|Home | About | Journals | Submit | Contact Us | Français|
First identified in the early 1980s as retroviral oncogenes, the Raf proteins have been the objects of intense research. The discoveries 10 years later that the Raf family members (Raf-1, B-Raf, and A-Raf) are bona fide Ras effectors and upstream activators of the ubiquitous ERK pathway increased the interest in these proteins primarily because of the central role that this cascade plays in cancer development. The important role of Raf in cancer was corroborated in 2002 with the discovery of B-Raf genetic mutations in a large number of tumors. This led to intensified drug development efforts to target Raf signaling in cancer. This work yielded not only recent clinical successes but also surprising insights into the regulation of Raf proteins by homodimerization and heterodimerization. Surprising insights also came from the hunt for new Raf targets. Although MEK remains the only widely accepted Raf substrate, new kinase-independent roles for Raf proteins have emerged. These include the regulation of apoptosis by suppressing the activity of the proapoptotic kinases, ASK1 and MST2, and the regulation of cell motility and differentiation by controlling the activity of Rok-α. In this review, we discuss the regulation of Raf proteins and their role in cancer, with special focus on the interacting proteins that modulate Raf signaling. We also describe the new pathways controlled by Raf proteins and summarize the successes and failures in the development of efficient anticancer therapies targeting Raf. Finally, we also argue for the necessity of more systemic approaches to obtain a better understanding of how the Ras-Raf signaling network generates biological specificity.
The first raf (rapidly accelerated fibrosarcoma) gene was described in 1983 as a retroviral oncogene, v-raf, transduced by the murine sarcoma virus isolate 3611.1 A year later, an avian homolog, v-mil, was found in the MH2 retrovirus.2 These 2 transforming retroviruses encoded the first oncogene to be discovered with serine/threonine kinase activity.3 After the cellular proto-oncogene homologs, c-raf 4 and c-mil,5 had been cloned, studies focused on elucidating the function of Raf proteins. They showed that c-Raf (also known as Raf-1) plays a critical role in mediating the cellular effects of growth factor signals.6-8 Later on, Raf proteins were identified as direct activators of MEK9,10 and effectors of Ras.11-15 Thus, Raf proteins were placed as essential connectors between Ras and the MEK-ERK pathway (Fig. 1). Most subsequent work focused on understanding this role and the regulation of Raf proteins in detail, until new functions of Raf-1 in the regulation of apoptosis16-18 and cell migration19 emerged in the last decade.
Three different Raf isoforms originating from 3 independent genes can be distinguished in mammals, Raf-1/c-Raf, B-Raf, and A-Raf. Raf-1 was the first isoform to be identified4 and for 20 years was the principal focus of attention on the proteins of the family. After the discovery 8 years ago of B-Raf mutations in different types of tumors,20 B-Raf moved into the limelight, resulting in a rapid increase of our knowledge of the biological functions of this isoform. On the other hand, still very little is known about A-Raf, and although it seems to share many of the properties of the other isoforms, its biological functions remain a mystery. All Raf proteins share MEK1/2 kinases as substrates. MEK1/2 in turn activate ERK1/2, and this pathway regulates many cellular functions such as cell proliferation, differentiation, migration, or apoptosis (for extensive reviews, see Wellbrock et al.,21 Leicht et al.,22 and Dhillon et al.23).
In recent years, it has become clear that the initial view of the ERK pathway as a linear pathway is not accurate but that there are many different proteins interacting with the proteins of the pathway. These proteins regulate the pathway by mediating the crosstalk with other signaling pathways and the regulation of positive and negative feedback mechanisms.24 In this review, we focus on the mechanisms of Raf family regulation and the biological roles of Raf family kinases especially in cancer and with relation to Ras signaling. We also explore the role of the Raf proteins in the context of the coordinated signaling networks that ultimately are responsible for cellular responses both in normal and tumor cells. We end on how systems biology can help us to integrate the information gathered from the many years of research in the Ras-Raf pathway and how it can be used to address open questions.
There are no Raf kinases in yeasts, and the phylogenetic oldest isoform is B-Raf, which appears in invertebrates. Mammals possess 3 Raf isoforms (Raf-1, B-Raf, and A-Raf), which share a common modular structure consisting of 3 conserved regions (CR) with distinct functions (Fig. 2). CR1 contains a Ras-binding domain (RBD), which is necessary for the interaction with Ras and with membrane phospholipids required for membrane recruitment, and a cysteine-rich domain (CRD), which is a secondary Ras-binding site and also necessary for the interaction of CR1 with the kinase domain for Raf autoinhibition.25 CR2 contains important inhibitory phosphorylation sites participating in the negative regulation of Ras binding and Raf activation.26 CR3 features the kinase domain, including the activation segment, whose phosphorylation is crucial for kinase activation.27 Unfortunately, the tertiary structure of a Raf holoenzyme has remained elusive, although the structures of the RBD and extended CR1 domains of Raf-128-30 and the CR3 domain of B-Raf31 and Raf-132 were solved. Functionally, the Raf structure can be split into a regulatory N-terminal region, containing the RBD, which is critical for activation as well as inhibitory phosphorylation sites, and a catalytic C-terminal region, which includes phosphorylation sites necessary for the kinase activation. The regulatory domain restrains the activity of the kinase domain,25,33,34 and its removal results in constitutive oncogenic activation.35 However, the activity of the isolated Raf-1 kinase domain is subjected to further regulation and can be stimulated by phorbol esters, v-Src, and phosphorylation.34,36 This observation is in keeping with the finding that the most common oncogenic mutation in B-Raf, V600E, activates B-Raf kinase activity by mimicking phosphorylation of the activation loop that releases its inhibitory interaction with the ATP-binding domain.31
In the inactive state, Raf-1 is thought to exist in a closed conformation in which the N-terminal regulatory region folds over and occludes the catalytic region.37 This conformation is stabilized by a 14-3-3 dimer binding to an N-terminal site, phospho-S259 (pS259), and a C-terminal site, pS621. Although the activation process of Raf-1 is not completely understood, we can assume the following sequence of events (Fig. 3).
A-Raf is generally thought to be regulated similarly to Raf-1, but important differences have emerged. A-Raf is only weakly activated by oncogenic H-Ras and Src and also displays low kinase activity towards MEK.106 The reason for the lower responsiveness to H-Ras is the exchange of an arginine for a lysine at position 22 in the A-Raf RBD, which weakens the binding of A-Raf to H-Ras.107 In addition, the low kinase activity may be unique nonconserved amino acid residues in the N-region. Mutation of Y296 in the N-region led to a constitutively active kinase, and molecular modeling showed that Y296 promotes a tighter interaction between the N-region and the catalytic domain, which may stabilize the closed conformation.108 Subsequently, a systematic phosphorylation site analysis revealed several interesting findings95: S432 located between the ATP-binding domain and activation loop was found critical for MEK binding and A-Raf signaling. Surprisingly, activation loop phosphorylation did not contribute to mitogen-induced activation. Finally, a cluster of phosphorylation sites between amino acids 248 and 267, which stimulated activation, facilitated A-Raf dissociation from the plasma membrane. These findings raise some intriguing aspects. First, activation loop phosphorylation, which is a widespread mechanism in the catalytic activation of kinases,109 may be less critical in Raf regulation, as corresponding structural reorganizations may be caused by 14-3-3 proteins binding to the Raf kinase domain (see below). Second, the fact that activating phosphorylation events release A-Raf from the plasma membrane suggests that while initial activation may occur at the membrane, downstream signaling proceeds in other subcellular compartments. In this context, both A-Raf and Raf-1 have been found in different subcellular compartments including mitochondria, endosomes, and the Golgi apparatus.110 It is unknown whether A-Raf is regulated by PKA.
B-Raf activation appears much simpler. In fact, Ras and 14-3-3 binding are likely to be the only major requirement for B-Raf activation.106,111 The N-region is already negatively charged because of the presence of aspartate at the position corresponding to Raf-1’s YY340/1 (DD448/9) and the constitutive phosphorylation of S446 (corresponding to the S338 of Raf-1). Phosphorylated S446 neutralizes the inhibitory role of N-terminal domain towards the catalytic region and in conjunction with D449 allows the catalytic domain to adopt a stabilized 3-dimensional conformation.25 Phosphorylation of S365 (corresponding to S259 in Raf-1) impairs B-Raf activity, and its mutation to a nonphosphorylatable residue can even overcome the debilitating effects of charge-neutralizing mutations in the N-region.112,113 In contrast to Raf-1 S259, B-Raf S365 is unlikely to be phosphorylated in cells upon cAMP stimulation, but another site, S429, is a potential target for PKA.114 Importantly, B-Raf can be inhibited or activated by PKA depending on the levels of 14-3-3 expression, which need to be high for permitting activation.115 A second pathway may involve the PKA-mediated activation of Rap1, which was reported to bind and activate B-Raf.116,117 However, these results are disputed, as many physiological signals that induce Rap1 activity fail to activate B-Raf.118 Possible reasons for this discrepancy are unclear and are discussed below.
