Functional analysis of a phosphatidic acid binding domain in human Raf-1 kinase: mutations in the phosphatidate binding domain lead to tail and trunk abnormalities in developing zebrafish embryos.

Previously, we and others identified a 35-amino acid segment within human Raf-1 kinase that preferentially binds phosphatidic acid. The presence of phosphatidic acid was found to be necessary for the translocation of Raf-1 to the plasma membrane. We have now employed a combination of alanine-scanning and deletion mutagenesis to identify the critical amino acid residues in Raf-1 necessary for interaction with phosphatidic acid. Progressive mutations within a tetrapeptide motif (residues 398-401 of human Raf-1) reduced and finally eliminated binding of Raf-1 to phosphatidic acid. We then injected zebrafish embryos with RNA encoding wild-type Raf-1 kinase or a mutant version with triple alanine mutations in the tetrapeptide motif and followed the morphological fate of embryonic development. Embryos with mutant but not wild-type Raf-1 exhibited defects in posterior axis formation exemplified by bent trunk and tail structures. Molecular evidence for lack of signaling through mutated Raf-1 was obtained by aberrant in situ hybridization of the ntl (no tail) gene, which functions downstream of Raf-1. Our results demonstrate that a functional phosphatidate binding site is necessary for Raf-1 function in embryonic development.

The protooncogene Raf-1 kinase plays a crucial role in several normal and pathologic cellular processes including proliferation, differentiation, development, senescence, programmed cell death, cell cycle progression, immune responses, and carcinogenesis (1)(2)(3). Raf-1 functions downstream of p21 Ras (4,5) and serves as an upstream regulator of the Ras-Raf-MEK 1 -MAP kinase signal transduction cascade that is activated in response to a wide variety of signals, including growth factors, differentiation hormones, tumor promoters, inflammatory cytokines, calcium mobilization, DNA-damaging agents, and oxygen radicals. A common aspect of Raf-1 activation is its translocation to the plasma membrane, which is composed predominantly of acidic (ϳ30%) and zwitterionic (70%) phos-pholipids (6,7). Membrane lipids also function as second messengers for several intracellular signal transduction events. In the case of Raf-1, lipids such as ceramide and leukotriene D4 are related to its activation in select experimental systems (8,9). Based on the requirement for membrane translocation of Raf-1 prior to activation, we investigated whether membrane lipids might be involved in a functional interaction with Raf-1 kinase that can be a precursor to subsequent activation. In vitro analysis of Raf-1-lipid interaction reveals two distinct phospholipid binding sites within Raf-1 kinase (site I and site II; see Refs. 10 and 11). Site I is located between amino acid residues 139 and 184 of human Raf-1 kinase and consists of a zinc-coordinating cysteine-rich domain analogous to domains present in protein kinase C and other proteins (reviewed in Ref. 13). Site I interacts with anionic phospholipids such as phosphatidylserine, predominantly via electrostatic interactions, driven by a cluster of basic amino acid residues (12). A second phospholipid binding site (site II) is located between residues 390 and 423 of human Raf-1. This region displays preferential interaction with phosphatidic acid, and the binding is not dependent on ionic interactions alone.
