Regulation of RAF family kinases: new insights from recent structural and biochemical studies Open Access

The RAF kinases, consisting of ARAF, BRAF and CRAF/RAF1, are essential members of the RAS pathway, functioning as direct effectors of the RAS GTPases to initiate the phosphorylation cascade that activates the downstream kinases MEK and ERK [1]. This pathway transmits signals required for normal growth and development, and when dysregulated, is a major contributor to human cancer and the RASopathy developmental syndromes [2]. In particular, somatic mutations in BRAF are observed in the majority of melanomas and are common in thyroid and colorectal cancer [3,4], whereas germline mutations in BRAF or CRAF are causative for certain RASopathy syndromes [5,6].

All RAF kinases can be divided into two functional domains: a C-terminal catalytic (CAT) domain and an N-terminal regulatory (REG) domain [1,7] (Figure 1A). The REG domain contains a hypervariable N-terminal segment (N-segment), a RAS binding domain (RBD), a cysteine-rich domain (CRD), and a serine/threonine-rich (S/T-rich) segment. The RBD serves as the primary high-affinity binding site for RAS [8,9], and the preceding N-terminal segment contributes to the RAS binding preferences of the various RAF family members [10]. The CRD is a zinc finger domain that interacts with the plasma membrane and RAS to promote RAF activation [9,11–15]; however, it also plays a role in maintaining RAF in a pre-signaling, inactive state [16–18]. The CAT domain encompasses the negative charge regulatory region (NC-region), the kinase domain (consisting of the N- and C-lobes), and the C-terminal tail. The NC-region preceding the kinase domain contributes prominently to the differences observed in the basal enzymatic activity of the RAF kinases, with BRAF having the highest activity and possessing a constitutive negative charge due to aspartic acid residues in this region, whereas CRAF and ARAF have significantly reduced activity and contain residues that become negatively charged via signal induced phosphorylation [19–21]. In addition, all RAF family members contain two high affinity phosphoserine binding sites for 14-3-3 dimers, one in the serine/threonine rich region of the REG domain (N′ site) and one in the C-terminal tail of the CAT domain (C′ site) [22]. It is important to note that members of the 14-3-3 family exist as obligate dimers, with each protomer of the dimer containing an independent phosphoserine/phosphothreonine binding pocket [23].

Years of research from many laboratories has revealed that the RAF activation process is particularly complex, involving changes in RAF phosphorylation, conformation, and subcellular localization, as well as intramolecular and intermolecular interactions [1,10]. In quiescent cells, the RAFs localize to the cytosol as inactive monomers [24]. When signaling cues are received, the RAFs are then recruited to the plasma membrane via interactions with GTP-bound RAS, where they form active dimers, with BRAF/CRAF heterodimers predominating in many cell types [25–28]. Subsequently, signaling is attenuated through mechanisms involving inhibitory feedback phosphorylation, disruption of active RAF complexes, and dephosphorylation events [27,29,30].

Although much has been learned regarding RAF regulation, knowledge gaps still exist, and to effectively target RAF in human disease states, these gaps must be filled. This review will focus on the recent structural and biochemical studies that have advanced our understanding of RAF activation and how disease-associated mutations impact various steps in this process. We will also discuss the mounting regulatory differences observed between BRAF and CRAF, the primary disease drivers of the RAF family.

Under normal signaling conditions, dimerization is essential for RAF activation [31,32] and typically requires binding to active RAS, interactions with the plasma membrane, certain phosphorylation and dephosphorylation events, and an intact C′ 14-3-3 binding site [25–28,33]. In 2019, the first cryoEM structure of a full-length BRAF dimer complex was published by Kondo et al. [34], showing dimerized BRAF kinase domains in complex with a 14-3-3 dimer (Figure 1B). Although no structural information regarding the BRAF REG domain was obtained, likely due to the flexibility of this region in the absence of membrane binding partners, the kinase domains were well-resolved and formed back-to-back dimers. Both protomers were in the active kinase conformation, as defined by the ‘in’ position of the N-lobe α-C helices, and a single 14-3-3 dimer bridged the BRAF protomers by binding to the C′ phosphosites (Figure 1B).

