Classification of KRAS activating mutations and the implications for therapeutic intervention

Abstract

The family of RAS proto-oncogenes, discovered just over 40 years ago, were among the first cancer initiating genes to be discovered. Of the three RAS family members, KRAS is the most frequently mutated in human cancers. Despite intensive biological and biochemical study of RAS proteins over the past four decades, we are only now starting to devise therapeutic strategies to target their oncogenic properties. Here, we highlight the distinct biochemical properties of common and rare KRAS alleles, enabling their classification into functional subtypes. We also discuss the implications of this functional classification for potential therapeutic avenues targeting mutant subtypes.

KRAS is a proto-oncogenic GTPase

The first RAS oncogenes were isolated from retroviruses that induced spontaneous soft-tissue tumors in mice (1,2). Initially, viral Kras and Hras (v-Kras and v-Hras) proteins were described as GTP-dependent kinases based on their apparent autophosphorylation activity on Thr59 (3–5). On the contrary, later work demonstrated that mammalian RAS proteins (KRAS, HRAS, and NRAS) do not have threonine at position 59 and do not undergo autophosphorylation. A second difference between viral and cellular RAS proteins is a Gly12 mutation that inhibits the GTPase activity of the viral proteins. Together, these studies demonstrated that the mammalian RAS proteins normally function as GTP hydrolases and that alteration of this biochemical activity is central to their oncogenic potential (6–14).

As in the case of other proteins in the RAS superfamily, the GTP-hydrolase activity of KRAS is what allows it to function as a binary switch downstream of cell surface receptors during signal transduction (15) (Figure 1A). At the simplest level, the ability of KRAS to transmit a signal is a function of its nucleotide binding state (Figure 1A,B). In the absence of mitogenic signals, KRAS proteins are primarily bound to GDP and in an inactive conformation. This inactive state is maintained through an intrinsic GTP hydrolysis activity and by interactions with GTPase Activating Proteins (GAPs), which accelerate the conversion GTP to GDP. When cell surface receptors become activated by mitogenic signals, they recruit Guanine Nucleotide Exchange Factors (GEFs), which bind to inactive KRAS, catalyze the expulsion of GDP from the active site, and enable the passive loading of GTP. Binding of GTP to KRAS allows the active site to transition from an open to a closed conformation, and this closed conformation promotes subsequent interactions between KRAS and various effector proteins (16,17) (Figure 1C). While numerous RAS effector pathways have been reported, it is important to note that not all RAS-effector interactions are supported by the same degree of evidence (18). For instance, biochemical evidence for the KRAS-SIN1 interaction is strong, but its role in AKT regulation is unclear (19,20) (Figure 1C, red arrow).

Figure 1.

The RAS GTPase cycle and its spectrum of effector interactions. A, Simplified schematic of the canonical RAS GTPase cycle. B, RAS proteins transition from their INACTIVE and GDP bound state (grey and yellow surface; PDB code 6MBT) to the GTP bound state through the activity of GEFs, such as SOS (brown surface; PDB code 1NVX), which in turn can be allosterically regulated by a second molecule of RAS (pink surface) (101,102). In the GTP bound state, RAS proteins are in a dynamic equilibrium between OPEN (PDB code 4EFL) and closed ACTIVE (PDB code 6XI7) conformations (103,104). Inactivation of RAS proteins is performed by GAPs, such as NF1 (brown surface, PDB code 6V65) and p120GAP (brown surface, 1WQ1), which accelerate GTP hydrolysis and return RAS to the GDP bound state (31,105). C, In the ACTIVE and GTP bound state, RAS proteins are capable of binding to a number of different effector proteins to influence signal transduction (cyan surfaces). These interactions include PI3K (PDB code 1HE8), SIN1 of mTORC2 (PDB code 7LC1), RAF kinases (PDB code 6XI7), PLCε (PDB code 2C5L), RALGDS (PDB code 1LFD), and RASSF5 (PDB code 3DDC) (19,104,106–109). Note that the dynamic active site of RAS (yellow surface) acts as a binding surface for both regulators and effectors. The affinities and selectivities of the different RAS proteins for any given effector has not been systematically analyzed.