An increase in membrane phosphatidic acid via activation of phospholipase D by the tumor promoter 12-O-tetradecanoyl phorbol-13-acetate (14) correlates with a net translocation of Raf-1 from the cytosol to the plasma membrane of Madin-Darby canine kidney cells (11). When the generation of phospholipase D-derived phosphatidate is inhibited by ethanol (via formation of phosphatidylethanol), a specific, dose-dependent loss of Raf-1 translocation is observed. These results suggest that agonist-induced Raf-1 translocation is coupled to the generation of phospholipase D-derived phosphatidic acid. Because translocation is a prerequisite for Raf-1 activation, we postulated that phosphatidic acid might regulate Raf-1 activation by enabling translocation. Subsequently Rizzo et al. (15,16) demonstrated that in Rat-1 fibroblasts overexpressing the human insulin receptor (HIRcB cells), the stimulation of the MAP kinase pathway by insulin is dependent on phospholipase D activation and is mediated via an induction of Raf-1 translocation to the plasma membrane and early endosomes by PA. The generation of PA is essential for Raf-1 translocation and brefeldin A, an inhibitor of the ADP-ribosylation factor (required for phospholipase D activation), prevents the translocation of Raf-1 in a dose-dependent manner. This inhibition of Raf-1 translocation can be reversed by exogenously added PA. Additionally, Raf-1 translocation in response to PLD-derived PA is also observed in Rat-1 cells expressing constitutively activated p21Ras (Q61L mutant) suggesting that PA and Ras may act concurrently and by mutually independent pathways to promote Raf-1 translocation to the plasma membrane (reviewed in Ref. 36). Based on the results from our experiments and that of Rizzo et al. (15,16), we postulate that one of the mechanisms by which PA activates the MAP kinase cascade is via the induction of Raf-1 translocation to cellular membranes. However, whether the proposed PA-Raf-1 interaction does indeed play a functional role in the biology of an intact organism is currently unknown. The present study addresses this question by first identifying the molecular nature of PA-Raf-1 interaction and then determining the biological consequences of disrupting such interaction in a model system of vertebrate development.

EXPERIMENTAL PROCEDURES
Chemicals 96-well microtiter plates (Probind) were obtained from BD Biosciences. Glutathione-agarose matrix (sulfur-linked), phosphate-buffered saline, and bovine serum albumin (fraction V) were purchased from Sigma. All phospholipids tested were purchased from Avanti Polar Lipids. All other chemicals used were of the highest available commercial grade.

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-1 plasmid template (a gift from Dr. Roger Davis, University of Massachusetts, Worcester, MA). Details of the PCR, cloning, bacterial expression, and purification of the fusion proteins have been described previously (10). Site-directed mutagenesis on a phosphatidate binding fragment of Raf-1 kinase was performed via PCR as described (17). The PCR primers shown in Table I were used for the generation of the GST-Raf-1 fusion proteins in this study (alanine mutations are denoted by the code GCT). Following affinity purification of the expressed GST-Raf fusion proteins over glutathione-agarose columns, the fusion proteins were analyzed for purity by SDS-PAGE and stored in aliquots at Ϫ80°C. All fusion proteins were expressed in the Escherichia coli strain, BL-21, obtained from Novagen (Madison, WI).
Enzyme-linked Immunosorbent Assay Format Assay-The interaction of all GST-Raf-1 fusion proteins with different lipids was assessed by the enzyme-linked immunosorbent assay format assay essentially as described earlier (35).
Maintenance and Breeding of Zebrafish-Adult zebrafish were obtained from local pet stores and were maintained on a 14-h light/10-h dark cycle at 28.5°C. For breeding, single male and female fish were placed in a breeding tank consisting of a 1 gallon plastic tank that was placed inside a 2.5-gallon tank. The bottom of the plastic tank had been removed and replaced with plastic mesh so that the eggs would pass through the mesh and would be collected at the bottom of the larger tank.
In Vitro Transcription and RNA Injection-Raf constructs were cloned in pT7TS (gift of Paul Krieg, University of Arizona), a pGEM4z derivative that contains Xenopus ␤-globin 5Ј-and 3Ј-untranslated sequences and an A 30 C 30 sequence inserted downstream from a T7 polymerase promoter site. Constructs were linearized downstream from the A 30 C 30 sequence and were transcribed in vitro using a mMessage mMachine kit (Ambion), following the recommended protocol of the manufacturer. Following DNase treatment, phenol extraction, and ethanol precipitation, the RNAs were quantitated either spectrophotometrically or by RiboGreen (Molecular Probes, Inc.). RNAs were injected into embryos in 0.1 M KCl, 0.2% phenol red at a concentration of 100 ng/l to 1 g/l using an Eppendorf 5242 pressure injection apparatus and sterile Femtotips.
