Regulation and Role of Raf-1/B-Raf Heterodimerization

Abstract

The Ras-Raf-MEK-extracellular signal-regulated kinase (ERK) pathway participates in the control of many fundamental cellular processes including proliferation, survival, and differentiation. The pathway is deregulated in up to 30% of human cancers, often due to mutations in Ras and the B-Raf isoform. Raf-1 and B-Raf can form heterodimers, and this may be important for cellular transformation. Here, we have analyzed the biochemical and biological properties of Raf-1/B-Raf heterodimers. Isolated Raf-1/B-Raf heterodimers possessed a highly increased kinase activity compared to the respective homodimers or monomers. Heterodimers between wild-type Raf-1 and B-Raf mutants with low or no kinase activity still displayed elevated kinase activity, as did heterodimers between wild-type B-Raf and kinase-negative Raf-1. In contrast, heterodimers containing both kinase-negative Raf-1 and kinase-negative B-Raf were completely inactive, suggesting that the kinase activity of the heterodimer specifically originates from Raf and that either kinase-competent Raf isoform is sufficient to confer high catalytic activity to the heterodimer. In cell lines, Raf-1/B-Raf heterodimers were found at low levels. Heterodimerization was enhanced by 14-3-3 proteins and by mitogens independently of ERK. However, ERK-induced phosphorylation of B-Raf on T753 promoted the disassembly of Raf heterodimers, and the mutation of T753 prolonged growth factor-induced heterodimerization. The B-Raf T753A mutant enhanced differentiation of PC12 cells, which was previously shown to be dependent on sustained ERK signaling. Fine mapping of the interaction sites by peptide arrays suggested a complex mode of interaction involving multiple contact sites with a main Raf-1 binding site in B-Raf encompassing T753. In summary, our data suggest that Raf-1/B-Raf heterodimerization occurs as part of the physiological activation process and that the heterodimer has distinct biochemical properties that may be important for the regulation of some biological processes.

Mitogen-activated protein kinase (MAPK) pathways are important within cells for the transduction of extracellular signals to the nucleus, allowing the appropriate response to be produced. The main components of these pathways are a small G-protein and three kinases; MAPK is activated by a MAPK kinase (MAPKK), which is in turn activated by a MAPKK kinase (MAPKKK). One such pathway is the extracellular signal-regulated kinase (ERK) pathway, where the G protein is Ras, MAPKKK is a Raf protein, MAPKK is MEK (MAPK/ERK kinase), and MAPK is ERK. This pathway relies on the signal being passed via phosphorylation and activation of the kinases ultimately to ERK, which can then affect transcription factors such as ELK, SAP, and other substrates involved in cell growth, differentiation, and survival. The ERK pathway has been found to be hyperactivated in around 30% of all cancers; two members of this pathway, Ras and Raf, are known oncogenes. Ras is frequently mutated in many cancers, including cancers of the pancreas, lung, colon, thyroid, liver, and testis (12). The B-RAF gene has recently been reported to have somatic mutations in 66% of malignant melanomas (7), as well as being implicated in many other human malignancies, including colon cancer, papillary thyroid cancer, and serous ovarian cancer (13, 22).

Raf proteins are activated by Ras, a small GTPase that resides at the cell membrane and becomes activated in response to a large variety of mitogens. The binding of Raf to Ras causes Raf to be translocated from the cytosol to the cell membrane where activation takes place. Activated Raf can then phosphorylate the dual-specificity kinases MEK1 and MEK2, which in turn activate ERK, starting the appropriate response to the extracellular signal. (For recent reviews, see references 5, 8, and 39).

The Raf family consists of three members—A-Raf, B-Raf, and Raf-1. While Raf-1 has been widely studied, less is known about the roles of A-Raf and B-Raf. Although all three Raf isoforms share Ras as a common activator and MEK as a common substrate, differences in their biological function have emerged, mainly through studies with Raf knockout mice (27). A-Raf knockout mice survive to birth but have gastrointestinal and neurological defects (29). Ablation of the B-Raf gene was reported to disturb neuroepithelial differentiation and cause massive apoptosis of endothelial cells, resulting in vascular hemorrhage and death of the embryos in midgestation (41). Raf-1^−/− mice die in utero, due to widespread apoptosis (15, 23). Interestingly, Raf-1^−/− fibroblasts display no alterations in ERK activation, most likely because in this respect B-Raf can fully compensate for the loss of Raf-1 (15, 23). However, Raf-1^−/− cells exhibit an increased susceptibility toward apoptosis against which B-Raf cannot protect and which is due to Raf-1 being required for the suppression of the proapoptotic kinases mammalian sterile 20-like kinase (MST2) (28) and apoptosis signal-regulating kinase 1 (ASK-1) (42).

The structures of the three Raf proteins are similar, each containing three conserved regions (CRs; CR1, CR2, and CR3). CR1 and CR2 are situated in the regulatory N terminus with CR1 containing the Ras binding domain (RBD), while CR3 roughly corresponds to the C-terminus catalytic kinase domain (5). Other regions, however, are more divergent and are likely to account for the differences seen in the regulation and role of the Raf proteins. For instance, in the rat pheochromocytoma PC12 cell line, Raf-1 and B-Raf may play different roles in proliferation and differentiation. Raf-1 is thought to be responsible for the transient activation of the ERK pathway required for proliferation upon epidermal growth factor (EGF) stimulation, while the sustained activation of ERK required for differentiation in response to stimulation with nerve growth factor (NGF) is mediated by B-Raf (17, 34).

There are also differences in the way the Raf proteins are activated (5, 8, 39). Raf-1 activation involves a complex series of changes in phosphorylation, which include the dephosphorylation of an inhibitory site, S259 (1, 9), and the phosphorylation of a critical activating site, S338 (10), as well as phosphorylation of the activation loop (4) for maximal activation. These sites are conserved in A-Raf, and activation seems to follow a pattern similar to that of Raf-1. However, in B-Raf the equivalent of Raf-1 S338 is constitutively phosphorylated. Additionally, Ras alone is sufficient to activate B-Raf, whereas Raf-1 requires other factors (20). In fact, in biochemical purification experiments B-Raf itself was found to be a component of the activator complex required to activate Raf-1, and depleting B-Raf from the complex had a negative effect on Raf-1 activation (25). These results suggested that B-Raf could be involved in the regulation of Raf-1. Weber et al. (38) showed that Ras induced overexpressed Raf-1 and B-Raf to form heterodimers in HEK293 cells. Further, it was demonstrated that low-kinase-activity mutants of B-Raf, which are found in tumors, can activate ERK by signaling through Raf-1. This was assumed to occur via heterodimerization, although the kinase activities of the heterodimers were not examined (36).

