The path to the clinic: a comprehensive review on direct KRASG12C inhibitors

The RAS oncogene is both the most frequently mutated oncogene in human cancer and the first confirmed human oncogene to be discovered in 1982. After decades of research, in 2013, the Shokat lab achieved a seminal breakthrough by showing that the activated KRAS isozyme caused by the G12C mutation in the KRAS gene can be directly inhibited via a newly unearthed switch II pocket. Building upon this groundbreaking discovery, sotorasib (AMG510) obtained approval by the United States Food and Drug Administration in 2021 to become the first therapy to directly target the KRAS oncoprotein in any KRAS-mutant cancers, particularly those harboring the KRASG12C mutation. Adagrasib (MRTX849) and other direct KRASG12C inhibitors are currently being investigated in multiple clinical trials. In this review, we delve into the path leading to the development of this novel KRAS inhibitor, starting with the discovery, structure, and function of the RAS family of oncoproteins. We then examine the clinical relevance of KRAS, especially the KRASG12C mutation in human cancer, by providing an in-depth analysis of its cancer epidemiology. Finally, we review the preclinical evidence that supported the initial development of the direct KRASG12C inhibitors and summarize the ongoing clinical trials of all direct KRASG12C inhibitors.


Background
Since the discovery of KRAS in the 1960's, little progress has been made recently in treating patients with KRASdriven cancers. After a period when directly targeting KRAS was considered unlikely, hence deeming KRAS "undruggable, " the Shokat lab reinvigorated the search for a direct KRAS inhibitor by discovering the KRAS G12C switch II pocket in 2013. In May 2021, the United States Food and Drug Administration (FDA) approved sotorasib (Lumakras ™ ) as the first treatment for adult patients with non-small cell lung cancer (NSCLC) harboring the KRAS G12C genetic mutation who have received at least one prior systemic therapy. Thus, sotorasib became the first FDA-approved therapy to directly target the KRAS oncoprotein in tumors, particularly those with KRAS G12C mutations. As of early August 2021, there have been twenty registered clinical trials investigating nine different direct KRAS G12C inhibitors, with seven inhibitors currently being investigated.
The key purpose of this review is to convey a thorough account of the KRAS G12C preclinical and clinical research landscape that has led to our present understanding of treating KRAS G12C -mutant cancers. We aim for this review to serve as a comprehensive primer for any physician or scientist interested in pursuing KRAS G12C investigation. To our knowledge, we are not aware of any published review that has provided as extensive a coverage of all the landmark preclinical and clinical research regarding KRAS G12C inhibition specifically.

Discovery of KRAS
Discovered by its namesake in 1967, the Kirsten murine sarcoma virus (Ki-MSV) was first isolated as a sarcomainducing retrovirus during the passaging of murine leukemia viruses in rats [1]. In 1975, Scolnick et al. established Ki-MSV to be a recombinant virus with incorporated sequences derived from the rat genome [2]. Soon after, these sequences were molecularly characterized [3], and Ki-MSV was established to contain the rat cellular gene Kras [4].
Around the same time the retrovirus oncogene studies were conducted, DNA transfection studies were initially performed to unearth genes capable of inducing morphologic cellular transformation. In 1979, Weinberg et al. first reported the morphological transformation of NIH/3T3 mouse fibroblasts using genomic DNA isolated from chemically transformed rodent fibroblasts [5], which was followed by additional transformation studies with NIH/3T3 cells transfected with different human tumor cell lines by other groups [6][7][8][9][10].
In 1982, perhaps serendipitously, the Cooper, Weinberg, Barbacid, and Aaronson groups [11][12][13] implemented DNA hybridization with retroviral oncogene probes to determine that the human oncogenes which transformed the NIH/3T3 mouse fibroblast cell lines in the earlier transfection studies were homologous to the RAS genes identified in the Kirsten and Harvey sarcoma viruses. In fact, it was later revealed that the first NIH/3T3 transformants reported by the Weinberg group possessed an activated KRAS oncogene [14]. By the end of 1982, a single missense mutation in codon 12 was found to be the molecular basis of HRAS gene activation in the EJ/T24 bladder carcinoma cell line [15][16][17]. Moreover, the codon 12 mutation was found to be the activating mutation in KRAS in lung and colon tumor cells [18]. Mutant RAS genes were later identified in patient tumors but not in normal tissue, which validated the fact that the mutant RAS genes from aforementioned tumor cell lines were real and not artifacts of in vitro cell passage [19][20][21]. Thorough reviews describing the discovery of the RAS family have been previously published [19,22,23].

Structure of RAS
Three genes (HRAS, NRAS and KRAS) encode the four major RAS isoforms: HRAS, NRAS, KRAS4A, and KRAS4B. KRAS encodes two variants due to alternative splicing of exon 4, which results in divergent C-terminal sequences. Both variants are expressed in human cells, but KRAS4B is the predominant isoform expressed in human cells while KRAS4A has low expression levels and is more similar to viral KRAS [19]. The amino-terminal residues 1-165 of the RAS proteins share 92-98% sequence identity. The remaining 23-24 carboxy-terminal residues diverge substantially in sequence and are thus defined as the hypervariable region (HVR) [24]. Although the protein sequences among RAS isomers are highly conserved, KRAS exhibits rare (i.e., genomically underrepresented) codon bias at the third (i.e., degenerate) position of codons: KRAS displays an A/T bias, resulting in relatively poor protein translation, while HRAS demonstrates an G/C bias. The degree of rare codon bias appears to be correlated with the mutation frequency among RAS isomers [25].
The remaining residues in the carboxy-terminal constitute the HVR and lipid tail that appear to be poorly structured in solution. The HVR includes the CAAX (C: cysteine amino acid, A: aliphatic amino acid, X: amino acid dictating whether farnesylated or geranylated) motif, which is a membrane anchor sequence. Upon its initial synthesis in the cytosol, the RAS protein undergoes post-translational modifications to enable its association with the plasma membrane, which is essential for activation of its various downstream pathways. The subcellular localization of RAS proteins is determined by the specific lipid modification, composition of local membranes, and electrostatic nature of the isoform-specific HVR's [27]. Once synthesized, RAS first undergoes prenylation, which involves farnesylation of the cysteine residue in the CAAX motif by fanesyltransferase. Only HRAS is exclusively prenylated by farnesyltransferase, while KRAS and NRAS (to a lesser extent) can undergo alternative prenylation by geranylgeranyl transferase-I when farnesyltransferase activity is blocked [28]. This alternative prenylation was discovered after the disappointing results of the promising clinical trials involving farnesyltransferase inhibitors (FTI's) when targeting KRAS-driven tumors [29]. After farnesylation, the -AAX motif is cleaved by RAS-converting enzyme 1 (RCE1), and the farnesylated cysteine residue is then carboxymethylated by isoprenylcysteine carboxymethyltransferase (ICMT) [24].
All RAS proteins require a second signal for proper plasma membrane localization. Except for KRAS4B, this second signal includes palmitoylation at a second cysteine in the HVR, which creates a second hydrophobic anchor. Because KRAS4B lacks a second cysteine in the HVR, KRAS4B is not palmitoylated. Instead, the HVR of KRAS4B contains a polylysine region that anchors the protein to the plasma membrane via electrostatics. Similarly, KRAS4A contains a bipartite polybasic sequence that also contributes to membrane attachment via electrostatic interactions in addition to palmitoylation [24].
RAS signaling is regulated through a balance between activation by guanine nucleotide exchange factors (GEF's) and inactivation by GTPase-activating proteins (GAP's). RAS proteins intrinsically have a relatively slow off-rate for GDP (t 1/2 = 6 min, k off = 2 × 10 − 3 s − 1 at 20°) [36]. Thus, GEF's, such as son of sevenless homologue (SOS) and RAS guanyl nucelotide-releasing protein (RAS-GRP), are needed to accelerate the GDP/GTP exchange reaction by several orders of magnitude. Upon binding to GEF, RAS undergoes conformational changes in the switch regions and P loop that weaken the GDP affinity, ultimately resulting in GDP release and replacement by GTP [37]. Because RAS and GEF each have similar affinities for GDP or GTP, increased levels of GTP-bound RAS occur mainly due to the ~ 10-fold higher cellular GTP concentration of GTP relative to GDP. Binding of GTP dissociates the GEF and leads to the formation of an active GTP-bound RAS that can bind effectors. Moreover, RAS proteins have a relatively slow rate of intrinsic GTP hydrolysis (t 1/2 = 16 min, k off = 6 × 10 − 4 s − 1 ) [36]. Therefore, for efficient GTP hydrolysis to occur, GAP's, such as p120GAP and neurofibromin, function by accelerating the cleavage step by several orders of magnitude.

