The free energy landscape in translational science: how can somatic mutations result in constitutive oncogenic activation?

Abstract

The free energy landscape theory has transformed the field of protein folding. The significance of perceiving function in terms of conformational heterogeneity is gradually shifting the interest in the community from folding to function. From the free energy landscape standpoint the principles are unchanged: rather than considering the entire protein conformational landscape, the focus is on the ensemble around the bottom of the folding funnel. The protein can be viewed as populating one of two states: active or inactive. The basins of the two states are separated by a surmountable barrier, which allows the conformations to switch between the states. Unless the protein is a repressor, under physiological conditions it typically populates the inactive state. Ligand binding (or post-translational modification) triggers a switch to the active state. Constitutive allosteric mutations work by shifting the population from the inactive to the active state and keeping it there. This can happen by either destabilizing the inactive state, stabilizing the active state, or both. Identification of the mechanism through which they work is important since it may assist in drug discovery. Here we spotlight the usefulness of the free energy landscape in translational science, illustrating how oncogenic mutations can work in key proteins from the EGFR/Ras/Raf/Erk/Mek pathway, the main signaling pathway in cancer. Finally, we delineate the key components which are needed in order to trace the mechanism of allosteric events.

Introduction

In 1991, Frauenfelder, Sligar and Wolynes proposed the free energy landscape description for the ensemble of folded protein states.¹ An energy landscape is a mapping of all possible conformations of a molecule, or the spatial positions of interacting molecules in a system, and their corresponding energy levels. The free energy landscape’s funnel-like shape indicated that folding is driven by the hydrophobic effect;^2,3 that proteins can populate a large number of conformational substates; and that strong energetic conflicts are minimized in the most populated native states, satisfying the principle of minimal frustration.⁴ The free energy landscape concept had a vast impact on the protein folding field, since it provided a simple, quantitative description of the conformational ensemble of protein states.^5–13 In 1999, we suggested that this concept can help explain protein function,^14,15 including binding events,¹⁶ aggregation,¹⁴ catalysis,¹⁷ allostery,¹⁸ and signaling across the cell,^19,20 via ‘conformational selection and population shift’. The concept of population shift reasons that all conformational substates pre-exist, and that evolution has exploited and optimized them for function, including rare states.^21–27 Population shift is now recognized as the basis for allostery, and thus of signaling across protein interfaces,²⁸ multimolecular complexes, pathways and the entire cellular network.²⁰ It takes place within biomacromolecules and across them. It occurs in proteins, nucleic acids^29,30 and lipids,^31,32 including cholesterol³³ and phosphatidylinositol triphosphates. It further provides a mechanistic explanation for cooperativity. Collectively, the free energy landscape underscores the linkage between fundamental physicochemical principles and protein behavior under normal physiological conditions and its dysfunction in disease, and broadly, the cellular network and its regulation.

Fig. 1 illustrates the concept of population shift. In the free form of the protein (left hand-side of the figure), the distributions of the conformations among the states are governed by the equilibrium constant. The conformational selection theory posits that a ligand will preferentially bind one of these states, whose binding site is most complementary. Binding will stabilize this state, resulting in a population shift of the ensemble toward this conformation, which in its bound state presents the deepest minimum (right hand-side). Induced fit takes place to optimize the interaction.³⁴

Fig. 1.

An overview of the free energy landscape, conformational selection and population shift. The left hand-side of the figure depicts the populations of the various states in the free form of the protein. The conformation populating the middle basin is the most stable. Ligand binds the most compatible conformation which populates the left-most basin. This leads to a population shift toward this, now most stable state. Mutations similarly redistribute pre-existing conformations.⁸³

