Raf-1 kinase possesses distinct binding domains for phosphatidylserine and phosphatidic acid. Phosphatidic acid regulates the translocation of Raf-1 in 12-O-tetradecanoylphorbol-13-acetate-stimulated Madin-Darby canine kidney cells.

Previous studies demonstrated that the cysteine-rich amino-terminal domain of Raf-1 kinase interacts selectively with phosphatidylserine (Ghosh, S., Xie, W. Q., Quest, A. F. G., Mabrouk, G. M., Strum, J. C., and Bell, R. M. (1994) J. Biol. Chem. 269, 10000-10007). Further analysis showed that full-length Raf-1 bound to both phosphatidylserine and phosphatidic acid (PA). Specifically, a carboxyl-terminal domain of Raf-1 kinase (RafC; residues 295 648 of human Raf-1) interacted strongly with phosphatidic acid. The binding of RafC to PA displayed positive cooperativity with Hill numbers between 3.3 and 6.2; the apparent Kd ranged from 4.9 +/- 0.6 to 7.8 +/- 0.9 mol % PA. The interaction of RafC with PA displayed a pH dependence distinct from the interaction between the cysteine-rich domain of Raf-1 and PA. Also, the RafC-PA interaction was unaffected at high ionic strength. Of all the lipids tested, only PA and cardiolipin exhibited high affinity binding; other acidic lipids were either ineffective or weakly effective. By deletion mutagenesis, the PA binding site within RafC was narrowed down to a 35-amino acid segment between residues 389 and 423. RafC did not bind phosphatidyl alcohols; also, inhibition of PA formation in Madin-Darby canine kidney cells by treatment with 1% ethanol significantly reduced the translocation of Raf-1 from the cytosol to the membrane following stimulation with 12-O-tetradecanoylphorbol-13-acetate. These results suggest a potential role of the lipid second messenger, PA, in the regulation of translocation and subsequent activation of Raf-1 in vivo.

The protooncogene product Raf-1 is a ubiquitously expressed Ser/Thr kinase which plays an essential role in signal transduction from a large number of growth factor receptors, cytokine receptors, several membrane-bound oncogenes, and certain mitogenic peptides with G-protein-linked receptors (1)(2)(3). Truncation of an amino-terminal fragment (first 305 amino acids) renders Raf-1 highly oncogenic (4). Based on in vitro direct binding assays and immunoprecipitation assays from cultured cells, the amino terminus of Raf-1 is now known to associate directly with another protooncogene product, p21Ras, and the regions on Raf-1 critical for high affinity interaction with p21Ras have been identified (5)(6)(7)(8)(9)(10). This association between Raf-1 and p21Ras appears to be an essential prerequisite for the normal pathway of Raf-1 activation. As a result, the relative binding affinity of specific effector domain mutants of p21Ras protein for Raf-1 is directly correlated with their ability to transform murine fibroblasts (7). Conversely, disruption of Raf-1:Ras interaction by a point mutation (R89L) in human Raf-1 prevents activation of Raf-1 kinase in Sf-9 insect cells (11) and also affects late zygotic functions of the Drosophila D-Raf protein (12).
However, a simple association between Raf-1 and p21Ras is not sufficient for the activation of Raf-1 kinase. It appears that both p21Ras and Raf-1 require to be present at the plasma membrane for expression of their biological activities. For example, the transforming activity of an oncogenic variant of the p21ras (K-ras 4B), is dependent on its membrane localization (13). On the other hand, the requirement for p21Ras in Raf-1 activation is overcome by targeting Raf-1 to the plasma membrane by addition of a membrane localization signal to Raf-1 (14,15). These results suggest that the function of p21Ras, at least for Raf-1 activation, is to recruit inactive, cytosolic Raf-1 kinase to the plasma membrane where additional protein-protein and/or protein-lipid interactions 1 may regulate the activation of Raf-1.
The inner leaflet of the plasma membrane of eukaryotic cells is composed of ϳ30% acidic and ϳ70% zwitterionic phospholipids (16). A growing body of evidence suggests that several proteins enhance their plasma membrane association by binding to the membrane lipids through electrostatic as well as hydrophobic interactions (37). Some examples include protein kinase C (17,18), the myristoylated, alanine-rich protein kinase C substrate, MARCKS (19), the human immunodeficiency virus Gag protein (20), and v-and c-Src (21). For protein kinase C, a zinc-containing cysteine-rich domain is believed to play a key role in the activation of the enzyme by binding to the acidic phospholipid, phosphatidylserine, and to diacylglycerol (22). A similar zinc-coordinating cysteine-rich region exists in the amino terminus of Raf-1 and is capable of binding to PS 2 * The work was supported in part by National Institutes of Health Grants CA-48995 and CA-43297 (to L. D.) and GM-38737 and DD-20205 (to R. M. B.). 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 National Institutes of Health Signal Transduction Mechanisms and Cell Functions Training Program (CA-09422).