Overview of the RAS family
Numerous studies have reported mutant RAS as a key driver of transformation and tumor progression [19,30,38]. Cross-referencing data from the four leading cancer mutation databases Catalogue of Somatic Mutations in Cancer (COSMIC) [39], The Cancer Genome Fig. 1 The G domain corresponds to the initial 166-168 residues. Bordering the nucleotide-binding pocket are four main regions: the phosphate-binding loop (P-loop, residues 10-17), switch I (residues 30-38), switch II (residues 60-76) and the base-binding loops (residues 116-120 and 145-147). The two switch regions regulate all known nucleotide-dependent interactions between RAS and its binding partners. The remaining residues in the carboxy-terminal constitutes the HVR, including the CAAX motif, a membrane anchor sequence. Abbreviations: HVR, hypervariable region; CAAX, C: cysteine amino acid, A: aliphatic amino acid, X: amino acid dictating whether farnesylated or geranylated Atlas (TCGA) [40], the Memorial Sloan Kettering Cancer Center (MSKCC) cBioPortal [41], and the International Cancer Genome Consortium (ICGC) [42], as well as the private Foundation Medicine dataset, Prior and colleagues estimated ~ 19% of cancer patients to have a RAS mutation, which is equivalent to ~ 260,000 incident cancer cases in the United States alone [43]. Globally, there are ~ 18 million new cancer diagnoses per year [44], from which Prior and colleagues extrapolated to be ~ 3.4 million new cancer incidents with RAS mutations worldwide. Among mutations in the three RAS genes, gain-of-function missense mutations in KRAS account for the majority of RAS gene mutations (75%), followed by NRAS mutations (17%) and HRAS mutations (7%) [43,45]. In pan-cancer analyses, mutations in any of these RAS family genes is associated with poor patient prognosis [41,46].
RAS genes are further distinguished by their distinctive mutation frequency differences at three codon-specific hotspots, which also vary by tumor type. The most frequent sites of oncogenic mutation in RAS are residues G12, G13, and Q61. Residues G12 and G13 are located in the P-loop region of RAS, while residue Q61 is located in the switch II region. Single base missense mutations at these three key residues account for 99% of all RAS mutations [47]. Among KRAS mutations, G12 mutations are predominant (81%), followed by G13 mutations (14%) and Q61 mutations (2%) [43]. Among NRAS mutations, Q61 mutations are the majority (62%), followed by G12 mutations (23%) and G13 mutations (11%). HRAS mutations are more evenly split in residue preferences, but the preferences are still relatively apparent. Among HRAS mutations, Q61 mutations are the most prevalent (38%), followed by G12 mutations (26%) and G13 mutations (23%).
Furthermore, there are differences in the specific amino acid that is mutated at each of the three hotspots (G12, G13, and Q61), which also vary by tumor type (see below). Nineteen different activating mutations in codon 12, 13, or 61 can be created in each RAS isoform by a single base change. Among those three codons, five mutations (G12D, G12V, G12C, G13D and Q61R) account for 70% of all RAS-mutant cancers [43].

Mutant KRAS in human cancer
Globally, among the ~ 18 million new cancer diagnoses each year, ~ 2.6 million cancer patients are estimated to harbor a KRAS mutation (i.e., ~ 14% of all cancer cases) [43]. KRAS mutations alone account for ~ 1 million annual deaths worldwide, similarly to that for malaria and tuberculosis [28]. As stated earlier, gain-of-function missense mutations in KRAS account for the majority of RAS gene mutations in human cancer (75%). Among the 29 cancer types with KRAS mutations that were included in Table 1 of the study by Prior and colleagues, 43% of all KRAS mutations in human cancer are found in colorectal adenocarcinoma, followed by pancreatic adenocarcinoma (20%) and NSCLC adenocarcinoma subtype (14%). Within a particular cancer type, the three cancers in which KRAS mutations are the predominant RAS mutations are pancreatic cancer (~ 88%), colorectal adenocarcinoma (50%) [colon adenocarcinoma (50%) and rectal adenocarcinoma (50%)], and lung adenocarcinoma (32%). Because these three cancers alone are responsible for the majority of KRAS mutations found in human cancer (77%), much of the current focus on KRAS-mutant cancer therapeutics is targeted against these cancers. However, as illustrated in Fig. 3, KRAS mutations are also noteworthy in a variety of other cancer types including small intestine adenocarcinoma (26%), biliary tract cancers [cholangiocarcinoma (23%), gallbladder carcinoma (16%)], plasma cell myeloma (18%), and anaplastic thyroid carcinoma (8.6%).
Differences are observed in codon-specific and amino acid-specific mutation patterns that further vary depending on tumor type. Among all residue-specific KRAS mutations, the G12 mutation is the predominant mutation (81%), followed by the G13 mutation (14%) and Q61 mutation (2%) [43]. Among KRAS mutations in pancreatic adenocarcinoma, residue G12 mutations (94%) predominate, while G13 and Q61 mutations (1.2 and 2.4%, respectively) are rare [43]. On the other hand, among KRAS mutations in colon and rectal adenocarcinoma, G13 mutations are observed in relatively high frequencies (20 and 21%, respectively). Analyzing amino acid-specific KRAS mutations in the three cancer types responsible for most KRAS mutations, the most common KRAS mutation in NSCLC adenocarcinoma subtype is KRAS G12C (glycine to cysteine) (33.4%), while the most common KRAS mutation in colon and rectal adenocarcinoma and pancreatic ductal adenocarcinoma is KRAS G12D (glycine to aspartate) (32.7, 32.5, and 46.1%, respectively) [39].
Focusing on the KRAS G12C mutant, this mutation is present in 12% of all KRAS-mutant cancers [39]. Prior et al. gathered data from COSMIC and American Cancer Society (ACS) 2018 cancer incidence statistics to estimate the number of KRAS G12C mutations found in newly diagnosed cancer patients each year in the U.S. alone. They estimated that 32.8% of all KRAS G12C mutations in human cancer are found most commonly in lung adenocarcinoma, followed by colorectal adenocarcinoma (18.8%) and pancreatic adenocarcinoma (4.6%) [43]. As a comparison, they estimated that KRAS G12D mutant is most commonly found in pancreatic adenocarcinoma (29.9%), followed by colorectal adenocarcinoma (25.8%) and lung adenocarcinoma (5.9%) [43]. Analyzing the KRAS G12C mutation frequencies in each of these three cancer types separately, Nassar et al. identified KRAS G12C mutations in 13.8% of NSCLC (7.0% of lung adenocarcinoma-subtype [39]), 3.2% of colorectal cancer, and < 1% of pancreatic cancer patients using data extracted from the registry of the American Association for Cancer Research Project Genomics Evidence Neoplasia Information Exchange (GENIE), version 8.0 [48,49]. In summary, these codon-and amino acid-specific mutation patterns that differ depending on the tumor tissue of origin must be taken into consideration when designing novel anticancer therapeutics since they may implicate distinct oncogenic properties and thus clinical consequences.