The common and key step in the activation of all 3 Raf isoforms is membrane recruitment by a Ras family protein. Membrane translocation triggers further activation events, such as the binding of PP2A to dephosphorylate the inhibitory pS259 site in Raf-1 (and presumably the corresponding sites in A-Raf and B-Raf) and the colocalization with the kinases responsible for the multiple activating phosphorylations as discussed above. The sequences forming the binding interface are well conserved in the Raf as well as Ras family. Hence, it is not surprising that several members of the Ras family can bind Raf kinases. A systematic comparison of the ability of different Ras family members to activate Raf isoforms119 showed that H-Ras, N-Ras, and K-Ras could stimulate all 3 Raf isoforms and were the only Ras proteins that could activate B-Raf. In contrast, A-Raf also could be activated by R-Ras3, while Raf-1 was the most promiscuous isoform, responding also weakly to R-Ras3, Rit, and TC21. In contrast, Rap1/2, Rin, and Rheb were ineffective. The ability to activate Raf generally corresponded with the binding affinity. Only H-Ras, N-Ras, and K-Ras led to a stimulation of the endogenous ERK pathway in HEK293 cells, while lower affinity interactions only could stimulate ERK when either Raf or ERK was overexpressed. These results suggest that cell type–specific expression stoichiometry of Ras isoforms and Raf-ERK pathway components potentially could generate a rich variety of ERK activation dynamics that could allow cells to respond to different growth factors with precisely tuned ERK activation. An example is Rheb, which was reported to interact with Raf-1 and B-Raf and to suppress their kinase activity by reducing N-region phosphorylation and heterodimerization.120,121 Interestingly, PKA-mediated phosphorylation of Raf-1 on S43 increases Raf-1 affinity for Rheb and thereby could contribute to the inhibitory effects of PKA by diverting Raf-1 from H-Ras to Rheb.122
Although this hypothesis is conceptually appealing, the experimental evidence for specific engagement of Raf isoforms by different Ras family members is often controversial. A case in particular is Rap1, which was initially isolated as a repressor of K-Ras transformation.123 Rap1 binds to Raf-1,28 but the functional consequences are disputed. Constitutively active Rap1 can inhibit Raf-1 when overexpressed,124 while at normal expression levels and in response to physiological stimulation, Rap1 did not regulate Raf-1.118,119 In addition, Rap1 was reported to activate B-Raf116,117 and mediate cAMP stimulation of ERK and neuronal differentiation of PC12 cells.117 However, this finding was not reproduced in other studies,118 and the issue remains open. A possible explanation for the differential effect of Rap1 and Ras on B-Raf and Raf-1 is the difference in affinity to the CRD domains. Rap1 has high affinity for the Raf-1 CRD domain and low affinity for the B-Raf CRD, whereas Ras has low affinity for both Raf-1 and B-Raf CRD domains. Swapping the CRD domains showed that the B-Raf CRD conveyed susceptibility to activation by Rap1, while the Raf-1 CRD abolished it.116 An alternative, but not mutually exclusive, mechanism could be the failure of Rap1 to target Raf-1 to the membrane compartment, where S338 phosphorylation can take place.125 Correct targeting needs Y341 phosphorylation or the negative charge that B-Raf carries at this position. Introducing a negatively charged residue at this position or redirecting Rap1 to lipid rafts could overcome the deficiency of Rap1 to activate Raf-1.125 A recent study using statistical modeling based on Bayesian inference revealed a role for Rap1 in EGF-stimulated ERK activation by supporting Raf-1 and B-Raf heterodimerization.126 The exact mechanism needs to be elucidated. Apart from direct Rap1 binding to B-Raf, it also could involve indirect effects such as a recruitment of a common scaffold by Rap1. In natural killer lymphocytes, Rap1 was recently shown to bind to the scaffolding protein IQ motif containing GTPase-activating protein 1 (IQGAP1) and to assemble a signaling complex that activated B-Raf, Raf-1, and ERK.127 Thus, the contradictory findings on Rap1 regulation of Raf proteins may eventually be reconciled when considering indirect mechanisms such as the involvement of scaffolds.
All 3 Raf isoforms are regulated at the level of protein expression. Alternative splicing gives rise to multiple B-Raf isoforms differentially expressed in various tissues.128,129 B-Raf activity is also regulated by splicing. B-Raf isoforms containing exon 8b are more phosphorylated on the inhibitory S365 site, leading to an increased interaction with 14-3-3 and strengthening the inhibitory interaction between N-terminal regulatory domain and kinase domain, altogether resulting in lower kinase activity.112 With respect to A-Raf, the 2 splice isoforms described so far, DA-Raf1 and D-Raf2, lack the kinase domain and act as dominant inhibitory mutants of Ras130 and ARF GTPases.131 As a consequence, DA-Raf1 is a positive regulator of myogenic differentiation by mediating the inhibition of the ERK pathway required for differentiation.130 Raf-1 also has a known splice variant preferentially expressed in the muscle and brain.132
Key regulatory phosphorylation sites in Raf kinases are also binding sites for the scaffolding protein 14-3-3. 14-3-3 is an obligatory dimer forming a rigid half-barrel structure that interacts with other proteins in a phosphorylation-dependent, bidentate way, constraining the conformation of the binding partner.133 In Raf kinases, 14-3-3 can stabilize both the inactive and the activated state. This dual property confounds the analysis of 14-3-3 effects on Raf kinase regulation and can explain many of the controversies in the literature.
In the inactive configuration, 14-3-3 binds to conserved phosphorylation sites in the N- and C-terminus of the Raf kinase domain (pS259 and pS621 in Raf-1; pS365 and p729 in B-Raf; pS214 and pS576 in A-Raf).36,134-138 This bidentate interaction is thought to physically clamp the regulatory domain to the kinase domain. Thus, it is not surprising that the 14-3-3 binding sites are targeted by kinases that inhibit Raf activation. Both protein kinase A (PKA) and B (PKB/Akt) were reported to induce phosphorylation of the N-terminal 14-3-3 binding site of Raf-1 and B-Raf.26,34,98-100,103,114,139-141 However, the relevance of Akt phosphorylation remains disputed, as Akt does not phosphorylate S259 in most physiological scenarios.104,105 While PKA is a bona fide S259 kinase, it is not responsible for the constitutive phosphorylation of S259,99 and the kinase that maintains basal S259 phosphorylation in cells is still unknown. The C-terminal 14-3-3 binding residue S621 is targeted by PKA,101 AMP-activated protein kinase (AMPK),142 autophosphorylation,101,102,143 and probably other yet unidentified kinases. The exact role of S621 is still not clear. It was reported to inhibit Raf-1 kinase101 but also to be essential for kinase activity137,138 and for the stability of the Raf-1 protein.143 Recent data102 show that mutations that preclude 14-3-3 binding to S621 inactivate Raf-1 by specifically disrupting its capacity to bind to ATP rather than by gross conformational alteration, as indicated by the intact ability to bind MEK. Phosphorylation of S621 inhibits Raf-1 catalytic activity in vitro, but addition of 14-3-3 proteins completely reverses this inhibition. 14-3-3 binding requires the phosphorylation of S621, and this interaction is essential for Raf-1 kinase activity, but S621 phosphorylation in the absence of 14-3-3 does not support kinase activity. These data explain the dual role of S621 phosphorylation and suggest that 14-3-3 may serve as a switch that can convert an inactive Raf-1 population phosphorylated on S621 into a kinase-competent state.102 Although B-Raf catalytic activity is less dependent on 14-3-3 binding, its biological activity is dependent on 14-3-3 binding,144,145 and 14-3-3 binding can switch inhibitory phosphorylation of B-Raf by PKA into activation.115 An additional role of 14-3-3 in Raf kinase activation is related to its enhancement of Raf-1 homodimerization and Raf-1/B-Raf heterodimerization, which elevates Raf-1 kinase activity82,146 and is discussed below.
Dimerization is a common motif in the activation of kinases. Homodimerization was initially highlighted as a potentially important step of Raf-1 activation by 2 studies showing that a forced interaction of Raf-1 monomers tagged with inducible dimerizing tags robustly induced kinase activity.147,148 Both studies proposed that active Ras would promote the formation of dimers. This hypothesis was later extended to heterodimerization between Raf-1 and B-Raf, which was found to be inducible by active Ras.146 As mutation of S621 abrogated Raf heterodimerization, the authors speculated that 14-3-3 binding to pS621 was necessary for Raf heterodimerization. Although these initial studies showed that homodimerization and heterodimerization can activate Raf kinases, they failed to show that the interaction can take place between endogenous proteins when stimulated by physiological mitogens. This was achieved by Rushworth et al.,82 who demonstrated that endogenous B-Raf and Raf-1 heterodimerize in multiple cell lines in response to mitogens. Biochemical fractionation of Raf heterodimers from homodimers and monomers showed that Raf-1–B-Raf heterodimers accounted for the majority of the mitogen-induced kinase activity. Remarkably, the heterodimers represented less than 1% of the total B-Raf pool but exhibited approximately 30-fold elevated kinase activity. Heterodimerization was enhanced by 14-3-3, but not a dimerization-negative 14-3-3 mutant, suggesting that the 14-3-3 dimer crosslinks Raf-1 and B-Raf by binding to the C-terminal sites on each kinase. This observation suggests a mechanism for how 14-3-3 can stabilize both inactive and active Raf-1 conformations. In the inactive conformation, 14-3-3 clasps the Raf-1 regulatory to the kinase domain via intramolecular binding to pS259 in the N-terminus and p621 in the C-terminus. Binding to activated Ras displaces 14-3-3 from pS259, leaving one 14-3-3 arm free to contact a 14-3-3 binding site in B-Raf to facilitate heterodimerization. In addition, Raf heterodimerization is also regulated by ERK-mediated feedback phosphorylation of B-Raf.82,92 The feedback phosphorylation mainly serves to limit the lifetime of B-Raf–Raf-1 heterodimers, and mutation of the relevant sites enhances ERK signaling and the associated biological activities. Other regulators of Raf heterodimerization include KSR1149 and MLK3,150 which both enhance heterodimerization. MLK3 was originally described as an activator of JNK, but by activating B-Raf, it may serve as an integrating hub between the ERK and JNK pathways.150
Heterodimerization also may play a pathophysiological role in cancer. When B-Raf mutations were discovered in cancer,20 a puzzling observation was that while the most frequent mutation, V600E, massively stimulated B-Raf kinase activity, several less frequent mutations activated B-Raf only mildly or not at all.31 However, even the low-activity B-Raf mutants could hyperstimulate the ERK pathway. Intriguingly, this activation was dependent on the presence of Raf-1, suggesting that low-activity B-Raf mutants require Raf-1 to activate the ERK pathway. A subsequent study81 confirmed the transactivation hypothesis by demonstrating that the low-activity B-Raf mutants found in cancer indeed could activate Raf-1 merely by forming dimers. While physiological heterodimerization is induced by Ras activation,81,82 oncogenic B-Raf mutants constitutively dimerized with Raf-1.81 The mechanism of activation by heterodimerization is incompletely understood. Mutational analysis showed that it does not depend on N-region phosphorylation but requires the activation loop phosphorylation sites.81 This finding is difficult to reconcile with the observation that Ras-mediated activation of Raf-1 requires N-region phosphorylation.70,75,106,151 The ability of Ras to induce Raf heterodimerization should overcome the requirement for N-region phosphorylation for Raf activation. Thus, more work will be needed to elucidate the mechanism of how heterodimerization activates Raf kinase activity. Regardless of the exact mechanism, dimerization clearly has biological effects. Increasing the lifetime of Raf heterodimers by mutating the T753 ERK feedback phosphorylation site in B-Raf augmented the ability of nerve growth factor (NGF) to induce neuronal differentiation of PC12 cells.82 Furthermore, the mutation of ERK feedback phosphorylation sites and the concomitant increase in Raf heterodimer levels enhanced the transforming potential of oncogenic B-Raf with intermediate or low kinase activity, which depend on Raf-1 for transformation, without affecting transformation by the high-activity B-Raf V600E mutant.144
An important role for Raf heterodimers was recently described in response to Raf kinase inhibitors. The original observation that Raf kinase inhibitors paradoxically could hyperactivate the Raf pathway152 is now explainable by the ability of Raf kinase inhibitors to promote Raf heterodimerization and activation.32,153,154 These findings are discussed below.