Mutational activation of KRAS

KRAS is among the most commonly mutated oncogenes in cancer, with a frequency greater than 20% in adenocarcinomas of the pancreas (PDAC), lung (LUAD), and small (SBAD) and large bowels (COAD), as well as multiple myeloma (MM) (Figure 2A). While several different mutant alleles of KRAS have been identified, their observed frequencies vary according to the cancer tissue-of-origin (Figure 2B) (21). In PDAC, virtually all mutations occur at Gly12. In contrast, mutant alleles are more heterogenous in LUAD, and even more so in COAD and MM, with many mutations affecting residues such as in Gly13, Ala59, Gln61, Lys117, and Ala146 (Figure 2B). An even broader range of mutations is observed in a group of developmental syndromes known as RASopathies. Our current understanding of the impact of different amino acid substitutions on the biochemical activity of KRAS is the result of extensive structure-function studies performed over the past four decades (22–24), while biological validation of functional differences has only occurred over the last five years (21,25–28). Similar to the recent classification of oncogenic BRAF mutants into different classes (29), our current knowledge of the core biochemical properties of KRAS mutants allows for their division into several classes, based on how they promote activation and how they influence interaction with different effectors (Figure 2C–D).

Figure 2.

Classification of KRAS mutations. A, KRAS is frequently mutated in common cancers. The graph shows cancers with at least 200 new cases in 2021 (American Cancer Society Facts and Figures) and with >5% KRAS mutation frequency. *Frequency and codon mutations are from a separate published dataset (21). This dataset contains only alleles Gly12, Gly13, Gln61, Lys117, and Ala146. B, Distribution of mutations at different KRAS codons in pancreatic cancer (n=3,816), non-small cell lung cancer (n=5,041), colorectal cancer (n=5,247), and multiple myeloma (n=1,262). ‘Other’ represents structural changes, such as truncations and insertions. Data were collected from AACR Project GENIE. C, Structure of KRAS showing the location of the top five sites of mutation. Residues lining the active site are colored yellow, while the rest of the protein is shown as a grey surface. The color coding of each mutation site is determined by their similarities in biochemical effect. D, Distribution of mutant alleles from panel B, re-classified based on their similarity of known or predicted biochemical function. Class 1 mutations are composed of Gly12 alleles. Class 2 mutations are composed of Gly13, Lys117, and Ala146 alleles and alleles predicted to be similar (mutations at codons Ala18, Leu19, Thr20, Gln22, Leu23, Phe28, Asp57, Asn116, Asp119, and Lys147). Class 3 mutations are composed of Ala59 and Gln61 alleles and alleles predicted to be similar (mutations at codons Gly60 and Ala66). Class 4 mutations are mutations on the surface or in the core of KRAS that are predicted to impact protein conformation. Details for each class are expanded upon in the text and in Figure 3.

Gly12 mutations are the most common in human cancers and it has been shown that an amino acid change of Gly12 to any other residue strongly impacts the ability of GAPs to accelerate GTP hydrolysis (Class 1, Figure 2C,D) (30,31), although there are exceptions in which noncanonical GAP activity on KRAS has been demonstrated (32). Many Gly12 mutants of KRAS also inhibit intrinsic hydrolysis (30). Considering the frequency of KRAS mutations affecting the GTPase function in different cancers, inhibition of GTP hydrolysis appears to be the most efficient mechanism of oncogenic activation, presumably by increasing the steady-state levels of RAS-GTP and, as a result, effector binding and activation. Thus, resurrection of intrinsic GTP hydrolysis using GTP-analogues designed to allow hydrolysis by Gly12 mutants – in an effort to reduce the levels of functionally activated protein – was one early avenue for targeting oncogenic KRAS (33,34). Nevertheless, GTP analogues sensitive to mutant-induced hydrolysis never gained traction as a therapeutic strategy, primarily due to the pM affinity of RAS for GTP, as well as its high cellular concentration (35). Alternatively, it has been proposed that the design of small molecules that bind outside the active site of RAS could allosterically promote or rescue intrinsic GTP hydrolysis in deficient mutants (36). These efforts are hindered by our understanding of the GTP hydrolysis mechanism, however, which remains incomplete and controversial (37–39).