Here, we present an analysis of the regulation and function of the Raf-1/B-Raf heterodimers. We show for the first time an interaction between endogenous B-Raf and Raf-1 proteins and that this interaction is mitogen regulated, is enhanced by 14-3-3 proteins, and can be identified in several different cell types. Mapping of the binding sites reveals a complex interaction surface. Heterodimer formation is stabilized by MEK inhibition, suggesting that the disruption of the B-Raf/Raf-1 complex is under negative feedback regulation emanating from ERK or a downstream molecule. We confirm that this regulation is from ERK-induced phosphorylation of T753 in the C terminus of B-Raf. Mutation of T573 results in a longer persistence of Raf-1/B-Raf heterodimers and, in a physiological context, augments the differentiation of PC12 cells induced by NGF. Finally, we show that the B-Raf/Raf-1 heterodimer has vastly elevated kinase activity compared to the respective monomers or homodimers.

MATERIALS AND METHODS

Maintenance and transfection of cell lines.

COS-1 cells were maintained in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum (FBS) and 1% l-glutamine and grown at 37°C in 5% CO₂. COS-1 cells were transfected with Effectene transfection reagent (QIAGEN) according to the manufacturer's instructions. PC12 cells were maintained in Dulbecco's modified Eagle medium supplemented with 10% horse serum, 5% FBS, and 1% l-glutamine and grown at 37°C in 5% CO₂. For differentiation experiments, PC12 cells were plated on polylysine-coated dishes and treated with 50-ng/ml NGF in serum-free medium. Jurkat T cells were maintained in RPMI supplemented with 10% FBS and 1% l-glutamine and grown at 37°C in 5% CO₂. Sf9 cells were maintained in TC100 medium supplemented with 10% FBS, 1% l-glutamine, 0.1% pluronic solution, and 0.1% amphotericin B at 27°C.

Preparation of lysates and immunoprecipitation.

Cells were washed with ice-cold phosphate-buffered saline and lysed in lysis buffer (20 mM HEPES [pH 7.5], 150 mM NaCl, 0.5 mM EGTA, and 0.5% NP-40) supplemented with protease and phosphatase inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 mM Na₃VO₄, 10 mM β-glycerolphosphate, 2 mM sodium pyrophosphate, 5-μg/ml leupeptin, and 2.2-μg/ml aprotinin). Cells were scraped from the plates, and lysates were centrifuged at 12,000 rpm for 10 min at 4°C. Supernatants were added to Sepharose beads and incubated with the appropriate antibody overnight at 4°C. Proteins were separated on 7.5% or 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels.

Site-directed mutagenesis.

B-Raf mutants were generated using a mutagenic PCR kit from Stratagene according to the manufacturer's instructions. The mutants made were S749A, T753A, and G465E using hemagglutinin (HA)-tagged B-Raf as a template.

Peptide arrays.

Synthetic 23-mer peptides offset by four amino acids representing the whole Raf-1 and B-Raf sequences were arrayed on nitrocellulose membranes (Cancer Research United Kingdom Central Service Unit). The arrays were blocked in bovine serum albumin and then incubated with the required Raf isoform (produced in SF9 cells) for 1 h. Bovine serum albumin was used to wash the filter before incubation with the appropriate primary antibody for 1 h. After being washed in phosphate-buffered saline-Tween (0.1%), the filter was incubated with the appropriate secondary antibody for 1 h and washed before being developed with enhanced chemiluminescence. To reprobe the filter, the previous protein was stripped off by incubation in 50% ethanol-10% glacial acetic acid for 20 min and a buffer containing 50 mM Tris-HCl (pH 8), 10 mM β-mercaptoetanol, and 10% SDS for 20 min.

In vitro transcription-translation (IVT) and Raf-1/B-Raf binding assays.

In vitro translation was performed using the Promega TNT T7 Quick Coupled Transcription/Translation system. Cold 50-μl reaction mixtures were incubated at 30°C for 90 min as described in the kit. Proteins were used immediately and without any purification. For binding reactions, 10 μl (each) of HA-B-Raf and Flag-Raf-1 transcription-translation reaction mixtures were mixed with 30 μl of lysis buffer and incubated at 30°C for 20 min to allow heterodimer formation. After this incubation, the binding reactions were diluted with 350 μl lysis buffer, and immunoprecipitations were carried out with either HA or Flag antibodies at 4°C for 3 h. The immunoprecipitates were washed three times in lysis buffer and separated by SDS-PAGE. For binding reactions in the presence of 14-3-3 proteins, eluted glutathione S-transferase (GST)-tagged 14-3-3ζ proteins purified from Escherichia coli were added to the binding reaction mixture incubated at 30°C before immunoprecipitation as described above. The dimerization-negative 14-3-3ζ (33) contains mutations (E5K, L12AE to Q12QR, Y82Q, K85N, and E87Q) in the dimerization interface and was kindly provided by G. Tzivion.

Heterodimer-homodimer kinase assays.