Despite much effort to identify Raf substrates, so far, the only bona fide physiological substrates of Raf are MEK1 and MEK2. Activated Raf kinases phosphorylate both MEK isoforms on 2 residues in the activation loop (S217 and S221). Phosphorylation on these 2 sites increases MEK activity, which in turn can bind, phosphorylate, and activate ERK (Fig. 1). Although all family members have the ability to bind and phosphorylate MEK in vitro, the activities towards MEK differ widely. B-Raf has the strongest activity towards MEK, followed by Raf-1 and A-Raf, whose MEK kinase activity is barely detectable.106 B-Raf possesses higher basal activity partially because it is “primed” for activation due to the aforementioned constitutive phosphorylation of the S445 residue and the negatively charged amino acids at positions in the N-region that need to be phosphorylated in Raf-1 and A-Raf to achieve activation. Furthermore, of all Raf isoforms, B-Raf has the strongest binding affinity for MEK.155 MEK is present in B-Raf protein-protein complexes even in starved cells, representing a preassembled complex ready to be activated.
The interaction of MEK with Raf-1 is also regulated by scaffolding proteins, such as KSR (see below), and by phosphorylation. The Rac effector kinase PAK1 can phosphorylate MEK1 on S298, which enhances the interaction with Raf-1156 as well as ERK2.157 Feedback phosphorylation of T292 by activated ERK prevents the phosphorylation of S298 and limits MEK activation by Raf-1.158 This modulation is likely part of the biochemical basis for the cooperation between Ras and Rac proteins in cell transformation.159,160 Structural studies revealed an interesting role for T292.161 T292 only occurs in MEK1, but it can also downregulate the activity of MEK2 when T292-phosphorylated MEK1 heterodimerizes with MEK2. The implications of this differential regulation are not yet fully understood but indicate that MEK1 and MEK2 are not equivalent in terms of their regulation, and in a wider context, that dimerization of kinases also can exert negative control.
An unresolved question is why the activities of the comparatively poor MEK kinases Raf-1 and A-Raf are regulated in a much more complicated manner than the activity of the superior MEK kinase B-Raf. One possibility is that Raf-1 and A-Raf have other, yet unidentified substrates that are the real targets of their elaborate regulation. A more subtle possibility is that Raf-1 and A-Raf act as modulators that fine tune the ability of B-Raf to activate the ERK pathway. The discovery of B-Raf–Raf-1 heterodimers and their unfolding biological roles supports the latter view.
A wealth of experimental data suggests that MEK is the only bona fide Raf substrate and that B-Raf is the main MEK kinase in vivo. This assessment is corroborated by phylogenetic comparisons. The single Raf homologs in invertebrates (lin-45 in Caenorhabditis elegans and D-Raf in Drosophila) are much closer to B-Raf in terms of sequence, suggesting that B-Raf is the archetypal MEK kinase, whereas A-Raf and Raf-1 may have been evolved towards MEK-independent functions.162,163
Although MEK is the only commonly accepted Raf substrate, several other potential Raf-1 substrates were described. The adenylyl cyclases (AC) type 6, 5, and 2 were proposed to be phosphorylated and activated by Raf-1 independent of MEK.164-166 As ACs generate cAMP, which activates PKA, their stimulation would initiate a negative feedback to Raf-1. Another proposed Raf-1 substrate is the retinoblastoma tumor suppressor protein (Rb)167 (Fig. 4). The inactivation of Rb by cyclin-dependent kinases 4 and 6 marks the irreversible commitment of the cell to divide.168 Raf-1 was reported to directly phosphorylate and inactivate Rb, leading to cell cycle progression.167 Disruption of the Rb–Raf-1 complex by interfering peptides169 or small molecules170 suppressed the growth of experimental tumors and associated angiogenesis in nude mice. Two other potential Raf-1 substrates regulate myosin contractility. One is myosin phosphatase (MYPT), which binds to Raf-1 and is phosphorylated by Raf-1, leading to MYPT inhibition and enhanced cell motility.171 Raf-1 can phosphorylate MYPT on the same site as the Rho effector kinase Rok-α and myotonic dystrophy protein kinase (MDPK).171 Interestingly, both kinases are regulated by Raf-1; Rok-α via binding19 and MDPK were reported to be phosphorylated and activated by Raf-1 directly.172 Finally, Raf-1 was described to phosphorylate cardiac troponin T, which regulates the contractile function of cardiomyocytes.173
However, while the functional role of these putative Raf-1 substrates is not yet clear, new Raf effector pathways were discovered in which Raf-1 kinase activity is dispensable and the regulation occurs through association (Fig. 4). These discoveries were facilitated by the availability of conventional and conditional raf knockout mice.174 a-raf knockout mice are born alive but show neurological and intestinal defects of different severity depending on the genetic background.175 In contrast, b-raf–deficient embryos die around midgestation because of vascular defects in the placenta. Epiblast-restricted ablation, which leaves b-raf intact in the placenta but knocks the gene out in the embryo, resulted in live born animals that within 3 weeks succumb to a violent neurodegenerative disease.176 raf-1–deficient embryos show increased apoptosis of embryonic tissues or, more selectively, of the fetal liver depending on the genetic background.177,178 These divergent phenotypes show that Raf-1, B-Raf, and A-Raf serve distinct essential functions in embryonic development. It subsequently became clear that a main function of Raf-1 was to restrict caspase activation in response to selected stimuli, notably Fas stimulation, pathogen-mediated macrophage apoptosis, and erythroid differentiation.179,180 The ERK pathway can antagonize apoptosis in a number of ways, including the expression of caspase inhibitors and the neutralization of proapoptotic Bcl-2 family members.181 A further prominent prosurvival molecule, the transcription factor NFκB, was proposed as a downstream target of Raf-1.182-185 It is still unclear how Raf kinases activate NFκB, but it is likely through the induction of autocrine factors.184,185 Importantly, neither MEK/ERK nor NFκB activation is altered in raf-1– or a-raf–deficient cells and embryos, indicating that the prosurvival roles of Raf-1 and A-Raf do not depend on these functions. What are then the essential downstream targets of Raf in apoptosis?
A mitochondrial pool of Raf-1 was shown to protect cells from apoptosis.186-188 Raf-1 could be targeted to the mitochondria via interaction with Bcl-2 when Bcl-2 was overexpressed.189 In addition, selected growth factors were reported to promote Raf-1 translocation to the mitochondria via p21-activated kinase (PAK)–induced phosphorylation on S338.190,191 However, it still is unclear how mitochondrially localized Raf-1 could prevent apoptosis. Raf-1 facilitates the phosphorylation and inactivation of the proapoptotic Bcl-2 family member BAD,189 although this is likely an indirect effect mediated by ERK-activated RSK.192,193 In addition, Raf-1 serves as a scaffold to recruit protein kinase C theta (PKCθ) to phosphorylate BAD.194 Another described mechanism is the direct interaction between Raf-1 and mitochondrial voltage-dependent anion channels (VDACs), which may prevent the release of cytochrome C from the mitochondria.195
Gene ablation experiments in mice demonstrated that Raf-1 is required for survival and protection against apoptosis.177,178 Interestingly, reconstituting Raf-1–/– mice with a mutated Raf-1 (Raf-1 YY340/1FF), which has no detectable kinase activity towards MEK, fully rescued the apoptotic phenotype and produced viable mice.178 Tracking the cause revealed several mechanisms, which may operate in a tissue-specific manner, but none of which requires Raf-1 kinase activity. One is the control of the Rho effector kinase Rok-α, which is hyperactivated and mislocalized to the membrane in Raf-1 knockout cells.18,19 Hyperactive Rok-α causes a defect in the internalization of the Fas death receptor, which maintains high levels of Fas in the plasma membrane, leading to increased Fas sensitivity.18 The other targets of Raf-1 in apoptosis suppression are 2 proapoptotic kinases, ASK116 and MST2,17 which are inhibited by Raf-1 through direct binding (Fig. 4). These inhibitions do not require Raf-1 kinase activity but are solely mediated by binding. ASK1 is a protein kinase that works upstream of JNK and p38 to promote apoptosis induced by stress or by death receptors, such as the TNF-α receptor or Fas.179,196 It was reported that in human endothelial cells, Raf-1 mediates the protective effect of basic fibroblast growth factor (bFGF) against doxorubicin-induced apoptosis by binding to and inhibiting ASK1 at the mitochondria. Mutation of S338/339 in the N-region abolished association and protection.197 The mechanism of inhibition is not known, but the pathophysiological relevance of ASK1 inhibition by Raf-1 was demonstrated in a mouse model of heart disease.198 Knocking out the Raf-1 gene specifically in the heart muscle resulted in ventricular dilation and fibrosis caused by an increase in cardiomyocyte apoptosis. These pathological changes could be prevented by also knocking out ASK1.