An alternative way that oncogenic mutations increase the steady state levels of active KRAS is through enhanced nucleotide exchange, the primary feature that defines Class 2 (Figure 2C,D). It is possible to activate KRAS through this mechanism because nucleotide release is an active process, while nucleotide loading is a passive process. Since the cellular concentration of GTP is typically higher than GDP, increasing the rate of active nucleotide release increases the likelihood that a given molecule of KRAS will passively bind GTP. Enhancement of nucleotide exchange is most prominent with mutations at Gly13, Lys117, and Ala146 (Figure 2C) (25,26,40). The influence of mutations at these sites is consistent with enhanced nucleotide exchange (i.e. hyperexchange), as they interfere with active site closure while reducing affinity for the nucleotide itself (25,26,41,42). Furthermore, these rapid exchange mutants can synergize with GEF function to produce extraordinarily high levels of exchange (25,41,43). Nevertheless, hyperexchange mutations may enhance nucleotide exchange to a point where the activity of GEFs is expendable, rendering cancer cells resistant to GEF inhibition (25). Thus, a more mechanistic understanding of intrinsic nucleotide exchange is needed, both for WT and oncogenic mutants of this class.

Several of the Class 2 hyperexchange mutants retain intrinsic and GAP-accelerated hydrolysis activity (25,42). Interestingly, in the case of G13D, GAP-sensitivity appears to be specific for NF1 (44). The biochemical properties of hyperexchange variants are consistent with the location of their mutations, which are removed from the immediate vicinity of the γ-phosphate of GTP (Figure 2C). Unlike mutants that inhibit hydrolysis, hyperexchange alleles have received far less attention as drug targets due to their comparably lower frequencies (Figure 2B). This class may be bigger than previously acknowledged, however, as mutations of KRAS that are not present in the active site can unexpectedly promote KRAS activation by means of enhanced nucleotide exchange (45–47). Thus, when grouped together based on their biochemical activities, hyperexchange (Class 2) alleles represent a sizable fraction of mutant alleles found in cancers (Figure 2D).

While Gly12, Gly13, Lys117, and Ala146 mutants fit into the two classes of oncogenic mutants described above, not all KRAS mutants neatly fit this classification, instead exhibiting a hybrid phenotype affecting both parts of the GTPase cycle. For instance, mutations of Ala59 and Gln61 are located in switch II of the active site. Mutations at these residues inhibit nucleotide hydrolysis and enhance exchange due their proximity to the γ-phosphate of GTP and their ability to disrupt closure of the active site (Class 3, Figure 2C,D). Class 3 mutations can also unexpectedly interfere with normal effector and regulator interactions, such as in the case of Q61K (30,48,49). Q61K inhibits both GAP and intrinsic GTP hydrolysis and modestly increases nucleotide exchange. It also inhibits the binding of SOS to KRAS, preventing synergistic increases in nucleotide exchange characteristic of Class 2 mutations. Finally, Q61K enhances the affinity of KRAS for RAF kinases and, at the same time, reduces interaction with PI3K. Together, these biochemical changes alter cell signaling in a way that is specific to the Q61K mutant and, accordingly, cancer cells expressing KRAS^Q61K are highly dependent on the MAPK signaling pathway (48). Interestingly, the Q61L mutant retains sensitivity to GEF function, demonstrating that distinct mutations at the same residue can influence the activity of the mutant protein in unique ways (49).