COS-1 cells were transiently transfected with Flag-Raf-1 and HA-B-Raf or corresponding mutants. Cells were serum starved overnight and stimulated for 20 min with 100-ng/ml tetradecanoyl phorbol acetate (TPA). Cells were lysed as described above. To isolate B-Raf-Raf-1 heterodimers, Flag-Raf-1 was immunoprecipitated for 4 h with Flag-M2 agarose (Sigma). Lysate was removed from the beads and retained, and beads were washed three times with lysis buffer. Flag-Raf-1 was eluted from the beads with Flag peptide (200 ng/ml; Sigma) on ice with two 30-min incubations. Eluates were pooled and diluted to 0.5 ml with lysis buffer containing fresh protease and phosphatase inhibitors. Subsequently, B-Raf proteins were immunoprecipitated from the eluates overnight using HA antibody. Beads were washed three times in lysis buffer to give the purified heterodimer sample. The B-Raf homodimer-monomer sample was prepared from the Flag-Raf-1-depleted lysate. The lysate was diluted 1:5 and immunoprecipitated with HA antibody overnight. Beads were washed three times with lysis buffer. The Raf-1 homodimer-monomer sample was prepared from HA-B-Raf-depleted lysate. Cell lysate was diluted 1:5, and B-Raf was depleted by HA immunoprecipitation for 4 h. After this time, the supernatant was removed, fresh protease and phosphatase inhibitors were added, and Raf-1 protein was immunoprecipitated overnight with Flag-M2 agarose. Beads were washed three times with lysis buffer to give purified Raf-1 homodimers-monomers. Heterodimer and homodimer samples were adjusted to contain similar B-Raf levels; generally, this involved a 1:30 dilution of homodimer beads. In vitro kinase assays were performed incubating heterodimer and homodimer beads, GST-MEK, 100 μM ATP, and 10 mM MgCl₂ in kinase assay buffer (50 mM Tris [pH 7.5], 150 mM NaCl, 5 mM EGTA, 5 mM MgCl₂) at 32°C for 20 min. Reactions were resolved on 7.5% SDS-PAGE gels and blotted. MEK phosphorylation was detected by Western blotting with a phospho-specific MEK antibody.

RESULTS

B-Raf and Raf-1 form stimulation-induced heterodimers.

Previous studies have demonstrated that growth factors and activated Ras stimulate heterodimerization of exogenously expressed B-Raf and Raf-1 (36, 38), but they did not investigate the nature of the interaction and its biochemical and physiological consequences. Setting out to address these open questions, we first confirmed the previous findings using Flag-tagged Raf-1 and HA-tagged B-Raf overexpressed in COS-1 cells. Raf-1 was immunoprecipitated with the Flag antibody, and the immunoprecipitate was blotted for B-Raf (Fig. 1A). Clearly, B-Raf and Raf-1 could interact. Notably, this interaction could be enhanced by mitogen stimulation, such as TPA.

FIG. 1.

B-Raf and Raf-1 form stimulation-inducible heterodimers. (A) Coimmunoprecipitation of overexpressed Raf proteins. COS-1 cells were transiently transfected with Flag-tagged Raf-1 and HA-tagged B-Raf. Cells were starved overnight and stimulated with TPA for 10 min. Lysates were immunoprecipitated with the Flag antibody and probed for associated B-Raf or immunoprecipitated with the HA antibody and probed for associated Raf-1. (B) Endogenous heterodimers are formed in COS-1 cells. Starved COS-1 cells were stimulated with 100-ng/ml TPA for 10 min or EGF for 5 min. B-Raf/Raf-1 heterodimers were detected by immunoprecipitation with a B-Raf antibody, followed by immunoblotting for Raf-1 and vice versa.

As previous work has been done with overexpressed proteins, we examined whether endogenously expressed Raf-1 and B-Raf proteins also could interact. Endogenous Raf-1/B-Raf heterodimers were detected in several different cell lines including COS-1 (monkey embryonic kidney), PC12 (rat pheochromocytoma), Jurkat T cells (human T cells), HCT116 (human colon), and MCF7 (human breast) (Fig. 1B; see Fig. 5A and B and data not shown), suggesting that this is a general phenomenon. In all cases, heterodimer formation could be enhanced by growth factors such as TPA or EGF. However, the kinetics and extent of heterodimerization were different, probably reflecting cell type-specific differences.

FIG. 5.

Heterodimer persistence is regulated by ERK signaling. (A) Treatment of PC12 with the MEK inhibitor U0126 causes an increase in heterodimer association. PC12 cells were serum starved overnight and then incubated with 10 μM U0126 for 1 h before stimulation with NGF for the times indicated. Endogenous heterodimers were detected by immunoprecipitation with a Raf-1 antibody, followed by blotting for B-Raf. Lysates were immunoblotted with antibodies against phospho-ERK to show the inhibition of ERK phosphorylation by U0126 and with antibodies against ERK, B-Raf, and Raf-1 to assure equal loading. (B) The same experiment as that in panel A is shown, with Jurkat T cells stimulated with 100-ng/ml TPA. (C) Dominant negative MEK blocks ERK activation and promotes Raf-1 heterodimerization with B-Raf. COS-1 cells were transfected with vector or MEKAA, a dominant negative MEK-1 mutant where the activating phosphorylation sites S218/S222 are mutated to alanines. Heterodimers were detected by immunoprecipitation of endogenous B-Raf and immunoblotting for endogenous Raf-1. MEKAA efficiently blocked ERK activation, as shown by blotting lysates for phospho-ERK. Lysates were also immunoblotted for the expression of endogenous total ERK, Raf-1, and B-Raf, as well as transfected MEKAA, as indicated. (D) ERK induces phosphorylation of T753 in B-Raf. COS cells were transfected with HA-B-Raf or HA-B-Raf T753A. Cells were serum starved overnight and then incubated with 10 μM U0126 for 1 h before stimulation with 20-ng/ml EGF for the times indicated. B-Raf proteins were immunoprecipitated with HA antibody. Phosphorylation of T753 was detected by immunoblotting with a phosphothreonine-proline-specific antibody. B-Raf proteins were visualized with HA antibody. (E) Sustained heterodimerization of B-Raf T753A with Raf-1. COS-1 cells were transiently transfected with Flag Raf-1 and either HA-B-Raf or the HA-B-Raf T753A mutant. Cells were starved overnight and stimulated with 20-ng/ml EGF over a time course of 0 to 120 min. Cells were lysed, immunoprecipitated with an HA antibody, and blotted for associated Raf-1.

The kinase activity of the Raf-1/B-Raf heterodimer.