Within this broad framework, here we draw attention to the usefulness of the free energy landscape in helping to understand the mechanism of allosteric mutations, and drug resistance. In light of the increasing focus of the community on translational science, we focus on mutations that result in oncogenesis through loss of normal physiological controls of protein activation, and on those that lead to drug resistance. We illustrate that interpreting mutations in terms of the free energy landscape helps appreciate the significance of the structural underpinnings of how oncogenesis develops and persists in drug resistance. Allosteric mutations typically work by shifting the protein population from an inactive to an active state (or, vice versa for a repressor) even in the absence of a physiological triggering event. Under normal conditions, the shift is regulated by a concentration-dependent ligand binding event, or a temporal post-translational modification; in disease, a constitutive mutation can shift the landscape and maintain the protein in its active state. An allosteric binding event which drives activation can be described by a double-well, two-state model where the states are separated by a sizeable but surmountable free energy barrier.²⁹ The protein populates one of the two states, the inactive or the active, with a bi-stable switch from the first to the second. Fig. 2 illustrates such a switch in terms of the free energy landscape. The figure emphasizes two key features of the bi-stable active–inactive switch: first, the system is populated dominantly by only one of the two states; and second, the switch from one state to the other is under control only within a narrow ligand concentration window. In the example of the epidermal growth factor receptor (EGFR) in Fig. 2, the intracellular kinase domain is activated by EGF binding to the extracellular domain. The population shift in favor of an active kinase domain is complete within a narrow range of concentration increment of EGF. This is highlighted in the plot by the sigmoid transition from the inactive to the active state. In the presence of constitutive mutations, the landscape is shifted from the inactive to the active state even in the absence of EGF, which as we describe below can take place via three distinct mechanisms (Fig. 3).

Fig. 2.

A bi-stable switch from an inactive to an active state, drawn in terms of the free energy landscape. In the figure, the bi-stable switch has two key features: (1) only one of the states is populated dominantly, with the states separated by a high but surmountable free energy barrier: (2) the switch is under control only within a narrow window of ligand concentration. In the figure the system relates to EGFR. The EGF ligand binds to the extracellular domain resulting in activation of the intracellular kinase domain. The population shift toward the active kinase domain takes place within a narrow range of EGF concentration increment. This is depicted in the activation plot that illustrates a single transition from the inactive to the active state.

Fig. 3.

A schematic free energy landscape representation of aberrant gene activation by oncogenic mutations. Similar to Fig. 2, prior to activation, the population of a protein encoded by a particular gene is always favorable for the inactive conformation (left funnels in the three panels), as shown by the free energy curves in green color. Due to oncogenic mutations, the relative populations become more favorable for the active conformation (right funnels in the three panels), as depicted by the red curves. This shift can occur through the mechanism by destabilizing the inactive conformation (left panel), by stabilizing the active conformation (middle panel), or by a combination of both mechanisms (right panel). The mechanisms are highlighted by the black arrows. In all three mechanisms, the result is a reversal of the relative populations between inactive and active conformations.

To illustrate the link between the energy landscape and mutations leading to dysfunction in oncogenesis, we chose the cellular pathway of EGFR (a receptor tyrosine kinase, RTK)/Ras/Raf/Mek/Erk, a key pathway which is often deregulated in cancer. Fig. 4 provides an overview.

Fig. 4.

There are two major cellular signaling pathways downstream of the activated receptor tyrosine kinase (RTK) and G protein coupled receptor (GPCR), Ras–RAF–MEK–ERK and PI3K–AKT. The pathways transduce signals at the cell surface and lead to protein synthesis, cell proliferation, and cell survival, reflecting the hallmarks of cancer cells in cell growth and survival. One of these, Ras–RAF–MEK–ERK, is shown here.

Not all mutations can be described by the free energy landscape

Not all mutations can be described by the free energy landscape; only those working via a population shift. The major mutations at the active site of Ras, a key protein in the EGFR/Ras/Raf/Mek/Erk pathway which is involved in over 30% of human cancers provide examples of mutations which are not characterized by the free energy landscape.³⁵ Ras is a GTPase. GDP-bound Ras is inactive; GTP-loaded Ras is active. GEF (GDP → GTP exchange factor) exchanges GDP for GTP, and thereby activates Ras. The active GTP-bound Ras binds and activates Raf, which is discussed below. GTP → GDP hydrolysis in the presence of GAP (GTPase activating protein) switches Ras back to its inactive state. Mutations, such as those involving Gly12, Gly13 and Gln61, abolish GAP’s binding and catalysis, thus keeping Ras in its GTP-bound active state.³⁶ GTP → GDP hydrolysis takes place via transfer of a proton from the attacking water molecule to a second water molecule; a different proton is then transferred from this second water molecule to the GTP. Gln61 stabilizes the transient OH⁻ and H₃O⁺ molecules.³⁷ The Gln61 → Leu mutant cannot activate the nucleophilic water. On the other hand, the larger side chains in substitutions such as Gly12 → Arg, Gly12 → Val, interfere with GAP binding and/or hydrolysis.³⁸ GAP’s arginine which is inserted into the catalytic site, together with the rearrangement of Gln61, lower the barrier of the transition state. Mutations of Gly12 and 13 to larger residues prevent the GAP arginine from inserting into the active site, hindering hydrolysis, keeping Ras in an activated GTP-bound state. While it is clear that all mutations abolish the GTP → GDP reaction, the mechanism is not always unclear. Preeminent among these still enigmatic mutations is the chief oncogenic mutation of Gly12 → Asp, and Gly13 → Asp, although it can be expected that charge plays a role. Nonetheless, even if the mutations are in the active site, they may have auxiliary significant conformational consequences as well, assisting in retaining the active (GTP-bound) signaling state. Currently, this appears to be the case at least for some of the mutations above.

