Mechanisms of Resistance to KRAS^G12C-Targeted Therapy

The KRAS gene is the prototypical oncogene and is among the most frequently mutated in human cancer. Indeed, three of the five leading causes of cancer-related death in the United States, non–small cell lung cancer (NSCLC), colorectal cancer, and pancreatic cancer (PDAC), are also among the most frequently associated with KRAS mutations, at rates of approximately 30%, 42%, and 80%, respectively (1–3). In addition, KRAS mutations, when present, are associated with poorer prognosis from these cancers than non-KRAS oncogenic drivers (4–6).

Nearly all oncogenic mutations in KRAS are activating missense mutations that center on three codons: 12, 13, and 61. Of these, missense mutations in the glycine residue at codon 12 are far and away the most common (7). Interestingly, tissue of origin predicts the likely KRAS missense mutation, with KRAS^G12D representing approximately 25% to 40% of all KRAS mutations in colorectal cancer and PDAC, and, in contrast, KRAS^G12C representing approximately 40% of all KRAS mutations in lung adenocarcinomas (1, 3, 7).

As the primary intracellular secondary messenger of EGFR, KRAS constantly cycles between the inactive GDP-bound and the active GTP-bound states (7, 8). When activated by a ligand-bound receptor tyrosine kinase (RTK) such as EGFR, KRAS triggers multiple proliferative signaling cascades, including the MAPK/ERK and PI3K pathways, to induce cell growth, division, and differentiation (9–11). In cases of oncogenic activating KRAS mutations, GTP hydrolysis is impaired and the KRAS protein is preferentially held in the active GTP-bound state (7, 12), which, in turn, drives proliferative MAPK/PI3K signaling and ultimately carcinogenesis, as classically modeled by Vogelstein and Fearon (7, 12). We note that although KRAS was historically considered to be a necessary stage in the adenoma–carcinoma sequence in colorectal cancer, this concept has been revolutionized with the identification of microsatellite instability (13) to now encompass tissue-, mutation-, and pathway-specific mechanisms of carcinogenesis.

Unfortunately, targeting mutated KRAS has been unsuccessful despite 40 years of sustained research and development, with generally limited response rates and short durations of response (14, 15). Accordingly, KRAS had long been considered “undruggable” due to high affinity for GTP and lack of large binding pockets for allosteric inhibitors to occupy (14, 15). However, the groundbreaking discovery by Shokat and colleagues of small molecules that covalently bind to the acquired cysteine residue within the switch II region in KRAS^G12C laid the first viable foundational steps to therapeutic KRAS blockade (16). Although these molecules selectively target KRAS^G12C alone, their potential impact is significant across many common cancer types, including approximately 12% of all lung adenocarcinomas and 3% of all colorectal cancers. Furthermore, these agents provide an option to patients for whom there had been a lack of targeted treatments.

Recently published and ongoing early-phase clinical trials of the KRAS^G12C inhibitors sotorasib (AMG 510) and adagrasib (MRTX849) demonstrate clear clinical benefit, with tumor response rates approximating 30% to 40% with little toxicity (17, 18). However, the duration of response for most patients is short, with the most recent data from the CodeBreaK100 trial showing median progression-free survival of only 6.3 months (17). Now, in this new era of targeting KRAS^G12C, the next research obstacle will be to understand and overcome mechanisms of resistance. Thus, we will review the current preclinical understanding of resistance to KRAS^G12C-targeted therapies, and present possible treatment approaches to combat such resistance.