Whole Mount in Situ Hybridization-Digoxigenin-labeled ntl antisense RNA and zebrafish phospholipase D sense and antisense RNAs were synthesized in vitro and used as probes for whole mount in situ hybridization, following previously published methods (19), except that the chorions were removed after fixation in 4% paraformaldehyde, and the incubation in phosphate-buffered saline with Tween 20 ϩ 2 mg/ml glycine was omitted. For detection, embryos were incubated in staining solution containing nitro blue tetrazolium chloride and 5-bromo-4chloro-3-indolylphosphate.

RESULTS AND DISCUSSION
Identification of Key Residues That Mediate Raf-1-Phosphatidate Interaction-Employing deletion mutagenesis, we GTGGCTGTTCTGCATGTGAACATTCTGCCTTTTC AATGTTCACATGCAGAACAGCCACCTCATTCCT

Raf-1: PA Binding and Embryonic Development
previously identified a 35-amino acid segment within the carboxyl terminus of Raf-1 kinase (site II) that interacts with phosphatidic acid (11). Analysis of site II amino acid sequences of Raf-1 isoforms (A-Raf, B-Raf, Raf-1) and Raf-1 from different species reveals two subdomains of significant sequence homology. The first subdomain includes a cluster of basic amino acids corresponding to residues 398 -401 of human Raf-1 mRNA (GenBank TM accession number X03484; Locuslink number 5894). The second subdomain contains predominantly hydrophobic residues encoded by amino acids 404 -407. A Kyte-Doolittle hydrophilicity profile analysis of site II identified two distinct regions of hydrophilic and hydrophobic characters that matched the two sequence homology domains (data not shown). We analyzed the relative contribution of each homology domain toward interaction of Raf-1 with phosphatidic acid via alaninescanning mutagenesis of individual amino acids. A total of 18 different site II mutants were generated by PCR-based mutagenesis and expressed as GST fusion proteins in E. coli. Fig.  1a is a schematic depiction of all the mutants used in this study. We then tested the ability of the mutated GST-Raf-1 fusion proteins to interact with phosphatidic acid via an in vitro assay as described previously (10). The results are shown in Fig. 1, b-d. Fig. 1b shows the binding curves generated with proteins that contain mutations within the basic residues of the first homology domain of site II. Single site R398A or R401A replacements did not affect phosphatidate binding when compared with the wild-type protein. K399A replacement resulted in reduced binding to phosphatidic acid. The two-site replacement mutants (R398A,R401A) also displayed reduced binding to PA compared with the wild-type protein. The reduction in binding was more severe in the two-site mutants containing mutations in Lys-399 (R398A,K399A, and K399A,R401A, respectively). Finally, the binding was severely compromised (80 -95% reduction compared with wild-type) in a triple site mutant (R398A,K399A,R401A). These results suggested that the basic cluster, RKTR, constitutes a critical determinant for Raf-1-PA interaction with Lys-399 probably providing the major contribution. We then focused on the second region of homology within site II, encompassing the hydrophobic residues, and created single site alanine mutants for residues 402-410 of human Raf-1. The mutated proteins were again tested for their relative ability to bind phosphatidic acid by the enzyme-linked immunosorbent assay format assay described previously. The results, shown in Fig. 1c, demonstrate that site-directed mutagenesis within this domain had only a modest effect on interaction with phosphatidic acid. Replacement of Val-403, Leu-406, Leu-407, and Phe-408 to alanine did not change the binding profile to phosphatidic acid, compared with the wild-type sequence. Mutations of His-402, Ile-405, and Met-409 resulted in slightly reduced binding to phosphatidic acid. The greatest reduction in phosphatidate binding (about 40% of control) was observed with the Asn-404 and Gly-410 mutants. A comparison of the effects of mutations within the first and second homology domains of site II suggests that the charged tetrad sequence, RKTR, probably constitutes a major determinant for binding to phosphatidate and may provide an initial, electrostatic-driven clustering around one or more phosphatidate head groups. The hydrophobic region, adjacent to RKTR, may subsequently function to associate with phosphatidic acid via non-ionic interactions, resulting in a more stable complex. However, in the absence of the primary interaction driven by the charged amino acids, the hydrophobic residues are not sufficient to form stable association with phosphatidate. Additional support in favor of the critical role of the RKTR sequence in binding to phosphatidate is provided by a site II mutant lacking the tetrapeptide motif altogether (RafDel398 -401). As expected, the corresponding mutant protein failed to interact with PA. A comparison of relative PA binding by all the site II mutants employed in this study is shown in Fig. 1d.