The other Raf-1–inhibited proapoptotic kinase, MST2, was identified in a proteomics screen of Raf-1–associated proteins.17 The mechanism of how Raf-1 regulates MST2 was elucidated. MST2 activation involves dimerization and autophosphorylation of the activation loop.199 Raf-1 binds to the SARAH domain of MST2, thereby interfering with dimerization, and also recruits a phosphatase that dephosphorylates MST2.17 Raf-1 kinase activity is not required, and kinase-dead Raf-1 mutants also can inhibit MST2 activation. Consequently, MST2 activity is constitutively elevated in Raf-1 knockout cells and hyperactivatable by Fas stimulation17 or expression of RASSF1A.200 RASSF1A can disrupt the Raf-1–MST2 complex and promote the assembly of a proapoptotic signaling complex consisting of RASSF1A, MST2, LATS1, and YAP.200 In this context, MST2 phosphorylates LATS1, which phosphorylates YAP, thereby enabling YAP to interact with p73. The YAP-p73 complex binds to the promoter and activates the expression of the proapoptotic BH3 gene PUMA, culminating in the induction of apoptosis. This pathway was mapped using proteomics to track protein interactions that change in response to proapoptotic signals.200 This study flagged important gaps in our understanding of MST2 signaling. First, it showed that the MST2 pathway in mammalian cells shares the core kinase module MST2-LATS but differs from the orthologous Hippo pathway described genetically in Drosophila melanogaster in regard to upstream regulators and downstream effectors. This finding was subsequently confirmed by other studies.201,202 Second, it revealed the double nature of YAP as an oncogene as well as tumor suppressor. While in the liver, YAP is a potent oncogene,203,204 YAP also can suppress oncogenesis by stimulating apoptosis in response to DNA damage205-207 or expression of RASSF1A.200 RASSF1A is a major tumor suppressor gene that is altered in the majority of human cancers usually by gene silencing due to promoter methylation and much less frequently by mutation.208,209 Thus, being regulated by a major tumor suppressor pathway and a major mitogenic pathway, MST2 may have a critical role in coordinating apoptotic and transforming signals. Therefore, it is not surprising that MST2 is also targeted by the phosphatidylinositol 3-kinase (PI3K)/Akt survival pathway. Akt activation is required to curtail MST2 activation during growth factor stimulation. Akt phosphorylation of MST2 stimulates the dissociation of MST2 from RASSF1A and rebinding of MST2 to Raf-1.210 The role of Raf-1 is reminiscent of the role of the Myc proto-oncogene, which can stimulate transformation and apoptosis.211 Ras binding to Raf-1 enables Raf-1 to activate the MEK-ERK pathway and promote proliferation but at the same time dissociates the MST2–Raf-1 complex and promotes apoptosis.212 Coupling cell proliferation to the risk of cell death seems paradoxical but makes perfect sense for a multicellular organism in which the unlicensed proliferation of cells can cause severe diseases including cancer. Interestingly, B-Raf fails to bind and regulate MST2.17 Therefore, MST2 regulation by Raf is absent in Drosophila melanogaster, whose single Raf gene is most closely related to B-Raf. These observations suggest that MEK kinase activity was the primary function of Raf and that the ability to inhibit MST2 was acquired later in evolution.
Another defect observed when Raf-1 was knocked out specifically in keratinocytes was retarded wound healing and migration of keratinocytes.19 Again, this novel function of Raf-1 can also be carried out by a kinase-dead mutant, and just like prosurvival, it involves the inhibition of another kinase. The target of Raf-1 in motility is Rok-α.19 Raf-1 knockout fibroblasts and keratinocytes have a contracted appearance, have a defective cytoskeleton characterized by tight cortical actin bundles, and fail to migrate. Chemical inhibition of Rok-α or expression of a dominant-negative Rok-α mutant rescues all these defects of the Raf-1–deficient cells, indicating that Rok-α is the only target of Raf-1 in motility.19 Interestingly, investigating the mechanism of Rho-α inhibition by Raf-1 revealed a critical role for the Raf-1 CRD. Rok-α, like Raf-1, is regulated by autoinhibition, and its C-terminal regulatory region features a domain highly homologous to the CRD found in Raf-1. Indeed, dependent on an intact CRD, the Raf-1 regulatory domain could crossregulate Rok-α by binding to the Rok-α kinase domain and repressing its function.213 The biological relevance of this interaction was borne out in a Ras-induced skin tumor model in mice.214 In this model, the inhibition of Rok-α by Raf-1 was required for Ras transformation by maintaining the dedifferentiated state of the tumor cells.
A-Raf is the family member with the poorest kinase activity towards MEK and hence is likely to have functions outside of the classic ERK pathway. A-Raf strongly interacts with and inhibits MST2, again independently of kinase activity. Interestingly, this inhibition is contingent on the splice factor hnRNP H maintaining the expression of a full-length A-Raf protein.215 hnRNP H is often overexpressed in tumors.216 Downregulation of hnRNP H results in alternative splicing of the a-raf transcript that abolishes the expression of full-length A-Raf protein. Both A-Raf and MST2 are localized at the mitochondria in tumor cell lines as well as primary tumors.215
In a yeast 2-hybrid screen, pyruvate kinase M2 (PKM2) was identified as an A-Raf binding partner.217 PKM2 is an embryonic splice form of PKM that is aberrantly re-expressed in cancer and responsible for the aerobic glycolysis, also known as the Warburg effect, in cancer cells.218 A-Raf–mediated transformation increased the activity of PKM2 by promoting the transition of PKM2 from the low-activity dimeric to the highly active tetrameric form, and PKM2 enhanced A-Raf–induced cell transformation.217,219 These findings potentially link A-Raf to the regulation of energy metabolism and cell transformation, a topic that is increasingly recognized as critical for tumorigenesis.220
For a long time, the Ras/ERK signaling pathway was depicted as a linear pipeline. Over the years, it became clear that signaling pathways form networks consisting of multiprotein nodes at various subcellular compartments.24,221,222 A main question is how do these networks generate biological specificity? Much of this coordination depends on controlling protein-protein interactions by scaffold proteins that regulate the intensity, amplitude, and spatial specificity of the ERK pathway signal.223,224 Scaffolds act as docking platforms and anchors of the signaling components, bringing together the different modules of the cascade. Thus, by facilitating interactions between their clients, they decrease reaction rates to the first order, and they also reduce the number of tiers of the cascade, causing the input/output responses to become more linear.225 They insulate the clients from other pathways but also can connect pathways by binding components of different pathways. They can target their clients to different localizations, thereby increasing the variety of signals regulated by the cascade. There is now experimental evidence that scaffolds can link different localizations of Ras activation with the phosphorylation of specific ERK substrates.226,227 Feedback phosphorylation of the EGF receptor (EGFR) by ERK involved the IQ motif containing GTPase-activating protein (IQGAP) scaffold, while the phosphorylation of cytosolic phospholipase A2 (cPLA2) utilized KSR1 or Sef-1 when ERK was activated by Ras localized at the plasma membrane or Golgi, respectively.226 In addition, scaffolds seem to preferentially bind dimerized ERK and direct ERK to cytosolic substrates, whereas ERK dimerization is not required for the phosphorylation of nuclear substrates.227 The requirement for ERK dimerization is likely related to the overlap between the binding sites for substrates and scaffolds, implying that a dimer is necessary to simultaneously engage the scaffold and substrate.227
Therefore, scaffolds can have a huge impact on the biochemical and biological behavior of the ERK pathway.24,224 However, our knowledge of their role in the functional modulation of the pathway and their exact mechanism of action is still limited. One problem is that scaffolds are quite difficult to study experimentally. Their function is highly dependent on concentrations and the stoichiometric ratios with respect to their client proteins, and both downregulation and overexpression have similar effects, as both conditions reduce the number of functional complexes. Scaffold proteins of the ERK pathway were extensively reviewed.24,223,224,228-234 We, therefore, only describe selected examples that allow us to outline salient functions of scaffolding proteins in the regulation of the ERK pathway (Fig. 5). For convenience, we have classified them by their major known functions. However, occasionally, these functions overlap and will expand as more details become known.
The best-characterized scaffold of the ERK pathway is kinase suppressor of Ras 1 (KSR1). Initially identified as a suppressor of an activated Ras phenotype in Drosophila melanogaster 235 and Caenorhabditis elegans,236,237 KSR1 has a kinase domain with high homology with Raf-1 but mutations in residues critical for catalytic activity. Whether KSR1 has remaining kinase activity or whether it is a pseudokinase is still discussed in the literature.234,238,239 However, it is now accepted that the main function of KSR1 is as a scaffold of the ERK pathway, which regulates the intensity and duration of the ERK signal independent of catalytic function. KSR1 can interact with all kinases of the ERK pathway. MEK is constitutively bound, while Raf (Raf-1 or B-Raf) and ERK are recruited to KSR1 upon mitogen stimulation.80,240,241 However, KSR1 only binds less than 5% of endogenous Raf-1,242 indicating that KSR1 affects only a subset of Raf functions, and Raf members might be present in other protein complexes. KSR1 can activate Drosophila melanogaster Raf (which is closely related to mammalian B-Raf) allosterically by dimerization. In addition, KSR1 facilitates N-region phosphorylation of Raf-1 and B-Raf by recruiting CK2.77 Thus, KSR does not only regulate substrate availability but also catalytic activity, suggesting complex kinetic effects. In the context of cancer biology, KSR1 regulates Ras-mediated signaling, in particular differentiation, proliferation, and cellular transformation.243-246 Gene deletion of KSR1 in the mouse had little effect on viability but decreased the oncogenic effects of the polyoma virus middle T antigen247 and blunted oncogenic Ras-mediated tumorigenesis.248,249 KSR1 depletion also accelerated the immortalization of cells and led to resistance to cisplatin-meditated apoptosis, while overexpression of KSR1 sensitized tumor cells to this anticancer agent and other drugs.250,251 Taken together, the functions of KSR1 in the context of Ras/ERK signaling vary dramatically and depend, like other scaffold proteins, on the level of expression. At low and physiological levels, KSR1 seems to work as a positive regulator of signaling. By contrast, overexpression of KSR1 has an inhibitory function on the activation of the ERK cascade.233,234 Recently, KSR2, a homolog of KSR1, was shown to participate in the calcium-mediated activation of ERK.252 This regulation is exerted by the calcium-dependent phosphatase calcineurin, which binds to KSR2 and dephosphorylates it, resulting in increased membrane translocation and ERK signaling. Another proteomic study showed that KSR2 preferentially interacts with A-Raf in response to TNF-α.253 The biological relevance of this interaction remains to be elucidated, but as KSR2 knockout mice have a striking metabolic dysregulation that includes obesity,254,255 it is tempting to speculate that KSR2 may be involved in mediating A-Raf effects on metabolism.