KRAS is a small, 21-kDa monometric guanosine 5′-triphosphatase. It consists of six beta strands and five alpha helices with a G-domain and a C-terminal membrane-targeting region (9, 10). Wild-type KRAS constantly cycles between active GTP-bound and inactive GDP-bound states depending upon stimuli from upstream RTKs, most importantly EGFR. Upon activation, KRAS interacts with a complex set of downstream effectors in intricate and, in many cases, redundant pathways (ref. 7; Fig. 1). Notable interactions include those with proliferation-associated pathways such as RAF–MEK–ERK and PI3K–AKT–mTOR, which reinforce cyclin/cyclin-dependent kinase (CDK)–dependent RB phosphorylation, drive cellular differentiation and growth, and oppose apoptosis (7, 19). The clinical relevance of KRAS signaling to human cancer is highlighted not only by frequent mutations in KRAS, but also by frequent (targetable) alterations in nearly every other downstream protein across multiple cancer types. Unfortunately, the redundancy in KRAS signaling and its centrality to cancer development foreshadows secondary resistance to KRAS blockade, as was seen with EGFR-mutated lung adenocarcinoma treated with osimertinib (20), and with BRAF-mutated melanoma and colorectal cancer treated with a combination of targeted BRAF inhibitors (21–23).

The oncogenicity of various KRAS mutations, including KRAS^G12C, arises from chronic KRAS activation due to reduced GTPase activity and prolonged residence in the GTP-bound active state. However, the specific targetability of KRAS^G12C relies upon the placement of the acquired cysteine within the P2 pocket of the switch II region. The resulting protein conformation with this specific missense mutation is accessible to small molecules that covalently bind the cysteine residue and hold KRAS^G12C in the inactive GDP-bound state, irreversibly switching off downstream signaling and inducing apoptotic cell death (16, 24, 25). Multiple small molecules have been developed against KRAS^G12C, including ARS-1620, sotorasib (AMG 510), and adagrasib (MRTX849). ARS-1620, the first KRAS^G12C inhibitor, has little clinical activity, but remains an important translational research tool to study mechanisms of resistance (8).

In contrast, sotorasib, developed by Amgen, and adagrasib, developed by Mirati, were the first and second KRAS^G12C inhibitors to reach the clinic, with recently completed or ongoing phase I clinical trials (17, 18, 26). Both agents were demonstrated in vitro to covalently bind the acquired cysteine within the switch II region and inhibit downstream MAPK signaling, as evidenced by diminished phosphorylation of ERK (pERK), S6 (pS6), and, in the case of sotorasib, MEK (pMEK). In addition, both drugs diminished the viability of KRAS^G12C human cancer cell lines, including, and most critically, both lung and pancreatic cancer cell lines. Notably, a non-KRAS^G12C mutation was insensitive to treatment, as evidenced by lack of effect on pERK or on cell viability, highlighting the specificity of these inhibitors to KRAS^G12C. When tested in vivo in murine models, both agents inhibited downstream MAPK effectors and shrank tumors (18, 26).

Finally, the recently completed phase I CodeBreaK 100 trial investigated the initial safety and efficacy profile of sotorasib in patients with locally advanced or metastatic KRAS^G12C NSCLC (n = 59), colorectal cancer (n = 42), and other solid tumors. This trial, which enrolled patients who progressed after at least one line of systemic therapy, but excluded those with active brain metastases, showed that sotorasib is well tolerated, with no dose-limiting toxicities or grade 4 therapy-related adverse events, and also efficacious, with overall response rate (ORR), disease control rate (DCR), and median progression free survival (mPFS) of 32.2%, 88.1%, and 6.3 months, respectively, in NSCLC, and 7.1%, 73.8%, and 4.0 months, respectively, in colorectal cancer (17). Remarkably, this result was achieved even in heavily pretreated patients, with nearly all patients with NSCLC having previously progressed on both platinum-doublet and anti–PD-1/anti–PD-L1 immunotherapy and patients with colorectal cancer having previously failed at least two lines of systemic therapy. A phase I/II clinical trial (NCT03785249) of adagrasib is ongoing; however, preliminary results presented at the EORTC-NCI-AACR Annual Symposium in 2020 similarly demonstrated little drug toxicity and ORR and DCR of 45% and 96%, respectively, in NSCLC, and 17% and 96%, respectively, in colorectal cancer.