To examine whether the RafAAA mutation would affect normal development, synthetic RNAs encoding wild-type human Raf, Raf K375M , or Raf AAA were injected into one-two cell zebrafish embryos. Previous work in Xenopus has shown that injecting the Raf K375M allele into one-two cell embryos results in embryos lacking trunk and tail structures (26). The Raf K375M mutation replaces a critical lysine residue at the ATP binding pocket of Raf-1 resulting in an inactive protein. As shown in Fig. 2 injection of wild-type Raf had no effect on normal zebrafish development, whereas about 30% of embryos injected with either Raf AAA or Raf K375M produced a phenotype similar to that seen in Xenopus; in 24 hours post-fertilization (hpf) embryos the tail is absent or severely shortened, and the overall length of the anterior-posterior axis is shortened. Examination of injected embryos at 12 hpf frequently revealed that the notochord was bent or shortened (data not shown). Measurement of the steady state levels of human Raf-1 in injected embryos by immunoblotting showed comparable protein levels in embryos injected with the different forms of Raf-1(data not shown).
One of the known downstream targets of FGF signaling during vertebrate gastrulation is the product of the Xbra or ntl gene. Expression of a dominant negative FGF receptor or a mutant allele of one of several different MAP kinase cascade components results in a reduction or elimination of normal Xbra expression. Conversely, overexpression of FGF or a constitutively active MEK or treatment of Xenopus animal caps with FGF causes an increase in Xbra expression (20). To assess the effects of injecting the various forms of Raf-1 on ntl expression, whole mount in situ hybridization was performed using a digoxigenin-labeled sense and antisense probe to ntl. ntl expression normally begins at 5.5 hpf. At this time (germ-ring stage) the germ ring forms and is considered the start of gastrulation. In normal embryos, ntl expression is restricted at the germ-ring stage to a circumferential belt of cells that have newly involuted from the epiblast layer inwards toward the yolk, forming the mesendodermal cell layer (Fig. 3a). In embryos injected with either Raf AAA or Raf K375M but not wild-type Raf-1 this belt of expression is disrupted, resulting in embryos with a discontinuous band of cells expressing ntl (Fig. 3b). Thus, injection of Raf-1 with a mutated phosphatidate binding site is sufficient to block normal signaling events that are required for the expression of ntl in the newly formed mesendodermal layer.
Phosphatidic acid can be generated by two different mechanisms. In one mechanism, glycerol is phosphorylated by enzymes belonging to the diacylglycerol kinase family to yield PA. The other mechanism involves the hydrolysis of phospholipids such as phosphatidylcholine to PA by the action of the phospholipase D group of enzymes. The relative contribution of these reactions toward the generation of PA in zebrafish em- bryos is not clear. However, previous experiments suggest that phospholipase D-derived PA is involved in the regulation of Raf-1 kinase (11, 14 -16). Based on these reports, we sought to determine whether phospholipase D transcripts were indeed present in the zebrafish embryos at the time when the dominant negative effects of Raf AAA were evident. Other studies have shown that phospholipase D isoforms (PLD1 and PLD2) are selectively and differentially expressed in the developing mouse brain (38) or ubiquitously expressed in Drosophila embryos (39). Based on the sequences of two PLD isoforms in mammals (PLD1 and PLD2), we designed degenerate primers and isolated, via RT-PCR, a ϳ1-kb fragment encoding part of a zebrafish phospholipase D using shield stage RNA as template. The amplified fragment was sequenced and found to be more closely related to mammalian PLD1. We did not obtain zebrafish phospholipase D sequences corresponding to phospholipase D2. A review of the sequence similarity of available zebrafish expressed sequence tags revealed several with similarity to vertebrate phospholipase D1 but none with similarity to the D2 form (data not shown). At present, it is not known if a zebrafish phospholipase D2 exists, and, if so, its pattern of expression. Based on the sequence of the amplified fragment we synthesized primers and probes to the zebrafish phospholipase D and performed real time, quantitative PCR analysis via Taqman (21), using RNA obtained from embryos at various stages of development. As shown in Fig. 4, zebrafish phospholipase D transcripts were found to be present at all stages of development that were examined, with highest expression at the 12 somite stage. The results obtained from the Taqman analysis demonstrate the presence of zebrafish phospholipase D transcripts in developing embryos consistent with the generation of PA during a time course when Raf-1 plays a crucial role in the process of mesoderm induction and anterior-posterior specification. However, these results provide no information about the status of activation of PLD protein during these stages of zebrafish development.