Another group of Raf scaffolds is the connector enhancer of KSR (CNK) family of proteins. First identified as a modifier of KSR signaling in Drosophila melanogaster dCNK,256 mammals possess 3 isoforms that lack kinase activity but feature different protein-protein interaction domains that can bind a variety of client proteins including Raf-1 and B-Raf.224,234 Thus, CNK proteins seem to be superscaffolds that may integrate different signaling pathways. It is beyond the scope of this review to discuss CNK function in detail, and hence, we only will focus on the role of CNKs in Raf regulation. In Drosophila melanogaster, a multiprotein complex formed between dCNK, Raf, KSR, and a small adaptor protein HYP mediates Ras-induced activation of Raf.257,258 HYP has no mammalian homolog, and mammalian CNKs lack the Raf regulatory domain found in dCNK, suggesting a different mode of regulation. Mammalian CNK2 participates in the NGF-induced sustained ERK activation that is required for the neuronal differentiation of PC12 cells.259 The mechanism was not defined but, by analogy to CNK1, may involve facilitation of Raf activation. CNK1 can augment Raf-1 activation by increasing tyrosine phosphorylation of the N-region through recruiting c-Src.260 Interestingly, CNK1 also can bind RASSF1A and enhance apoptosis in a MST1/2-dependent manner.261 Thus, it is tempting to speculate that CNK1 may play a role in balancing apoptosis and proliferation by coordinating MST2 binding to RASSF1A and Raf-1, respectively.
In addition to KSR1, other ERK pathway scaffolds were implicated in tumor progression. Among them are the IQ motif containing GTPase-activating proteins (IQGAPs), a family of multidomain proteins.262,263 IQGAP1 directly interacts and modulates the functions of B-Raf, MEK, and ERK.264,265 Furthermore, IQGAP1 is required for the activation of B-Raf by EGF.266 As a result, IQGAP1 increases proliferation and reduces cellular differentiation. Thus, it comes as no surprise that IQGAP1 is involved in carcinogenesis.230,263 Augmented expression of IQGAPs was reported for several malignancies including cancers of the stomach, colon, lung, and prostate.263 Overexpression of IQGAP1 in human breast epithelial cells increased the formation and invasion of tumors, whereas reducing IQGAP1 expression had the opposite effect.267 Therefore, IGGAP1 is considered as a putative oncogene.268 Furthermore, IQGAPs are linked with metastasis, as IQGAP1 promotes cell migration and invasion via direct interactions with Cdc42, Rac1, actin, and calmodulin.263
Prohibitin (PHB) facilitates the displacement of 14-3-3 from Raf-1 by activated Ras, thereby promoting plasma membrane localization and phosphorylation of Raf-1 at the activating S338.269 Interestingly, PHB binds only to Raf-1, but not to Ras, and may function as a chaperone of Raf to enable interaction with Ras. In the context of cancer, PHB function was connected to immortalization, aging, and cell cycle regulation. Furthermore, PHB overexpression was reported in breast cancer, human endometrial adenocarcinoma,270 gastric cancer,271 and bladder cancer,272 although it is not known if this is related to aberrant Raf function.
Scaffold proteins are also crucial for the localization of the members of the ERK pathway to different subcellular signaling platforms. One such scaffold is similar expression to FGF (Sef-1), also known as interleukin-17 receptor (IL-17RD), which is situated at the Golgi apparatus.273 This transmembrane protein binds to activated MEK and facilitates activation of ERK but prevents ERK translocation to the nucleus. Therefore, ERK can only activate cytosolic targets. Interestingly, loss of Sef-1 expression is associated with high-grade metastatic prostate cancer.274 In clathrin-coated pits, the β-arrestins were proposed to augment ERK activation by scaffolding Raf-1, MEK, and ERK.275,276 The β-arrestins seem to act in a similar fashion as Sef-1, preventing ERK nuclear translocation and therefore restricting Ras signaling to the cytoplasmic effectors of the pathway.
The small scaffold MEK partner-1 (MP1) is an obligatory heterodimer with p14, and this complex interacts with MEK and ERK, targeting them to late endosomes.277,278 Recent results suggest that an additional adaptor, p18, is involved in specifying this subcellular localization.279 In vitro results indicated that MP1 also may facilitate MEK activation by B-Raf, although the mechanism is unknown.280 While decreased MP1 levels reduce ERK activation, overexpression of MP1 increases the binding of ERK to MEK and thus enhances the efficiency of ERK signaling.278,280 The speculation that the specific localization directed by MP1 generates signaling specificity was confirmed by the finding that MP1 specifically regulates PAK1-mediated ERK activation during cell adhesion and spreading but is not required for ERK activation by PDGF.281 Enhanced MP1 expression in several melanoma cell lines could be linked with a genetic translocation (4q23),282 suggesting a mechanism for the enhanced MAPK signaling in melanomas. In addition, MP1 may target MEK-ERK to high molecular weight protein complexes.283 Such complexes may be organized by MAPK organizer 1 (MORG1), a member of the WD-40 protein family, which was identified as an interaction partner of MP1 as well as Raf-1, B-Raf, MEK, and ERK.284 There is evidence that MP1 and MORG1 are part of a larger network built from nested scaffolds,283 as MEK binding to MORG1 is stabilized by MP1, Raf-1, and ERK. MORG1 acts like a classic scaffold with enhanced activation of ERK at low concentrations and being inhibitory at higher concentrations.284 Furthermore, MORG1 promotes ERK activity in response to serum or other signals. Interestingly, MORG1 was also shown to act as a scaffold with hypoxia-inducible factor prolyl hydroxylase 3 (PHD3) and downregulation of MORG1-augmented HIF-1 activity, suggesting MORG1 as a connection to other signaling networks.285
The multidomain protein paxillin is a component of focal adhesions, providing a structural and signaling link between the actin cytoskeleton and the extracellular matrix (ECM).286 Paxillin constitutively interacts with MEK but in response to growth factors also binds to activated Raf and ERK, directing activated ERK to sites at the focal adhesions.287 The most significant impact of paxillin is on developmental processes and on tissue morphogenesis,287-289 but it also plays a role in tumor cell invasion.290 Elevated levels of paxillin, together with enhanced Src activity, contribute to the high metastatic potential of human osteosarcomas, and in gastric cancer, high levels of paxillin correlated with advanced tumor stage and invasiveness.291
Raf kinase inhibitor protein (RKIP) is receiving sharply increasing attention as a modulator of the ERK pathway and several other signaling pathways including G protein signaling and NFκB signaling.224,292-294 Initially identified as phosphatidylethanolamine-binding protein-1 (PEBP-1), RKIP was later identified as a negative modulator of Raf-1.91 RKIP binds Raf-1, MEK, and ERK. While RKIP can bind MEK and ERK simultaneously, binding to Raf-1 and MEK is mutually exclusive, disrupting the Raf-1–MEK complex and activation of MEK by Raf-1.90 RKIP also interferes with Raf-1 activation by preventing the interaction with PAK1 and Src kinases and the phosphorylation of the N-region.295 In this study, B-Raf activation was not affected. However, another study found that RKIP inhibited B-Raf activation in cells as well as its ability to phosphorylate MEK in vitro.296 Interestingly, in an RKIP-related protein, hPEBP4, the Raf- and MEK-binding sites, which overlap in RKIP, are separated by an insertion converting hPEBP4 into a scaffold for the Raf-1–MEK complex.105 Consequently, hPEBP4 enhances the activity of the ERK pathway in growing human myoblasts. Upon induction of differentiation, hPEBP4 expression rises to levels that exceed the optimal stoichiometric relationship to its client proteins and contribute to the inhibition of the ERK pathway observed during myoblast differentiation.105 This regulation highlights not only that scaffolds can assume both stimulating and inhibitory roles under physiological conditions but also that a primary function of such proteins is the fine tuning of the activation kinetics of signaling pathways. RKIP has a main function in causing switchlike activation behavior of ERK and in supporting oscillations of the pathway caused by the negative feedback from ERK to Ras activation.297 Physiologically, RKIP plays a major role as a suppressor of cancer invasiveness and metastasis in various cancers294 including common cancers of the prostate,298 breast,299 colon,300 and liver.301
Other negative modulators of Raf-mediated signaling are members of the Sprouty (Spry) family, comprising various Spry and Spred (Spry-related proteins with an EVH1 domain) isoforms.302-306 These proteins are negative feedback regulators of the ERK signaling pathway, and their expression is regulated by the cascade. Depending on the context, Spry proteins inhibit ERK signaling by binding and sequestering the Grb2-SOS complex, thus preventing Ras activation.307 Additionally, both Sprouty and SPRED can physically interact with Raf-1 and B-Raf, interfering with the phosphorylation of Raf on activating sites.308-310 Spry genes and proteins were shown to be deregulated in different tumor types. Spry1 and Spry2 are downregulated in breast, prostate, and liver cancer.223,310,311 This downregulation seems to be due to hypermethylation of the promoter region, indicating that Sprys are putative tumor suppressors.