Collectively, these data suggest that targeting KRAS^G12C is efficacious and well tolerated, and have prompted the development of multiple new KRAS^G12C agents, as summarized in Table 1. Nonetheless, these clinical responses, although significant and exciting relative to historical attempts at KRAS targeting, are highly variable between different tumor types, with markedly different overall response rate between NSCLC and colorectal cancer. In addition, no patient in any study achieved complete response, and the observed clinical responses are not durable, lasting only 4 to 6 months for most patients. Thus, resistance to treatment is evident and necessitates further investigation to guide future treatment approaches.

Preclinical studies have hinted at multiple possible mechanisms of resistance including innate, acquired, and adaptive tumor responses that diminish the therapeutic efficacy of KRAS^G12C inhibitors. One frequently identified mechanism is induction of bypass MAPK signaling to overcome KRAS blockade. Indeed, multiple studies have now demonstrated that KRAS^G12C inhibition can be overcome via feedback activation of either upstream or downstream mediators of the RTK–KRAS–MAPK cascade, as was observed with selective targeting of BRAF and EGFR (8, 18, 26, 27).

One of the first suggestions of this bypass signaling was made by Hallin and colleagues in initial studies of adagrasib (MRTX849; ref. 18), which demonstrated that cell line–derived mouse xenografts can be highly sensitive (i.e., PaCa-2, H1373), partially sensitive (i.e., H358, H2122), or refractory to KRAS^G12C inhibition. Gene set enrichment analysis on all tumor models regardless of response demonstrated that the most differentially expressed genes encompassed those mediating KRAS signaling, including MYC and MTOR, confirming that the drug specifically and efficiently inhibits KRAS^G12C and its downstream effectors. However, RNA sequencing of MAPK feedback pathways revealed that KRAS^G12C inhibition also elicits significant suppression of DUSP, SPRY, and PHLDA family genes, which are known negative regulators of MAPK signaling (28). Indeed, this finding was further corroborated by IHC for pERK and pS6, which significantly diminished (by >90%) in both highly sensitive (PaCa-2 and H1373) and partially sensitive (H358 and H2122) tumor models soon after exposure to adagrasib, only to subsequently recover in the latter, but not in the former, even after five days of continuous treatment. Together, these experiments demonstrate that ERK-dependent signaling is reactivated to bypass KRAS^G12C treatment. A survey of in vitro and in vivo models using a CRISPR/Cas9 knockout screen with short guide RNAs targeting approximately 400 genes revealed that in the partially sensitive H2122 xenograft model, guide RNAs targeting SHP2 (a phosphatase that mediates signaling between activated RTK and KRAS), MYC, and MTOR pathway genes, all mediators of the RTK–KRAS–MAPK/PI3K cascade, were among the most depleted after two weeks of exposure to MRTX849, whereas guide RNAs targeting KEAP1, a tumor suppressor, were notably enriched.

Further clarification of resistance mechanism was provided by Xue and colleagues (8), who hypothesized that as novel KRAS^G12C inhibitors solely inhibit the inactive GDP-bound conformation of KRAS^G12C, only those cells with KRAS^G12C in the inactive conformation would be strongly inhibited in any population of cells with nonuniform rates of inactive to active KRAS^G12C cycling. As such, those cells with KRAS^G12C preferentially held in the active conformation would be insensitive to treatment and could mediate reactivation of MAPK signaling (8). In human lung adenocarcinoma cell lines previously shown to be either partially sensitive or refractory to adagrasib, namely H358, H2122, and SW1573, ARS-1620 was found to induce a quiescent (G₀) state in most, but not all, cells, as defined by abundant expression of p27 and as analyzed by single-cell RNA sequencing. However, those cells with low-level expression of p27 do not become quiescent and express active GTP-bound KRAS more abundantly, and are not eliminated by rechallenge with ARS-1620. Differential expression analysis and genome-wide knockout screening subsequently revealed two candidate genes that can mediate escape from KRAS^G12C inhibition: heparin-binding EGF (HBEGF) and aurora kinase (AURKA). Specifically, HBEGF is downregulated soon after exposure to ARS-1620, but is then rapidly upregulated after 48 hours within a subpopulation of quiescent cells, suggesting a role in adaptive resistance. Corroborating this, small interfering RNA (siRNA) knocking down HBEGF augment the antiproliferative effect of ARS-1620. Conversely, stimulation with EGF induces KRAS activation in quiescent, ARS-1620–treated cells, strongly suggesting that EGFR signaling mediates adaptive resistance to KRAS^G12C drugs. On the other hand, AURKA accumulates in adapting cells as opposed to quiescent ARS-1620–treated cells, suggesting a relationship with overcoming quiescence. Alternatively, induction of AURKA in ARS-1620–treated H358 cells elicits accumulation of KRAS-GTP and pERK, and lowers the potency of ARS-1620 as assessed by cell viability.

