As the scientific community’s understanding of protein signaling pathways grows, there has been a transition from identifying and describing individual proteins along a pathway towards dimers, nanoclusters, and complexes that allow signals to propagate through dynamic conformational changes and productive high-affinity recruitment.[1] The mitogen activated protein kinase (MAPK) pathway is one of the most studied signaling pathways due to its role in cell proliferation and tumorigenesis.[2–4] This has led to the development of small molecule inhibitors that target nearly every step in the MAPK pathway.[5, 6] Despite this progress, drug resistance arises through other, pre-existing or emerging, mutations, paradoxical activation, negative feedback, and activation of alternative pathways.[7, 8] One of the methods currently being investigated to overcome drug resistance is through the development of inhibitors that either prevent complex formation[9] or lock protein complexes in an inactive state.[10–12] Simulations of these complexes can describe how dimerization/complex formation influences protein dynamics, and promotes conformational changes. They can also provide the activation mechanism and insight for drug development.
Dimerization of Raf and the ensuing cascading phosphorylation events are key to the MAPK pathway.[13, 14] Clustering of GTP-bound Ras recruits Raf family proteins to the membrane, increasing their local concentration. The high affinity interaction of Raf with active Ras shifts the equilibrium from a populated autoinhibited state toward the active state,[15, 16] where Raf assumes a side-by-side dimerization of its kinase domain.[17] Active Raf dimer phosphorylates and activates mitogen activated protein kinase kinase (MEK) in a transient MEK/Raf/Raf/MEK quaternary complex.[18, 19] Exactly how the quaternary complex forms is still unclear. Depending on the relative populations and the affinities of the interactions, multiple routes are possible. Inactive Raf and MEK are known to form face-to-face heterodimers in the cytosol, and two of these can assemble once Raf autoinhibition is relieved.[20] Recent work has shown that MEK and Raf family proteins interact only weakly at basal levels and the epidermal growth factor (EGF) strongly promotes the Raf/MEK interaction, suggesting that Raf dimerization is favored to precede the Raf/MEK interaction.[21] Once Raf has been activated it can phosphorylate MEK,[20, 22] or promote a face-to-face MEK homodimerization[23] and autophosphorylation.[24] A MEK dimer then activates extracellular signal-regulated kinase (ERK).[24] Activated ERK is transported to the cell nucleus where it activates transcription factors, leading to cell proliferation, survival, and growth.[25] Additional proteins (including Hsp90/Cdc37 chaperone complex, 14-3-3 proteins, and kinase suppressor of Ras (KSR)) are integral to proper protein folding, complex stabilization, and allosteric activation along the MAPK pathway.[26] KSR is particularly interesting due to its ability to interact with Raf, MEK, and ERK.[27] Yet, to date its role has been a matter of contention.
The Raf family of serine/threonine kinases consists of A-Raf, B-Raf, and C-Raf (Raf-1).[28] Each consists of three conserved regions, CR1, CR2, and CR3 (Fig. S1). CR1 is further divided into a Ras binding domain (RBD) and cysteine-rich domain (CRD). CR2 is a loop that contains a phosphorylation site that can interact with 14-3-3 proteins to stabilize an inactive form of monomeric Raf. CR3 is the kinase domain.[29] In inactive Raf, CR1 and CR3 interact to form an autoinhibited state.[30, 31] Active Ras recruits Raf to the membrane through interactions with the RBD[14, 32] while the CRD binds to the membrane[33–35]. This relieves Raf autoinhibition and allows for dimerization and activation. Of the three isoforms, B-Raf has the highest activity[18, 36] and is frequently mutated in cancer.
MEK1 and MEK2 are dual-specificity threonine/tyrosine kinases and the only known activators of ERK1 and ERK2.[37] They contain one conserved region, the kinase domain (Fig. S1). N-terminal to the kinase domain is an ERK recognition sequence, a nuclear export sequence, and an autoinhibitory helix.[38] The Raf family of kinases are the best studied MEK activators; however, MEK can also be activated by other MAPK kinases.[39–43] Therefore, MEK can be considered a “gate keeper kinase,” processing signals from multiple upstream activators to control ERK activation. MEK inhibitors (MEKi) have been developed to block signaling in cancers driven by mutations in Ras and Raf. These MEKi, however, can lead to increased signaling through parallel pathways controlled by non-Raf MEK activators, including the c-Jun N-terminal kinase (JNK), p38, ERK5, and nuclear factor-κB (NF-κB) pathways.[44]
The mammalian KSR family consists of KSR1 and KSR2. KSR1 contains five conserved regions, CR1-CR5 (Fig. S1), while KSR2 lacks the CR1 domain. CR1 is a coiled coil sterile α motif (CC-SAM) domain that is involved in KSR1 membrane recruitment. It is also able to bind to an N-terminal B-Raf specific (BRS) region, a region unique to B-Raf.[45] CR2 is a proline-rich (P-rich) domain. CR3 is a CRD acting in membrane recruitment. CR4 contains an ERK binding motif. CR5 is a kinase or pseudokinase domain.[27] KSR has been considered an active kinase capable of phosphorylating MEK,[46, 47] a scaffolding protein involved bringing Raf, MEK, and ERK together, [48] and an allosteric activator of B-Raf.[45] How and when KSR interacts with the MAPK pathway is key to understanding cancer progression and acquired inhibitor resistance.
