Membrane Localization of Raf Assists Engagement of Downstream Effectors*

We have previously described a small molecule-directed protein dimerization strategy, using coumermycin to juxtapose Raf fusion proteins containing the coumermycin-binding domain GyrB. Oligomerization of cytoplasmically localized Raf-GyrB fusion proteins leads to an increase in the kinase activity of both Raf and its substrate Mek. Surprisingly, more distal targets, such as Erk1 and Erk2, are not activated using this approach. Here we report that coumermycin-induced oligomerization of a membrane-localized Raf-GyrB fusion protein potently activated Erk1 and Erk2, up-regulated Fos protein levels, and induced expression of many immediate-early response genes. Thus, both membrane localization and oligomerization of Raf-GyrB are required to target Raf signals to downstream effectors. The ability to activate the entire Raf signal transduction cascade conditionally, using coumermycin-induced oligomerization, should prove useful for dissecting Raf-mediated effects on gene expression and cellular differentiation.

All cells sense and respond to changes in the external environment. They accomplish this, in part, by expressing a diverse set of transmembrane receptors on their surfaces that are capable of recognizing and responding to a large number of distinct ligands. Stimulation of a specific receptor typically leads to the activation of several signal transduction pathways; activation of these distinct pathways effects changes in gene transcription, thereby shaping the response to a particular stimulus. One such pathway, the ubiquitous Ras signal transduction pathway, regulates proliferative responses emanating from a variety of receptors (reviewed in Ref. 1). Ras itself was initially identified based on its potent ability to induce neoplastic transformation of a variety of cell types (2)(3)(4). This oncogenic potential is underscored by the finding that over 25% of all human tumors bear activated alleles of ras (reviewed in Ref. 5). More recently, the Ras signaling pathway has been shown to play an important role in regulating many developmental processes, e.g. vulval development in Caenorhabditis elegans, ommatidial development in Drosophila, and lymphocyte development in mice (reviewed in Refs. 6 -9). Thus, understanding the molecular mechanisms by which Ras transduces signals will provide important insights into both the process of tumorigenesis and the regulation of cellular differentiation.
Ras activates a number of distinct effector molecules including the serine/threonine kinase Raf, the lipid kinase phosphatidylinositol 3-OH-kinase, the guanine nucleotide exchange factor for Ral, Ral GDP dissociation stimulator, and the Ras GTPase-activating protein (10). A key question concerns the role that these specific effectors play in directly modulating the expression of genes entrained by Ras. To evaluate the effects of stimulating a specific pathway, a method capable of activating discrete downstream effector molecules is required. We sought to engineer such an approach through coumermycin-mediated intracellular multimerization of target proteins, focusing in particular on Raf.
Raf is normally activated through recruitment to the plasma membrane mediated by an interaction with GTP-bound Ras (11)(12)(13)(14)(15). This membrane localization leads to Raf activation via a process that is not yet well understood. To study Raf-specific signal transduction events, we developed an inducible dimerization strategy that allows for the conditional activation of Raf (16). Our approach uses the symmetrically dimeric antibiotic coumermycin to promote either membrane translocation or dimerization of Raf fusion proteins that bear the coumermycinbinding domain GyrB. During the course of these initial studies, we found that Raf kinase activity can be induced by Raf dimerization, or more likely oligomerization, even in the absence of membrane localization. This activation of Raf by oligomerization is not unique to coumermycin-based strategies, as a similar result was obtained using FK1012-induced dimerization of Raf-FKBP chimeras (17). We have now examined more distal effects of Raf activation following coumermycininduced oligomerization, and here report that, although coumermycin-induced clustering of cytoplasmic forms of Raf-GyrB does lead to robust activation of Mek, it does not activate downstream signaling effectors such as Erk1 or Erk2, nor does it modulate Raf-regulated gene transcription. In contrast, coumermycin-mediated oligomerization of membrane-localized forms of Raf-GyrB, which are not by themselves constitutively active, leads to dramatic increases in downstream MAP 1 kinase activation, Fos protein levels, and Raf-dependent gene transcription. These results show that membrane localization is not necessarily required for Raf activation, but is crucial for targeting Raf signals to distal effectors. In addition, our results demonstrate that coumermycin-induced activation of membrane-localized Raf-GyrB can be a useful tool for identifying both protein effectors and gene targets regulated by Raf.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Novobiocin, coumermycin A1, and phorbol 12-myristate 13-acetate (PMA) were purchased from Sigma and dissolved in Me 2 SO before use. Epidermal growth factor (EGF) was purchased from Life Technologies, Inc.