Mechanistic and clinical implications of missense mutation-specific hotspots
Missense gene mutations at amino acid residues G12, G13, and Q61 have unique structural and functional consequences on the RAS protein, respectively [36,38]. Substitution of glycine at G12 or G13 by any amino acid except proline are believed to cause a steric block that prevents the arginine finger of GAP from entering the GTPase site of RAS, thus discouraging hydrolysis as a result [50]. Because Q61 is part of the GTP hydrolysis mechanism, Q61 mutations abolish both intrinsic and GAP-mediated GTP hydrolysis. Overall, missense mutations that affect these residues result in increased RAS activation via GTP-loading, thus bypassing normal physiological activation by upstream input, for example, from receptor tyrosine kinases [28].
Mutagenesis studies first demonstrated the idea that certain amino acid substitutions at a given mutation hotspot can lead to varied functional consequences. In a study with HRAS as the model, a wide range of oncogenic outcomes was shown with all 19 possible amino acid substitutions at the G12 codon [51]. Additionally, in a similar study that examined the outcomes of 17 different mutations on HRAS Q61, all 17 mutants shared comparable defects in GTP hydrolysis activity in vitro, while the Q61P and Q61E mutations did not show increased transforming ability relative to HRAS WT [52]. In other studies, G12V mutation decreased GAPmediated hydrolysis, Q61L mutation reduced both GAPmediated and intrinsic hydrolysis and increased intrinsic nucleotide exchange, and G13D mutation reduced GAPmediated hydrolysis and noticeably increased intrinsic nucleotide exchange [53]. In addition to specific mutations impacting direct RAS protein function, expanding evidence suggest mutation-specific consequences on the effector signaling pathways of RAS as well [54,55].
To underscore the impact of the mutation-specific consequences on direct RAS protein function and its signaling pathways, Westover and colleagues profiled the biochemical and biophysical properties of commonly occurring KRAS mutants (G12A, G12C, G12D, G12R, G12V, G13D, Q61L, and Q61H) [36]. They analyzed the intrinsic and GAP-mediated GTP hydrolysis rates, GTP and GDP-binding kinetics, relative affinities for RAF kinase, and high-resolution crystal structures. Overall, mutations that affect the G12, G13 and Q61 residues resulted in decreased intrinsic and GAP-stimulated GTP-hydrolysis rates. Additionally, they showed that the G13D mutant expressed an increased rate of nucleotide exchange independent of GEF (e.g., SOS) regulation relative to the wild-type protein, which resulted in increased auto-activation and thus provided another potential mechanism of increased aberrant cellular RAS signaling.
Mutations impacting KRAS activity can be classified by having either a high (WT, G12C, G12D, G13D) or low (G12A, G12R, G12V, Q61L, and Q61H) level of intrinsic GTPase activity. Mutations can be further distinguished by having either high (WT, G12A, G12C, G13D, and Q61L) or low (G12R, G12V, and G12D) RAF affinity, based on their relative affinity for RAF kinase Ras-binding domain (RBD). With these criteria in mind, Westover and colleagues proposed a predictive model for the relative dependence on (i.e., activation of ) the RAF kinase pathway compared with other canonical RAS signaling pathways such as PI3K and RalGDS among tumors with KRAS mutations. The Westover model predicts tumors with G12A and Q61L mutations preferentially signal through the RAF kinase pathway due to their high affinity for RAF kinase and relatively lower rates of intrinsic hydrolysis. In contrast, the model predicts G12D to have the lowest RAF activation levels due to its low affinity for RAF and faster hydrolysis rate. G12V and G12R would have moderate activation of RAF kinase due to their slow intrinsic hydrolysis rate coupled with a low RAF affinity. Likewise, G12C and G13D would also be predicted to have moderate activation of RAF kinase due to their high affinity but more rapid intrinsic GTPase activity that likely results in an attenuated duration of RAF kinase activation compared with G12A and Q61.
While the therapeutic response to farnesyltransferase inhibitors appears to be dependent on the specific RAS isoform mutated [29], there is increasing evidence that mutations at specific hotspots can also impact treatment outcomes among cancer patients. For example, based on clinical trials and various studies examining the efficacy of anti-epidermal growth factor receptor (EGFR) therapies such as cetuximab and panitumumab in treating colorectal cancer patients [56][57][58][59][60][61], the FDA revised their recommendation to exclude patients with KRAS G12 or G13 mutations from such treatment [62]. Currently, the National Comprehensive Cancer Network (NCCN) advises that colorectal cancer patients with any KRAS or NRAS mutation will not benefit from anti-EGFR therapy [63].
With increasing preclinical and clinical evidence supporting the importance of RAS isoform-and residue-specific mutation differences that are also both cellularly and genetically context-dependent [54], it appears for now that developing a single, universal anti-RAS therapeutic approach that specifically targets RAS mutants, without affecting RAS wild-type, in cancer is a challenging task. Thus, a pan-KRAS [64] or pan-RAS [65] inhibitor may have distinct advantages. For the foreseeable future, a targeted therapeutic approach that includes combinatory strategies and is dependent on both the RAS mutation subset and its environmental context seems to be the next step in optimally treating patients with RAS-driven cancers.

Introduction
Prior to the seminal discovery by Shokat and colleagues in 2013, KRAS was considered undruggable for several reasons. Firstly, the molecular structure of KRAS has intrinsically shown high resistance to small molecule modulation. Small molecule inhibitors, by design, primarily bind to identifiable pockets within a protein. However, KRAS is a small protein with a relatively smooth surface. Except for its GTP/GDPbinding pocket, the KRAS protein does not appear to have any well-defined, hydrophobic pockets viable for small molecule inhibitor binding. Furthermore, under the physiological conditions, GTP almost exclusively occupies GTP/GDP-binding pocket with extremely high affinity in the picomolar range. Coupled with the relatively high cellular GTP concentrations in the micromolar range (~ 500 μM), designing a small molecule inhibitor that can achieve adequate intracellular concentration to compete with GTP at the GTP/ GDP-binding pocket appeared formidable. Examining the successful advent of tyrosine kinase inhibitors, as a comparison, the binding affinity with their natural cosubstrate ATP is in the micromolar range (a millionfold difference in binding affinity), and ATP cellular concentrations range from 1 to 10 mM. Taking these factors into consideration, direct targeting of KRAS by small molecule inhibitors has been a challenging task [24,31,34,47,[66][67][68][69][70].
Because of these challenges, therapeutic approaches primarily focused on targeting KRAS indirectly. These indirect therapeutic strategies include targeting downstream or upstream RAS signaling pathways, metabolic pathways, and synthetic lethality approaches, all of which have yielded relatively disappointing results. Time and time again, indirect inhibition strategies have resulted in disappointing clinical trial outcomes in treating KRAS-mutant cancers [30,31,47,[66][67][68][71][72][73][74][75][76][77][78][79][80]. Efficacy with these treatments were lacking due to subsequent unexpected resistance mechanisms, as observed with the farnesyltransferase, BRAF, and EGFR inhibitors. However, in 2013, the groundbreaking foundation set by Shokat and colleagues renewed hope for direct KRAS inhibition, paving way for pharmacotherapeutics such as AMG510 and MRTX849 now currently being investigated in clinical trials with hopeful preliminary outcomes.