Below, we first illustrate how somatic mutations in EGFR result in kinase constitutive activation. Consistent with the requirement of a robust bi-stable switch from the inactive to the active state upon EGF binding (Fig. 2), the EGFR structural dimer model delineates the decisive role of the conformational change at the extracellular domain in stabilizing the asymmetric active intracellular kinase dimer. The details of these events underscore the vital importance of structural guidelines in studies of oncogenesis.

A case study: epidermal growth factor receptor (EGFR)

Members of the epidermal growth factor receptor family (EGFR/ErbB1/HER1, ErbB2/HER2, ErbB3/HER3, and ErbB4/HER4) constitute one of the 20 subfamilies with a total of 58 currently known human receptor tyrosine kinases (RTKs).³⁹ These cell-surface receptors play important roles in cell growth, differentiation, and migration⁴⁰ and are often misregulated in a variety of human cancers.⁴¹ In the normal cell, the binding of epidermal growth factor (EGF) to these receptors activates their cytoplasmic kinase domains, resulting in phosphorylation of tyrosine residues in their C-terminal tails and subsequent recruitment of downstream effectors.^42,43 There are two major transduced signaling pathways, PI3K/Akt and Erk1/2 MAPK which correspondingly are primarily responsible for cell proliferation and survival. The second is illustrated in Fig. 4. The bi-stable switch from an inactive to an active state via EGF binding at the extracellular domain can be formulated as a population shift in favor of an active kinase domain. Fig. 2 depicts the population shift to an active conformation in terms of the free energy landscape, where the shift is elicited by an activation event. Fig. 5A illustrates the domain organization of EGFR and the structural dimer model⁴⁴ is depicted in Fig. 5B. The dimer model clarifies the robust EGFR bi-stable activation switch. Upon EGF binding, the extracellular domain undergoes a conformational change which propagates, resulting in the cooperative binding of the transmembrane helix and the juxtamembrane segment, stabilizing the activated asymmetric kinase dimer.⁴⁵

Fig. 5.

The domain organization and structural dimer model of EGFR. EGFR, with a total of 1210 residues (in the coding region), is composed of a ligand-binding extracellular region, a membrane-spanning region, a juxtamembrane region, a kinase domain, and a C-tail that can be auto-phosphorylated. In Fig. 5A, the phosphorylation sites are indicated above the EGFR C-terminal tail domain. Also listed under the domain organization are the more frequent somatic cancer mutations labeled with their count in cancer samples, and the genetic alteration regions of insertion and deletion from COSMIC dataset. The regions involving asymmetric kinase dimer interface are also indicated. Fig. 5B depicts the EGFR dimer model proposed by Jura, et al.⁴⁷ Excluding the C-terminal tail, the structural dimer construction is based on three crystal structures. The orientation of domain IV when mediating between the EGF-bound extracellular domain (PDB 1ivo) and the transmembrane segment is based on PDB structure 1nql. The construction of intracellular asymmetric kinase dimer including the juxtamembrane segment is based on PDB 3qop.