Whole mount in situ hybridizations were performed using a digoxigenin-labeled antisense RNA probe. As shown in Fig. 5, phospholipase D1 transcripts appear to be present at uniform levels in different regions of embryos at the 64-cell, shield, and 12-hpf stages. Colley et al. (38) have also shown a widespread distribution of PLD1 and PLD2 transcripts, with expression detected in all mouse tissues examined, as assayed by North-ern analysis. However, unlike our findings, expression levels were not uniform, varying 10 -100-fold in the tissues examined. In addition, in situ hybridization analysis of the expression pattern within the E11.5 embryonic brain and in the adult brain revealed that both PLD1 and D2 transcripts are present in a restricted pattern, with PLD1 transcripts detected in ventricular cells lining the brain and spinal cord, in the retina, and in a portion of the nasal neuroepithelium of the E11.5 embryo. The 12-hpf zebrafish embryo brain lacks any overt morphological subdivisions, so our analysis of the transcript pattern of zebrafish PLD1 was done at embryonic stages that are earlier than an E11.5 mouse embryo. It is possible that as development proceeds, the pattern of expression of zebrafish PLD1 may also become more restricted. It is also possible that sectioning of the zebrafish embryos shown in Fig. 5 may reveal an expression pattern that is restricted to specific cell types or specific regions of these embryos. These results indicate that phospholipase D transcripts are present before the onset of zygotic transcription at the mid-blastula transition and are present in both ectodermal and mesendodermal cells at later stages. CONCLUSIONS We previously identified a segment within human Raf-1 kinase (amino acid residues 390 -423) that binds PA. We also showed that PA is required for the translocation of Raf-1 from the cytosol to the plasma membrane in Madin-Darby canine kidney cells stimulated with the phorbol ester, 12-o-tetradecanoyl phorbol-13-acetate. Rizzo et al. (15) demonstrated the requirement for PA in the translocation of a GFP-Raf-1 fusion protein in response to insulin stimulation. In their studies, PA did not directly activate Raf-1 in vitro or in vivo strongly arguing for a role of PA primarily in facilitating the translocation of Raf-1 kinase.
The involvement of PA in binding Raf-1 and mediating its agonist-dependent translocation is consistent with the effect of lipid second messengers on the MAP kinase pathway. Many of the signals that activate the MAP kinase pathway also activate phospholipase D (22). Additionally, phospholipase D is also regulated by heterotrimeric G proteins, G 13 and G q (23). This is suggestive of a role of PA as a lipid second messenger (24,25). The data generated from our group and that of Rizzo et al. (15) identifies a molecular mechanism by which PA might be exerting its regulatory function.