In addition to the cysteine-rich domain, a detailed analysis of Raf-1-phospholipid interaction revealed a second site for association with acidic phospholipids, notably phosphatidic acid, in the carboxyl-terminal domain of the protein. In the present article, we have identified, by deletion mutagenesis, a 35-amino acid fragment of human Raf-1 kinase that is capable of interacting with PA. This finding suggests that, besides binding to p21Ras, Raf-1 also associates with plasma membrane lipids through multiple domains within the amino-and carboxyl-terminal regions of the protein, resulting in a firm anchor. One might postulate that these protein-lipid interactions result in a conformational switch in Raf-1 from an inactive to an activable state. Consistent with this postulate, we further show that an inhibition of the generation of phosphatidic acid in Madin-Darby canine kidney (MDCK) cells results in a decrease in translocation of Raf-1 kinase from the cytosol to the membrane following agonist stimulation.

Chemicals
Ninety-six-well microtiter plates (ProBind variety) were obtained from Becton Dickinson, Franklin Lakes, NJ. Glutathione-agarose matrix (sulfur-linked), phosphate buffered saline (PBS) and bovine serum albumin (fraction V) were purchased from Sigma. All phospholipids tested were purchased from Avanti Polar Lipids with the exception of bis-phosphatidic acid, cardiolipin, and bovine brain phosphatidylserine which were obtained from Serdary. All other chemicals used were of the highest available commercial grade.

Experimental Methods
Expression and Purification of GST Fusion Proteins-All the GST fusion proteins used in this study were obtained from DNA generated by PCR using a Flag-Raf plasmid template (a gift from Dr. Roger Davis, University of Massachusetts, Worcester, MA). The details of the PCR, cloning, expression, and purification of the fusion proteins have been described previously (23,24). The following primers were used for synthesis of the different Raf-1 fragments: Raf   .In all cases, the first sequence represents the upstream PCR primer followed by the downstream primer. A fragment of Raf-1 containing a mutation in the ATP binding site (K375M) was generated by PCR from a template encoding an ATP-binding mutant of Raf-1 (a gift from Dr. Deborah Morrison, National Cancer Institute, Frederick, MD). Following purification over glutathione-agarose columns, the fusion proteins were analyzed by SDS-polyacrylamide gel electrophoresis and stored in aliquots at Ϫ80°C. All fusion proteins were expressed in the Escherichia coli strain BL-21 (Novagen).

Assays for Measuring Raf-1:Lipid Interactions
ELISA Format Assay-All procedures were performed at room temperature. Phospholipid solutions in chloroform were diluted in methanol and loaded onto the wells of a 96-well titer plate (1-10 g lipid for each well). The lipids were allowed to bind to the plate overnight. The remaining binding sites on the wells were blocked with 3% bovine serum albumin in PBS for 1 h. GST fusion proteins, encoding different domains of Raf-1, were diluted in dilution buffer (0.3% bovine serum albumin in PBS), added to the wells, and incubated for 1 h. The plate was then washed with PBS containing 0.1% Tween-20 for 5 min each, for a total of three washes. Anti-GST rabbit polyclonal antiserum (1: 2000 dilution in dilution buffer) was added to the wells and allowed to react for 1 h. Excess primary antibody was removed by washing as before, and the plate was subsequently incubated with goat anti-rabbit IgG-alkaline phosphatase conjugate (Sigma) diluted 1:2000 in dilution buffer for 1 h. After washing off excess secondary antibody, a 10 mg/ml solution of 4-nitrophenyl phosphate in buffer (10 mM Tris-HCl, pH 9.5, 100 mM NaCl, 5 mM MgCl 2 ) was added to the wells for color development. The reaction was terminated by the addition of 33 mM disodium-EDTA (final concentration). The resulting absorbancies were quantitated at 405 nm on an ELISA plate reader (Bio-Rad). The magnitude of the absorbance was directly related to the amount of GST fusion protein bound. By using the GST fusion proteins in serial dilution, a set of binding curves were generated by plotting absorbance against GST fusion protein concentration. For quantitative estimates of association, the binding curves were fit to the modified Hill equation y ϭ a ϩ b(x n /(k n ϩ x n )) (25), where y is the measured binding of RafC, a is the background binding, b is the maximal phospholipid stimulated binding, x is the concentration of PA, k is the concentration of PA resulting in half-maximal binding, and n is the Hill coefficient.