A final class of amino acid substitutions (Class 4, Figure 2D) involve those that do not directly interact with the guanine nucleotide, but instead are located on the surface of the protein or are distal from the active site (Figure 3A). The biochemical properties of KRAS proteins with these substitutions are not well characterized (or to be determined, TBD), but there is potential for such changes to influence a wide range of KRAS functions, including affinity for effector and regulator proteins, rates of nucleotide exchange and the ability of KRAS to attain the catalytically competent conformation. Without a doubt, the effect of these substitutions, if any, is subtle and therefore they are rarely found in cancer (Figure 3B). Instead, these mutations are found far more often in germline variants of KRAS, in particular non-pathogenic variants (45,50,51). If/how these variants influence the activation of KRAS will certainly lead to a more mechanistic understanding of KRAS function.

Figure 3.

Breakdown of KRAS mutational classes. A, Location of mutated codons in each class in reference to the sequence and functional domains of KRAS. Each dash refers to an amino acid, from 1 to 189. B, Each row describes general characteristics of each class and their contribution to somatic mutations in cancer (n = 18,399), germ line mutations in RASopathies (n=80), and single nucleotide polymorphisms in the human genome (SNPs, n=37). Data were collected from AACR Project GENIE, NSEuroNet, and gNOMAD databases (51,72,73). Arrows represent the direction and magnitude of change in the different biochemical properties of KRAS for each mutational class. VAR indicates that the effects of different mutations is variable. ? indicates that the biochemical properties of most Class 4 substitutions are not well characterized. TBD refers to the biochemical activities of Class 4 mutants that are To Be Determined.

Therapeutic targets downstream from KRAS

In the absence of direct KRAS inhibitors, efforts to combat oncogenic KRAS have focused on downstream pathways. Early identification of potential RAS effectors relied on the use of yeast two-hybrid screens and in vitro interaction assays, which identified RAF and RALGDS as direct RAS interactors (52). Later experiments identified proteins in cell extracts that bound to RAS in a GTP-dependent manner, such as p110α, that catalytic subunit of PI3K. The putative effectors identified through these techniques were further validated through extensive structural, biochemical, and genetic experiments, including the identification of point mutants unable to bind to different effectors (53,54) Amongst the plethora of RAS effector pathways that have since been identified, MAPK and PI3K stand out as particularly critical for RAS function and have been the focus of multiple efforts to develop therapeutics for cancers expressing mutant KRAS.

The activation of any given effector pathway in a KRAS mutant cell depends upon (i) the amount of KRAS-GTP in the cell, (ii) the affinity of the mutant KRAS to the effector, and (iii) features that control flux through the pathway, such as expression of the pathway components and the activities of positive and negative feedback loops. At the level of direct effector binding by KRAS, both oncogenic somatic mutations and inherited developmental mutations (discussed below) have been shown to alter effector binding (25,26,47,55), sometimes with differences between alleles of the same codon (e.g. G12V, G12D, G12C) (30). By extension, each KRAS mutant has a unique capacity to alter downstream signaling. For example, G13D is a weaker oncogenic allele than G12D and induces an attenuated hyperproliferative phenotype in the mouse colonic epithelium (26). Although its hyperexchange feature allows G13D to increase the steady state levels of KRAS bound to GTP, the mutation limits KRAS oncogenicity by reducing the affinity for RAF kinases (26). Similarly, G12R has been shown to have a dramatically reduced ability to initiate tumors in the mouse pancreas relative to G12D because of an inability to drive macropinocytosis via binding and activation of p110α (27,28).