Having shown that Raf-1 and B-Raf can heterodimerize in response to mitogens raised the question whether this would affect kinase activity. In previous work, the catalytic activity of heterodimers was inferred indirectly by measuring MEK or ERK phosphorylation induced in cells in response to the overexpression or small inhibitory RNA knockdowns of Raf-1 and B-Raf (36, 38). We deemed it important to measure the activity of the heterodimer directly and devised a sequential immunoprecipitation strategy to isolate Raf-1/B-Raf heterodimers and the corresponding homodimers or monomers (Fig. 2A). COS-1 cells were transiently transfected with HA-tagged B-Raf and Flag-tagged Raf-1, starved, and stimulated with TPA. In the first step, Flag-Raf-1 was immunoprecipitated with the Flag antibody and eluted from the beads with Flag peptide. The Flag eluate was immunoprecipitated with HA antibody to obtain the Raf-1/B-Raf heterodimer. HA-B-Raf homodimers or monomers were immunoprecipitated with the HA antibody from the Flag-Raf-1-depleted lysates. Flag-Raf-1 homodimers or monomers were immunoprecipitated from lysates that had first been depleted with an HA antibody to remove B-Raf. Then, the kinase activity of these three Raf fractions was measured by testing their ability to phosphorylate recombinant purified MEK in vitro. These fractions were not corrected for the content of Raf proteins (Fig. 2B, top). Under these conditions, the B-Raf fraction had the highest basal activity, and the Raf-1/B-Raf heterodimer had the highest stimulated activity. The activity of the Raf-1 fraction was very low, suggesting that the main contribution to MEK activation was made by B-Raf and the Raf-1/B-Raf heterodimer. However, Western blotting revealed that the Raf protein content was unequal, with the heterodimer containing much less B-Raf and Raf-1 than the other fractions. As B-Raf appeared to be the strongest contributor to kinase activity, the fractions were diluted to adjust to similar B-Raf levels. Under these conditions (Fig. 2B, bottom), the Raf-1/B-Raf heterodimer had a substantially higher level of kinase activity than B-Raf homodimers-monomers containing similar amounts of B-Raf, suggesting that B-Raf and Raf-1 cooperate to drastically elevate their kinase activity when forming heterodimers. These results further suggest that Raf-1/B-Raf heterodimers are a main source of Raf kinase activity in cells, although of much lower abundance than the B-Raf homodimers-monomers. As estimated by averaging five experiments, the ratio between Raf-1/B-Raf heterodimers and B-Raf homodimers-monomers was approximately 1:300. A similar share of Raf kinase activity was contributed by the more abundant B-Raf homodimers-monomers. Thus, these results underscore the importance of heterodimerization but are also consistent with the findings that B-Raf can effectively activate the ERK pathway in Raf-1 knockout cells (15, 23).

FIG. 2.

Raf-1/B-Raf heterodimers possess increased kinase activity. (A) Experimental scheme for the isolation of Raf-1/B-Raf heterodimers and the corresponding homodimers or monomers. (B) Comparison of the specific in vitro kinase activity of Raf-1/B-Raf heterodimers and the corresponding monomers-homodimers. COS-1 cells were transfected with the indicated expression plasmids and serum starved overnight before treatment with 100-ng/ml TPA for 30 min as indicated. Raf protein complexes were isolated as described in panel A and Materials and Methods. Their kinase activity was determined by in vitro kinase assays with recombinant kinase-negative GST-MEK as substrate. MEK phosphorylation was detected with a phosphospecific MEK antibody. The levels of Raf-1 and B-Raf proteins in the immunoprecipitates were visualized by Western blotting with HA or Flag antibodies. (Bottom) The concentrations of the protein complexes were adjusted to similar B-Raf levels. (C) COS-1 cells were transiently cotransfected with Flag-Raf-1 or Flag-Raf-1 K375M (kinase negative) and the HA B-Raf wild-type, G465E, or K482M (kinase-negative) mutants. Heterodimers and B-Raf monomers-homodimers were isolated; after B-Raf levels were balanced, their kinase activity was determined in vitro. MEK phosphorylation was detected with a phosphospecific MEK antibody.

Therefore, we sought to determine the source for the increased kinase activity of the Raf-1/B-Raf heterodimer by coexpressing different Raf-1 and B-Raf mutants (Fig. 2C). Raf-1/B-Raf heterodimers and monomeric-homodimeric B-Raf were isolated from TPA-stimulated COS cells by the sequential immunoprecipitation steps described above. Levels of B-Raf were balanced, and activities were assayed. Raf-1/B-Raf heterodimers formed by the kinase-negative mutants Raf-1 K375M and B-Raf K482M lacked any detectable ability to phosphorylate MEK, indicating that MEK phosphorylation is solely due to Raf kinase activity and not mediated by a coprecipitating kinase. B-Raf has a much higher specific catalytic activity than Raf-1 (Fig. 2B) (2), implicating B-Raf as the main source of kinase activity in the heterodimer. Therefore, we investigated the effect of reducing or eliminating B-Raf activity from the heterodimer. For this purpose, we coexpressed Flag-Raf-1 with wild-type HA-B-Raf, the B-Raf low-activity mutant G465E (36), and the B-Raf kinase-negative K482M mutant. B-Raf G465E and B-Raf K482M monomers/homodimers had severely reduced kinase activity. However, when heterodimerized with Raf-1, both B-Raf G465E and K482M mutants displayed kinase activity to a level similar or slightly reduced from that of wild-type B-Raf/Raf-1 heterodimers. This indicates that Raf-1 makes a major contribution to the kinase activity of the Raf-1/B-Raf heterodimer. B-Raf catalytic activity is not essential, and it is the presence of B-Raf that is required, not its kinase activity. The same was true when the activity of a heterodimer between the kinase-negative Raf-1 K375M mutant and B-Raf was assayed. Thus, either a kinase-competent Raf-1 or B-Raf isoform can confer high activity to the heterodimer even when its dimerization partner is defective in catalytic activity.

Interaction stoichiometries and mapping of interaction domains.