Liposome Association Assay-Experiments to determine the liposome association of GST fusion proteins were based on previously established protocols described elsewhere (26,27).
PA and Phosphatidylethanol (PEt) Formation in MDCK Cells-MDCK cells were grown at a concentration of 5 ϫ 10 5 cells/ml overnight at 37°C. Media were removed and replaced with 1% serum, DMEM containing 15 mCi/ml [ 3 H]-20:4-arachidonic acid for an additional 24 h. Unincorporated radiolabel was removed by washing the cells in serumfree DMEM. Cells were subsequently stimulated with 10 nM TPA with varying concentrations of ethanol for 15 min. The reaction was stopped by scraping the cells directly from the plates into 2 ml of methanol, 2% acetic acid. Cellular lipids were extracted by a modification of the procedure of Bligh and Dyer (28) as described previously (29). Extracted lipids were dried under nitrogen and resuspended in chloroform/methanol (9:1, v/v). PA and PEt were resolved on Silica Gel 60 plates (EM Science, Merck) developed in ethyl acetate/isooctane/acetic acid/water (80/50/20/100, v/v). Lipids were located by autoradiography using EN-HANCE spray (DuPont NEN) and Kodak SB 5 film. Lipids were scraped from the plates and further quantitated by scintillation counting.
Translocation of Raf-1 and PKC-␣ in MDCK Cells-MDCK cells were grown for 2-3 days (or until 80% confluent) at 37°C. Cells were then incubated with serum-free DMEM overnight at 37°C prior to stimulation. Cells were stimulated with 10 nM TPA in the presence of varying concentrations of ethanol (0 -2%). After stimulation, cells were washed with cold PBS and scraped into buffer containing 10 mM Hepes, pH 7.4, 2 mM EDTA, 1 mM Na 3 VO 4 , and 1 mM phenylmethylsulfonyl fluoride. Cells were pelleted and lysed by sonicating in 500 l of buffer (above) plus 50 mM NaF, 10 g/ml aprotinin, and 10 g/ml leupeptin (three 15-s bursts). Unbroken cells were pelleted (6,000 rpm, 2 min, 4°C), and the remaining supernatant was recentrifuged at 36,000 rpm for 80 min at 4°C. Membrane pellets were solubilized by sonicating as described previously in buffer (above) plus 50 mM NaF, 10 g/ml aprotinin, 10 g/ml leupeptin, 100 mM NaCl, and 1% Triton X-100 and placed on ice for an additional 30 min. Membrane sonicates were centrifuged as before at 36,000 rpm for 80 min at 4°C. One hundred micrograms of total protein from cytosol and membrane were separated by SDSpolyacrylamide gel electrophoresis (5% stacking gel, 7% running gel) according to Laemmli (30) and transferred to a nitrocellulose membrane (Schleicher & Schuell) at 100 mA, at room temperature overnight. Membranes were blocked with PBS containing 5% non-fat dry milk and 1% Tween-20 for 1 h and then incubated with polyclonal rabbit anti-Raf-1 antibody (C12) for an additional hour. After washing, the membranes were incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG for 1 h. Raf-1 was detected using enhanced chemiluminescence (DuPont NEN) according to the instructions of the manufacturer. Anti-Raf-1 antibody was removed by stripping the membrane in buffer containing 62.5 mM Tris, pH 6.8, 2% SDS, and 100 mM 2-mercaptoethanol for 30 min at 50°C. The membranes were then reprobed for PKC-␣ using an anti-mouse PKC-␣ monoclonal antibody as the primary antibody and horseradish peroxidase-conjugated goat antimouse IgG for detection via enhanced chemiluminescence.