For almost 2 decades, research focused on Raf-1 as the critical Ras effector of the Raf family. However, this changed when Davies et al. described mutations of B-Raf in 66% of melanomas and at a lower frequency in a wide range of human solid cancers.20 Further research revealed that approximately 2% of human malignancies carry a mutation in B-Raf,312 with highest frequencies observed in melanoma and carcinomas of the colon, thyroid gland, ovary, and biliary tract.21,313,314 Currently, more than 100 different B-Raf mutations were described, with V600E (formerly labeled as V599E) being by far the predominant lesion (Catalogue of Somatic Mutations in Cancer: www.sanger.ac.uk/genetics/CGP/cosmic). Most of the B-Raf mutants associated with cancer are located in exons 11 or 15 in the kinase domain. The biggest group of mutations (including V600E) affects residues that normally stabilize the kinase in the inactive form. Mutations of these amino acids disrupt this conformation, usually resulting in a significantly increased B-Raf kinase activity that leads to the constitutive activation of the ERK pathway.31 However, even impaired kinase activity mutants can constitutively activate the ERK pathway because of their ability to heterodimerize with and activate Raf-1.31,81 Interestingly, B-Raf mutations normally do not coexist with oncogenic mutations in Ras in human tumors, arguing that they are equivalent in their transforming effects.315 This conclusion indeed highlights B-Raf as a critical effector of Ras in cell transformation and cancer. In further support of this interpretation, mutant B-Raf has also been shown to be a critical step in tumorigenesis in mouse models of melanoma.316,317 Melanocyte-specific, conditional expression of the V600E mutation inserted into the endogenous BRAF gene locus resulted in the development of both benign nevi and malignant melanoma. This is in line with observations from human tumors, where B-Raf V600E is detected in approximately 44% of melanoma cases,316 but with even higher frequency in benign nevi, which do not progress into a malignant melanoma.318 This dormancy is probably due to the potent induction of senescence by B-Raf V600E, which suppresses tumorigenesis.316,319 These data suggest that additional secondary alterations cooperating with mutant B-Raf may be required. Further support of this hypothesis comes from observations demonstrating that B-Raf V600E induces senescence and needs secondary events like p16INK4A loss to overcome it.319 Interestingly, melanomas in the above-mentioned mouse model did not show any alterations in p16INK4A, suggesting that the pathogenetic mechanisms differ between the mouse and human or else that there are other, hitherto unknown possibilities for secondary genetic events. However, there are also B-Raf mutations associated with human cancer, which display impaired kinase activity, with D594V being the most frequent one. These mutations require Raf-1 to activate the ERK pathway,31 relying on the ability of catalytically compromised B-Raf to activate Raf-1 by heterodimerization.81,82 As Raf heterodimerization is augmented by activated Ras, these low-activity B-Raf mutants (in contrast to the high-activity mutants) can be found coexpressed with mutant Ras.154
Although quite rare, cancer-associated mutations also were reported in Raf-1. They were first described in a mouse model of chemically induced lung cancer,320 in human cancer cell lines,321 and finally in patients with therapy-related acute myeloid leukemia (t-AML).322 Interestingly, the latter mutations were detected in the germline of affected patients but still exhibited weakly transforming and antiapoptotic properties. Therefore, they might constitute a hereditary predisposition to solid neoplasms and t-AML. This hypothesis was supported by the observation that constitutive activation of the ERK pathway in affected patients was only observed in malignant but not in the surrounding normal tissues. Indeed, further studies identified a leukemia-specific, somatic loss of RKIP as a genetic second hit that further promoted malignant transformation in these patients.323
Besides mutations, other alterations of Raf genes were described in human malignancies as well. Rearrangements and fusions of B-Raf and Raf-1 to a variety of other genes have been described in thyroid cancer, pilocytic astrocytoma, prostate cancer, gastric cancer, and melanoma.324-328 They seem to be particularly frequent in sporadic pilocytic astrocytoma, with more than 60% of cases demonstrating B-Raf rearrangements.324 Usually, the resulting fusion products lose the regulatory N-terminal region but retain an intact Raf kinase domain. They can activate the ERK pathway, transform transfected cell lines, and induce tumors in nude mice, underlining their functional role in the pathogenesis of human malignancies. Mutated B-Raf further was found amplified in human melanoma with a gain of chromosome 7q. This amplification is a frequent event in melanoma, suggesting that B-Raf mutations are one of the factors driving its selection.329 Elevated levels of A-Raf mRNA and protein were observed in a number of malignancies including head and neck squamous cell carcinomas and colon carcinomas.215 This study also showed that the splice factor heterogeneous nuclear ribonucleoprotein H (hnRNP H) is required for the correct splicing and expression of full-length A-Raf. Elevated expression of A-Raf was found in testicular germ cell tumor–derived cell lines caused by the duplication of the X chromosome.330
Raf mutations are not only critical steps for tumorigenesis but also for the pathogenesis of rare developmental disorders, such as neurofibromatosis type 1 and Costello, Noonan, LEOPARD, and cardiofaciocutaneous (CFC) syndromes, which are reviewed in detail elsewhere.331-333 Affected individuals present with overlapping yet distinct phenotypes that include a variable degree of mental retardation, cardiac defects, facial dysmorphisms, short stature, macrocephaly, and skin abnormalities. Germline mutations in Raf were first linked to these disorders when 2 groups simultaneously reported mutations in B-Raf causing CFC.334,335 The B-Raf mutations in CFC comprised mainly hitherto unknown mutations, which were more widely distributed across B-Raf as compared to their counterparts detected in human cancer. However, a few mutations are shared between cancer and CFC. Some of the CFC germline mutations resulted in increased B-Raf kinase activity and constitutive activation of the ERK pathway.336 Compared to the V600E mutation, the kinase activity and transforming capacity of CFC B-Raf mutants seem to be lower.333,337
Germline mutations in Raf-1 were recently described in both Noonan and LEOPARD syndromes.338,339 As with B-Raf, one of the mutations (S427G) was found in both Noonan syndrome and human malignancies, and again, some of the germline variants increase Raf-1 kinase activity and transforming ability.322,340 One might expect that carrying an oncogene in the germline increases the risk for the development of malignancies. Indeed, patients with neurofibromatosis type 1, Costello syndrome, and Noonan syndrome are at increased risk for developing a wide range of solid tumors and hematological malignancies.331-333 Whether CFC and LEOPARD syndromes result in a predisposition to tumor development is an open question. The numbers of patients affected by these disorders are too low for a thorough statistical analysis, and descriptions of patients developing a malignant disorder are limited to case reports.341-345 However, close monitoring of patients with all germline Ras/MAPK disorders, including LEOPARD and CFC syndromes, for the occurrence of neoplasias is often suggested.
The efforts to develop drugs targeting the Raf family and their downstream effectors were increased after strategies implemented to inhibit Ras signaling failed in the preclinical and clinical studies.346 Different approaches included the inhibition of Raf kinase activity by small molecule inhibitors, decreasing Raf protein levels using antisense oligonucleotides, and targeting Raf protein-protein interactions, especially the Raf-Ras interaction. The first drugs were developed against Raf-1, but the discovery of B-Raf–activating mutation in tumors20 shifted the efforts toward the inhibition of this protein and of MEK1/2.
Sorafenib (BAY 43-9006) was the first Raf inhibitor to progress into clinical trials. Preclinical studies showed that sorafenib inhibited Raf-1 and B-Raf in tumor cell lines and xenograft models for Ras-dependent tumors.347,348 Today, sorafenib is approved for the treatment of advanced renal cell carcinoma (RCC) and unresectable hepatocellular carcinoma (HCC). However, sorafenib monotherapy failed to be clinically effective against other tumors such as melanoma, although it increased progression-free survival in combination with other treatments.349,350 Sorafenib was developed as a specific Raf-1 kinase inhibitor, and while it poorly inhibits mutant B-Raf, it is highly effective against several other kinases such as VEGF and PDGF receptors.351 In fact, sorafenib is now regarded as a multikinase inhibitor, and the success in HCC and RCC is probably primarily because of the inhibition of VEGF and PDGF receptors rather than Raf kinases. The results of the sorafenib clinical trials raised questions about the suitability of Raf proteins as therapeutic targets and the low predictive properties of the preclinical models. One possible explanation for the unexpected therapeutic target spectrum of sorafenib is that it is a poor B-Raf inhibitor. This conclusion led to the development of a new generation of inhibitors active against B-Raf. RAF265 is active against all Raf isoforms, mutant B-Raf, and VEGF-2. RAF265 inhibits cell proliferation in mutant B-Raf and N-Ras melanoma cells but has no effect in cell lines that express the normal genes.352 This finding led to the initiation of a series of clinical trials for the treatment of advanced melanoma in which patients were evaluated for B-Raf and N-Ras mutation before treatment. The results from these clinical trials are eagerly awaited and are expected to be published shortly. XL281 is another pan-Raf inhibitor that is active against mutant B-Raf and currently is in phase I clinical trials. Finally, PLX4032 is a potent B-Raf V600E selective kinase inhibitor that suppresses the activation of the ERK pathway and cell proliferation in melanoma xenografts.353,354 PLX4032 does not inhibit the ERK pathway in cells that do not express mutant B-Raf. A recent clinical phase I study reported a spectacular response rate in 81% of melanoma patients with mutant B-Raf.355 These results demonstrate that mutant B-Raf is an excellent therapeutic target in melanoma. However, this success did not come without cost, as 31% of patients developed skin tumors, keratoacanthomas, and squamous cell carcinomas. This side effect may be due to the enhancement of aberrant ERK pathway activation by drug-induced B-Raf–Raf-1 dimerization as discussed below. Although the skin tumors can be easily recognized and surgically removed, the appearance of malignancies in other less accessible organs remains a concern, and further studies are advocated before PLX4032 is approved for the treatment of metastatic melanomas.
The generation of drugs that target Raf interactions with other proteins is not as advanced but conceptually promising. MCP-110 is a small molecule that inhibits the Raf-Ras interaction. This compound decreased anchorage-independent growth in cell lines expressing oncogenic Ras but did not affect cell lines with a constitutively active Raf-1, indicating that MCP-110 is working specifically at the level of the Ras-Raf interaction.356 However, this agent did not progress into preclinical development. Another Raf interaction that was targeted is the binding to the Rb tumor suppressor protein. As mentioned above, Raf-1 was reported to bind to and phosphorylate Rb, resulting in Rb inhibition and S phase progression.357 Small synthetic peptides that interrupt the Raf-1–Rb interaction in cell lines suppressed the growth of A549 xenograft tumors.169 The use of peptides as drugs is limited by their short half-life and problems in their delivery, but similar results were obtained using the small-molecule drug RRD-251. This drug inhibited cell proliferation in vitro and suppressed the growth of xenograft tumors as well as tumor angiogenesis in a manner dependent on the expression of intact Rb.170 Although no clinical data are available, these results indicate that targeting the Raf-1–Rb interaction may be a successful antitumoral strategy. It also suggests that targeting other Raf interactions may be a good strategy. Of special interest would be the disruption of Raf-1 binding to Rok-α, which may promote the differentiation of epidermal skin tumor cells,214 and the dissociation of Raf-1 from MST1/217 and ASK1,16 which should activate the proapoptotic potential of these kinases. More interestingly, the disruption of the Raf-1/A-Raf–MST1/2 complex should compensate for the frequent loss of the RASSF1A tumor suppressor, which normally promotes the disruption of the MST–Raf-1 complex and activation of MST1/2.200 Such drugs would restore a natural tumor suppressor function and hence may be expected to be specific, efficacious, and without severe side effects.