The rabbit polyclonal antibody used to detect the phosphorylated, active form of Erk1 and Erk2 was obtained from Promega (Madison, WI). Rabbit polyclonal antibodies used to detect either phosphorylated, active Mek or total Mek were obtained from New England Biolabs (Beverly, MA). 7D3 is a murine monoclonal antibody (IgG2b) that recognizes the 24-kDa amino-terminal fragment of the B subunit of bacterial DNA Gyrase (GyrB), and was generously provided by Dr. Martin Gellert (18) (currently available from Lucent Ltd.). Rabbit polyclonal IgG specific for Erk1 and Erk2 was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). The anti-Raf monoclonal antibody (IgG1) was obtained from Transduction Laboratories (Lexington, KY). Alkaline phosphatase-conjugated goat anti-rabbit and rabbit antimouse polyclonal antibodies were obtained from Amersham Pharmacia Biotech as part of their Vistra enhanced chemifluorescence Western blotting kits.
Cells and Cell Culture-NIH3T3 cells were obtained from ATCC (Manassas, VA) and grown in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (Gemini Bio-Products, Calabasas, CA), 2 mM L-glutamine (Life Technologies, Inc.), 50 units/ml penicillin, and 50 g/ml streptomycin (Life Technologies, Inc.). NIH3T3 cells were transfected with various Raf-GyrB constructs by the calcium phosphate method as described previously (19). Individual clones were obtained by limiting dilution of bulk transfected cells.
Plasmid Construction-Construction of the Raf-GyrB plasmid was described previously (16). Raf-GyrB constructs bearing distinct amino termini were generated by PCR-directed mutagenesis. In brief, specific primers (listed below) encoding either the Myc epitope tag or myristylation sequences derived from Lck were used in conjunction with a common 3Ј-oligonucleotide to amplify an approximately 500-base pair fragment encompassing the amino terminus of Raf with the desired amino-terminal extension. PCR amplification was carried out for 25 cycles under the following conditions: 30 s of annealing at 59°C, 1 min of extension at 72°C, and 30 s of denaturation at 94°C. PCR fragments were digested with HindIII and subcloned into pCDNA3-Raf-GyrB. For transfection studies, the various Raf-GyrB chimeras were excised from pCDNA3 by AseI and NotI digestion, the ends were blunted with Klenow, and the resulting fragment subcloned into the bicistronic expression vector EMCV.neo.
Primers-Primers are as listed below.  Erk and Mek Phosphorylation Assays-Cells were plated in 100-mm tissue culture dishes at 5 ϫ 10 5 /dish and allowed to adhere overnight. The following day, cells were washed two times with phosphate-buffered saline (PBS) and then incubated for an additional 18 h in 5 ml of serum-free DMEM containing 2 mM L-glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin. Cells were subsequently stimulated by adding 5 ml of DMEM containing the appropriate amounts of Me 2 SO, novobiocin, coumermycin, or PMA. At indicated time points, stimulation was stopped by washing cells twice with ice-cold PBS and then lysing them in ice-cold buffer H (20) containing 1% Triton X-100. Lysates were scraped into 1.5-ml Eppendorf tubes (Beckman) with a plastic cell scraper (Costar) and cellular debris removed by centrifugation at 45,000 rpm in an Optima TLX 120 ultracentrifuge (Beckman). Supernatants were removed and aliquots mixed with 4ϫ Laemmli buffer. Samples were fractionated by SDS-PAGE and transferred to PVDF membranes. Membranes were blocked for 1 h in Tris-buffered saline (TBS) containing 0.15% Tween 20 (TBS-Tween: 10 mM Tris, pH 7.5, 0.9% NaCl, 0.15% Tween 20) and 3% milk powder and then incubated overnight at 4°C in TBS-Tween, 3% milk containing either a 1:20,000 dilution of rabbit anti-phospho-Erk1/Erk2 (for the phospho-Erk assay) or a 1:1000 dilution of Rabbit anti-phospho-Mek (for the phospho-Mek assay). Membranes were washed in TBS-Tween and then incubated for an additional hour at room temperature in TBS-Tween, 3% milk containing a 1:10,000 dilution of alkaline phosphatase-conjugated goat anti-rabbit polyclonal antibody. The membranes were then washed, with frequent changes, for 2 h in TBS-Tween and then developed in Vistra detection reagent (Amersham Pharmacia Biotech) for 15 min. Results were quantitated using a Storm 840 fluorimager (Molecular Dynamics, Sunnyvale, CA) and ImageQuant software.