Based on the results shown above, the Raf-1/B-Raf heterodimer seems able to contribute substantially to MEK kinase activity in cells. Thus, we wanted to characterize the stoichiometries and interaction domains. To evaluate the fraction of Raf proteins participating in heterodimers formation, we performed in vitro binding assays using B-Raf and Raf-1 proteins produced by coupled IVT (Fig. 3A). Raf-1/B-Raf heterodimerization was readily detectable in the IVT system, with approximately 8% of Raf-1 and 7% of B-Raf produced being recovered in heterodimers. In Raf complexes isolated from mitogen-stimulated cells, these ratios were substantially lower (Fig. 3B), with Raf-1/B-Raf heterodimers containing, on average, 0.1% of the total cellular B-Raf and 0.3% of the total Raf-1 pool, suggesting that this interaction is tightly controlled in cells. Both Raf-1 (37) and B-Raf (L. K. Rushworth, E. O'Neill, and W. Kolch, unpublished data) exist in multiprotein complexes. Given the tendency of Raf proteins to become insoluble when purified away from their chaperones, it is thus extremely difficult to determine the stoichiometries of Raf proteins engaged in heterodimer complexes. Sucrose gradient fractionation showed that cotransfection of Raf-1 shifted the size distribution of B-Raf complexes and vice versa, but there was no simple relationship that would have permitted estimation of the stoichiometries of Raf-1 and B-Raf in heterodimers (data not shown). A better estimation was possible from the separation of HA-B-Raf and Flag-Raf-1 heterodimers and B-Raf homodimers-monomers prepared from TPA-stimulated COS-1 cells by blue native gel electrophoresis (31). The observed sizes of Raf-1/B-Raf complexes migrating slightly above 250 kDa and B-Raf complexes spreading between 150 to 250 kDa makes it unlikely that B-Raf (90 kDa) and Raf-1 (74 kDa) form complexes of more than a 1:1 (i.e., dimeric) stoichiometry (data not shown).

FIG. 3.

Binding stoichiometries and mapping of interaction domains. (A) Raf proteins were produced by coupled IVT. As indicated, Flag-Raf-1 or HA-B-Raf proteins were immunoprecipitated and analyzed by Western blotting alongside the IVT reactions (input) with Flag-specific (Raf-1) and HA-specific (B-Raf) antibodies. The blots were quantified by laser densitometry, showing that approximately 8% of Raf-1 and 7% of B-Raf produced were recovered in heterodimers. (B) Heterodimers were isolated from TPA (100 ng/ml; 10 min)-stimulated COS cells transfected with Flag-Raf-1 and HA-B-Raf as described in Fig. 2A and analyzed by Western blotting alongside a dilution series of cell lysates. Quantification by laser densitometry showed that Raf-1/B-Raf heterodimers contained, on average, 0.1% of the total cellular B-Raf and 0.3% of the total Raf-1 pool. Shown are representatives of three experiments. (C) B-Raf/Raf-1 interaction requires the Raf-1 kinase domain and is modulated by mutation of K375M or S621. COS-1 cells were cotransfected with B-Raf and the indicated Flag-tagged Raf-1 constructs. FL, full-length; C-term, C-terminal kinase domain. B-Raf immunoprecipitates were stained with anti-Flag antibody to detect bound Raf-1 proteins, followed by the B-Raf H145 antibody, to assure similar loading of B-Raf proteins. To control for equal expression of Flag-Raf-1 proteins, cell lysates were immunoblotted with Flag antibody. (D) Peptide arrays that represent the entire Raf-1 and B-Raf coding sequences as 23mer peptides offset by four amino acids were probed with Raf-1 and B-Raf proteins produced in Sf9 insect cells. Raf-1 and B-Raf proteins are schematically shown with the following motifs indicated: CR1 to CR3, conserved regions; ATP, ATP binding loop; activation loop, delineated by DFG and APE amino acid sequences; RBD, Ras binding domain; CBD, cysteine-rich domain; RKIP, minimal RKIP binding domain. Phosphorylation sites are shown in blue. Proteins used to probe the arrays are boxed, and the mapped sites of sites of interaction are indicated by red (high-affinity) and green (medium-affinity) squares with amino acid numbers.

To gain better insight into the nature of the interaction between Raf-1 and B-Raf, we then determined the binding domains. In cotransfection experiments with Raf-1 mutants and truncation proteins, we identified the Raf-1 kinase domain as sufficient to mediate the interaction with B-Raf (Fig. 3C). As previously reported (38) the mutation of S621 in the Raf-1 C terminus compromised heterodimerization, suggesting that 14-3-3 binding to this site may be required for efficient heterodimerization. Mutation of the adjacent S624 had no effect. On the other hand, mutation of S259, another major 14-3-3 binding site in Raf-1, had no effect on heterodimerization. As Raf-1 S621A is also almost completely devoid of kinase activity (24, 26) and the true kinase-negative mutant Raf-1 K375M also was equally compromised to associate with B-Raf, an intact catalytic activity may be required for efficient heterodimerization. However, mutation of SS338/339, which also severely compromises Raf-1 kinase activity (10), did not affect binding to B-Raf, suggesting that reduction of heterodimerization probably cannot be attributed only to the loss of kinase activity.

These results indicated that the interaction between Raf-1 and B-Raf is regulated in a more complex fashion. To gain more insight into the interaction sites, we fine mapped the interaction domains for homo- and heterodimerization in vitro by peptide arrays that represent the entire Raf-1 and B-Raf coding sequences as 23mer peptides offset by four amino acids. These arrays were probed with Raf-1 and B-Raf proteins produced in Sf9 insect cells (see Fig. S1 in the supplemental material). The results confirmed that Raf-1 and B-Raf can form hetero- and homodimers, both modes of interactions involving several sites of contact (Fig. 3D). As expected from the cotransfection results (Fig. 3C), regions for high- and medium-affinity heterodimerization were found in the kinase domains. B-Raf binds strongly to amino acids (aa) 301 to 315 in Raf-1, which constitute a nonconserved region just upstream of the N region. These are the N-terminal boundaries of the minimal Raf kinase inhibitor protein (RKIP) binding domain in Raf-1 (43), suggesting that RKIP may regulate heterodimer formation. However, under the conditions tested we could not obtain conclusive evidence that RKIP regulates Raf heterodimerization (data not shown). An additional high-affinity binding site where B-Raf binds to Raf-1 is located between the ATP binding and activation loops (aa 365 to 387). An equivalent region in B-Raf is also prominently involved in mediating B-Raf homologous interactions. Further contact regions mediating both Raf-1 homologous and B-Raf/Raf-1 interactions are situated just upstream of the activation loop. Another region downstream of the activation loop in Raf-1 and B-Raf seems to be involved in heterodimerization. Raf-1 binding to itself and B-Raf also involves the C-terminal 14-3-3 binding domain. This result is consistent with the observation that the mutation of Raf-1 S621, the 14-3-3 docking site in this domain, impairs heterodimerization with B-Raf (38). Importantly, Raf-1 strongly bound to aa 737 to 766 in B-Raf, which contains T753, an ERK-induced phosphorylation site that destabilizes the heterodimer (see Fig. 5). Rather surprisingly, the peptide arrays also detected high-affinity interaction sites in the RBD and cysteine-rich domains in the N-terminal region. Both these regions are implicated in regulating Ras binding and Ras-mediated activation (6, 19, 35). Judged by their density distribution, these binding sites seem to be more important for homodimerization than heterodimerization. This would support a model where Ras could break up homodimers to promote heterodimerization. Activated Ras has been reported to induce Raf-1/B-Raf heterodimerization (38), which is in keeping with our findings (data not shown), but also Raf-1 homodimerization (16).