The kinase domains of B-Raf, KSR1, and MEK1 all exhibit typical kinase structures. They consist of a small N-lobe and larger C-lobe connected by a short hinge. Between these two lobes is an ATP binding pocket. The N-lobe contains five β-strands and an α-helix, called αC-helix. The C-lobe is mostly made up of α-helices (αD through αI) and contains a catalytic HRD motif and activation loop (A-loop).[49, 50] A sequence alignment of the kinase domains of B-Raf, MEK1, and KSR1 reveals that they are homologues (Fig. S2). The inactive state is characterized by an “outward” position of the αC-helix and a “collapsed” A-loop. In the active state, the αC-helix moves to an “inward” position and the A-loop is “extended.” The “collapsed” A-loop contains an N-terminal “inhibitory helix” that prevents the inward motion of the αC-helix. Activation involves phosphorylation of A-loop residues, which disrupts this “inhibitory helix” and allows the A-loop to extend.[50, 51] In B-Raf the phosphorylated residues are Thr599 and Ser602[52] and in MEK1 they are Ser218 and Ser222.[53] KSR1 is known to undergo autophosphorylation, however it is not known if any residues of the KSR1 A-loop phosphorylate.[47] In addition to these general features of protein kinases, MEK1 also contains a P-rich loop that contains several serine residues (Ser286, Ser292, and Ser298) that undergo phosphorylation and are involved in regulating MEK1 activation/deactivation.[54, 55] KSR1 has long been considered a pseudokinase due to its limited kinase activity.[56] The cause of the low kinase activity of KSR1 is often attributed to synonymous changes in key conserved residues found in typical kinases. These include Arg639 in the β3-strand instead of lysine as in B-Raf and Lys732 in the catalytic motif instead of arginine as in MEK1 and B-Raf, i.e., HKD instead of HRD.
The ability of MEK1 to form face-to-face heterodimers with both B-Raf and KSR1 is an important MAPK feature. This face-to-face recognition is centered around the C-lobe αG-helix. Mutation of a key residue in any of the three proteins abrogates their ability to form the complex (MEK1 F311S, B-Raf I666R, KSR1 W831R). B-Raf is known to phosphorylate MEK1 through a face-to-face interaction.[20, 48] The face-to-face interaction between MEK1 and KSR1 has been implicated in serving multiple roles. Evidence suggests that under certain circumstances KSR1 can directly phosphorylate MEK1, acting as a true kinase.[47] The direct phosphorylation of MEK1 by KSR1, however, appears to be a low probability event compared to MEK1 phosphorylation by B-Raf. Instead, the primary role of the KSR1/MEK1 heterodimer is to act as either a scaffold or an allosteric activator. As a scaffold, when in a KSR1/MEK1 heterodimer, KSR1 interacts with B-Raf through a side-to-side interface resulting in a B-Raf/KSR1/MEK1 ternary “scaffolding unit.” MEK1 from this unit is then translocated to an active B-Raf dimer nearby and phosphorylated.[48, 57] As an allosteric activator, MEK1 interacts with KSR1 in the cytoplasm forming a “transactivation unit”. This blocks an autoinhibited KSR1 state and enables a side-to-side heterodimer with a B-Raf monomer that has already been recruited to the membrane by Ras. The stabilized active configuration of B-Raf is able to phosphorylate a second MEK1 kinase (that is, not the MEK1 involved in the transactivation unit).[45]
The interactions between B-Raf, MEK1, and KSR1 in the assembly offer a unique system to explore how dynamic and allosteric effects impact protein complex formation. In this study, we performed molecular dynamics (MD) simulations of active B-Raf/MEK1, inactive B-Raf/MEK1, and KSR1/MEK1 heterodimers. Coupled with the available data, these simulations allow us to investigate (i) why B-Raf activation is necessary to phosphorylate MEK1, (ii) what occurs at the interface between active B-Raf and MEK1 that leads to phosphorylation, and (iii) why B-Raf is more potent at activating MEK1 than KSR1. Our results show that the P-rich loop of MEK1 moves in concert with the B-Raf A-loop which influences the flexibility of the MEK1 A-loop. The collapsed A-loop in inactive B-Raf draws MEK1 P-rich loop towards MEK1 A-loop, reducing the A-loop flexibility. When the B-Raf A-loop is extended, the MEK1 P-rich loop moves with it, repositioning it towards the bottom of the MEK1 C-lobe. Once this has occurred, the MEK1 A-loop becomes more flexible and is able to orient Ser222 towards the ATP in B-Raf. The increased flexibility in the MEK1 A-loop also allows B-Raf αG-helix residue Arg662 to move from a position within the A-loop of MEK1 to a position outside of the A-loop “inhibitory helix.” The motion of Arg662 allows the MEK1 A-loop to reorient, bring Ser218 closer to ATP. Our results also show that additional residues in KSR1 compared to B-Raf lead to steric clashes at the KSR1/MEK1 interface and result in different dynamics in the two complexes. This creates a large gap between the N-lobes of the two proteins and has implications in KSR1’s ability to function as an active kinase or scaffold.