Discovery of the KRAS G12C inhibitor and the switch II pocket (S-IIP)
In 2013, Ostrem, Shokat, and colleagues made a breakthrough by successfully developing a series of compounds that covalently and irreversibly bound to the cysteine residue of the KRAS G12C (glycine to cysteine substitution) mutant, the most potent compound being compound 12 ( Fig. 4) [24]. Creatively, they used a disulfide tethering approach that enabled identification of small molecular fragments that could noncovalently bind to shallow pockets with low affinity under reducing conditions and covalently tether with cysteine residues. With this approach, they identified compounds that bound to a newly discovered allosteric pocket beneath the switch II region near the mutant cysteine, which they named the switch II pocket (S-IIP). This highly dynamic allosteric pocket was not apparent in the apo-crystal structures but appeared to be induced or stabilized by the compound. These compounds preferentially bound to RAS in the GDP-bound state, decreased RAS affinity for GTP relative to GDP, impaired SOS-catalyzed nucleotide exchange from GDP to GTP, and blocked RAS-RAF effector interaction in the KRAS G12C mutant cells. The blockade of RAS-RAF effector interaction may be attributed to a combination of inhibitor effects on nucleotide exchange, relative nucleotide affinities, and perhaps most critically, conformation effects on switch I and switch II. Importantly, by depending on the mutant cysteine residue for binding, the compounds specifically target the KRAS G12C mutant protein while sparing KRAS wild-type protein, an observation supported by the selectively inhibition of growth of cancer cell lines expressing KRAS G12C . Specifically targeting KRAS mutants over KRAS wild-type would, in theory, minimize the toxicities and side effects associated with a potential pharmacotherapeutic agent, thus greatly enhancing its therapeutic index when treating cancer patients.
Examining the mechanism of action of the mutant-specific inhibitors in detail, by covalently binding to mutant cysteine in the S-IIP, the compounds were found to displace glycine 60, located in the switch II region (residues 60-76), towards the switch I region (residues 30-38). The displacement resulted in the disordering of the switch I region and loss of Mg 2+ in several co-crystal structures produced by the Shokat lab. The GTP-bound state of RAS is quite sensitive to conformational disturbances in the regions of glycine 60 (switch II region) and threonine 35 (switch I region), demonstrated by the dominantnegative effects of the conservative mutants RAS G60A and RAS T35S . A co-crystal structure in the GTP state was unable to be obtained to confirm the conformational changes of the switch I and switch II regions upon inhibitor binding, but the presence of the inhibitor in the S-IIP would have pronounced effects on the positions of these two described residues [24].
Preferential binding to GDP-bound RAS may appear to be an issue because it contradicts the widely held perception that RAS mutants in cancer are deficient in both intrinsic and GAP-stimulated GTP hydrolysis activity, resulting in RAS mutants to be in the constitutively active, GTP-bound state independent of GEF activity at steady state. After all, codon 12 mutations are known to disable the activation of intrinsic RAS GTPase activity by GAP's. However, this perception has been gradually dispersed as new evidence emerged in recent years. KRAS G12 mutants have relatively high intrinsic GTPase activity (t 1/2 = ~ 23-27 min) that is similar to that of wildtype RAS proteins [36,81,82]. Biochemical and biophysical analyses by Westover and colleagues have shown that different RAS mutants exhibit variable rates of intrinsic and GAP-stimulated GTP hydrolysis, as well as variable intrinsic and GEF-mediated nucleotide exchange [36]. These discrepancies result in concentration differences in GTP-bound KRAS between KRAS mutations, even at a specific residue. Moreover, additional studies have strengthened the possibility that the nucleotide-bound state of RAS is in constant flux, instead of a single, constitutively active state (see below) [82,83]. These data may explain why the compounds are able to specifically bind to the inactive, GDP-bound KRAS mutants to ultimately reduce the level of the active, GTP-bound KRAS mutants in the cell.

ARS-853
Despite its initial success, the initial lead compound (compound 12) developed by the Shokat lab had suboptimal pharmacologic properties, with indeterminate potency against the KRAS G12C mutant in a cellular context. Indeed, two subsequent studies by Patricelli et al. and Lito et al. demonstrated that compound 12 was not capable of engaging KRAS G12C in cells even at a relatively high dose and long incubation time [82,83]. Compound 12 was eventually licensed to Wellspring Biosciences, a subsidiary of Araxes Pharma (both cofounded by Kevan Shokat in 2012). Further optimization of this scaffold resulted in an improved inhibitor, ARS-853, the first direct KRAS inhibitor shown to selectively inhibit KRAS in cells with potency in the range of a drug candidate [82].
ARS-853 had a very similar chemical structure to compound 12, which further supported the potential to inhibit the S-IIP [24], and also only covalently bound KRAS G12C in the GDP-bound state. Target engagement by ARS-853 was demonstrated to be both time-and concentration-dependent, as expected for irreversible covalent interactions. In comparison to compound 12, the newly improved inhibitor rapidly bound KRAS G12C , increasing the reaction rate in a biochemical assay by 600-fold; bound KRAS G12C at lower micromolar concentrations in both a biochemical and cellular context; and blocked nucleotide exchange (GDP to GTP) with much higher efficacy. ARS-853 potently inhibited the growth of KRAS G12C mutant cells, but not non-KRAS G12C cells, in ultra-low adherent 3D culture. Using mass spectrometry, Patricelli et al. and Lito et al. demonstrated specific binding of ARS-853 to KRAS G12C in cells and showed subsequent blockade of downstream signaling through the RAF-MEK-ERK and PI3K-AKT pathways, which is most likely due to failure of GTP loading onto RAS. The authors found no effects of ARS-853 on RAS signaling or growth in non-KRAS G12C cells at concentrations up to 10-fold higher than its KRAS G12C potency. Moreover, across the 2740 surface-exposed cysteine residues profiled from 1584 proteins, KRAS G12C was the most potent target of ARS-853, with only two off-targets (FAM213A and Reticulon-4) that also showed significant engagement with ARS-853. In summary, the studies established ARS-853 to be a selective, covalent inhibitor with low micromolar potency in cancer cells harboring G12C mutations in KRAS [82,83].
As mentioned earlier, there is increasing evidence that KRAS mutants are not in a constitutively active, GTPbound configuration at steady state. Rather, RAS mutants exhibit variances in intrinsic nucleotide exchange and GTPase activity, and GEF-and GAP-mediated cycling, that ultimately create concentration differences of GTPbound (and GDP-bound) RAS within a cell. These variances depend on the RAS codon mutation, even down to the specific residue that is mutated. The studies on ARS-853 and direct KRAS G12C inhibition conducted by Patricelli et al. and Lito et al. provide further support for the plasticity of the nucleotide cycle of KRAS mutants. When second-site substitution mutations were introduced that accelerated intrinsic nucleotide exchange, a decrease in ARS-853-mediated KRAS G12C inhibition was observed. Furthermore, when second-site substitution mutations were introduced that decreased the intrinsic rate of nucleotide hydrolysis, ARS-853-mediated inhibition of KRAS G12C was prevented.
Observing that the KRAS mutant had intrinsic GTPase hydrolysis and nucleotide exchange activity, Patricelli et al. and Lito et al. assessed if increased cellular GEF activity regulated by upstream receptor tyrosine kinases (RTK's) attenuated ARS-853 engagement with KRAS G12C . Treatment with the EGFR inhibitor erlotinib increased ARS-853 engagement with KRAS, which one would expect since treatment would result in an increased GDP/GTP-bound ratio of KRAS mutants due to the decrease in GEF activity, given that KRAS mutants truly do not exist in a constitutively active GTP-bound state. Furthermore, the impact of inhibiting downstream effectors of the RAS signaling pathway on ARS-853-mediated KRAS G12C inhibition was investigated. Treatment with the MEK inhibitor trametinib increased GTP-bound KRAS G12C levels and decreased the efficacy of ARS-853 in cells, possibly through the release of ERK-mediated negative feedback inhibition of the SOS1 GEF. Taken together, these findings further strengthen the possibility that KRAS G12C retains appreciable intrinsic GTPase activity and nucleotide exchange such that it is not fully GTP bound in vivo. Moreover, the studies revealed a potential mechanism of acquired resistance against direct KRAS G12C inhibitors via increased upstream signaling by RTK-mediated GEF activity, hence supporting combination therapy with direct or indirect GEF inhibitors, such as EGFR inhibitors, in KRAS G12C mutant cancers.