In summary, the peptide array data show that both Raf homodimer and heterodimer formation uses a complex interface with multiple contact sites. Many of these sites participate both in homo- and heterodimerization, and only a few seem uniquely involved in either homo- or heterodimerization, suggesting that homo- and heterodimers are mutually exclusive and may present Raf proteins in different conformations.

Heterodimer formation is regulated by 14-3-3 proteins.

Raf proteins associate with 14-3-3 proteins, and the mutation of the 14-3-3 docking site S621 in Raf-1 compromises heterodimerization with B-Raf (38). Therefore, we investigated whether 14-3-3 can enhance heterodimerization. As 14-3-3 proteins typically use phosphoserine or phosphothreonine as docking sites, such interactions would not be detected by the peptide array analysis. Coexpression of increasing amounts of GST-14-3-3 with Raf-1 and B-Raf in COS cells resulted in a dose-dependent stimulation of Raf heterodimerization (Fig. 4A). This correlated with an increase in 14-3-3 binding to Raf proteins. Similar results were observed in vitro. The addition of purified GST-14-3-3 to Raf-1 and B-Raf proteins produced by in vitro transcription-translation also augmented heterodimerization and concomitant 14-3-3 binding (Fig. 4B). 14-3-3 proteins are stable, obligatory dimers. A dimerization-negative 14-3-3 mutant (33) failed to enhance Raf protein heterodimerization, consistent with the interpretation that 14-3-3 functions as a cross-linker.

FIG. 4.

14-3-3 proteins enhance Raf heterodimerization. (A) COS cells were cotransfected with Flag-Raf-1, HA-B-Raf, and increasing amounts of GST-14-3-3. Lysates from serum-starved cells were immunoprecipitated with Flag or HA antibodies as indicated and examined for the presence of Raf-1, B-Raf, and 14-3-3 by immunoblotting with the indicated antibodies. As a control for protein expression, lysates were immunoblotted as well. (B) Raf proteins were produced by coupled IVT. As indicated, increasing amounts of purified GST-14-3-3 produced in E. coli were added. Flag-Raf-1 or HA-B-Raf proteins were immunoprecipitated and analyzed by Western blotting with Flag- and HA-specific antibodies. DN 14-3-3, dimerization-negative 14-3-3 mutant.

ERK-mediated phosphorylation of B-Raf harnesses B-Raf:Raf-1 interactions.

In the course of our experiments, we noticed that the formation of Raf-1/B-Raf heterodimers was enhanced by mitogenic stimulation independent of ERK but that ERK activation could induce a negative feedback that promoted the dissociation of the heterodimer. The kinetics of this feedback seemed to vary, depending on the cell line and stimuli used (data not shown). However, we consistently observed that two different Raf kinase inhibitors (data not shown) and the MEK inhibitor U0126 stabilized heterodimer levels; an example is shown in Fig. 5A. PC12 cells were treated with U0126 and then stimulated with NGF for 5 or 30 min. Western blot analysis of lysates with a phospho-ERK antibody showed that U0126 was effective at inhibiting ERK activation. Notably, U0126 increased the levels of heterodimers in NGF-treated samples, suggesting that ERK activation exerts a negative feedback that can destabilize Raf heterodimers. Similar results were observed with Jurkat T cells (Fig. 5B). U0126 completely ablated the TPA induction of phospho-ERK and markedly increased heterodimerization. Furthermore, the expression of a dominant negative MEK mutant in growing COS-1 cells downregulated ERK activation and promoted Raf-1/B-Raf heterodimerization (Fig. 5C). These results suggest that ERK signaling is not required for Raf-1/B-Raf heterodimerization but can destabilize mitogen-induced heterodimers.

The C terminus of B-Raf contains threonine 753 within a typical consensus motif for ERK phosphorylation and has been described as ERK phosphorylation site by Brummer et al. (3). We found that this residue is phosphorylated in cells in an ERK-dependent manner, as detected by staining HA-B-Raf immunoprecipitates with a phosphothreonine-proline-specific antibody (Fig. 5D). Phosphorylation is induced by EGF and blocked by U0126. Importantly, no phosphorylation of the B-Raf T753A mutant was detectable, demonstrating the specificity of the antibody and correct assignment of the site. To investigate the contribution of ERK phosphorylation of B-Raf to dissociation of the Raf-1/B-Raf heterodimer, we transiently transfected COS-1 cells with either wild-type B-Raf or B-Raf T753A and monitored heterodimer formation in response to EGF over a time course of 0 to 120 min (Fig. 5E). In response to stimulation, wild-type B-Raf bound to Raf-1, but this association declined quickly and was back to basal levels between 30 and 60 min. Raf-1 binding to B-Raf T753A was also stimulated by EGF, showing that this site is not required for heterodimerization. However, the heterodimer persisted much longer and was still present at the end of the observation period. This finding indicates that ERK phosphorylation of B-Raf at T753 exerts a negative feedback on the persistence of the Raf-1/B-Raf heterodimer.