In the present work, we have identified, via alanine-scanning mutagenesis, individual amino acid residues within Raf-1 that are critical for its interaction with PA. In vitro binding assays, employing GST-Raf-1 fusion proteins (containing either the wild-type PA binding fragment or mutated versions thereof), have identified a tetrad of charged residues (RKTR, residues 398 -401) required for PA binding. Mutation at the lysine residue (Lys-399) in the tetrad significantly inhibits PA binding. Conversion of all three charged residues to alanine, or their deletion, leads to total loss of binding. We conclude that the RKTR tetrad is the major contributor toward the interaction of site II within Raf-1 with PA. In this context it is worthwhile to compare the PA binding site mutations employed by Rizzo et al. (16) to those used in the current study. Rizzo et al. (16) observed that conversion of the first arginine of the RKTR tetrad into alanine (R398A) in full-length Raf-1 was sufficient to block the translocation of the mutant protein to endosomes in response to insulin. They also expressed a GFP fusion protein linked to a 36-amino acid sequence from Raf-1 containing the PA binding site (GFP-PABR, equivalent to site II in this study). The native GFP-PABR blocked Raf-1 translocation to endosomes (by sequestration of PA) and also blocked MAP kinase activation. However, GFP-PABR containing a single mutation equivalent to R398A was still capable of interacting with PA and prevented the translocation of native Raf-1. This result is unexpected in light of the behavior of full-length Raf-1 harboring the R398A mutation. It is only when a second mutation was introduced (equivalent to R401A), that the GFP-PABR lost its ability to prevent Raf-1 translocation and MAP kinase activation. In our hands, a single mutation at R398A in a GST-site II fusion protein does not significantly reduce binding to PA. We find instead that the point mutation K399A produces a dramatic reduction in PA binding, and any dual mutations in the RKTR tetrad that include K399A cause significant loss of binding.
Lu et al. (27) described a study involving the isolation of temperature-sensitive mutations in the catalytic domain of Raf-1 and the expression of conditionally active and dominant-defective forms of Raf-1 in cultured mammalian cells. The authors introduced pairwise alanine-scanning mutations into the entire ATP binding subdomain of a mutant Raf-1 that was truncated for the first 334 amino-terminal residues and displayed a kinase activity comparable with that of oncogenic v-raf (28). Two of the alanine-scanning mutations were for residues R398A,K399A and K399A,R401A, the same residues identified for PA binding. Upon electroporation into the rat fibroblast cell line, TGR-1, each of the mutant proteins exhibited a behavior similar to wild-type Raf-1 as measured by two independent biological assays for v-Raf function (focus formation and growth in soft agar). This suggests that Arg-398, Lys-399, and Arg-401 are not involved in ATP binding, and, significantly, mutations at those sites do not interfere with the catalytic activity of Raf-1. These results further suggest that unlike full-length Raf-1, where loss of PA binding leads to a dominant negative effect, the constitutively activated, NH 2truncated version of Raf-1 is not subject to regulation by PA (or p21Ras) and consequently does not display a dominant negative phenotype when the PA binding sites are mutated.
The RKTR tetrad is positioned within the ATP binding domain of Raf-1 (343-426). X-ray crystallographic studies in other kinases such as protein kinase A, Cdk2, MAP kinase, and twitchin kinase (29 -32) have indicated that the ATP binding domain is a small, compact, and conserved structure and is largely independent of other structural entities within the kinase such as the substrate binding domain and other regulatory domains. Therefore, it is unlikely that mutations in Arg-398, Lys-399, and Arg-401 would affect the conformation of the MEK binding and p21Ras binding domains of Raf-1, which are situated in the upstream amino-terminal portion of the protein.
This study elucidates the amino acid residues primarily responsible for mediating an interaction between site II of Raf-1 kinase and PA. The functional consequences of mutations at these amino acid residues have been followed in developing zebrafish embryos. Our results indicate that in the presence of Raf-1 kinase deficient in binding PA via site II, the developmental program in zebrafish embryos is severely compromised, leading to embryos with bent tail/trunk structures and a shortened anterior-posterior axis. Our work is consistent with previous work in a number of animal model systems that have shown that Raf is required for normal embryonic development. Thus, a mutation in the ATP binding pocket of Raf (RafK375M) acts as a dominant negative in Xenopus embryos, resulting in embryos with shortened body axes and defects in the processes by which mesoderm is formed and specified. Similarly, mice lacking A-Raf, B-Raf, or c-Raf-1 exhibit a number of defects in development and organogenesis (33). Drosophila D-Raf acts downstream of a number of receptor tyrosine kinases that are required for normal development, including the EGF, FGF, torso, and sevenless receptors. Similar work with other components of the MAP kinase cascade indicates a role for this pathway in the normal specification of the anterior posterior and dorsal ventral axes. Studies indicate that similar pathways control germ layer formation and axis specification in zebrafish (34). Thus, the normal processes by which mesoderm is formed and the anterior posterior axis is specified provides a readout for perturbations in the normal signaling events, including the overexpression of dominant negative forms of c-Raf-1 kinase.