Results were normalized for total Erk2 or Mek levels by stripping the membranes and reprobing with either anti-Erk1 and -Erk2 polyclonal antisera or anti-Mek antisera, respectively. Membranes were first washed in methanol for 30 min to remove Vistra substrate and then washed for 30 min in TBS or PBS to remove methanol. Membranes were stripped by incubating in 62.5 mM Tris, pH 6.7, 2% SDS, 100 mM 2-mercaptoethanol for 30 min at 50°C. Membranes were washed in TBS-Tween for 30 min, blocked as above, and then incubated for 1 h at room temperature with TBS-Tween, 3% milk containing a 1:1000 dilution of either anti-Erk1 and -Erk2 rabbit polyclonal antibody or rabbit anti-Mek polyclonal antibody. Membranes were washed, incubated with TBS-Tween, 3% milk containing alkaline phosphatase-conjugated goat anti-rabbit polyclonal antibody, and then washed and analyzed as above.
c-Fos Assay-Cells were plated and serum-starved as for the Erk phosphorylation assay. Cells were stimulated for 75 min with novobiocin, coumermycin, or PMA. Cell nuclei were obtained and c-Fos protein levels assayed by immunoblotting as described previously (21).
Subcellular Fractionation-Cells were seeded into two T-175 tissue culture flasks and grown until they were 70 -80% confluent. Cells were then harvested by trypsin treatment and washed once in media and once in PBS. Cells were lysed by Dounce homogenization in homogenization buffer (0.25 M sucrose, 50 mM Tris, pH 7.5, 25 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, and 100 g/ml phenylmethylsulfonyl fluoride) and subcellular fractionation carried out by differential centrifugation as described (22). Cytosolic and membrane fractions were normalized such that aliquots containing equal amounts of Raf protein were analyzed (thus membrane fractions contain about 15-fold more material, by volume, than cytosolic fractions). Normalized samples were fractionated by SDS-PAGE, followed by transfer to PVDF membranes. Membranes were subject to immunoblotting with a monoclonal antibody that recognizes an epitope common to both Raf and Raf-GyrB polypeptides, followed by a secondary alkaline-phosphatase conjugated anti-mouse antibody and developed with Vistra enhanced chemifluorescence substrate. The amount of Raf and Raf-GyrB was quantitated using a Storm 840 Fluorimager. -Fold enrichment was calculated as the ratio of Raf-GyrB to Raf in the membrane fraction divided by the ratio of Raf-GyrB to Raf in the cytosolic fraction.
Gene Arrays-Global changes in gene transcription were examined using Atlas Gene arrays (CLONTECH, Palo Alto, CA). Cells were plated in six 150-mm tissue culture dishes. Upon reaching 80% confluence, cell monolayers were washed twice with PBS and incubated for an additional 18 h in serum-free DMEM containing 2 mM L-glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin. Two plates each were stimulated for 40 min with novobiocin, coumermycin, or PMA, respectively. Stimulation was stopped by washing cells twice with ice-cold PBS. Cell monolayers were harvested by trypsin treatment, and trypsin was removed by washing in PBS. Total RNA was harvested using CLONTECH's Atlas Pure RNA isolation kit and hybridization probes prepared as per the manufacturer's instructions. Mouse expression array membranes were hybridized overnight with probes derived from novobiocin, coumermycin, or PMA-stimulated cells. Membranes were washed and exposed on PhosphorImager cassettes and subsequently analyzed on a Storm 840 PhosphorImager using AtlasImage software. ␤-Actin and glyceraldehyde-3-phosphate dehydrogenase mRNA levels were used to normalize for differences in the total amount of mRNA probe used to hybridize to each gene array blot.