Surprisingly, even though 14-3-3 and the BRAF kinase domains each formed symmetric dimers, the 14-3-3 dimer was in an asymmetric orientation with respect to the BRAF dimer. Another unexpected feature of the structure was that the C-tail of one BRAF protomer appeared to be inserted into the active site of the other. The C-tail insertion was initially hypothesized to be a mechanism for promoting the transactivation of the C-tail-donating protomer and to cause the BRAF₂:14-3-3₂ asymmetry. However, further experimentation refuted this model, showing that the BRAF and 14-3-3 dimers assume an asymmetric orientation even in the absence of the C-tail and that the C-tail is dispensable for RAF activation [35–37]. In contrast, 14-3-3 binding to the C′ site of the BRAF protomers was required for RAF activation, functioning to maintain the kinase domains in a dimerized state when ATP is bound [36,38]. Interestingly, subsequent structures have indicated that the degree of asymmetry observed between the 14-3-3 and BRAF dimers is influenced by whether RAF's substrate MEK is bound to the BRAF kinase domains, [35], suggesting that flexibility between the BRAF and 14-3-3 dimers may be needed for movement involved in substrate binding and catalysis [39].

In 2024, Dedden et al. [40] published the first cryo-EM structures of dimerized CRAF kinase domains bound to a 14-3-3 dimer as well as active CRAF^KD₂:14-3-3₂ complexes that containing MEK. The molecular architecture of the CRAF dimer complexes was highly similar to the reported MEK-free and MEK-bound BRAF dimer complexes [34,35,37], indicating a conserved dimeric state for these critical disease drivers.

The idea that autoinhibition contributes to RAF regulation was first proposed based on studies showing that when expressed as isolated proteins, the RAF REG domain could bind and suppress the signaling activity of the CAT domain [16–18,41]. Mutational analysis further revealed that an intact CRD was required for the inhibitory effect of the REG domain [16,42]. In addition, 14-3-3 dimers had also been implicated in maintaining RAF in an inactive state as both the RAF N′ and C′ docking sites are highly phosphorylated under quiescent conditions and preventing phosphorylation of the N′ site is known to increase RAF activity [27,29,43]. Thus, it had been suggested that a 14-3-3 dimer might stabilize the autoinhibitory interactions between the REG and CAT by binding simultaneously to the N′ and C′ sites.

This model for an inactive RAF complex was finally confirmed in 2019, when Park et al. [35] published the first cryo-EM structure of a full-length BRAF monomer complex containing a 14-3-3 dimer and MEK1. The monomeric BRAF structure demonstrated the pivotal role of the CRD in RAF autoinhibition, revealing that the CRD makes contacts with the BRAF kinase domain and both protomers of the 14-3-3 dimer (Figure 1C). Moreover, the structure showed that a single 14-3-3 dimer could indeed bind the N′ and C′ phosphoserine sites and in doing so obscure two key regions involved in RAF activation — the plasma membrane binding loops of the CRD and the dimer interface of the kinase domain. Although the RBD was not resolved, the Park et al. [35] structure did confirm that BRAF can exist in pre-formed complexes with its substrate MEK. The active sites of the BRAF and MEK kinase domains interacted in a face-to-face manner, and both kinase domains were in the inactive configuration with their α-C helices adopting the ‘out’ position.

Subsequently, Martinez Fiesco et al. [37] reported the cryo-EM structures of two monomeric BRAF complexes that had been isolated from serum-depleted human epithelial cells. These complexes contained a 14-3-3 dimer and assumed a similar autoinhibited conformation as was observed for the Park et al. [35] complexes. However, in the Martinez Fiesco et al. [37] structures, one contained MEK and one was MEK-free, indicating that BRAF can assume the autoinhibited conformation in the presence or absence of MEK. In addition, the critical RBD was well-resolved in both the MEK-bound and MEK-free structures and was positioned adjacent to the BRAF kinase domain and on top of the 14-3-3 protomer bound to the BRAF C′ site (Figure 1C). The contact surface between the RBD and 14-3-3 was significant (∼435 Ų), and mutations that perturbed this interface increased the biological activity of BRAF, suggesting that the RBD provides additional contacts that further stabilize the autoinhibited conformation.