RESULTS AND DISCUSSION
Expression of GST-Raf-1 Fusion Proteins-Fusion proteins consisting of different fragments of human Raf-1 attached to GST were expressed in E. coli BL-21 strain using the pGEX 4T-2 vector, with the exception of full-length Raf-1 (RafFull), cysteine-rich domain of Raf-1 (RafCys), and the carboxyl-terminal fragment of Raf-1 (RafC), which were expressed from the pGEX-2T vector. The details of the expression protocol employed have been described (23). The quality of the expressed fusion protein was routinely monitored by SDS-PAGE followed by staining with Coomassie Brilliant Blue. A schematic diagram of the different GST-Raf-1 fusion proteins used in this study is shown in Fig. 1. The level of purity of the fusion proteins, assessed by scanning densitometry of the stained gels, was typically greater than 80%.
Binding of a COOH-terminal Fragment of Raf-1 to PA-In a previous article (23), we reported that residues 128 -196 of human Raf-1 kinase, containing the cysteine-rich domain, was capable of translocating to PS-enriched liposomes in vitro. Specifically, three GST-Raf-1 fusion proteins which contained the cysteine-rich domain, either wholly or in part, showed significant lipid-dependent association with PS-enriched liposomes (40 -50%). Full-length Raf-1 showed a lower level of PS-dependent association (15%), probably owing to steric constraints imposed on the cysteine-rich domain by additional structural elements in Raf-1. The interaction of the cysteine-rich domain of Raf-1 with PS is remarkably similar to that observed for the cysteine-rich domain of protein kinase C-␥. Since PKC-␥ also interacts with phosphatidic acid and diacylglycerol (27,31), we investigated whether the same was true for Raf-1 kinase. Diacylglycerol and PA were generated in situ from phosphatidylcholine coated on the wells of a microtiter plate by treatment with either phospholipase C (Bacillus cereus) or phospholipase D (cabbage), respectively. The resulting lipid products were incubated with different fragments of Raf-1 and screened for binding by the ELISA format assay. No binding was observed for either untreated PC or for diacylglycerol generated from PC treated with phospholipase C. However, RafC (residues 295-648 of human Raf-1) was found to bind to the PA generated from PC treated with phospholipase D (Fig. 2A). This suggested that besides the cysteine-rich domain, a second lipid-binding site exists within the COOH-terminal domain of Raf-1 kinase. To determine if this site was the preferred locus for interaction with PA, pure phosphatidic acid was coated on microtiter plates and incubated with GST fusions of RafFull, RafCys, and RafC. Both RafFull and RafC interacted strongly with PA, but the binding of RafCys to PA was considerably weaker (Fig. 2B). Thus, the cysteine-rich domain is likely to be the preferred PS binding site while a distinct PA binding site exists within the carboxyl-terminal domain of Raf-1.
Characterization of RafC-PA Interaction-The RafC-PA interaction was further characterized with respect to several different biochemical parameters. First, we investigated the nature of the association between RafC and PA. RafC (150 nM) was incubated with varying concentrations of dioleoyl PA (0 -30 mol %) coated on the wells of microtiter plates in the presence of dipalmitoyl-PC such that the total lipid content in each well was invariant at 10 g (Fig. 3A). In a separate experiment (Fig. 3A, inset), different concentrations of RafC (150 -9.4 nM) were tested against 0 -50 mol % PA. At all concentrations of RafC tested, a significant increase in binding between 5 and 10 mol % PA was observed. The data was plotted and fit to the Hill equation for receptor-ligand binding (31). From the sigmoidal fit, the half maximal binding affinity (apparent K d ) of RafC for PA was estimated to vary between 4.9 Ϯ 0.6 and 7.8 Ϯ 0.9 mol % PA. This value is comparable to that observed for the interaction between protein kinase C and PS which was estimated between 5.5 Ϯ 1.1 and 6.1 Ϯ 0.6 mol % PS (31). From the RafC-PA binding plots, a Hill number between 3.3 and 6.2 was obtained, suggestive of positive cooperativity. Again, this value is similar to that obtained for PKC-PS interaction (Hill number between 4 and 6) (31, 32). This observed cooperativity may reflect a cooperative sequestering of a domain of PA around RafC. Alternatively, it is also possible that the apparent cooperativity arises from a combined effect of electrostatics and reduction in dimensionality when RafC binds to the PA coated on the solid surface, in which case the interaction could be suitably described by the Gouy-Chapman-Stern theory of the diffuse double layer (33).