A greater advance was made in targeting RAF expression using antisense oligonucleotides. The use of oligonucleotides for therapy is restricted by their susceptibility to degradation by nucleases, but the substitution of oxygen for sulfur in the phosphodiester linkages confers stability to these molecules.358 Using these chemical modifications, different compounds were generated, such as ISIS 5132 and ISIS 2503. ISIS 5132 is a 20-base phosphorothiate antisense oligonucleotide against Raf-1 that inhibited tumor progression in clinical trials. However, subsequent phase II clinical trials demonstrated that this compound was of no benefit as a single agent, and hence, it was not further developed.356 Another Raf-1–directed antisense agent that entered clinical trials is LErafAON, a liposome-entrapped derivative of a 15-mer antisense oligonucleotide.356 Packaging the antisense oligonucleotide within liposomes was expected to protect the oligonucleotide against nucleases and increase cell delivery. Unfortunately, several phase I clinical trials demonstrated a lack of objective responses but adverse side effects due to the liposomal formulation.359,360 The failure of the antisense drugs suspended the development of this approach for Raf inhibition. However, improvements in delivery methods,361 and the observation that small interfering RNA (siRNA) against mutant B-Raf362,363 and Raf-1364 can inhibit cell proliferation in melanoma and breast cancer xenograft models, respectively, may rekindle the interest in this strategy.
A decade ago, a report was published showing that the Raf inhibitor ZM 336372 produced a massive paradoxical activation of Raf kinases when cells were treated with the inhibitor, and then, Raf kinases were isolated and their activity measured in vitro.152 Now, 3 publications finally shed light on this apparent contradiction.32,153,154 The key is Raf-1 homodimerization or heterodimerization with B-Raf, which is driven by mutant Ras and facilitated by the Raf inhibitor drugs. The activation conferred by dimerization is not compromised if the kinase activity of one of the Raf dimerization partners is destroyed and the kinase activity of Raf-1–B-Raf heterodimers is higher than that of Raf-1 homodimers.82 Dimerization, induced by Ras or by Raf inhibitors, of mutant B-Raf V600E and Raf-1 actually dampens the overall kinase activity.365 These constellations exacerbate the activation effect when B-Raf–specific inhibitors are used and Raf dimerization is ushered by Ras mutations. Thus, Raf inhibitors are rather effective when B-Raf is mutated but ineffective when Ras is mutated.32,153,154 Beyond this shared theme, the 3 studies differ in mechanistic details.
In cell lines expressing high-activity B-Raf mutants, Raf inhibitors function as expected and efficiently reduced signaling through the ERK pathway.154 Consequently, cell proliferation was inhibited in vitro as well as in xenografts when cells were treated with B-Raf inhibitors. Surprisingly, in cell lines lacking an activating B-Raf mutation and expressing mutant Ras, the effects were opposite.32,154 Specific B-Raf inhibitors, such as PLX4720, enhanced ERK phosphorylation, and the cells demonstrated a high drug tolerance in xenograft models and proliferation assays. Moreover, the inhibitors initiated Raf heterodimerization, which enhanced the kinase activity and downstream signaling. Pan-Raf and less specific inhibitors, on the other hand, enhanced ERK signaling at a lower concentration, while higher doses reverted this effect and caused inhibition.153 These data suggest that both Raf-1 and B-Raf kinase activities present in a dimer need to be inhibited in order to shut down the signaling efficiently. Consistently, the inhibitor-induced ERK activity is Raf-1 dependent when B-Raf–specific inhibitors are used.32,153 Interestingly, B-Raf inhibitors are able to activate Raf-1 independently of B-Raf.153 PLX4720 robustly induces ERK phosphorylation in B-Raf–/– fibroblasts, which led to the conclusion that association with the inhibitor and drug-induced Raf-1 homodimerization is sufficient to enhance Raf kinase activity and ERK signaling. Heidorn et al.154 could also demonstrate in a mouse model that mutant K-Ras and kinase-dead B-Raf cooperated in the induction of melanomas. This is a worrying discovery, and it would be interesting to know if B-Raf inhibitors would have the same effect as kinase-dead B-Raf in promoting melanoma in a mutant K-Ras model.
In light of these data, the development of keratoacanthomas and squamous cell carcinoma in 31% of melanoma patients treated with PLX4720355 may be ascribed to the paradoxical activation of Raf signaling by drug-induced heterodimerization. These results have direct implications for clinical practice. Firstly, the patient population, which should receive B-Raf–specific inhibitors, has to be carefully selected, as only tumors expressing activating B-Raf mutations will respond to the treatment, while mutant Ras expression may even worsen the condition. Furthermore, a combination therapy including pan-Raf and MEK inhibitors should also be considered in order to enhance the efficacy (see below) and limit adverse side effects and the fast onset of drug resistance.
The first highly selective and effective inhibitors against the ERK pathway were directed against MEK1/2. Work from different groups in preclinical models indicated that MEK inhibitors are highly effective if the pathway is activated by B-Raf mutations but rather ineffective against cells harboring mutant Ras.366,367 CI-1040, the first MEK inhibitor to proceed to clinical phase I trials, had limited side effects and reduced ERK phosphorylation in more than 65% of the tumor biopsies.368 However, a subsequent phase II trial did not show clinical responses in a variety of tumor entities,369 and consequently, the development was stopped. Nevertheless, these studies suggested that targeting MEK was relatively safe, and new generations of MEK inhibitors were developed.370 PD0325901, one of these new-generation MEK inhibitors, is 50 times more potent than CI-1040 and has better pharmacological properties. However, the results from 2 clinical trials were disappointing, showing a high level of toxicity, and PD0325901 was dropped from further development.348 Currently, 7 MEK inhibitors are in different phases of clinical trials, and many more are in preclinical development.370 The results published so far show different antitumoral efficacies and also different levels of toxicity.
Thus, the development of drugs against the Raf family and MEK has produced both successes and disappointments, which have helped to develop new strategies that may lead to better compounds or better use of compounds. Most of the Raf and MEK inhibitors show limited effect as single agents, but they may still be effective in combination with classic chemotherapeutic agents or drugs that specifically target other signaling pathways, such as PI3K or growth factor receptor inhibitors. Deriving effective drug combinations will be helped by the rational prediction of drug responses using mathematical and computational modeling (see below) and by the identification of better biomarkers that allow discriminating, which patients would benefit from the treatment with Raf or MEK inhibitors, leading to the creation of personalized protocols for individual patients.
Despite the 25 years of research on Ras and Ras effectors, we are still bewildered by the diverse functionalities of this pathway, which can specify a multitude of often contradictory biological outcomes. While we have identified a large number of components and thoroughly characterized the individual functions of many of them, we still lack an understanding of how the Ras-Raf network processes signals to generate specific biological responses. Emerging evidence shows that much of this specificity is encoded by the pathway structure and dynamic changes in the connections between the proteins.24,126,297,371,372 These so-called emergent properties are difficult to understand by experimentation alone. Therefore, mathematical and computational modeling approaches were developed that allow us to analyze and predict network responses including the behavior of the Ras-Raf pathway.24,373
The “backbone” of the Raf pathway consists of a 3-tiered kinase cascade. This design allows for a larger repertoire of regulation by feedback, crosstalk, and scaffolding to increase the number of signaling functions a single pathway might perform. It also allows for signal amplification, turning a low-abundance, noisy signal into a higher abundance, clearer signal at each tier of the cascade. Measured in terms of active molecules, the amplification factor from active Ras to active ERK can be 6- to 30-fold.374-376 Not only does this design amplify signals, but it also amplifies the sensitivity of the cascade output (ppERK) to the cascade input (RasGTP).377,378 This property was also shown experimentally in Xenopus oocytes, where the effective cooperativity (Hill coefficient) increases with each tier of the cascade, so that a defined increase in the stimulation of the first component causes successively larger increases downstream.379 The activation mechanism by multisite phosphorylation adds further to this sensitivity amplification, if the phosphorylations occur in a distributive manner, where each phosphorylation requires a separate binding event between kinase and substrate.380,381 In vitro ERK phosphorylation by MEK and dephosphorylation by MAPK phosphatase 3 (MPK3) follow such a distributive mechanism,382,383 which based on theoretical considerations can give rise to switchlike behavior, bistability, and oscillations.380,384 While bistability of ERK activity was experimentally observed in several settings,371,372,379,385 oscillations may be confined to certain experimental conditions and pathway topologies.297,386-388 Bistability means that the system is either “off” or “on,” even after the initiating stimulus has ceased. Thus, bistability is advantageous in biological processes that require irreversible decisions, such as cell cycle progression, differentiation, or cell death. In contrast, the biological role of oscillations of activities in signaling pathways is less obvious but may facilitate the temporal synchronization of processes.
The properties arising from signal and sensitivity amplification provide advantages, for instance, the ability to respond to small input signals, while filtering out noise through the switchlike behavior. However, there are also tradeoffs. Amplifiers may overshoot the desired output levels, and the more sensitive the cascade becomes, the quicker the output will saturate, and it will respond to a smaller range of inputs. Biology has found ways to avoid these tradeoffs, and they include feedback mechanisms. Modeling is very useful in analyzing how feedback loops change the behavior of a biological system, as it is usually difficult to isolate these feedbacks in an experimental setting but easy to do so within a mathematical model. The Ras/ERK pathway features a number of negative and positive feedback loops (Fig. 6), which can generate a wide variety of dynamic behavior.
The light blue lines in Figure 6 denote short-term negative feedback loops, which begin to act nearly immediately upon activation of ERK. Activated ERK leads to the phosphorylation and inactivation of the RasGEF SOS,46,47 and modeling suggests that several phosphorylation sites on SOS all independently mediate strong negative feedback.389 ERK feedback phosphorylation affects targets at each upstream activation step. The first is the EGFR. Phosphorylation of T669 by ERK triggers multiple effects including decreases in receptor internalization, kinase activity, and the phosphorylation of selected substrates.390,391 Blocking this feedback inhibition by MEK inhibitors increased EGFR activity and enhanced epithelial-to-mesenchymal transition and migration.392 The next layer of negative feedback targets comprises adaptor protein working immediately downstream of the EGFR, that is, the Grb2-SOS complex,46,47 and Gab1, a scaffolding protein involved in PI3K and RasGEF recruitment to the plasma membrane.393 Further downstream, Raf-1,92,385 B-Raf,82,144,394 and MEK158 are phosphorylated, leading to decreased pathway activity as discussed above.