RESULTS
We have previously demonstrated that coumermycin-mediated oligomerization of Raf-GyrB fusion proteins leads to robust activation of the direct target of Raf, the dual-specificity kinase Mek (16). To explore more fully this stimulation of the Raf signaling cascade induced via oligomerization, we examined the activation of more distal effectors. We focused initially on the MAP kinases Erk1 and Erk2; these MAP kinases are activated by Mek via phosphorylation of specific threonine and tyrosine residues (23)(24)(25). To examine whether coumermycininduced oligomerization of Raf-GyrB could activate Erk1 and Erk2, Raf-GyrB expressing NIH3T3 stable transfectants were stimulated with either coumermycin or EGF. The presence of activated MAP kinase was detected using antibodies that recognize the threonine-and tyrosine-phosphorylated forms of Erk1 and Erk2. As shown in Fig. 1A, EGF treatment led to a dramatic increase in phosphorylated Erk1 and Erk2 within 5 min of stimulation (17.8-fold induction of phospho-Erk2). In contrast, coumermycin stimulation led to a very modest (2.7fold induction of Erk2) and transient increase in Erk phosphorylation. Similarly, PMA treatment, which is known to potently activate the MAP kinase pathway, gave a greater than 20-fold increase in Erk1 and Erk2 phosphorylation while coumermycin addition stimulated only weak MAP kinase activation (Fig.  1C). This result was not due to ineffective Mek activation as coumermycin stimulation of Raf-GyrB transfectants led to Mek phosphorylation (and hence activation) at comparable levels to those observed for PMA stimulation of either these same cells ( Fig. 1B and Ref. 16) or non-transfected NIH3T3 cells (data not shown). These results indicate that, despite robust activation of Mek (Ref. 16 and Fig. 1B), coumermycin-induced Raf-GyrB oligomerization does not lead to significant activation of the downstream targets Erk1 and Erk2.
Activation of Raf normally occurs following its recruitment to the plasma membrane, and this localization may be important for coupling Raf/Mek activation to efficient Erk1 and Erk2 phosphorylation. To test this hypothesis, we generated a series of Raf-GyrB chimeric proteins that localized either to membrane or cytosolic compartments (Fig. 2, A and B). The membrane-localized form of Raf-GyrB was created by adding an amino-terminal myristylation and palmitylation sequence derived from the Src family kinase Lck (referred to as Lck-Raf-GyrB). Previous work has established that this sequence is sufficient for targeting heterologous proteins to the membrane compartment, and that both myristylation and palmitylation are required for this effect (26,27). We also generated three control constructs in which either a Myc tag or mutated myristylation/palmitylation sequences were added to the amino terminus of Raf-GyrB. In the first mutant (C3S,C5S), the cysteine residues required for palmitylation were changed to serines. In the second mutant (G2A), the glycine residue at position 2 was changed to alanine. The latter mutation prevents both myristylation and palmitylation, since prior myristylation is required for palmitylation (27). Using these constructs, we generated a panel of stable NIH3T3 transfectants. The Lck-Raf-GyrB constructs were expressed at levels equivalent to endogenous Raf, and the other three constructs were expressed at 2-3-fold higher levels, as judged by immunoblotting (data not shown). Subcellular fractionation studies demonstrated that the Lck-Raf-GyrB construct was predominately localized to the membrane fraction, while the other three constructs co-localized with endogenous Raf and were found predominately in the cytoplasm (Fig. 2B).
These cell lines were then tested for the ability of coumermycin to induce Mek and Erk phosphorylation. As seen in Fig.  3 (panels A-D), coumermycin stimulation leads to an increase in Mek phosphorylation in all cell lines tested. The level of Mek activation seen in Myc-Raf-GyrB, C3S,C5S, and Lck-Raf-GyrB transfectants was comparable to that observed when stimulating these cell lines with PMA. The G2A transfectant responded to coumermycin to a lesser but still significant extent. In contrast, neither the Myc-Raf-GyrB construct nor the G2A or C3S,C5S constructs exhibited notable, sustained increases in Erk activation following coumermycin stimulation (Fig. 3, panels E-G). However, stimulation of Lck-Raf-GyrB transfectants with coumermycin resulted in a rapid and dramatic increase in the level of phosphorylated Erk1 and Erk2 (Fig. 3, panel H). Appreciable activation could be seen within 1 min of stimulation, with maximal phosphorylation occurring after 5 min. Membrane localization mediated by the Lck myristylation/ palmitylation sequence did not lead to constitutive activation of the Raf pathway (see Fig. 3, panels D and H, Me 2 SO (DMSO) treatment). Moreover, none of the cell lines tested responded to novobiocin (a monomeric form of coumermycin), indicating that the response was truly dependent on coumermycin-induced oligomerization of Lck-Raf-GyrB, and not just due to nonspecific effects of drug binding to the GyrB domain (data not shown). All cell lines tested, irrespective of the construct with which they were transfected, remained responsive to PMA.