The sigmoidal nature of the association of RafC with PA was also confirmed by liposome association experiments. PA/PC liposomes containing varying mole fractions of PA were prepared as described previously (22) and incubated with RafC. The amount of RafC associated with the liposomes was quantitated as a function of the concentration of PA. Apparent association, in the absence of PA was taken as background (RafC does not bind significantly to PC as discussed later). The results of the liposome association experiment are summarized in Fig. 3B. Notably, an enhancement in RafC-PA association was observed between 10 and 20 mol % PA which is comparable to the results obtained in the ELISA format assay. The quantitative differences that exist between the ELISA format assay and the liposome association assay regarding the mole percent PA required for maximal Raf-1 association arise from two effects. First, in the ELISA format assay, the phospholipids are immobilized on the microtiter plates, thus providing a high, invariant, local concentration of PA which results in a high surface charge density even at a low mole percent. In the case of liposomes the total surface area is greatly increased owing to their spherical shape consequently requiring a higher mole percent PA to attain a similar surface charge density. Second, there is a gain of dimensionality in the liposome association assay compared to the ELISA format assay for Raf-lipid interactions. This additional degree of freedom increases the entropy of the system and consequently diminishes the chances of productive association between the liposomes and Raf-1. This effect is only counteracted to some extent when the concentration of the reagent (in this case, PA) is increased. Hence a greater mole percent PA is required in the liposome association assay to obtain saturable binding to Raf-1.
The interaction of Raf-1 with PA was next characterized with respect to pH and ionic strength. RafFull, RafCys, or RafC were allowed to bind to PA at acidic, neutral or alkaline pH and the binding was quantitated by the ELISA format assay (Fig. 4A). RafFull and RafC bound to PA with the same profile; lower binding was observed at acidic (4.5) or alkaline (9.0) pH values, with maximal binding occurring at near neutral pH values (6.0 and 7.5). In contrast, RafCys displayed maximal PA binding at acidic pH which was drastically reduced with increases in pH. This difference in pH dependence between RafC and RafCys explains why, under the standard conditions of the ELISA format assay (pH 7.2), RafCys binds with about 10-fold lower affinity to PA compared to RafC (Fig. 2A). The identical pH profiles observed for RafC and RafFull further support the hypothesis that the binding of full-length Raf-1 to PA is mediated primarily by its COOH-terminal domain.
A possible mechanism for the RafC-PA interaction may be postulated as arising from electrostatic interactions where the negatively charged PA nonspecifically binds to positively charged amino acids on the protein. If such were the case, then the interaction would be highly dependent on ionic strength and would eventually be competed out at high ionic strength. We therefore determined the effect of increasing ionic strength on RafC-PA interaction and the results are shown in Fig. 4B. At all RafC concentrations tested (25, 50 and 100 nM), the binding to PA was weak in the absence of salt. Binding was progressively increased up to 250 nM NaCl after which additional increases in salt concentration, up to 1 M, had little effect on binding. The requirement for a certain concentration of salt for maximal binding may be attributed to shielding effects pro- The data were fit to the modified Hill equation as described in the text. B, association of RafC to PA/PC liposomes. Liposome association experiments were performed as described previously (26,27). Liposomes of different PA/PC composition were incubated with RafC (4 g of protein), and the extent of protein association with the liposomes was determined relative to the protein recovered in the absence of lipids. The protein bands obtained in the SDS-PAGE were quantitated by scanning densitometry, and the values are plotted as a function of PA concentration in the liposomes. vided by counter ions between adjacent negatively charged lipid molecules or between negatively charged regions of RafC and PA. Since the binding cannot be competed out even at 1 M NaCl, it suggests that the interaction is not entirely electrostatic in nature. Other forces (e.g. hydrophobic, hydrogen bonding, and van der Waals) are therefore inferred to play a part in stabilizing the RafC-PA interaction.
We next tested different potential competitors of the RafC-PA interaction. In all cases, RafC was first incubated with the test reagent in solution for 30 min and then the entire mixture was added to the wells of microtiter plates coated with PA and allowed to incubate for an additional 30 min. The results of the experiment are shown in Fig. 5. None of the reagents tested competed effectively for the binding of RafC to PA. At all concentrations of RafC tested, again a weaker binding was observed in the absence of salt. Addition of all reagents at 50 mM (10 mM for Ser-O-phosphate and ATP) enhanced the binding of RafC to PA by varying extent. Importantly, 10 mM ATP did not inhibit RafC-PA interaction, which suggested that the site within RafC involved in PA binding was distinct from the site involved in binding to ATP.