Efforts to target downstream pathways of KRAS, primarily through specific MAPK and PI3K pathway inhibitors, have largely been clinically unsuccessful due to resistance to monotherapies and extreme toxicity of combination therapies. For instance, clinically approved inhibitors of MEK (cobimetinib and trametinib) have shown poor efficacy in KRAS mutant cancers in part due to their concomitant inhibition of ERK-mediated feedback inhibition (56). In the presence of mutant KRAS, lack of ERK-mediated feedback on RAF paradoxically promotes hyperactivation of MEK, resulting in breakthrough of MEK inhibition and reactivation of ERK. Therapeutic strategies that account for the unique properties of individual KRAS mutations are likely to be more successful. For example, because G12R mutations are defective in PI3K activation and since PI3K signaling contributes to MAPK resistance, these tumors may be more sensitive to MAPK pathway inhibitors than tumors with other substitutions (28). Therefore, a rational approach to treat cancers with different KRAS mutations would be to determine their specific signaling capabilities – a function of their unique biochemical and biological properties – and to target downstream processes necessary for the oncogenic function of a given mutant.

While this approach is theoretically sound, it fails to account for the influence that tissue-specific biochemical and genetic features have on KRAS signaling (57). For example, KRAS^G12D induces dramatically distinct downstream signaling patterns depending on the tissue-of-origin and the presence or absence of additional mutations (58). These observations become directly relevant for the design of effector therapies because, as discussed previously, each cancer type exhibits a distinct spectrum of KRAS allele frequencies (Figure 2B). Moreover, individual KRAS alleles have distinct patterns of cooperating mutations that vary depending on the tumor’s tissue-of-origin (21). Hence, the biochemical properties of different KRAS mutants are reflected in the genetic networks that are unique to cancers arising from different cellular lineages. By extension, the activation of downstream pathways by oncogenic KRAS can be further refined as a function of (iv) the existing co-mutations and (v) the tissue from which the cancer arises. This complexity makes it difficult – perhaps impossible – to intuitively identify targeted effector therapies that will be effective across cancers. The solution is to develop direct inhibitors of KRAS.

Lessons from direct inhibition KRAS^G12C

While long considered the ‘holy-grail’ of cancer targets, direct targeting of KRAS has finally been realized. Recently, sotorasib, a covalent inhibitor for KRAS^G12C, was approved for the treatment of patients with advanced lung adenocarcinoma expressing this mutant and who have failed at least one systemic pre-treatment (59). A second inhibitor (adagrasib) has been evaluated in Phase II trials and has been granted an FDA breakthrough therapy designation (60). These inhibitors take advantage of cysteine at codon 12 to covalently bind to GDP bound KRAS, locking the active site in an inactive conformation. Surprisingly, while KRAS^G12C is resistant to canonical GAP-induced GTP hydrolysis, it retains intrinsic GTPase activity and is sensitive to non-canonical GAP-induced hydrolysis (e.g. RGS3), two functions that are necessary to maintain a pool of GDP bound KRAS^G12C and overcome its activation by receptor tyrosine kinases (32,61). This mechanism of inhibiting oncogenicity – by shifting the population of KRAS into the inactive state – was originally demonstrated in studies of KRAS post-translational modification; acetylation of Lys104 destabilizes the active site to prevent SOS-induced nucleotide exchange (62). More recently, molecules that target nucleotide exchange either directly or indirectly – for example inhibitors of SOS1 or SHP2 – have been developed (63,64). Since these inhibitors do not target KRAS directly, they could presumably work in a genotype-agnostic manner, although single agent efficacy has been difficult to achieve, similar to inhibitors targeting kinases in KRAS effector pathways. Nevertheless, SHP2 inhibitors have recently shown promise in combination with G12C inhibition; SHP2 inhibition is thought to increase the amount of KRAS bound to GDP, thereby enhancing the efficacy of the G12C inhibition (65).