Structural analyses of intracellular regions revealed that the kinase catalytic domain is allosterically activated by formation of an asymmetric dimer in which the C-lobe of the ‘activator’ kinase domain interacts with the N-lobe of a second ‘receiver’ kinase domain.⁴⁶ In contrast, in symmetric dimers, which are also present in crystal structures, both kinases are in inactive conformations. This regulatory mechanism via an allosteric switch from an inactive αC-helix-out state to an active αC-helix-in conformation (through a rotation and shift movement of the αC-helix accompanied by the formation of a salt-bridge between the β3-lysine and the αC-glutamate), has been observed in many protein kinase families with diversified allosteric activators.⁴⁷ Recent optical tracking studies of single, full-length EGFRs in living cells⁴⁸ indicated that unliganded EGFR fluctuates continuously between the monomer and dimer states. The negative-stain EM data of liganded EGFR lacking the C-terminal tail⁴⁹ also revealed that the equilibrium distribution between the active asymmetric dimer and the inactive symmetric dimer depends on several factors. These data led to the conclusions that (1) an inhibitor that stabilizes an active kinase conformation will favor the asymmetric dimer, (2) mutations destabilizing the asymmetric interface will favor the inactive symmetric dimer, and (3) those mutations in the juxtamembrane region that stabilize the dimer will favor the asymmetric dimer. With the fact that two separate oncogenic mutants (either Δ746–750 or L858R) in a near full-length purified EGFR (lacking part of the C-terminal tail) are fully active independently of EGF binding, the authors suggested that kinase activation in EGFR strongly depends on the formation of asymmetric kinase dimer.⁵⁰

In a cellular circuit, many nodes are required to operate as bi-stable switches in response to specific cellular functions. To be bi-stable, a folded protein is optimized by evolution to have a double-well conformational landscape at the bottom of its folding funnel; a double-well landscape facilitates a transition from one population to the other. As depicted in Fig. 2, the two local free energy minima of the EGFR kinase domain, one corresponding to the active conformation and the other to the inactive conformation, are separated by a surmountable free energy barrier; a small increment in the EGF concentration shifts the kinase domain population from a dominant inactive to a dominant active state. Given the available EGFR dimer model (Fig. 5B) and the recent cell imaging data discussed above, we may safely conclude that kinase activation is through an allosteric mechanism, via the formation of an asymmetric kinase dimer, with or without extracellular EGF ligand binding. As implied by a bi-stable switch, in normal cells the kinase domain of EGFR is mostly populated either in a stabilized autoinhibited monomer or in an inactive symmetric dimer, making the formation of the asymmetric dimer feasible, but not favorable. Upon the EGF ligand binding, the induced conformational change in the extracellular domain facilitates the kinase asymmetric dimer association through specific interactions both in the transmembrane domain dimer and in the juxtamembrane segment, with itself and with the kinase domain.^44,51 The EGFR dimer model shown in Fig. 5B not only provides the structural mechanism explaining how the shifting of the kinase population from an autoinhibited state to an active form is achieved through EGF binding; it also provides the foundation for accounting how individual genetic alterations frequently observed in somatic cancer mutations cause aberrant constitutively, ligand-independent, active kinase. Frequent cancer mutations in EGFR from the COSMIC database sample counts are summarized in Fig. 5A.

In principle, the breakdown of a bi-stable switch in tumor cells through genetic alterations can be attributed to oncogenic mutations which either destabilize the inactive conformation^46,52,53 and/or stabilize the active conformation.⁵⁴ The change in the relative populations of the inactive versus active conformations, due to oncogenic mutations in the absence of ligand binding, can be described by the free energy landscape as illustrated in Fig. 3. As we indicated above, in the case of the oncogenic EGFR, the kinase conformation is aberrantly activated through an allosteric mechanism resulting in formation of the asymmetric dimer; not simply via monomeric kinase activation. Therefore, to address the somatic mutation data summarized in Fig. 5A, we should look not only at the fact that mutations stabilize the active conformation and/or destabilize the inactive conformation of the monomer, which in turn will favor the formation of the asymmetric dimer; but also at the fact that mutations will enhance the asymmetric dimerization.

Approaches such as the structural energetic analysis⁵² and molecular simulation^54,55 can provide further, more detailed insight into the molecular mechanism of kinase activation in the presence of cancer mutations. Here, to illustrate how an individual mutation causes kinase constitutive activation based on the destabilizing–stabilizing concept described above (Fig. 3), we simply utilize the static structural details of the kinase domain of the wild type and mutant EGFRs captured in inactive (monomer and symmetric dimer) and active (asymmetric dimer) states. First, a structural model of the asymmetric kinase dimer was built in Fig. 6A with the active receiver kinase in pink ribbon and the inactive activator kinase in light green. In the model, the spatial locations of somatic cancer mutations and indels (listed in Fig. 5A) are highlighted, with the sidechains shown in space-fill model and marked by triangles. Zooming into the inactive structure, the hydrophobic environments surrounding both the L858 (Fig. 6B) and L861 (Fig. 6C) residues are destabilized by the substitution of the hydrophobic leucine by the hydrophilic residues, arginine and glutamine (L858R and L861Q), which are the most frequent cancer mutations. On the other hand, structural analysis of the wild type L858 in an active conformation⁴⁶ and the mutant 858R in an inactive monomer⁵⁶ reveals similar local environments surrounding in L858 and 858R, strongly suggesting that the oncogenic L858R mutation depends mainly on a mechanism of destabilizing the inactive kinase conformation. However, the fact that the intrinsically disordered N-lobe in wild type EGFR is suppressed by the L858R mutation to facilitate the asymmetric dimerization, as observed in a long-timescale molecular dynamics simulation,⁵⁴ indicates that the mechanism of stabilizing the active conformation may also play a role in the L858R mutation.