ARS-1620
Although ARS-853 was the first KRAS G12C inhibitor with cellular potency in the range of a drug candidate, it lacked adequate potency in vivo in mouse tumor models. A major drawback of ARS-853 was the short metabolic plasma stability and poor bioavailability. Thus, the inherent poor chemical and metabolic stability of ARS-853 series made it unsuitable for further preclinical development.
Subsequently, Janes et al. reported a new generation of KRAS G12C S-IIP inhibitors, ARS-1620 [84]. ARS-1620 was a direct KRAS G12C small molecule inhibitor that is potent, selective, orally bioavailable, and well-tolerated in mice. The compound exhibited both in vitro and in vivo potency with a therapeutic window in the range of a drug candidate. With improved potency and pharmacologic properties, ARS-1620 was highly efficacious as a single agent in multiple human cancer cell line-and patientderived mouse xenograft tumor models. Thus, ARS-1620 provided the first in vivo evidence that the S-IIP targeted approach may be a promising therapeutic strategy for patients with KRAS p.G12C mutant cancers.
As an interesting aside, Janes et al. noted that in vitro studies assessing KRAS dependency using monolayer (2D-adherent) cell cultures significantly underestimated KRAS dependence in vivo, since they found that 3D ultra-low adherent suspension spheroid cultures better predicted in vivo sensitivity of KRAS mutant cancer cells to ARS-1620. The authors were not aware of any approved oncology drugs that displayed differential activity between 2D and 3D cultures as substantial as KRAS inhibition. These findings have significant translational Fig. 3 A The pie chart depicts the percentage of each RAS isoform contributing to all RAS mutations. Gain-of-function missense mutations in KRAS account for the majority of RAS gene mutations (75%), followed by NRAS mutations (17%) and HRAS mutations (7%). The remaining 1% represents RAS mutations that are other than gain-of-function missense mutations. Among KRAS mutations, the G12 codon (81%) is the most frequently mutated, followed by G13 (14%) and Q61 (2%). B The bar graph portrays the proportion of RAS mutations that are KRAS mutations found in a given cancer type. KRAS mutations are the most common RAS mutations in pancreatic cancer (~ 88%), followed by colon adenocarcinoma (50%), rectal adenocarcinoma (50%), lung adenocarcinoma (32%), small intestine adenocarcinoma (26%), cholangiocarcinoma (23%), plasma cell myeloma (18%), gallbladder carcinoma (16%), and anaplastic thyroid carcinoma (8.6%) implications for interpreting in vitro synthetic lethal relationships of KRAS as a driving oncogene. However, the value of using 3D cultures to predict clinical response to KRAS G12C -directed inhibitors needs to be validated by patient response rates in clinical trials.

AMG510
Prior to the Shokat lab's initial disclosure of their efforts targeting KRAS G12C , Amgen had also independently initiated their own research program to identify covalent inhibitors of KRAS G12C [85]. In collaboration with Carmot Therapeutics, Amgen utilized Carmot's Chemotype Evolution platform, a custom library synthesis and screening platform, and structure-based design to identify a series of selective covalent inhibitors of KRAS G12C , which ultimately resulted in advanced lead 1 that potently inactivated KRAS G12C in biochemical and cellular assays [86]. Similar to ARS-1620, advanced lead 1 also attached to the S-IIP and covalently bound to mutant cysteine. However, cocrystallization of advanced lead 1 with GDPbound KRAS G12C revealed advanced lead 1 to occupy a previously unexploited cryptic pocket on the surface of KRAS [85], which was first independently identified earlier by Gentile et al. [87]. The hidden pocket is bordered by the H95, Y96, and Q99 side chains, and was unveiled by rotation of the H95 (histidine 95) side chain. Engagement of this cryptic pocket led to a multifold enhancement in cellular potency relative to ARS-1620. However, advanced lead 1 was not suitable for preclinical development due to its high clearance and low oral bioavailability in vivo in rodent model systems. Subsequently, superposition of binding modalities of advanced lead 1 and ARS-1620 led to the substitution of the quinazoline nitrogen (N1) of ARS-1620 as an alternative approach to exploit the H95/Y96/Q99 cryptic pocket. After further improvements, compound (R)-38 was created with enhanced potency and optimized pharmacokinetics suitable for preclinical development, and it was later nominated as a clinical candidate and coded as AMG510.
In a series of experiments, Canon et al. demonstrated AMG510 to be potent, efficacious, and selective in both cellular and mouse tumor models [88]. Treatment resulted in regression of KRAS G12C -specific tumors and decreased downstream p-ERK signaling. Sharing structural similarities with ARS-1620, AMG510 selectively bound to KRAS G12C , not KRAS wild-type, in the S-IIP region and covalently bound to mutant cysteine as well. However, unlike ARS-1620, AMG510 additionally bound to the His95 groove. One prominent liability of ARS-1620 is its suboptimal potency resulting from the small volume of the S-IIP that it occupies, which provided limited avenues for additional protein-ligand interactions. In contrast, the isopropyl-methylpyridine substituent of AMG510 occupying the His95 groove engaged in a continuous network of 25 ligand-protein van der Waals contacts, which extends from the backbone of helix 2 (His95, Tyr96) to the backbone of the flexible switch II loop. These interactions enhanced the potency of AMG510 by approximately 10-fold [mean half-maximum inhibitory concentration (IC 50 ) = 0.09 μM)] compared to ARS-1620 in a nucleotide-exchange assay with recombinant GDPbound KRAS G12C .
Additionally, Canon et al. examined various combination therapies with AMG510 to assess if combination therapy could overcome potential resistance and enhance tumor-cell killing. Referencing the clinically validated strategy of combining BRAF and MEK inhibitors in treating melanoma [89], they investigated combinations of AMG510 and various inhibitors of the MAPK and AKT signaling pathways, including inhibitors of HER kinases, EGFR, SHP2, PI3K, AKT and MEK. Multiple combinations resulted in synergistic activity, especially with the MEK inhibitor trametinib, a combination that demonstrated significantly enhanced anti-tumor activity in monolayer and spheroid cell culture models as well as in mouse tumor models compared to using either agent alone. Given the prevalence of KRAS G12C mutations in lung adenocarcinoma, the authors also investigated combination treatment of AMG510 with carboplatin, a standard-of-care chemotherapeutic for lung cancers. While both drugs alone significantly inhibited tumor growth in KRAS G12C -mutant NCI-H358 (NSCLC) xenograft mice models, combination therapy resulted in greater antitumor activity. In summary, these preclinical data provide support for the clinical combination of AMG510 with MAPK inhibitors or chemotherapeutic agents to potentially bypass compensatory mechanisms of resistance, increase potency of therapeutic effects, and create a durable response to treatment.
Moreover, the authors investigated the role of the immune system in affecting the efficacy of AMG510 in treating KRAS G12C -mutant cancers. They reported that AMG510 induced long-term cures in immunocompetent mice with KRAS G12C -mutant CT-26 (colorectal) xenograft mice models, and induced tumor regression but not cures in the same mice tumor models lacking T cells (immunocompromised). Thus, they examined a combination strategy with immunotherapy, specifically with anti-PD-1 immune checkpoint inhibition since immune checkpoint blockade involving programmed cell death protein 1 (PD-1) and programmed death-ligand 1 (PD-L1) has been clinically validated in multiple settings [90,91]. As monotherapy, AMG510 or anti-PD-1 inhibition caused complete tumor regression in only one of ten mice, respectively. However, combination treatment resulted in significantly improved survival, with complete and durable tumor regression in nine out of ten mice.
Analyzing the treatment effects on the immune cell composition of the tumor microenvironment, Canon et al. reported that AMG510, alone and in combination with anti-PD-1 inhibition, promoted a pro-inflammatory tumor microenvironment. Notably, there was increased T cell infiltration, primarily cytotoxic CD8 + T cells, into the tumor in both treatment arms relative to anti-PD-1 inhibition alone. Additionally, AMG510 treatment led to increased number of total and proliferating CD3 + T cells and total CD8 + T cells, further increased with combination therapy; and increased infiltration of innate immune system macrophages and dendritic cells, including CD103 + cross-presenting dendritic cells, which are critical in T cell priming and activation and are also implicated in T cell recruitment [92]. As an additional comparison, AMG510 was also compared with a MEK inhibitor. Both inhibitors regressed CT-26 KRAS G12C tumors in mice to similar level, but unlike AMG510, MEK inhibition did not significantly affect the numbers of infiltrating CD8 + T cells.
To study a potential mechanism of action of immune system upregulation in the tumor microenvironment, transcriptional profiling of tumor RNA post-treatment with AMG510 revealed increased interferon signaling, chemokine production, antigen processing, cytotoxic and natural killer cell activity, and markers of innate immune system stimulation. Notably, there was increased expression of chemokines Cxcl10 and Cxcl11, two key attractants of tumor-suppressive immune cells, that may explain the enhanced immunosurveillance after AMG510 treatment [92,93].
Finally, in contrast to previous data suggesting that non-tumor-selective MEK inhibition in combination with anti-PD-1 treatment blocked T cell expansion and priming in vivo despite anti-tumor activity [94] There was a ~ three-fold increase in IFNγ secretion, a marker of tumor-specific T cell priming and activity, when splenocytes collected from the cured mice were stimulated by the two KRAS-mutant tumor cell lines, but no IFNγ induction was observed in the non-KRAS-mutant cell line. Together, these data suggest that AMG510 and anti-PD-1 combination therapy can establish an adaptive immune response against shared antigens that can recognize and eliminate related but non-KRAS G12C tumors as well, a clinically relevant outcome since intratumoral KRAS mutation status can be heterogeneous within the same tumor and between primary and metastatic sites [95][96][97][98][99][100].