As with all other targeted inhibitors, resistance to G12C inhibition develops rapidly through a variety of mechanisms, including those that cause resistance to inhibitors of receptor tyrosine kinases, which function upstream of KRAS (66,67). These include amplification of the mutant allele, mutations in downstream members of the MAPK signaling pathway, acquisition of oncogenic fusions, and/or transdifferentiation to an alternative cellular state (68). In addition, de novo mutations in KRAS itself represent a common mechanism of resistance. These resistance alleles reveal important details about KRAS oncogenicity and, surprisingly, the relevance of inherited KRAS alleles. Secondary KRAS mutations can occur either in cis or in trans with the original G12C allele. The trans alleles frequently gain alternative activating mutations, such as Gly12 or Gly13 mutations, while the cis allele can be rendered resistant by mutations that promote hyperexchange – bypassing the mechanism of action of the inhibitors – or by mutating residues in the inhibitor binding pocket, such as Tyr96 or His95. Because sotorasib and adagrasib have unique chemical structures, mutations in the binding pocket can selectively influence the efficacy of one over the other (66,67). For example, mutations at His95 are associated with resistance to adagrasib, which contacts His95, while retaining sensitive to sotorasib (66). Intriguingly, H95N is a common germline single nucleotide polymorphism (SNP, rs1309399018), occurring in approximately one in every 30,000 individuals (51). Based on the analysis of somatic resistance mutations, the expectation is that cancers expressing KRAS^G12C that arise in the cis allele in patients carrying a germline H95N polymorphism would be intrinsically resistant to adagrasib.

Pathogenic germline KRAS variants

Germline expression of mutant forms of KRAS increases RAS/MAPK signaling during development, resulting in a number of multi-system syndromes with overlapping clinical features, including Noonan Syndrome (NS), Cardio-facio-cutaneous syndrome (CFCS), and, less commonly, Costello Syndrome (CS) (69). These syndromes represent a subset of a clinically defined group of developmental disorders known as “RASopathies”, each of which can result from mutation of any of a number of MAPK regulatory proteins, including SHP2, SOS1, RAF, MEK and the GAP proteins NF1 and p120 (70). This spectrum of syndromes is characterized by facial dysmorphia, heart defects, short stature, and skeletal defects, among others. In NS, the most common KRAS mutations are K5N, V14I, T58I, P34R, D153V (71). Although some of these mutants have been identified as somatic changes in cancer, they are rare (72,73). One mouse model of Noonan Syndrome, in which KRAS^V14I is expressed from the endogenous locus, models many of the clinical features observed in patients, including smaller size, facial dysmorphia, cardiac hyperplasia, and myeloproliferative disease (74). The severity of the features of this model are modified by the inbred background strain of the mice, emphasizing the role of genetic modifiers on the output of RAS signaling (75). This observation has larger implications for RAS biology, again highlighting the potential for inherited SNPs – in this case outside of the KRAS locus – to affect RAS function.

The KRAS variants identified in RASopathy patients result in a wide range of biochemical defects and are mostly grouped into Classes 2 and 4 (Figure 3B). Several variants (P34R, F156L, G60R) show a clear reduction in GTP hydrolysis, as is observed in some cancer-associated variants (45). Others (V14I, Q22E, F156L, K147E) exhibit WT levels of GTPase function, but high rates of nucleotide exchange and nucleotide dissociation (45,76). Indeed, V14I exhibits enhanced nucleotide exchange and shares structural similarities with another oncogenic hyperexchange mutant, A146T (47). Y71H is regulated much like WT KRAS by GAPs and GEFs, but has a higher affinity for RAF. Two particularly interesting variants, K5N and D153V demonstrate WT levels of hydrolysis and exchange, but instead have been proposed to affect MAPK signaling by shifting the orientation of KRAS on the underside of the plasma membrane, allowing for more efficient activation of effectors (45,77). Thus, as with clearly oncogenic KRAS mutations, developmental mutants can employ multiple mechanisms to increase the functional activation of KRAS.