Fig. 6.

Structural representations of somatic cancer-associated mutations in EGFR. In Fig. 6A, a structural model of the asymmetric kinase dimer was generated with the crystal structures of the active asymmetric dimer (PDB 2gs6) and the inactive symmetric dimer (PDB 2gs7). The association of the active receiver kinase in pink ribbon and the inactive activator kinase in light green ribbon is based on the asymmetric dimer of two active kinases (PDB 2gs6). The most prominent somatic mutations and indels reported in the COSMIC database (Fig. 5A) are spatially indicated by highlighted sidechains in space-fill models and triangle marks. The zoom-in local environments of the L858 and L861 residues in an inactive conformation are given in Fig. 6B and C respectively with the side chains which are in close contact shown by space-fill model.

The T790M mutation has been shown to activate the wild type EGFR and retain low-nanomolar affinity for the gefitinib, an ATP-competitive small molecule tyrosine kinase inhibitor.⁵⁷ Unlike the L858R mutation which destabilizes the inactive kinase, the gatekeeper residue mutation stabilizes the active kinase conformation by enhancing the stability of the hydrophobic R-spine through the methionine residue. It is the same stabilizing factor that establishes a stable active kinase conformation in L858R/T790M double mutations, which confers drug resistance to gefitinib and erlotinib by increasing the ATP affinity by more than an order of magnitude. The glycine located at the P-loop of the kinase stabilizes the inactive conformation. Mutation at the 719 position (G719S) destabilizes the P-loop conformation which in turn destabilizes the inactive conformation.⁵³ Deletions and insertions which occur in the region around the αC-helix correlate very well with the motions involved in the shift of the inactive αC-helix-out to active αC-helix-in conformation by pulling (deletions) at the N-terminal of αC-helix and pushing (insertions) at the C-terminal of αC-helix.

Oncogenic mutations in Raf

Dysregulation of the Ras–Raf–Mek–Erk pathway can lead to cancer with gain-of-function mutations in B-RAF (a Raf paralog) among the leading causes.⁵⁸ Raf activation takes place at the plasma membrane by binding to GTP-loaded Ras. This triggers phosphorylation and activation of Mek and Erk which then phosphorylates its many substrates, leading to cell proliferation and survival. Ras-mediated Raf activation involves homo- or heterodimerization of the kinase domain of Raf through a side-to-side interface.^59–62 Dimerization induces catalytic activity probably via allosteric switching of the Raf protomers.⁶⁰ In the inactive state, Raf monomers exist in the cytosol in a closed conformation with the regulatory region bound to the kinase domain, with this conformation stabilized by binding of the 14–3–3 scaffolding proteins.⁶³ The first step in the Raf activation process is plasma membrane recruitment by Ras-GTP. Ras-GTP shifts Raf’s population from a closed to an open state by weakening the interaction between the kinase domain and the regulatory region,^64,65 thus releasing its autoinhibition, and allowing dimer formation. Under normal physiological conditions, Raf functions only when in a specific asymmetric dimer configuration. Membrane recruitment results in dephosphorylation of Ser259, dissociation of 14–3–3, and stabilization of the open conformation.^58,66