MRTX849
In collaboration with Array BioPharma, Mirati Therapeutics conducted a covalent fragment screen of the Array BioPharma collection and identified a series of tetrahydropyridopyrimidines as irreversible covalent inhibitors of KRAS G12C , which led to the discovery of compound 4 [101]. Despite originating from an independent fragment screen, compound 4 markedly resembled the previously reported KRAS G12C acrylamide inhibitors ARS-1620 and AMG510 [102]. With structure-based design, compound 4 was further optimized to compound 13 [101]. While compound 13 potently inhibited KRAS G12C in animal models and led to tumor regression via intraperitoneal dosage, it suffered from rapid clearance and very low oral bioavailability intravenously. Subsequent tests identified the naphthol and acrylamide groups as the most metabolically sensitive positions in the molecule. Consequently, modifications of the groups were made that ultimately gave rise to MRTX849, which displayed much better stability and bioavailability [103].
Hallin et al. provided detailed characterization of MRTX849 activity, elucidation of response and resistance mechanisms, and identification of effective combinations [104]. Conducting cell viability assays using KRAS G12C -mutant cell lines in vitro, they found IC 50 values ranging between 10 and 973 nmol/L in the 2-D monolayer cultures and between 0.2 and 1042 nmol/L in the 3-D spheroid cultures, suggesting a differential degree of sensitivity to treatment they found to be at least partially attributed to KRAS-dependent ERK and S6 signaling. In vivo, in a panel of human KRAS G12C mutant cell linederived xenograft and patient-derived xenograft models, MRTX849 demonstrated broadly active tumor regression exceeding 30% volume reduction from baseline in 17 of 26 models (65%). In addition, the authors found that baseline RNA sequencing, reverse phase protein array analysis, nor co-occurring alterations were sufficient to predict activity or response variability to therapeutic intervention. To examine mechanisms of therapeutic response and resistance and explore possible synergistic vulnerabilities for combination therapy, the authors performed CRISPR/Cas9, targeting 400 genes, and pharmacologic screening on several KRAS G12C models. Both screens resulted in hits converging on regulators of the RAS pathway [(e.g., SHP2/PTPN11, receptor tyrosine kinases (RTK)], mTOR pathway, and cell cycle. These data were consistent with the more limited prior studies with the ARS-1620 KRAS G12C inhibitor [105,106]. In a subsequent series of in vivo experiments, combination therapy with MRTX849 was examined with afatinib (pan-HER inhibitor), palbociclib (CDK4/6 inhibitor), vistusertib (ATP-competitive mTOR inhibitor), and RMC-4550 (SHP2 inhibitor). Overall, results showed augmented responses that led to substantial tumor regression in multiple tumor models, including those that were refractory to MRTX849 single therapy.

JNJ-74699157 (ARS-3248)
In February 2013, Wellspring Biosciences, a subsidiary of Araxes Pharma (both co-founded by Shokat and colleagues), entered a drug discovery and development agreement with Johnson & Johnson's Janssen Biotech to further develop direct KRAS G12C inhibitors for the treatment of cancer. Their collaboration led to the creation of ARS-3248, a new generation of KRAS G12C inhibitor building upon ARS-1620. ARS-3248 is an investigational, orally available small molecule designed to potently and selectively inhibit KRAS G12C . JNJ-74699157 (ARS-3248) entered clinical trials in July 2019 (ClinicalTrials. gov number NCT04006301, Table 1), and the trial was completed in November 2020, with no results currently posted (Fig. 5).