Importantly, critical basic residues in the RBD that form high affinity ionic bonds with acidic residues in the switch I region of GTP-bound RAS were largely exposed, and Martinez Fiesco et al. [37] found that both the MEK-bound and MEK-free complexes could bind activated KRAS with high affinity. However, modeling of the RAS:RBD interaction indicated that RAS binding would generate steric clashes and electrostatic repulsion at the RBD:14-3-3 interface. In the model proposed by Martinez Fiesco et al. (Figure 2A), initial contact between RAS and the RBD would repel the 14-3-3 dimer, allowing for full RAS:RBD contact that in the context of the membrane environment, would ultimately result in the disassembly of the autoinhibited complex, including the release of 14-3-3 from the N′ site and its dephosphorylation by the SHOC2:MRAS:PP1C complex [44–46]. Release of 14-3-3 from the N′ site would also expose the dimer interface of the kinase domain and free the CRD such that it could rotate and interact with the plasma membrane and RAS, thereby facilitating the formation of active BRAF dimers.

In 2023, Park et al. [47] published additional structures of the autoinhibited BRAF monomer in which the RBD was bound to the isolated G-domain of KRAS. In contrast with the Martinez Fiesco et al. [37] structure, the RBD in the KRAS-bound structures did not contact 14-3-3. Moreover, when the activity of the purified monomeric complexes was assayed in vitro, the addition of GTP-bound KRAS was found to only have a small activating effect in the absence of membrane lipids [47]. Based on these observations, Park et al. hypothesize that KRAS:RBD interactions are insufficient to disrupt the autoinhibited conformation and that steric effects of the plasma membrane may be the major factor driving the structural changes needed for dimerization and activation. They also noted that the low resolution of the RBD in their RAS-free and RAS-bound structures may be due to mobility issues [35,47], and a possible explanation for the RBD differences observed between their structures and the Martinez Fiesco et al. [37] structures could lie in the expression systems used to generate these complexes. Indeed, the insect cell expression system used to produce the Park et al. [35,47] BRAF complexes may not provide the environment or the entire spectrum of post-translational modifications [48] needed to fully recapitulate all of the autoinhibitory interactions that occur in mammalian cells, the source of the Martinez Fiesco et al. [37] complexes. Thus, additional studies will be needed to fully elucidate the specific roles and relative importance of RAS binding, the plasma membrane, and perhaps other factors in facilitating BRAF's transition from an autoinhibited monomer to an active dimer.

Like many protein kinases, members of the RAF kinase family interact with the HSP90/CDC37 chaperone complex. CRAF and ARAF have been classified as strong clients of the chaperone complex, and although BRAF in its wild-type form is a weak client, certain oncogenic BRAF mutants, including BRAF^V600E, are strong clients [49–51]. Systematic analysis of various HSP90 clients has shown that CDC37 provides recognition of the kinase family, but that the structural stability of the kinase domain is a major determinant for the degree of binding within a kinase family, in that strong clients possess a more intrinsically unstable kinase domain [51].

The molecular details regarding how RAF interacts with HSP90 and CDC37 were revealed in 2022 with the publication of cryoEM structures for two RAF-containing HSP90/CDC37 complexes (RHC complexes). In the first study, Garcia-Alonso and colleagues reported the structure of CRAF bound to CDC37 and dimeric HSP90. Only portions of the CRAF kinase domain were resolved in this structure, showing a mostly folded C-lobe bound to HSP90 and CDC37 and an unfolded N-lobe threaded through the center of the HSP90 dimer, making extensive hydrophobic interactions with the HSP90 binding pocket and contacting CDC37 (Figure 1D). Further analysis of the purified CRAF RHC complex revealed that both the N′ and C′ 14-3-3 binding sites of CRAF were phosphorylated but not bound to 14-3-3, indicating that 14-3-3 dimers are unable to stably associate with CRAF while contained within the chaperone complex. However, when the CRAF RHC complexes were incubated in vitro with high levels of recombinant 14-3-3, HSP90 and CDC37 were released and 14-3-3-bound CRAF dimers were detected, suggesting that 14-3-3 may play a role in disbanding the RHC complex.