Specificity of RafC-PA Interaction-As mentioned earlier, many proteins are known to interact with acidic phospholipids in general with little selectivity for the phospholipid head groups (21, 34 -36). However, for certain proteins such as PKC, only PS is capable of enzyme activation, although both PS and PA bind to PKC (31). We wanted to determine whether the observed RafC-PA interaction is specific for PA or whether it could occur with other lipids as well. A series of lipids were tested by the ELISA format assay. The results are shown in Fig. 6A. The only lipids that bound significantly to RafC were PA and cardiolipin. Neutral phospholipids such as PC and PE did not bind RafC, suggesting that the RafC-PA interaction was not due to nonspecific aggregation of the protein on the lipid surface. Other acidic lipids such as PS, PI, PG, bis-PA, or gangliosides did not bind to RafC under conditions of high stringency binding, although at lower stringencies some binding was observed with PG and PI (data not shown). The phosphatidyl alcohols, phosphatidylmethanol and phosphatidylpropanol, did not bind to RafC, suggesting that the binding depended on the availability of two ionized oxygen atoms (Fig.  6B). This also explains why other phospholipids such as PC or PE did not bind RafC, since only one free oxy-anion is available in those lipids. For the same reason, bis-PA is unable to bind owing to a condensation of the phosphate group to a diacylglycerol moiety. The glycerol backbone appears also to be important for recognition since ceramide 1-phosphate and 2-aminophosphatidic acid also did not bind to RafC under the conditions of the assay (data not shown). The binding of RafC to cardiolipin is unlikely to be physiological since cardiolipin is exclusively localized in the inner mitochondrial leaflet (38) where Raf-1 is not known to be present. However, the preference of RafC for phosphatidic acid and cardiolipin is remarkably similar to the properties displayed by yeast and beef heart mitochondrial cytochrome oxidase. Using spin-labeled phospholipids, this protein was shown to select either cardiolipin or phosphatidic acid in its annular domain (the immediate lipid ring surrounding a protein), rather than PC (39). The activity of cytochrome oxidase was also directly correlated to the levels of cardiolipin (40,41). The specificity of the interaction with cardiolipin and PA was inferred not to be simply due to electrostatic interactions as the experiments were carried out at high ionic strengths and no preference for the acidic phospholipid PS was observed. It was suggested that the mitochondrial cytochrome oxidase was actually recognizing the shapes of the lipid molecules. Whether a similar recognition occurs for RafC-PA interaction awaits further elucidation. From the results obtained in this study, we conclude that the binding of RafC to PA has specificity for (a) the glycerol backbone and (b) the presence of two negatively charged oxygen atoms. Other acidic functional groups, all of which bear a single negative charge, do not bind to RafC under conditions of high stringency.
Deletional Analysis of the PA Binding Site in RafC-In order to determine whether a smaller peptide fragment within RafC was competent to bind PA, we undertook a deletion mutagenesis approach. A series of deletional mutants of RafC were generated by PCR (Fig. 1) as GST fusion proteins and tested for their ability to bind PA. The results of the experiments are summarized in Fig. 7, A-D. Panel A shows the results of an ELISA analysis of the different GST-RafC fragments used in the study. All the constructs tested reacted with the anti-GST polyclonal antiserum with relatively equal affinity. Therefore any differences that may be observed between the fusion proteins in the lipid-binding assay would be due to their differential binding to lipids and not due to a variation in their binding to the anti-GST antibody. The results with the first generation of deletion mutants are shown in Panel B. Raf(200 -307) did not bind to PA whereas Raf(200 -423) and Raf(200 -536) both bound, suggesting that residues between 307 and 423 of Raf-1 consist of the PA binding site. Based on this result, we generated the second generation of RafC mutants and tested for their ability to bind PA, as shown in Panel C. The constructs Raf(301-423) and Raf(360 -423) both bound PA, while Raf(301-367) did not, suggesting that the PA binding site was located between residues 367 and 423. Also Raf(200 -423; K375M), which codes for an ATP-binding site mutant of RafC (due to a lysine to methionine mutation at residue 375) was able to bind PA, suggesting that the PA binding site was distinct from the ATP binding site. This observation explained why ATP did not behave as a competitor of the RafC-PA interaction (see Fig. 5). Finally, based on the results in Panel B, we prepared the third generation of RafC mutants and tested them for PA binding. The results are shown in Panel D. Raf(360 -391) did not bind PA but Raf(389 -423) did, thereby identifying a 35-amino acid segment within the COOH terminus of Raf-1 that was competent to bind PA. Certain amino acids within this segment were found to be conserved among the Raf polypeptides from different species (Scheme 1A). Notably, two distinct domains of homology are present, one involving charged amino acids (residues 398 -402 of human Raf-1) and the other involving hydrophobic residues (amino acids 405-408 of human Raf-1). These regions may provide determinants for Raf-PA interaction. From the predicted secondary structure with an expected accuracy of Ͼ70% (based on the profile network prediction algorithm) (65), the PA binding domain of Raf-1 can be classified as a "mixed class" containing significant proportions of helices, strands, and loops (Scheme 1B). A Kyte-Doolittle hydrophilicity profile analysis of this region also shows distinct domains of hydrophilic and hydrophobic character (Scheme 1B). Interestingly, the major hydrophilic and hydrophobic domains within the segment belong to the two distinct regions of homology observed among the Raf polypeptides.