When strong negative feedback is combined with the amplifier, constituted by the Raf-MEK-ERK cascade, the pathway adopts characteristics of a negative feedback amplifier (NFA).385 This circuitry is widely employed in electronic systems to confer response linearization and robustness to noise.395 These properties also exist in the biological NFA, as predicted by mathematical modeling and experimental validation.385 First, the NFA rendered ERK activation to become more linear in response to input dose, countering the effects of multiple kinase tiers and multisite phosphorylations that amplify input/output sensitivity and cause switchlike behavior. Second, ERK activation became robust to internal perturbations; for example, when MEK activity was incompletely inhibited, the attenuation of the negative feedback permitted the input to rise and sustain MEK activity. Breaking the negative feedback, for example, by expression of Raf-1 mutants that are resistant to feedback phosphorylation by ERK, very effectively sensitized the system to MEK inhibitors. Several important conclusions emanated from the analysis of the NFA design. Proteins embedded in NFAs are difficult drug targets because the NFA design will buffer any inhibition that is not complete. In contrast, inhibition of inputs outside of the NFA module, such as growth factor receptors, produced a linear dose response curve. Even more interestingly, weakening the negative feedback by a Raf inhibitor drastically improved the efficacy of a MEK inhibitor. Thus, the mathematical model made a concrete suggestion for a highly efficient drug combination, that is, Raf and MEK inhibitors, which from the experimentalists’ point of view appears so counterintuitive that it probably never would have been tested. Thus, the analysis of design principles of signaling pathways by mathematical models can give very applicable results.
The next layer in time is the delayed negative feedbacks that emanate from ERK but that are mediated by the transcriptional induction of feedback inhibitors. They are depicted by the dark blue lines in Figure 6 and only begin to take effect approximately 30 minutes or longer after ERK activation. Active ERK induces the transcription of multiple cytoplasmic and nuclear dual-specificity phosphatase (DUSP) isoforms, which dephosphorylate and deactivate ERK.396,397 In some instances, the stability and/or the phosphatase activity of DUSPs is also controlled by ERK activity, resulting in a positive feedforward loop embedded into this negative feedback.398,399 In addition, ERK stimulates the transcription of other pathway inhibitors, such as the Sprouty and Spred proteins,304,306 which inhibit EGFR endocytosis, RasGEF recruitment, and Raf activation, and Mig6/RALT, which not only inhibits the activity of various receptor tyrosine kinases (RTKs)400,401 but also leads to increased ErbB1 receptor degradation in a manner apparently independent from the traditional ligand-stimulated pathway.402
Another function of negative feedback is adaptation or return of ERK activity at steady state to near prestimulus levels despite the persistence of stimulus.403,404 Which negative feedback(s) are responsible for adaptation was the topic of many theoretical studies, but there is still no consensus as to which are the most important in general or if other mechanisms such as receptor downregulation play a role. When ERK itself is responsible for the direct negative feedback, then the system may adapt but will not exhibit perfect adaptation, where Ras and ERK activity returns exactly to prestimulus levels.404 The perfect adaptive behavior is characteristic of an engineering design termed “integral negative feedback,” where the strength of the negative feedback is proportional to the time-integrated ERK activity. Recent modeling work suggests that the long-term, transcriptional negative feedbacks might act as such integral negative feedback circuits, as mRNA responses of active ERK-dependent genes are proportional to the total time active ERK spends in the nucleus.397 Thus, a function of the delayed negative feedback distinct from the short-term negative feedbacks may be to achieve adaptation, that is, a resetting of the system after the response has been executed. Adaptation of biochemical networks usually requires either a negative feedback that acts proportional to the input or a negative feedback combined with a positive feedforward (incoherent feedforward loop).405,406
The orange and red lines in Figure 6 denote short-term and long-term positive feedbacks, respectively. The short-term feedbacks include the ERK-mediated 1) phosphorylation and inactivation of RKIP,297 a protein that inhibits Raf’s ability to activate MEK; 2) phosphorylation of PEA-15, which releases ERK from this cytosolic anchor protein and allows the nuclear accumulation of ppERK407,408; and 3) phosphorylation of NF-1, a RasGAP whose ability to convert RasGTP back into the inactive RasGDP form is inhibited by ERK phosphorylation.408 In addition, RasGTP produced by the RasGEF SOS can bind to another site in SOS that allosterically stimulates GEF activity.409 In T lymphocytes, this positive feedback is key to establish a bistable Ras response that contributes to the establishment of memory cells, which maintain the ability to be rapidly reactivated by the specific antigens they have encountered before.410 The biological relevance of positive feedback includes the generation of bistability. Bistability typically is brought about by strong positive feedback, which maintains the response even after the input was removed. Thus, bistability in the Ras/ERK pathway underlies processes that require clear and sustained “on” and “off” signals, such as memory formation in individual neurons411 and cell fate decisions of neuronal371,408 and lymphoid cells.410,412 The long-term positive feedbacks involve the ERK-induced autocrine production or release of growth factors that entertain further ERK activity by stimulating receptors.413 Such autocrine mechanisms are widely implicated in development, cell differentiation, and tumorigenesis.414-416
Although we know that the Ras/ERK pathway is only part of a much larger network, most of our analysis treats it as an isolated entity. This approach has proven very effective, but we have to be aware that much of the distinction between pathways may simply reflect the historic sequence of discoveries that conveniently helps us to compartmentalize the network into conceptually and experimentally accessible entities. From this vantage point, we usually summarize connections between pathways as crosstalk. Much of the known crosstalk between ERK and other pathways is positive, but there are also modes of negative crosstalk. The assortment of crosstalks discussed here is certainly incomplete, and many of the crosstalk mechanisms are likely to be cell type dependent. However, a striking observation is that many crosstalk mechanisms are under the control of Ras or act on Ras, suggesting that Ras plays a major role in coordinating this crosstalk.
A prime example is PI3K, which can be activated by receptors and by Ras directly.417 PI3K phosphorylates PIP(4,5)2 to produce PIP(3,4,5)3, which binds with high affinity to proteins containing pleckstrin homology (PH) domains, such as the Gab scaffolds.418 This allows PI3K to facilitate activation of Ras through 2 mechanisms, as Gab1 recruits SOS complexes and also the phosphatase SHP2.419 SOS directly activates Ras, while SHP2 maintains RasGTP levels by dephosphorylating residues on RTKs that recruit RasGAP.420 Furthermore, PI3K also leads to activation of PAK, which phosphorylates Raf-1 on the activating S338 residue.72 A main effector of PI3K is Akt, which intersects with the ERK pathway on several levels. Akt was reported to inhibit Raf-1 by phosphorylation of S259,103 which however was not substantiated in subsequent work.104,105 More interestingly, Akt shares several substrates with ERK, where phosphorylation by ERK and PI3K acts synergistically. Examples include the proapoptotic protein BAD, which is jointly inactivated by Akt phosphorylation and ERK-dependent phosphorylation.421 Similarly, both Akt and ERK jointly phosphorylate PEA-15, a cytosolic anchor for ERK, thereby inducing the release of ERK and allowing ERK phosphorylation and nuclear accumulation.408 In this case, positive crosstalk is coupled with positive feedback, which is likely to yield highly nonlinear synergistic effects. Another example for positive crosstalk emanates from the PLCγ pathway, which is activated by many RTKs on the same time scale as Ras. Active PLCγ cleaves the phospholipid PIP(4,5)2 to produce diacylglycerol (DAG) that stays in the plasma membrane and soluble inositol triphosphate (IP3) that induces a calcium release from the endoplasmic reticulum (ER). DAG and calcium, alone and in combination, can activate 2 classes of proteins important for Ras/ERK signaling: a family of RasGEFs called the RasGRPs and various protein kinase C (PKC) isoforms. There is not yet evidence of whether the RasGRPs are subjected to negative feedback regulation as is SOS. However, RasGRP primes SOS for allosteric activation by RasGTP and thereby plays an essential role to cause the bistable Ras activation in T lymphocytes discussed above.410 However, calcium via its binding protein calmodulin also can inhibit Ras-mediated activation of ERK signaling.422,423 Activated PKC enhances signaling through the ERK pathway through several mechanisms including the inhibition of RKIP by phosphorylation on S153.371,424 RKIP can also be disabled by ERK phosphorylation on S99,297 and this event plays a major role in the crosstalk between the ERK and Wnt pathways.425
What is the purpose of all this crosstalk? Likely, the primary reason is the integration of various environmental cues by the cell to make appropriate decisions. For instance, it is known that under normal circumstances, adherent cells will not proliferate while not attached, and the crosstalk between adhesion/PI3K signaling and ERK signaling may coordinate this response.73,74 Similarly, the crosstalk between the cAMP and ERK systems, which was discussed above, can inhibit Raf-1 but stimulate B-Raf.426 Thus, the specific response depends on the expression of Raf-1 and B-Raf, which can differ between cell types and tissues.
What have we learned about the Raf-MEK-ERK pathway as an effector of Ras in the 17 years since this relationship was brought to light? A simple summary could be that Rafs are the “first in, last out” in Ras signaling. In accounting, this is a method of inventory valuation, which is based on the assumption that the cost of goods purchased first is the cost of goods sold last. In other words, in a growing enterprise, the value of the old stock rises with the value of new additions. Undoubtedly, exploring the Ras signaling network is a blooming industry. We now have entered a phase in which the mapping of the components of signaling networks is rapidly progressing and for the Ras network may soon approach completion. However, the result is a telephone book full of names rather than an ordnance survey type of map that in detail connects the names with pathway topologies. We also need to be able to trace the wanderers’ steps in time to understand the function of the pathways. Here again, the old stock is taking the lead in developing approaches to draw the missing lines, track the temporal relationships, and analyze the composite behavior of simple modules embedded in complex networks. The most important insights will be those that transcend the directory of names and provide an understanding of the design principles of biological systems and their evolution. This will enable us to address the next big challenge: How do biochemical signaling networks generate biological specificity? The current state of analysis of the Ras-Raf network offers a glimpse into this new world. The concepts and tools developed in this process hopefully will widen this glimpse into a window overlooking the whole Ras network.
The authors apologize to those whose works could not be cited because of space constraints.
The author(s) declared no potential conflicts of interest with respect to the authorship and/or publication of this article.
This work was supported by the Science Foundation Ireland [grant number 06/CE/B1129]; an EMBO long-term fellowship (A.Z.); a Marie Curie Fellowship [number 236758 (M.B.)]; and the European Union Seventh Framework Programme (FP7) [grant number LSHC-CT-2006-037731 (D.M.)].