Hence, the lack of response to coumermycin did not reflect a generalized defect in the Ras pathway in these transfectants (Fig. 3).
Typically, activation of Erk1 and Erk2 results in the phosphorylation of additional effector proteins including transcription factors such as the Ets family member Elk-1. Phosphorylated Elk-1, in turn, can bind to the promoters of a number of genes including those for c-fos. This leads to an increase in fos mRNA transcription and ultimately an increase in Fos protein abundance (28,29). Therefore, we examined whether coumermycin stimulation of Raf-GyrB or Lck-Raf-GyrB transfectants could lead to up-regulation of Fos protein levels. As shown in Fig. 4, coumermycin stimulation of Lck-Raf-GyrB transfectants resulted in a significant increase in Fos protein accumulation (3.6-fold induction). This result compares well with that obtained when stimulating cells with PMA. In contrast, Raf-GyrB transfectants showed no significant increase in Fos protein levels following coumermycin stimulation, although they still remained responsive to PMA (Fig. 4).
To examine this issue more broadly, we compared wholesale changes in mRNA abundance using gene macroarrays. Lck-

FIG. 2. Membrane localization of Raf-GyrB is required for coumermycin-dependent activation of Erk1 and Erk2.
A, unique amino-terminal extensions were added to Raf-GyrB consisting of a Myc tag, an 8-amino acid Lck-derived myristylation/palmitylation sequence, or two distinct mutants of the Lck myristylation/palmitylation sequence. These constructs were transfected into NIH3T3 cells, and stably expressing cell lines were generated. B, subcellular localization of Raf-GyrB constructs. Subcellular fractionation was carried out by differential centrifugation as described under "Experimental Procedures." The ratio of Raf-GyrB to Raf was determined in membrane (M) versus cytosolic (C) fractions. -Fold enrichment of Raf-GyrB constructs in the membrane fraction was determined relative to endogenous Raf and was calculated as the ratio of Raf-GyrB to Raf in the membrane fraction divided by the ratio of Raf-GyrB to Raf in the cytoplasmic fraction. Raf-GyrB transfectants were stimulated with novobiocin (as a negative control), coumermycin, or PMA (as a positive control). Transcript abundance in cells treated with novobiocin was equivalent to that seen in cells treated with carrier alone (Me 2 SO). In contrast, stimulation of Lck-Raf-GyrB cells with either coumermycin or PMA led to very similar changes in the pattern of gene expression. Specifically, stimulating Lck-Raf-GyrB transfectants with coumermycin for 40 min led to a greater than 2.5-fold change in mRNA levels for 71 out of 588 genes examined. These included a number of genes known to be induced downstream of the Raf pathway, most notably egr-1 (20-fold induction), c-fos (4.4-fold induction), and c-myc (5.5fold induction) (29 -31) (Fig. 5 (A and B) and Table I). In contrast, no changes in gene expression were observed when stimulating Raf-GyrB-transfected cells with coumermycin ( Fig.  5A and data not shown). Taken together, these results suggest that both oligomerization and appropriate membrane localization of Raf are critical for connecting Raf activation to changes in downstream gene transcription. DISCUSSION The Ras signal transduction pathway plays a key role in regulating both cell proliferation and differentiation. A large number of specific effectors of Ras have been identified that propagate Ras signals (reviewed in Ref. 10). However, the exact role that these different effectors play in regulating downstream protein function remains to be elucidated. In particular, it has been very difficult to uniquely and inducibly activate any one Ras effector and thereby investigate effects specific to that branch of the Ras signaling cascade. We have pursued the development of a system that allows one to study the activation of one such effector, the serine/threonine kinase Raf, in a specific and conditional manner.
Raf activation is initiated by translocation from the cytoplasm to the plasma membrane (32,33). Once at the plasma membrane, Raf becomes fully activated by a process that is still not completely understood. However, a growing body of evidence supports the hypothesis that oligomerization of Raf plays a role in this process. For example, it has recently been shown that Ras exists as a dimer and that driving Ras dimerization also activates the Raf pathway (34). Furthermore, in yeast the scaffolding protein Ste5 has been shown to organize signaling of the yeast homologs of Raf (Ste11), Mek (Ste7), and MAP kinase (Fus3) (35-37). Dimerization of Ste5 is both necessary and sufficient to activate Ste11 and its downstream targets

FIG. 4. Coumermycin-induced activation of membrane-localized Raf-GyrB increases nuclear Fos protein levels. Raf-GyrB or
Lck-Raf-GyrB transfectants were treated with novobiocin (Novo., 900 nM), coumermycin (Cou., 900 nM), or PMA (10 ng/ml) for 75 min. Cell nuclei were harvested and nuclear extracts prepared. Extracts were fractionated by SDS-PAGE, and transferred to PVDF membranes. Fos protein was detected by immunoblotting with anti-Fos antibodies. The y axis represents -fold induction relative to cells stimulated with novobiocin. Shown is a representative example of three experiments; error bars represent the range of duplicate experiments.