CodeBreaK100 (phase 1)
In August 2018, sotorasib (AMG510) became the first KRAS G12C inhibitor both to enter human clinical trials and to demonstrate clinical safety and efficacy (NCT03600883 , Table 1) [107,108]. Hong and colleagues reported findings of a phase 1, multicenter, open-label trial of sotorasib in patients with heavily pretreated advanced (locally advanced or metastatic) solid tumors harboring the KRAS p.G12C mutation [107]. A total of 129 patients (59 NSCLC patients; 42 colorectal cancer [CRC] patients; and 28 patients with other tumor types) were enrolled in the dose escalation and expansion cohorts. Enrolled patients had received a median of 3 (range, 0 to 11) previous lines of anticancer therapies for metastatic disease. Sotorasib was administered orally once daily with each treatment cycle lasting 21 days. Because of its highest response, the 960 mg dosage administered daily was identified as the dose for the expansion cohort.
The primary endpoint was safety that included incidence of dose-limiting toxic effects (defined as sotorasib-related toxic effects within the first 21 days after the first dose), adverse events during the treatment period, and treatment-related adverse events. No dose-limiting toxic effects nor treatment-related adverse events resulting in death were observed. A total of 73 patients (56.6%) had treatment-related adverse events, and 15 patients (11.6%) had grade 3 or 4 events. The most common grade 3 treatment-related adverse events were gastrointestinal or hepatic side effects. Adverse events of any cause that occurred during treatment were reported in 125 patients (96.9%), with the most common events being diarrhea (29.5%), fatigue (23.3%), and nausea (20.9%). Grade 3 or higher adverse events of any cause that occurred during treatment were reported in 68 patients (52.7%).
Clinical efficacy was investigated as a secondary end point. Efficacy was described as objective response (complete or partial response), disease control (objective response or stable disease), duration of objective response, duration of stable disease, and progression-free survival. Among NSCLC patients, 19 patients (32.2%) had a confirmed objective response, and 52 patients (88.1%) had disease control. The median progression-free survival was 6.3 months. Among CRC patients, 3 patients (7.1%) had a confirmed response, and 31 patients (73.8%) had disease control. The median progression-free survival was 4.0 months.
Overall, the authors concluded that sotorasib had a relatively tolerable safety profile that consisted mainly of low-grade toxic effects and no dose-limiting toxic effects. Furthermore, sotorasib showed promising anticancer activity in patients with heavily pre-treated KRAS p.G12C mutant solid tumors, particularly NSCLC and CRC to a lesser extent.

CodeBreaK100 (phase 2)
In June 2021, Skoulidis et al. reported the results of the multicenter, single-group, open-label, phase 2 trial of CodeBreaK100 (NCT03600883, Table 1) [108]. The phase 2 trial evaluated the efficacy and safety of sotorasib monotherapy in patient with locally advanced or metastatic KRAS p.G12C-mutated NSCLC previously treated with standard therapies for KRAS-mutated NSCLC [108]. Among the 126 enrolled patients, the majority (81.0%) had previously received both platinum-based chemotherapy and inhibitors of PD-1 or PD-L1. Ultimately, 124 patients had measurable disease at baseline and were thus evaluated for response.
As the primary endpoint, an objective response was observed in 46 patients (37.1%; 95% confidence interval [CI]: 28.6 to 46.2), which exceeded their benchmark response of 23% in results previously reported in the REVEL trial investigating ramucirumab plus docetaxel in this population of patients [109]. Among the 46 patients, 4 (3.2%) had a complete response and 42 (33.9%) had a partial response. Examining secondary endpoints, disease control occurred in 100 patients (80.6%; 95% CI: 72.6 to 87.2). Additionally, among the 46 patients with an objective response, the median duration of response was 11.1 months (95% CI, 6.9 to could not be evaluated). Among the 124 patients, the median progression-free survival was 6.8 months (95% CI, 5.1 to 8.2), and the median overall survival of the entire 126 patient cohort was 12.5 months (95% CI, 10.0 to could not be evaluated).
Consistent with the safety findings of the phase 1 study, treatment with sotorasib in the phase 2 trial produced mostly grade 1 and 2 side effects, primarily low-grade hepatic and gastrointestinal toxic effects. Treatment-related adverse events occurred in 88 of 126 patients (69.8%), including grade 3 events in 25 patients (19.8%) and a grade 4 event in 1 (0.8%). No new safety signals were reported.
Exploratory biomarker analyses were conducted to assess associations between response to sotorasib therapy, baseline tumor PD-L1 expression level, tumor mutational burden, and mutations in STK11, KEAP1, and TP53, which are among the most prevalent genes with co-occurring mutations in KRASmutated NSCLC [110]. Of note, among all other subgroups or all patients that could be evaluated, the subgroup with STK11-mutated tumors and wild-type KEAP1 represented the highest number of patients that responded to sotorasib. The current standard care for patients with newly diagnosed KRAS-mutated NSCLC commonly involves an immune-checkpoint inhibitor (e.g., PD-1/PD-L1 inhibitor) as monotherapy or in combination with chemotherapy [111][112][113]. Thus, this finding has important clinical implications since inactivating genomic alterations in STK11 result in primary resistance to PD-1 and PD-L1 blockade and docetaxel in patients with KRAS-mutated NSCLC [114,115].
Subsequent clinical trials are currently investigating sotorasib as monotherapy or in combination with various anticancer agents in patients with advanced or metastatic solid tumors, including NSCLC and CRC, with the KRAS p.G12C mutation (Table 1).
At the same conference, Johnson and colleagues reported KRYSTAL-1 preliminary safety data on 31 patients with CRC treated with adagrasib 600 mg BID [117]. The most commonly reported (> 20%) treatmentrelated adverse events include diarrhea (58%), nausea (52%), fatigue (42%), and vomiting (36%). Among the 18 CRC patients evaluable for clinical activity, 3 patients (17%) had a confirmed objective response, while 17 patients (94%) achieved disease control. Moreover, in the other solid tumor cohort with 6 evaluable patients for clinical activity, confirmed partial responses were observed in a patient with endometrial cancer and in another patient with pancreatic cancer. Unconfirmed partial responses were observed in two other patients with ovarian cancer and cholangiocarcinoma, respectively.
More recently, at the European Lung Cancer Virtual Congress 2021, Riely et al. presented updates on the NSCLC patients in the KRYSTAL-1 trial [118]. In addition to the data presented previously by Jänne et al., Riely et al. noted promising results in patients with STK-11 comutations who typically have a relatively poor response to immune checkpoint inhibition, as discussed earlier.
Among these patients, 9/14 patients (64%) had an objective response. Examining the mechanistic action of adagrasib, preliminary pharmacodynamic and mechanistic biomarker analyses on pre-and post-treatment tumor NSCLC biopsies of 3 patients demonstrated downregulation of KRAS/MAPK pathway genes including DUSP6 and SPRY4. Furthermore, in patients with tumors containing STK11-comutations, there was minimal expression of immune transcripts such as CD4 and CD8 at baseline. However, these immune transcripts were increased with adagrasib treatment, which suggests a potential immune response after therapy.
Subsequent clinical trials are currently investigating adagrasib as monotherapy or in combination with other anticancer agents in patients with advanced or metastatic solid tumors, including NSCLC and CRC, with the KRAS p.G12C mutation ( Table 1).

Other direct KRAS G12C inhibitors
Additional clinical trials investigating other direct KRAS G12C inhibitors are currently ongoing and are summarized below and in Table 1.

LY3499446, LY3537982
After the phase 1 clinical trial of LY3499446 was terminated due to an unexpected toxicity finding (NCT04165031), Eli Lilly and Company revealed LY3537982 at the AACR Annual Meeting 2021 [119]. According to Eli Lilly, LY3537982 is a potent, highly selective covalent inhibitor of the KRAS G12C protein that was discovered using structure-based design. In preclinical studies, LY3537982 inhibited KRAS-GTP loading in the KRAS G12C mutant H358 NSCLC cell line with an IC 50 value of 3.35 nM, while AMG510 and MRTX849 had IC 50 values of 47.9 nM and 89.9 nM, respectively. Moreover, LY3537982 selectively inhibited the growth of KRAS G12C mutant tumor cells but not KRAS wildtype or non-G12C mutant cells. In multiple xenograft or patient-derived xenograft models with the KRAS G12C mutation, LY3537982 had a range of anti-tumor activity ranging from complete regression to significant tumor growth inhibition. Furthermore, mechanismbased combinatory screens identified potential synergistic targeted therapies with LY3537982 that promoted enhanced anti-tumor activity in vitro and in vivo, including abemaciclib, LY3295668 (selective AurA inhibitor), and cetuximab. A phase 1 clinical trial evaluating the candidate LY3537982 is currently registered but not yet recruiting (NCT04956640). The trial is investigating LY3537982 alone and in combination with various therapies, including the aforementioned studied agents, in patients with KRAS G12C mutant solid tumors.