Given their biochemical similarities to known oncogenic variants, it is surprising that these germline mutations are not more frequently observed in cancer. Amongst germline variants, V14I has been observed most often (albeit at very low frequency) in patient samples. This is consistent with results from the V14I mouse model, which demonstrate that this allele can cooperate with either inflammation or loss of tumor suppressors (i.e. p53) to induce PanIN in the pancreas (78). Further analysis of many of these mutants demonstrated that, despite reduced hydrolysis and enhanced nucleotide exchange – both of which would contribute to an increase in steady state KRAS-GTP – these variants generally have reduced affinity for RAF and, potentially, other effector proteins (45,47). This loss of effector binding appears to keep these KRAS variants in check, preventing the full activation of downstream pathways that occurs with the common oncogenic alleles, therefore emphasizing the importance of effector binding on RAS allele oncogenic activity and the relevance of drug development strategics that target these interactions.

Mouse models have demonstrated that germline expression of G12D is developmentally lethal (79,80). Recently, however, an additional class of RASopathy, which includes Schimmelpenning-Feuerstein-Mims syndrome (SFMS) and cutaneous skeletal hypophosphatemia syndrome (CSHS), has been described. These syndromes present as mosaic expression of abnormally high levels of MAPK signaling, often resulting from a postzygotic activating mutation in a RAS gene. Unlike syndromes arising from germline variants, which would be expressed from the very earliest stages of development, these mosaic RASopathies frequently arise from expression of variants commonly associated with cancer in a subset of cells. Mosaic RASopathies frequently manifest as focal or segmental dysplasia, such as congenital nevus sebaceous, and can sometimes affect multiple body systems. These mosaic cases have been found to express oncogenic alleles, including G12D, G12V, Q61H, G13D, and A146T, which are virtually never observed in germline RASopathies (81–92).

Conclusion/perspectives

The conventional wisdom is that KRAS functions as a binary on/off switch by virtue of this GDP/GTP cycling (93). This concept oversimplifies the complex interplay between the nucleotide state of an individual molecule and the resulting dynamic protein conformations that arise from the nucleotide state, which are nearly unique for each isoform and variant (17,94). The variety of conformational states unique to each variant contributes to the distribution of activity seen for the entire population of KRAS protein molecules in the cell (95). Here we have described four classes of KRAS mutants, defined by their effects on GTP hydrolysis, nucleotide exchange, and effector interactions. This distribution of variant KRAS functions is further fine-tuned by tissue- and cell type-specific factors, such as coincident extracellular signals or patterns of GAP and GEF expression and localization, the expression pattern and binding affinity of effectors, or the distinct cellular distribution of the two KRAS splice forms (KRAS4A and KRAS4B) which will share the same mutation (55,77,95,96).

The recent successes in generating G12C-specific inhibitors have justified efforts to develop allele-specific therapies that target the unique biochemical features of each mutant. For example, a proof-of-concept strategy for developing G12D-specific inhibitors, through formation of an Asp12-dependent salt-bridge, has been described recently (97,98). Identification of common features within each functional class may more efficiently allow the extension of therapeutic strategies across mutants of the same class. For instance, in some studies, EGFR inhibitors exhibit efficacy against metastatic colorectal cancers harboring G13D, but not Gly12, mutants of KRAS (99). Recently, a potentially mechanism was put forward suggesting that the incomplete impairment of NF1 function against KRAS^G13D is responsible for this efficacy (100). However, this putative mechanism, which was partly derived using computational modeling approaches, relies on intrinsic nucleotide hyperexchange and does not incorporate the synergism seen between GEF and KRAS^G13D (25,41,43). It will be interesting to determine whether metastatic colorectal cancers expressing other class 2 mutants are also sensitive to EGFR mutants. Finally, mutant-specific or class-specific therapies will have the added benefit of specifically targeting only the mutant KRAS, rather than targeting upstream or downstream signaling proteins that are required by WT RAS signaling in non-cancer cells.

Significance.

Efforts in the recent past to inhibit KRAS oncogenicity have focused on kinases that function in downstream signal transduction cascades, although preclinical successes have not translated to patients with KRAS mutant cancer. Recently, clinically effective covalent inhibitors of KRAS^G12C have been developed, establishing two principles that form a foundation for future efforts. First, KRAS is druggable. Second, each mutant form of KRAS is likely to have properties that make it uniquely druggable.