FIG. 5. Coumermycin stimulation leads to changes in mRNA levels for multiple genes.
Lck-Raf-GyrB-or Raf-GyrB-transfected cells were stimulated for 40 min with novobiocin (Novo., 900 nM), coumermycin (Cou., 900 nM), or PMA (10 ng/ml). RNA was harvested and used to generate radiolabeled probes. The probes were hybridized to parallel Atlas mouse gene array blots to determine relative abundance of mRNA following stimulation with the above agents. Results shown for Lck-Raf-GyrB are the average of two separate experiments, while those for Raf-GyrB represent one experiment. A, Egr-1 mRNA levels in Lck-Raf-GyrB and Raf-GyrB transfectants following stimulation with novobiocin (Novo.), coumermycin (Cou.), or PMA. B, a representative example of changes in gene expression in Lck-Raf-GyrB transfectants following treatment with either coumermycin or PMA. The first two columns from an Atlas array are shown. Results are quantitated in Table I. Ste7 and Fus3 (38,39). Our initial results suggested that oligomerization of Raf not only leads to its activation but may, in fact, obviate the membrane localization step that typically occurs during this process (16). To investigate this in greater detail, we examined more distal events such as the activation of Erk1 and Erk2 and the accumulation of Fos protein levels. Following coumermycin-induced oligomerization of cytoplasmic Raf, we found that Erk1 and Erk2 were only minimally activated and that no accumulation of Fos protein could be detected. Furthermore, using gene macroarrays that included a number of genes thought to be regulated by the Raf pathway (i.e. fos, myc, and egr-1), we saw no changes in gene transcription. Thus, although cytoplasmic clustering of Raf was sufficient to activate Mek, this did not lead to efficient coupling of activated Mek to more distal effectors.
One possible explanation for this result, especially in light of what is known about Ste5 functioning in yeast, is that membrane localization of Raf is required to target Raf/Mek signals to appropriate substrates. To test this possibility, we generated Raf-GyrB fusion proteins that contained an amino-terminal myristylation and palmitylation signal sequence derived from the Src family kinase Lck. This resulted in a membrane-localized form of Raf-GyrB (Lck-Raf-GyrB). Previous studies have demonstrated that transiently expressing farnesylated versions of Raf (Raf-CAAX) in Cos cells results in membrane localization of Raf and constitutive activation of the Raf pathway (32,33). In contrast, we found that membrane localization of Raf, achieved by amino-terminal acylation, does not lead to constitutive activation of Raf in stably transfected NIH3T3 cells. It is unclear why membrane targeting of Raf via farnesylation should result in constitutive activation of Raf, while targeting via myristylation/palmitylation does not. This discrepancy may reflect differences in the expression level achieved in our stable NIH3T3 transfectants, in which Lck-Raf-GyrB is expressed at the same level as endogenous Raf, versus transient Cos cell transfectants, in which Raf-CAAX is most likely expressed at supraphysiological concentrations. Overexpression of Raf-CAAX could lead to fortuitous association, mimicking oligomerization, and hence activation, such as has been seen following the overexpression of other kinases (40) or of caspases (41). Supporting this hypothesis, Mineo et al. (42) have found that stable expression of Raf-CAAX in NIH3T3 cells also does not lead to constitutive activation of the Raf pathway. In addition, it is possible that farnesylation of Raf directly promotes interactions with effector substrates. For example, farnesylation of Ras2 (in yeast) leads to a 100-fold increase in its affinity for adenylyl cyclase (43). This result reflected a change in binding affinity, and was not merely an effect of enhanced proximity due to membrane association, as a similar change was observed using solubilized Ras2 and adenylyl cyclase. Regardless of the actual mechanism, our results suggest that the effect achieved by transient expression of Raf-CAAX in Cos cells most likely involves more than simply targeting Raf to the plasma membrane.