The B-Raf T753A mutant accelerates PC12 cell differentiation.

The differentiation of PC12 cells into neuron-like cells requires the sustained activity of the ERK pathway that is achieved by NGF but not EGF. Thus, this system is ideal to examine whether the stability of the Raf-1/B-Raf heterodimer is relevant for the regulation of a physiological process. PC12 cells were transfected with B-Raf or the B-Raf T572A mutant and treated with EGF or NGF for 36 h. GFP was cotransfected to identify transfected cells and monitor differentiation. Cells were counted as differentiated when they extended neurites longer or equal to the double diameter of the cell body. The differentiation in response to NGF was slightly (average, ∼8%) but statistically significantly (P = 0.016) enhanced by the expression of wild-type B-Raf. The B-Raf T753A mutant increased differentiation by a further ∼14% (P = 0.004). Neither wild-type B-Raf nor B-Raf T753A significantly affected the inability of EGF to differentiate the cells, showing that the B-Raf T753A mutant retained its biological specificity. These data complement findings by Brummer et al. (3) that replacement of S750 by phosphomimetic glutamic acid reduced the ability of B-Raf to differentiate PC12 cells (Fig. 6).

FIG. 6.

Mutation of B-Raf T753 enhances the differentiation of PC12 cells. (A) PC12 cells were transiently cotransfected with GFP plus either empty vector, B-Raf WT, or B-Raf T753A. Cells were serum starved overnight and stimulated with EGF or NGF for 36 h. Cells were fixed with formaldehyde and photographed under a fluorescent microscope. Six random pictures of each condition were taken, and the percentage of differentiated, GFP-positive cells was calculated. Cells with neurites at least two cell bodies long were considered differentiated. (B) The results of three independent experiments were plotted as a bar graph. Statistical significance was evaluated by Student's paired t test.

DISCUSSION

In this paper, we demonstrate for the first time an endogenous interaction between B-Raf and Raf-1 to form heterodimers. The basal level of association is enhanced by mitogen stimulation, and we found this in all cell types analyzed, including PC12, COS-1, HCT116, MCF-7, and T cells. The kinetics of heterodimer formation and persistence are variable, however, depending on the cell type and stimulus used. This suggests that the increase in heterodimerization is part of the activation process of Raf proteins.

Based on immunoprecipitation experiments, <1% of cellular Raf-1 and B-Raf is found in heterodimers. These low stoichiometries under the steady-state conditions of immunoprecipitation could be due to low affinity for each other, fast turnover, or sequestration into different complexes and subcellular compartments. Unfortunately, we were not able to perform quantitative measurements of affinities and on-off rates, as we currently cannot purify full-length Raf proteins to the purity and concentration required for such experiments. The expression yields of Raf-1 and B-Raf in commonly used expression systems, including Sf9 insect cells, Pichia pastoris, and E. coli, are very low, and the Raf proteins rapidly lose solubility when deprived of their copurifying chaperones HSP90, HSP70, and CDC37. Nevertheless, the heterodimer seems to be physiologically important, as suggested by its very high catalytic activity, its rather complex regulation, and the observation that a B-Raf mutant, B-Raf T753A, which causes accumulation of the heterodimer, enhances the differentiation of PC12 cells. Interestingly, the KSR1 scaffolding protein, which enhances ERK activation by MEK, also engages only a small fraction of MEK and ERK. Reconstitution experiments of KSR1 knockout cells with increasing amounts of transfected KSR1 showed that ERK activation became maximal when KSR1 was overexpressed ∼14 fold above endogenous levels. However, even under these optimal conditions <5% of MEK and ERK were bound to KSR1 (30). This example shows that even small changes in the efficiency of an activation step within the Raf-MEK-ERK kinase cascade can translate into a strong effect on ERK activity.

It has to be noted that in mitogen-stimulated cells, B-Raf becomes efficiently activated as a MEK kinase, independently of its participation in the heterodimer (Fig. 2B). Only when adjusted to similar B-Raf levels does the much higher catalytic activity of the heterodimer becomes apparent. Thus, B-Raf in cells may contribute significantly to MEK activation independently of Raf-1. This view is supported by the observation that in Raf-1 knockout cells the activation of the MEK-ERK pathway is apparently normal, presumably due to compensation by B-Raf (15, 23). Thus, it is attractive to speculate that Raf-1/B-Raf heterodimers modulate the signaling strength or fulfill specialized functions, rather than affecting ERK activation globally. This hypothesis is consistent with our observation that extending the lifetime of the heterodimer by transfecting B-Raf T753A selectively supported NGF-driven differentiation in PC12 cells, which relies on the sustained activation of ERK, but did not change the response to EGF, which produces transient ERK activation without differentiation.

It becomes increasingly apparent that the protein components of signaling pathways are organized in multiprotein complexes that can carry out context-dependent cellular functions, determined by their interaction partners and subcellular localization (18). Both Raf-1 (37) and B-Raf (Rushworth et al., unpublished) exist in multiprotein complexes, which when analyzed by mass spectrometry contain >40 proteins, including 14-3-3 proteins and the chaperones HSP90, HSP70, and CDC37, which are important to keep Raf proteins folded and soluble (32). According to sucrose gradient fractionation, Raf complexes are spread out between 70 and >500 kDa. Cotransfection of Raf-1 shifts B-Raf complexes to higher molecular weights, whereas cotransfection of B-Raf shifts Raf-1 complexes to smaller size (data not shown). Thus, presumably due to changes in other association partners, there is no simple relationship between Raf-1/B-Raf heterodimerization and the size of the respective protein complexes. This raises the question whether the association between Raf-1 and B-Raf is direct or mediated through other proteins. 14-3-3 proteins can increase heterodimerization, but there is also a constitutive or basal level of heterodimerization. Although the impediments to purify full-length Raf proteins to homogeneity prevented association studies with highly purified proteins, several lines of evidence indicate that heterodimerization involves direct interactions between Raf-1 and B-Raf. First, partially pure Raf-1 and B-Raf proteins produced by in vitro transcription-translation (Fig. 3A) or in Sf9 insect cells (data not shown) readily form heterodimers. Second, partially pure Raf proteins produced in Sf9 insect cells bind to peptide arrays representing the Raf-1 and B-Raf amino acid sequences (Fig. 3D). Third, the enhancement of heterodimerization conferred by 14-3-3 proteins indicates direct interactions between Raf-1 and B-Raf (see below).