Role of PA in the Translocation of Raf-1 in MDCK Cells-Traverse et al. (42) had suggested that binding of Raf-1 by Ras leads to activation of Raf-1 as a result of its recruitment to the plasma membrane. However, the interaction between Ras and Raf-1 by itself is not sufficient to stimulate Raf-1 kinase activity significantly in vitro (43). It has been suggested that the function of Ras in vivo is either to position Raf-1 at the plasma membrane near its appropriate activator, or an unidentified signal acts in conjunction with Ras in Raf-1 activation (44). Once Raf-1 is recruited to the membrane, Ras dissociates from Raf-1 (15). The mechanism by which Raf-1 associates with the membrane is separate from its activation (15). However, once located at the membrane, Raf-1 can be phosphorylated by protein kinases such as PKC. PKC-␣ has been demonstrated to FIG. 6. Specificity of RafC-PA interaction. A, different lipids (1 g each) were coated on the wells of microtiter plates in the presence of dipalmitoyl-PC (9 g). RafC (200 nM) was added to the lipids and the extent of association was quantitated by the ELISA format assay. The lipids used were PC, dipalmitoyl phosphatidylcholine; PS, bovine brain phosphatidylserine; PI, pig liver phosphatidylinositol; PA, dioleoyl phosphatidic acid; PE, bovine heart phosphatidylethanolamine; Bis-PA, bis-tetrapalmitoyl phosphatidic acid; PMt, dipalmitoyl phosphomethanol; PPt, dipalmitoyl phosphopropanol; PG, dioleoyl phosphatidylglycerol; DG, dioleoyl glycerol; ganglioside, bovine brain type II ganglioside; cardiolipin, bovine heart cardiolipin; sphingomyelin, pig brain sphingomyelin; ceramide, bovine brain type IV ceramide; sphingosine, bovine brain D-sphingosine. Results shown are the mean of two independent experiments. B, schematic representation of phosphatidic acid, cardiolipin, and bis-phosphatidic acid. The ionized oxygen atoms are depicted in bold type.
phosphorylate and activate Raf-1 in vitro and in vivo (45,46). It is possible that Ras and PKC act together in vivo to activate Raf-1 fully (47). However, the mechanism by which Raf-1 is activated and the constituents that anchor Raf-1 to the membrane when Ras dissociates is not known.
It has been shown that phorbol esters such as TPA activate phospholipase D via a PKC-␣-dependent mechanism in MDCK cells (48), resulting in the generation of PA, a potential second messenger. Since in vitro studies indicated a specific interaction between PA and Raf-1, we next determined if this was also the case in vivo. In order to determine if the PA generated by TPA-activated PLD has a role in Raf-1 translocation, we blocked the production of PA derived from PLD with ethanol.