In a second study, Oberoi et al. [52] reported the structure of an RHC complex containing the oncogenic BRAF^V600E mutant. The structure of the BRAF^V600E-containing complex was in general similar to that of the CRAF RHC complex, and as with the CRAF-containing complex, only portions of the BRAF^V600E kinase domain were resolved even though both complexes were formed using full length RAF proteins. In addition, Oberoi et al., as well as a study by Jamie-Garza et al. [53], reported structures of RHC complexes that contained the PP5 protein phosphatase (Figure 1D). Both groups found that PP5 binding did not alter the overall structure of the complexes but did result in widespread RAF dephosphorylation. Notably, PP5 was found to dephosphorylate both the N′ and C′ 14-3-3 docking sites, along with a critical activating phosphosite in the NC-region of the CAT domain (S338 in the CRAF; S446 in BRAF) [30,52,53]. In addition, PP5 could remove negative regulatory phosphosites [52], including targets of ERK-mediated feedback phosphorylation, which are known to disrupt active RAF signaling complexes and must be dephosphorylated prior to RAF reactivation [27,29,54]. Thus, these findings indicate that the RHC-PP5 complex is likely to contribute to the post-signaling inactivation of RAF and/or its recycling to a pre-signaling state (Figure 2B).

Of the RAF family members, BRAF is the main cancer driver (Figure 2C), with oncogenic mutations primarily observed in two key regions of the kinase domain — the P-loop of the N-lobe and the activation segment in the C-lobe [55]. Interactions between these two regions function to maintain the kinase domain in an inactive closed conformation, and cancer-associated mutations are thought to disrupt these contacts and, in turn, the inactive configuration [55]. Notably, perturbing the interactions between the P-loop and activation segment destabilizes the BRAF kinase domain such that the majority of these mutants, including BRAF^V600E, are classified as strong clients of the HSP90/CDC37 complex. In contrast, wild-type BRAF is a weak client that transiently associates with the HSP90/CDC37C complex during maturation [49–51]. Similar to CRAF, oncogenic BRAF mutants that are strong clients of the chaperone complex rapidly degrade in the presence of HSP90 inhibitors [49–51], prompting numerous pre-clinical and clinical studies evaluating the use of HSP90 inhibitors in combination with RAF-specific inhibitors as a cancer therapy [56,57].

Regarding the RASopathy developmental syndromes, germline mutations in either BRAF or CRAF can be disease drivers (Figure 2C). However, mutations in BRAF versus CRAF tend to occur in different regions of the proteins and are primarily associated with different RASopathy syndromes. For example, germline alterations in CRAF are mainly associated with Noonan syndrome, and these mutations typically lie near the N′ 14-3-3 docking site and disrupt 14-3-3 binding [6,58,59]. In contrast, BRAF RASopathy mutations are largely observed in cardiofaciocutaneous syndrome and often occur in the CRD [5,60], a domain that facilitates RAF activation by interacting with phosphatidylserine (PS) and RAS at the plasma membrane, but also contributes to RAF autoinhibition.

In 2022, the importance of the CRD in maintaining BRAF in an autoinhibited state was clearly demonstrated in experiments analyzing a panel of RASopathy-associated germline mutations in the BRAF CRD [61]. This work revealed that all of the CRD mutations increase the biological activity of BRAF by disrupting the autoinhibitory function of the CRD and/or enhancing its PS/membrane binding properties. Interestingly, mutations that enhanced the CRD's PS binding activity added additional positively charged residues to the surface of the CRD, which would be conducive for increased binding to the negatively charged head groups of PS, whereas mutations that disrupted the CRD's autoinhibitory function occurred in or adjacent to residues that contact the BRAF CAT domain and/or 14-3-3 in the autoinhibited state.

Further analysis of the BRAF-CRD mutants revealed that relief of autoinhibition alone was sufficient to significantly increase the overall levels of RAS-RAF binding, whereas enhanced CRD-PS binding contributed to the binding affinity. CRD mutants with impaired autoinhibitory activity also displayed an increased presence at the plasma membrane and exhibited a level of RAS-independent activity. In contrast, a mutant that was unimpaired for autoinhibition but displayed enhanced PS binding was fully dependent on RAS for its signaling function and exhibited the weakest gain-of-function activity in all biological assays. These results indicate that relief of autoinhibition is the primary factor determining the severity of the BRAF-CRD mutations, a finding consistent with the phenotypic analysis of RASopathy patients possessing these mutations. Moreover, these findings suggest that identifying compounds that can stabilize the autoinhibited state may have therapeutic benefit.