Gain-of-function mutations in the B-RAF gene are found in about 50% of human melanomas; the most prevalent of these is oncogenic variant B-RAF^V600E. B-RAF^V600E induces ERK activity independently of the normal EGFR-Ras signals. B-RAF^V600E essentially works as a monomer, with the allosteric mutation working by shifting the population similar to the activation taking place via dimer formation.⁶⁷ The vemurafenib inhibitor targets the Raf mutant, and in so doing disrupts ERK activity.⁶⁸ Acquired resistance to vemurafenib can take place through several mechanisms: in three of these, ERK signaling is activated through alternative pathways; in the fourth through an ERK-independent pathway. Recently a fifth mechanism was discovered, involving an abnormal messenger RNA transcript lacking specific exons, p61B-RAFV600E.⁶⁹ Drug resistant p61B-RAF^V600E – the truncated variant of B-RAF^V600E – has an increased propensity to form dimers. The deletion in this mutant allosterically shifts the kinase domain toward a dimer-favored state, a shift which normally takes place through Raf’s association with GTP-bound Ras via Raf’s Ras binding domain (RBD).

Of particular interest, upon increased Ras activity, Raf inhibitors, such as vemurafenib, promote Raf dimerization and activation also of the drug-free Raf monomer through its drugbound partner.^67,69 Because the deletion in p61B-RAFV600E similarly increased Ras signaling, an analogous mechanism could be involved here, with vemurafenib treatment enhancing p61B-RAF^V600E dimerization and activity. It is also possible that binding of the drug to one of the monomers in the p61B-RAF^V600E inhibits drug binding to the other monomer in the dimer, resulting in one active monomer per dimer and in this way abolishing drug resistance.

Although the structural details are yet to be worked out, current observations already point to the importance of the conformational behavior of Raf, modulated through its binding to Ras, dimeric associations, oncogenic mutations, conformational consequences of drug binding and drug resistant mutations. Together, these weave a fascinating and challenging account which can be reasoned through the free energy landscape. Going back to Fig. 3 and interpreting the mutational effects in this light, native Raf populates the inactive state. The RAF^V600E mutation destabilizes the inactive conformation, shifting the landscape toward the active state. The absence of a coding region corresponding to the regulatory domain in the p61B-RAF^V600E deletion mutation relieves Raf’s autoinhibition, shifting Raf’s conformation to a Ras bound-like state. Thus, by abolishing the autoinhibition, the p61B-RAF^V600E mutation both destabilizes the inactive conformation, and stabilizes the active conformation, presenting a combination of both mechanisms as can be seen in the Fig. 3 diagrams. Dimerization similarly shifts the equilibrium to the active state by stabilizing the active conformation. In both EGFR and Raf, one monomer activates the other, with the active state involving an asymmetric dimer complex.

Distinguishing between allosteric versus orthosteric mutations

How to distinguish between mutations that manifest their influence through population shift changes and those that do not? Data corresponding to cancer pathway activation via population shift relate to mutations away from the functional site; mutations that act without changing the free energy landscape are located at the functional site. Following the drug and post-translational modification nomenclature,⁷⁰ we call these orthosteric. Orthosteric mutations typically directly affect the stability of the binding. In the case of Ras, orthosteric mutations act by preventing the stabilization of the transition state in GTP hydrolysis. Orthosteric mutations can be predicted from sequence conservation at known functional sites across members of a protein family and SNPs (single nucleotide polymorphism); allosteric mutations are more challenging to predict since their locations are often known. Further, residues away from functional sites are less likely to be conserved. Molecular dynamics simulations and cross-correlation analyses, or construction of residue networks are among the more popular strategies for identification of allosteric mutations. The types of mutations and specific residue substitutions differ across and within proteins even for the same positions and may have different consequences. In the Ras example, mutation of Gly12 to Asp has worse prognosis than to Cys or Val. This is also the case in allosteric mutations; however the reasons are different. In the case of allosteric mutations, these may elicit different extents of energetic conflicts. Finally, mutations can lead to constitutive activation (or, in repressors, deactivation) in all proteins, acting by mimicking a binding event or a post-translational modification.⁷¹ Above, in addition to Ras, we have provided examples for two kinases, EGFR and Raf. Further discussion relating to kinases has been given earlier.^19,72 While here we chose to illustrate it for a particularly important example relating to the most common pathway in cancer, this mutational behavior is general and applies to all protein classes.

Finally, can a mutation also act through allosteric pathways without causing a disease? Because the number of mutations across the population is very large, and most of the time no disease is apparent, this is likely to be the case. While any mutation causes some perturbation, and thus some perturbation and population shift across the pathway, the question is its extent and the pathway through which it propagates.

Conclusions: what is needed to usefully employ the free energy landscape?

