies, the structures of 14.3.3 (24) and (25) crystals were reported. Both 14.3.3 isoforms exhibited dimeric structures; each monomer was composed of nine helical segments arranged in antiparallel arrays. The dimer interface is created by highly conserved, primarily hydrophobic residues from the four amino-terminal helices, suggesting that 14.3.3 heterodimers will form readily. The carboxyl-terminal five helices of each monomer (which provide the Raf binding domain as demonstrated in the present report) are folded so as to provide within the dimer a cavity whose internal face is composed of (primarily hydrophilic) residues that are highly conserved in all 14.3.3 isoforms and whose external surface is provided by nonconserved residues. Future studies will determine the contributions of specific residues on each of these surfaces to the interactions of 14.3.3 with its polypeptide partners and phospholipids.