GDC-6036
Created by Genentech, GDC-6036 is being investigated in a phase 1 clinical trial as monotherapy and in combination with other anti-cancer therapies in patients with advanced/metastatic solid tumors, including NSCLC and CRC, with a KRAS G12C mutation (NCT04449874).

D-1553
Designed by InvestisBio, D-1553 was presented at the AACR Annual Meeting 2021 [120,121]. According to the authors, D-1553 is an orally bioavailable, covalent small molecule compound that specifically inhibits KRAS G12C in vitro (e.g., lung, pancreatic, colorectal cancer cell lines) and in vivo (e.g., cell line-derived and patient-derived xenograft tumor models). In both cell line-derived (CDX) and patient-derived xenograft (PDX) tumor models with KRAS G12C mutation, D-1553 expressed anti-cancer activity as monotherapy and enhanced tumor growth inhibition or regression in combination with other targeted therapies (e.g., MEK inhibitor, SHP2 inhibitor) or chemotherapy. Specifically, in lung cancer PDX models, D-1553 exhibited tumor growth inhibition (TGI) ranging from 43.6 to 124.3%, with 4/8 models resulting in tumor regression. In colorectal cancer PDX models, TGI ranged from 60.9 to 105.7%, with 3/9 models showing tumor regression. An ongoing phase 1/2 clinical trial is evaluating D-1553 alone and in combination with other standard agents (not specified) in patients with advanced/metastatic solid tumors, including NSCLC and CRC, with the KRAS G12C mutation (NCT04585035).

JDQ443
Created by Novartis Pharmaceuticals, JDQ443 is being examined in a phase 1/2 clinical trial as monotherapy and in combination with TNO155 (SHP2 inhibitor) and/ or spartalizumab (anti-PD1 antibody) in patients with advanced or metastatic solid tumors with the KRAS G12C mutation (NCT04699188).

BI 1823911
Designed by Boehringer Ingelheim, BI 1823911 is the latest small molecule KRAS G12C inhibitor to enter clinical trials in July 2021 (NCT04973163). Initially presented at the AACR Annual Meeting 2021, BI 1823911 was found to have selective and potent anti-proliferative activity in a panel of KRAS G12C mutant cancer cell lines with higher or similar potency compared to AMG510 and MRTX849 [64]. In a panel of KRAS G12C NSCLC cell lines, treatment with BI 1823911 resulted in downregulation of MAPK pathway-responsive genes, such as DUSP6 and CCND1. In addition, at the protein level, strong and sustained inactivation of the MAPK pathway was observed using p-ERK as a pharmacodynamic biomarker. Furthermore, in a panel of NSCLC or CRC CDX or PDX mouse models, a daily oral dose of 60 mg/kg BI 1823911 showed in vivo efficacy comparable to 100 mg/kg AMG510 and 100 mg/ kg MRTX849, respectively. Finally, to examine potential synergistic anti-proliferative activity, the authors tested a large set of compounds in combination with BI 1823911 in a panel of selected KRAS G12C mutant cancer cell lines. Among other MAPK and PI3K pathway inhibitors, the SOS1::KRAS inhibitor BI 1701963 was shown by the authors to be a promising combinatorial partner. BI 1701963 is a first-in-class, small molecule, pan-KRAS inhibitor that binds to and prevents SOS1 from associating with KRAS. Together, in vitro and in vivo combination studies with BI 1823911 and BI 1701963 in NSCLC and CRC tumor models showed increased, sustained pharmacodynamic modulation and stronger anti-tumor activity. A phase 1 clinical trial is underway investigating BI 1823911 alone and in combination with BI 1701963 in patients with advanced/metastatic solid tumors (e.g., NSCLC, CRC, pancreatic cancer, cholangiocarcinoma), with the KRAS G12C mutation.

Clinical mechanisms of adaptive resistance to direct KRAS G12C inhibitors
Although mechanisms of resistance to direct KRAS G12C inhibition have been explored in depth using preclinical models [96,[122][123][124][125][126], the clinical mechanisms of resistance remain to be elucidated. Awad et al. have begun to characterize the clinical mechanisms of resistance to adagrasib with results from the KRYSTAL-1 trial (NCT03785249) [95]. In the study, the authors included patients with KRAS G12C mutant cancers who had disease progression while receiving adagrasib monotherapy. They defined acquired resistance to therapy as stable disease for at least 12 weeks or a partial or complete response followed by disease progression according to the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1. A total of 38 patients (27 NSCLC, 10 CRC, and 1 appendiceal cancer patients) were subsequently included in the study. Ultimately, putative mechanisms of resistance to adagrasib were detected in 17 patients (45%), of which 7 patients had multiple coincident mechanisms.
The authors categorized these putative resistance mechanisms into three main categories. First, secondary mutations or amplifications in KRAS were observed in patients. Specifically, acquired KRAS mutations were most notably found at codons 12, 13, 61, 68, 95, and 96; which included G12D/R/V/W, G13D, Q61H, R68S, H95D/Q/R, and Y96C. Codon 12 mutations prevented adagrasib binding, while the other secondary mutations occurred in the switch II drug-binding pocket.
Secondly, another resistance mechanism is alternative oncogenic alterations that activate the RTK-RAS signaling pathway but do not directly alter KRAS itself. These acquired bypass mechanisms of resistance included MET amplification; activating mutations in NRAS, BRAF, MAP 2 K1, and RET; oncogenic fusions involving ALK, RET, BRAF, RAF1, and FGFR3; and loss-of-function mutations in NF1 and PTEN. Thirdly, histologic transformation from lung adenocarcinoma to squamous-cell carcinoma was observed without identification of any other resistance mechanisms. Collectively, these data suggest that adagrasib treatments resulted in the evolution of a diverse set of adaptive mechanisms in KRAS G12C mutant cancers, instead of dominant adaptive mechanisms seen with other targeted therapies.

Conclusions
Thanks to the efforts of the Shokat lab in discovering the KRAS G12C switch II pocket in 2013, it appears that we are currently in a golden era of developing targeted therapy that can directly inhibit the KRAS G12C -driven human cancers. As of early August 2021, twenty registered clinical trials have been implemented to investigate nine different direct KRAS G12C inhibitors, which began with the first-in-human trial with sotorasib three years ago (Table 1). In addition to sotorasib (Lumakras ™ ) becoming the first FDA-approved drug to directly target tumors with any KRAS mutation in May 2021, adagrasib received breakthrough therapy designation from the FDA in June 2021, thus making adagrasib one step closer to receiving FDA approval as well.
However, as with other clinical trials involving targeted therapies, such as the MEK inhibitors in KRAS-mutant NSCLC patients [127][128][129][130][131][132][133] and the BRAF inhibitors in BRAF V600E mutant colorectal cancer patients [134][135][136], adaptive resistance to single-agent therapy with adagrasib and sotorasib eventually occurred in most patients. In the CodeBreaK100 phase 2 clinical trial with sotorasib, among the 126 patients with previously treated KRAS p.G12C-mutated NSCLC, 83 patients (65.9%) had discontinued treatment due to disease progression [108]. Analyzing results from the KRYSTAL-1 trial with adagrasib, Awad et al. initially characterized the clinical mechanisms of resistance to the direct KRAS G12C inhibitors [95]. Their study supports the advent of novel KRAS inhibitors with different modes of binding and allelic specificities to counter the diversity of on-target and off-target adaptive resistance mechanisms. Moreover, effective combination therapy appears to be necessary to overcome these adaptive mechanisms of resistance to direct