In the presence of ethanol, PLD catalyzes a transphosphatidyl- were generated as GST-fusion proteins as described in the text. All fusion proteins were serially diluted and tested for their ability to bind PA (1 g/well) in the ELISA format assay. SCHEME 1. A, multiple sequence alignment of the PA binding segment of human Raf-1 (amino acids 390 -426) to other Raf polypeptides. Conserved regions of charged and hydrophobic residues are shown in boxes. B, secondary structure prediction for the PA binding segment of Raf-1. The graph represents a Kyte-Doolittle hydrophilicity profile of the segment with hydrophilic regions lying on the (ϩ) side. The top panel indicates the predicted secondary structure according to the profile network prediction algorithm. The letters E, H, and L stand for "strand," "helix," and "neither strand nor helix," respectively. The bottom panel shows the amino acid sequence of the PA binding domain of Raf-1. ation reaction generating PEt at the expense of PA. MDCK cells were stimulated with 10 nM TPA with varying concentrations of ethanol for 15 min. Cytosol and membrane fractions were prepared and analyzed for Raf-1 as described under "Experimental Methods." The translocation of Raf-1 from the cytosol to the membrane in response to TPA was inhibited by blocking the formation of PA by PLD with ethanol (Fig. 8A). Inhibition of translocation of Raf-1 was dependent upon the concentration of ethanol. To determine if ethanol had an effect on the translocation of PKC-␣, the nitrocellulose membrane was stripped and reprobed for PKC-␣ (Fig. 8B). The translocation of PKC-␣ in response to TPA was not significantly affected by ethanol. In addition, increasing the concentration of ethanol correlated with the increase in PEt (Fig. 8C), and with a decrease in Raf-1 translocating to the membrane. These results suggest a role of PA derived from PLD in Raf-1 translocation in vivo. It is conceivable that PKC-␣ activates PLD, leading to the generation of PA which aids in associating Raf-1 to the membrane. Once positioned at the membrane, Raf-1 can be phosphorylated and activated by other kinases such as PKC-␣. CONCLUSIONS Our results demonstrate that RafC is capable of interacting with phosphatidic acid. This interaction between RafC and PA displayed a distinct pH dependence when compared to another lipid-binding motif within Raf-1, RafCys. At physiological pH, RafCys bound PA with markedly reduced affinity compared to RafC. Also, the binding of RafC to PA was not abolished in the presence of high salt concentrations, suggesting that the interaction was not purely electrostatic. Binding of RafC to PA exhibited a highly sigmoidal dependence on the mole % PA with Hill numbers between 4 and 6, under the experimental conditions described. This observed positive cooperativity in RafC-PA interaction may arise from two sources, (i) as a consequence of cooperative sequestering of PA molecules by RafC such as been reported for protein kinase C binding to PS and PA (31) and (ii) an electrostatics driven accumulation of positively charged regions of RafC to the negatively charged PA surface, followed by a reduction in dimensionality. Interactions of the second kind result in an apparent positive cooperativity (49).
Through deletional mutagenesis, we identified a segment within RafC (residues 389 -423) that was competent to bind PA. This region did not contain any characteristic motifs with the exception of a positively charged tetrapeptide sequence, RKTR (residues 398 -401), which could be involved in an initial electrostatic interaction with PA; additional forces would then stabilize this interaction which consequently becomes relatively insensitive to high ionic strengths.
The physiological relevance of the interaction of Raf-1 with PA is unclear. Raf-1 is translocated to the plasma membrane before it is activated. However, the events at the membrane that trigger the activation of Raf-1 are yet unknown. Although the highly conserved family of acidic proteins termed 14-3-3 was shown to associate with Raf-1 and implicated in Ras-dependent activation of Raf-1 (50 -52), Raf-1 mutants unable to stably interact with 14-3-3 could still be biologically activated in a Ras-dependent manner (53), suggesting that 14-3-3 is not a necessary agent for the activation of Raf-1. Our observation that inhibition of PA formation in intact cells can inhibit the translocation of Raf-1 from the cytosol to the membrane suggests that the presence of PA may facilitate the translocation and stabilization of Raf-1 at the plasma membrane. It is interesting to note that many of the signals that activate the mitogen-activated protein kinase pathway through Ras also activate phospholipase D leading to a transient generation of PA (54 -56). The role of PA as a second messenger has been documented in several systems (57,58). In the present context, one might postulate that a simultaneous activation of Ras (from Ras⅐GDP to Ras⅐GTP) and phospholipase D would create an environment at the membrane where translocated Raf-1 kinase can be firmly anchored by interaction with Ras⅐GTP and with the membrane lipids, notably PS (via the cysteine-rich domain) and PA (via RafC). Conversely, the hydrolysis of PA to diacylglycerol by phosphatidate phosphohydrolase, and the deactivation of Ras⅐GTP to Ras⅐GDP would sufficiently weaken the affinity of Raf-1 for the membrane causing a net return of Raf-1 to the cytosol. Recent evidence also suggests that Raf-1 may be deactivated at the membrane by the action of membrane-associated phosphatases (59). Whether lipids would function as additional activators of Raf-1 in a manner analogous to protein kinase C is at present unclear (60 -64). Thus, the exact role of membrane lipids in the regulation of Raf-1 kinase activity remains to be elucidated.