Interestingly, Raf-1/B-Raf heterodimerization is regulated. We describe here two mechanisms, 14-3-3 and ERK feedback phosphorylation. 14-3-3 proteins are ubiquitously expressed adaptor proteins that typically dock to phosphoserine or phosphothreonine residues, although phosphorylation-independent modes of binding have been previously described (40). Despite a wealth of literature describing the interaction of 14-3-3 proteins with Raf-1 and B-Raf, their role in Raf protein regulation is still unclear. A simplified consensus view suggests that 14-3-3 proteins can stabilize both inactive and activated conformations of Raf proteins (33). Our results show that 14-3-3 can enhance Raf-1/B-Raf heterodimerization. This is dependent upon the intactness of the dimerization site within 14-3-3 proteins, suggesting that in this scenario 14-3-3 acts as a true bridging molecule that cross-links Raf-1 to B-Raf. Consistent with this view is the fact that the mutation of S621, a 14-3-3 binding site in Raf-1, reduced heterodimerization. However, this reduction also could be related to an inability to bind ATP, as the kinase-negative Raf-1 K375M mutant (which is defective in ATP binding) also exhibited a similar reduction of association. S621 is also a Raf-1 autophosphorylation site (24), and its mutation renders Raf-1 inactive. Thus, the classical mutational approach breaks down; without structural information, it will be difficult if not impossible to distinguish between these possibilities.

Further, when studying what regulated the Raf-1/B-Raf interaction, we noticed that the MEK inhibitor U0126 and two different Raf inhibitors stabilized the concentration of Raf-1/B-Raf heterodimers in all cell lines examined, albeit with different kinetics and efficiencies (Fig. 5A and B and data not shown). Our findings suggest a dual role for mitogens in Raf-1/B-Raf heterodimerization. First, mitogens induce heterodimer formation independent of ERK, but once they are established they destabilize them via an ERK-mediated pathway. The destabilization of Raf-1/B-Raf heterodimers seems to be mediated at least in part by ERK phosphorylating B-Raf on T753, thus constituting a direct negative feedback. T753 is located within a main interaction site where Raf-1 binds to B-Raf (Fig. 3D); it is conceivable that phosphorylation at this site disrupts the interaction. The possibility that the phosphorylation of T753 and the adjacent S750 are part of a negative ERK feedback loop has been raised by Brummer et al. (3). Here, we show that a target of this feedback is the destabilization of the Raf-1/B-Raf heterodimer. This coupling of stimulation with a negative feedback could serve as a timing device that ensures that heterodimers are dissociated after an appropriate response time. It should be noted that ERK also phosphorylates Raf-1 as part of a negative feedback loop (11). Whether this may influence heterodimer formation has not been investigated here.

The fact that the formation of the Raf-1/B-Raf heterodimer is regulated on several levels suggests that it fulfils a physiological function that is either qualitatively or quantitatively different from that of the corresponding homodimers or monomers. One such function was inferred from results that B-Raf mutants with low levels of kinase activity could form heterodimers with Raf-1 and efficiently stimulate the ERK pathway if Raf-1 was present in the cell (36). As low-kinase-activity B-Raf mutations are found in human tumors, this observation indicates a potentially important role for Raf-1 in transformation by B-Raf mutants. In the study by Wan et al. (36), the kinase activity of the Raf-1/B-Raf heterodimers was not determined directly. Thus, it remained open whether this enhancement was due to heterodimerization or whether B-Raf stimulated Raf-1 activity through another mechanism. Our finding that the Raf-1/B-Raf heterodimer has enhanced kinase activity supports the former interpretation. Importantly, B-Raf mutants with low intrinsic kinase activity could stimulate the kinase activity of the heterodimer to an extent similar to that of wild-type B-Raf. The in vitro heterodimer kinase assays show that the biochemical effect of forming heterodimers is to vastly elevate the catalytic activity of the heterodimer and that Raf-1 makes a major contribution to this. Recent work has suggested that in many cells the major MEK phosphorylating protein is B-Raf (15, 23) and that the comparably low kinase activity of Raf-1 lends itself to a scaffolding role (28) or a role in signaling to other, as-yet-unidentified substrates (14). Our results suggest that Raf-1 can play an important role in signaling to MEK by a process that requires B-Raf, but in a purely interactive way rather than as a kinase. The detailed mechanism is not clear at present. According to our results, B-Raf activity seems not to be required for the superactivation of Raf-1 in the heterodimer, suggesting that B-Raf confers a conformational change on Raf-1 that allows its stimulation, or it brings a protein into the complex which can activate Raf-1. Interestingly, the same seems to hold true in the reverse, as the kinase-negative Raf-1 K375M mutant could efficiently induce MEK phosphorylation via B-Raf (Fig. 2C). Thus, either kinase-competent Raf isoform is sufficient to confer high catalytic activity on the Raf-1/B-Raf heterodimer. In Raf-1^−/− fibroblasts, ERK activation is normal in response to all stimuli tested, which has been attributed to compensation by B-Raf (15, 23). This finding suggests that wild-type B-Raf does not rely on Raf-1 or that it can use an alternative mechanism to enhance ERK activation. Interestingly, our recent proteomics experiments have shown that A-Raf can also be coimmunoprecipitated with both B-Raf and Raf-1 (unpublished data). Thus, other combinations of Raf heterodimers or indeed heterotrimers may form, and A-Raf may be able to substitute for Raf-1 as a heterodimerization partner for B-Raf. A-Raf and Raf-1 have recently been shown to cooperate in the transient activation of ERK, although sustained activation was unaffected (21). Interestingly, the Raf-1 and A-Raf double knockout slowed down proliferation, suggesting that Raf-1 or A-Raf are redundant for this function, possibly because either one can cooperate with B-Raf.

Thus, Raf isozyme heterodimerization begins to emerge as an important regulatory motif and opens a new avenue of research toward understanding the regulation of Raf signaling.