Our results indicate that in the presence of Raf-1 kinase deficient in binding PA via site II, the development program in zebrafish embryos is severely compromised, leading to embryos with bent tail/trunk structures and shortened anterior-posterior axis. This phenotypic effect of mutated Raf-1 is corroborated at the molecular level by a concomitant disruption in the expression of the ntl gene, signifying impaired FGF signaling.
In addition to the results reported here, Raf AAA also exhibits dominant negative effects in a cell-based assay for cyclooxygenase gene expression. 2 There are several possible explanations for this dominant negative effect. It is possible that the non-physiological levels of externally introduced RafAAA in the embryos may result in its direct interaction with its substrate, MEK, thereby making MEK unavailable for the endogenous Raf proteins. However, lack of an intact PA binding site does not allow the externally introduced RafAAA to be appropriately translocated and subsequently activated by Ras resulting in a net dominant negative effect. Alternatively, as reported previously (40), Raf-1 proteins are known to form homodimers. It is conceivable that RafAAA will oligomerize with endogenous Raf-1. However, because of its inability to bind PA, the RafAAA-endogenous Raf complex may not be effectively translocated to the membrane thereby preventing the activation of the endogenous Raf. Finally, it should be noted that a recent study (37) investigated the association of B-raf and Raf-1 with synthetic lipid vesicles and demonstrated high affinity binding of Raf-1 to ceramide, cholesterol, phosphatidylserine, and phosphatidic acid. The nature of the interacting lipids led the authors to suggest two possible microdomains on Raf-1, one interacting with lipid rafts (enriched with cholesterol and ceramides) and the other, nonraft binding domains (phosphatidylserine and PA). Interestingly, in the presence of liposomes, the recruitment of Raf-1 by activated, farnesylated Ras was minimal, suggesting that the observed Ras-Raf binding in response to activation of Rafcoupled receptors may utilize membrane-prebound Raf protein.
In a separate study, phosphatidic acid was shown to enable Raf-1 translocation to the membrane independent of Ras, although Ras was required for Raf-1 activation (16). Taken together, it appears that the direct, Ras-independent association of Raf-1 with specific membrane lipids might constitute a basal Raf translocation pathway. Subsequent activation of membrane-prebound Raf-1 would involve Raf recruitment and binding by GTP-Ras primarily via diffusion in the plane of the membrane such that Raf-1 is structurally and positionally reconfigured for interaction with activating kinases and substrates. If the initial interaction with the membranes is primarily driven via the raft-interaction domains (suggested in Ref. 37) then subsequent interactions with phosphatidylserine, PA, and GTP-Ras should influence the activation of Raf-1. In this light, a likely explanation for the dominant negative effects seen with RafAAA in this study is as follows. RafAAA has fully functional binding domains for phosphatidylserine, cholesterol, and ceramide (in its amino-terminal domain) and can populate the cell membranes when overexpressed in zebrafish embryos (if such lipids are indeed present at that stage in the embryos), at the expense of endogenous Raf-1. However, the lack of a functional PA binding site inhibits the proper membrane attachment of RafAAA required for its activation. Consequently, a dominant negative phenotype is observed. Whatever the mechanism by which RafAAA exerts its dominant negative effect, these results strongly implicate that the regulation of Raf-1 kinase by PA is an essential component of normal Raf-1 function in biological systems.