Because RASopathy-associated mutations are frequently observed in the BRAF CRD but not the CRAF CRD, Spencer-Smith et al. [61] also conducted experiments comparing the activities of the BRAF and CRAF CRDs (Figure 2C). Strikingly, both the autoinhibitory activity and the PS-binding activity of the CRAF CRD were reduced in comparison with the BRAF CRD. Further experimentation using proteins in which the CRDs of BRAF and CRAF had been swapped indicted that the increased autoinhibitory activity of the BRAF CRD may be uniquely required to suppress BRAF's high basal enzymatic activity and maintain BRAF in a non-signaling state. With regard to CRAF, RAS-mediated activation of CRAF is known to require an intact CRD [62–64], and recent crystal structures of the CRAF RBD-CRD in complex with RAS show that the CRAF-CRD makes significant contacts with RAS and that the RAS binding interface is distinct from the PS/membrane binding loops of the CRD [9,14]. Mutational studies further indicate that although mutations in the PS or RAS-binding interfaces of the CRAF CRD have a significant effect on CRAF activity, they have little impact on the overall levels of CRAF:RAS binding, leading to the model that the primary function of the CRD at the plasma membrane is to correctly orient the RAS:RAF complex for dimer formation.

Recent structures of the RAF kinases have provided significant insight regarding the opposing roles that 14-3-3 dimers and the RAF RBD and CRD domains play in RAF regulation, with each contributing to RAF activation and autoinhibition. For example, a 14-3-3 dimer can stabilize the RAF autoinhibited state by simultaneously binding to the N′ and C′ sites, but it can also facilitate the formation of active RAF dimers by bridging the C′ sites on RAF protomers and may promote the release of RAF from HSP90/CDC37 chaperone complexes. Moreover, recent studies have also revealed regulatory differences between the BRAF and CRAF family members, with CRAF and some oncogenic BRAF mutants being more dependent on the HSP90/CDC37 chaperone complex for their folding and stability than is wild-type BRAF. In addition, while autoinhibition is crucial for maintaining BRAF in a non-signaling state, it is unclear how important autoinhibition is for CRAF regulation given the reduced autoinhibitory activity of the CRAF CRD and the lack of structures for an autoinhibited CRAF monomer complex. However, evidence that CRAF, like BRAF, does assume a 14-3-3-bound autoinhibited state comes from the observations that mutations disrupting 14-3-3 binding to the N′ site result in increased CRAF activity and that this region is a hotspot for RASopathy-associated mutations. Thus, further structures and studies are needed to fully elucidate the regulatory cycle of all RAF family members, including ARAF for which there are no structures. Finally, recent studies have shown that PP5 associates with both the CRAF- and BRAF-containing RHC chaperone complexes and can induce widespread RAF dephosphorylation; however, the relative contributions of PP5 and the RHC complex to post-signaling inactivation and pre-signaling recycling of the RAF kinases remains unclear. Nonetheless, the utilization of cryo-EM technologies to solve the structures of large multi-protein signaling complexes, together with assays that measure interactions between pathway components in live cells, is greatly aiding our understanding of RAF regulation, which may inform new strategies to target RAF function in human disease states.

• The RAF kinases are essential components of the RAS pathway required for the transmission of normal growth signals. Mutations in BRAF and CRAF are also disease drivers.

• Regulation of the RAF kinases is complex and recent studies have revealed important regulatory differences between BRAF and CRAF.

• Further elucidation of RAF regulation will help in the development of more effective RAF therapies.

The authors declare that there are no competing interests associated with the manuscript.

This work has been funded in whole or in part with Federal funds from the National Cancer Institute, National Institutes of Health, under project number ZIA BC 010329.

CAT

catalytic

CRD

cysteine-rich domain

KD

kinase domain

PS

phosphatidylserine

RBD

RAS binding domain