The free energy landscape provides an effective description of protein and RNA folding;^73–75 it also successfully portrays the shift in the equilibrium between the inactive and active protein states around the bottom of the folding funnel following some allosteric event. Here we chronicle its usefulness in accounting for the mechanism of constitutive activation by somatic and drug resistance mutations. This leads us to suggest that three elements are needed to productively exploit the free energy landscape for mutations switching the protein from an inactive to an active state: first, the active conformation which populates one of the local free energy basins at the bottom of the folding funnel; second, the interacting residues that form an allosteric communication pathway between the active and allosteric sites; and third, data as to whether the mutation works by destabilizing the inactive state, stabilizing the active state or both. The first emphasizes that all conformations (basins in the landscape) pre-exist, and the constitutive mutation does not forge a new conformational state; it only shifts the population among existing states. The structure of the active state is essential in order to understand how the mutation acts to trigger the shift. The second maintains that the relative populations of the active versus inactive conformations is predetermined and linked to critical residues in the communication pathway. Mutations on these pathways will affect the allosteric propagation and thus population of the active versus the inactive species. The constitutive mutations in the EGFR and Raf examples fall into this category. The third provides insight into the conformational changes that explain the mutational consequences. These are useful in drug discovery. The examples provided above are selected from a critical signaling pathway, EGFR/Ras/Raf/Mek/Erk (Fig. 4).

Molecules exist in conformational ensembles, and their distributions are affected by the environment.^{15,17,34,76,77} Evolution has biased the conformational spread to favor a preorganization where the functional state is a minor conformational species. The switch takes place only upon certain physiological activation events; EGF binding to EGFR or Raf binding to Ras in our examples above. Constitutive mutations violate the evolutionary engineered temporal triggers, and lead to a long-lived binding favored (or for repressors, disfavored) states. A conformation designed by evolution to predominate only via transient binding or post-translational events overtakes its controls, acquiring a permanent status of complementarity. Population shift resulting from a mutational event follows the principle of minimizing strong energetic conflicts and relieves frustration.

Modulation of protein conformations is a universal regulatory mechanism to control protein function. Conformational changes in response to signaling cues can control patterns of interaction and cellular localization. The challenge is to resolve the relationship between specific mechanisms for switching between distinct protein conformational states and regulation of activation. Going back to our example, it is a challenging aim is to figure out how dimer formation activates Raf; further linking the unregulated conformational change to the physicochemical basis of the conformational behavior of single molecules, and ultimately relating it to cellular function compounds the challenge. Here, we emphasize that the free energy landscape can be useful also in translational science to understand constitutive and drug resistant mutations in cancer, and argue that beyond its theoretical value, it is a practical concept in a key applied area. While we focused on an oncogenic pathway, because it relates to the fundamental behavior of proteins it similarly concerns other proteins and diseases. One such example is amyloidogenesis in neurodegenerative diseases, such as Alzheimer. Amyloids present a heterogeneous population.⁷⁸ Mutations shift the landscape to more aggregation-prone conformations,^79–82 resulting in stabilized toxic oligomers which insert into the bilayer, resulting in calcium-permeating toxic channels.

Biography

Chung-Jung Tsai

Chung-Jung Tsai is in the Cancer and Inflammation Program, Frederick National Laboratory for Cancer Research, National Cancer Institute, Frederick, MD. Dr Tsai received his PhD in chemistry from the University of Pittsburgh in 1992, followed by a brief period as a Postdoctoral Fellow there with Professor K. D. Jordan. Since then he has been at National Cancer Institute. Broadly, Dr Tsai’s research interests focus primarily on the study of protein folding, binding, function and dysfunction. He has published over 80 papers in the fields of computational chemistry, physics, and biology.

Ruth Nussinov

Ruth Nussinov received her PhD in 1977, from the Biochemistry Department at Rutgers University and did post-doctoral work in the Structural Chemistry Department of the Weizmann Institute. Subsequently she was at the Chemistry Department at Berkeley, the Biochemistry Department at Harvard, and the NIH. In 1984 she joined the Department of Human Genetics, at the Medical School at Tel Aviv University. In 1985, she accepted a concurrent position at the National Cancer Institute, where she is a Senior Principal Scientist and Principle Investigator heading the Computational Structural Biology Group at the NCI. Her National Cancer Institute website gives further details. http://ccr.cancer.gov/Staff/Staff.asp?profileid=6892