The ability to target Raf to the membrane fraction, without concomitantly activating it, allowed us to examine the effects of oligomerization of Raf in this compartment. In striking contrast to what we observed with cytoplasmic forms of Raf, coumermycin-induced oligomerization of membrane-localized Raf led to dramatic increases in Erk1 and Erk2 phosphorylation, a significant increase in Fos protein accumulation, and the up-regulation of mRNAs for many genes. These changes paralleled those seen when cells were stimulated with PMA. Such overlap in response is not surprising since protein kinase C is thought to act in part via activation of the Ras/Raf pathway (44). Interestingly, we observed that mRNA levels for at least one gene, egr-1, are particularly sensitive to signals entrained by Raf alone. In addition, several of the genes induced by coumermycin-mediated activation of Raf-GyrB are involved in regulating cell proliferation (e.g. myc, Rb, fos; see Table I), thereby underscoring the role of Raf in this process. Finally, these findings allow us to separate Raf activation from the activation of Erk1 and Erk2. Neither oligomerization nor membrane localization of Raf is sufficient to potently engage the entire Raf signal transduction cascade; only the combination of these events leads to a robust activation of distal Raf effectors and an increase in Raf-dependent gene expression.
Why does oligomerization of membrane-localized Raf (Lck-Raf-GyrB) lead to complete activation of the Raf pathway, while oligomerization of cytoplasmic Raf (Raf-GyrB) does not? One potential explanation is that the strength of the signal delivered by clustering Raf in the cytoplasm could be insufficient to activate the pathway fully. This possibility seems unlikely since coumermycin-induced oligomerization of Raf-GyrB or Myc-Raf-GyrB activated Mek to the same degree as that seen when stimulating cells with PMA ( Figs. 1 and 3) or fetal bovine serum (16). Since both PMA and fetal bovine serum potently activate Erk1 and Erk2 (Figs. 1 and 3 and Ref. 16), we cannot attribute the failure of coumermycin stimulation to activate Erk1 and Erk2 to inadequate activation of Raf-GyrB and its downstream substrate Mek. An alternative explanation is that the simple addition of acyl groups to the amino terminus of Lck-Raf-GyrB might facilitate interaction with substrates, much in the same way that farnesylation of Ras promotes interactions with adenylyl cyclase and other targets (43,45). However, the finding that the myristylated C3S,C5S mutant cannot activate downstream effectors following coumermycininduced oligomerization demonstrates that simple amino-terminal acylation does not account for the effect we see using Lck-Raf-GyrB.
A more likely explanation is that membrane-localization is required to couple Raf effectively to downstream targets. The myristylation/palmitylation signal sequence used to target Lck-Raf-GyrB to the membrane compartment has been previously shown to target heterologous proteins to caveolae or glycosphingolipid-enriched membrane compartments (26). Separate studies demonstrated that in quiescent fibroblasts these membrane compartments contain all components of the Ras signaling pathway including inactive Erk1 and Erk2, and suggested that this may promote efficient activation of the MAP kinase pathway following stimulation with growth factors (46). In examining Lck-Raf-GyrB subcellular distribution, we have found that a substantial fraction can be detected in caveolae (data not shown). Thus, activating Lck-Raf-GyrB within caveolae could facilitate interactions of the activated Raf/Mek complex with the inactive MAP kinases Erk1 and Erk2 sequestered there. Additional membrane-localized activators of Raf (such as Src family or p21-activated kinases (47,48)), or potential membrane-proximal scaffolding proteins that could coordinate Erk signaling (such as those described for Jun kinase signaling (Ref. 49)), could contribute to activation of this pathway as well. Interestingly, such a process bears striking similarities to the activation of protein kinase A, wherein distinct protein kinase A-anchoring proteins target protein kinase A to different subcellular compartments and thereby determine accessibility to both upstream activators and downstream substrates (reviewed in Ref. 50).
The ability to regulate Raf activation in a dose-dependent and temporal manner should prove to be very useful for studying Raf's role in development. In particular, the magnitude of the induced Raf signal has been postulated to play a key role in determining the biological outcome of signaling via the MAP kinase pathway (51). Using the Lck-Raf-GyrB construct, Rafderived signals can be titrated, permitting direct correlation between the extent of kinase activation and changes in gene expression or protein function. Such information will prove invaluable in furthering our understanding of the molecular mechanisms that underlie cell fate decisions.