In an elegant experiment, quiescent/p27-expressing H358 cells were then engineered to inducibly express siRNA-resistant KRAS^G12C. To mimic the initial quiescence phase following exposure to ARS-1620, cells were treated with a KRAS^G12C-targeted siRNA, but were then induced to express siRNA-resistant KRAS^G12C to mimic the adaptive phase. As seen with cells exposed to ARS-1620, these cells became initially quiescent following exposure to siRNA targeting KRAS^G12C, but a subpopulation induced to express siRNA-resistant KRAS^G12C subsequently escaped this state. Accordingly, the adaptive response to KRAS^G12C inhibition was hypothesized to arise from newly synthesized KRAS^G12C that undergo immediate nucleotide change to an active GTP-bound conformation before being trapped by KRAS^G12C inhibitors, with EGF being the likely driver of new KRAS transcription and AURKA maintaining KRAS in the active GTP-bound conformation.

Moreover, Ryan and colleagues recently observed that reactivation of MAPK signaling after treatment with ARS-1620 coincided with increased expression of wild-type GTP-bound RAS (i.e., HRAS, NRAS) and phosphorylated RTK (i.e., EGFR, FGFR, HER2, c-MET) in KRAS^G12C-driven lung, pancreatic, and colon cancer cells, suggesting secondary resistance via upregulated RTK signaling to wild-type RAS isoforms (29). Notably, the RTK specifically activated by ARS-1620 was not the same across cell lines, which suggests that different RTKs may drive MAPK reactivation, and these differences may be histology-specific or even tumor-specific. Indeed, tissue of origin predicts responsiveness of KRAS^G12C inhibition, as seen in the divergent clinical responses to sotorasib between lung adenocarcinoma and all other advanced cancer types, likely via differences in adaptive resistance. In fact, Amodio and colleagues demonstrated that although both NSCLC and colorectal cancer cells treated with sotorasib exhibit equivalent reduction in cell viability, the latter show rapid upregulation of pMEK and pERK, suggestive of early development of adaptive resistance (30). Furthermore, colorectal cancer cells show increased basal phosphorylation (activation) of EGFR and respond to EGF stimulation by activating RAS–MAPK signaling even in the presence of an activating KRAS^G12C mutation, behaviors not seen in NSCLC cells. These findings strongly suggest that EGFR specifically mediates the adaptive resistance response in colorectal cancer, as previously observed in BRAF-mutant colorectal cancer (31), and could explain the poor ORR to single-agent KRAS^G12C inhibition.

Finally, Adachi and colleagues observed that among cell lines previously sensitive to sotorasib (i.e., H358), induction of epithelial-to-mesenchymal transition (EMT), either by treatment with TGFβ or conditional expression of Twist or Snail, was associated with intrinsic and acquired resistance to KRAS^G12C inhibition (32). Resistance via EMT occurred in conjunction with increased PI3K/AKT signaling due to upregulated insulin growth factor receptor (IGF1R) signaling, and led to increased MAPK signaling via FGFR.

Taken together, the data indicate that upstream RTK regulators (EGFR, HER2, FGFR, and SHP2), direct mediators of KRAS activation (AURKA), and/or effectors of MAPK and PI3K pathways (MYC and MTOR) may mediate escape from KRAS^G12C inhibition, with escape mechanisms being notably tissue-specific. Fortunately, many of these resistance mediators can be targeted with therapeutic agents already on the market or in development, enabling rapid preclinical testing and now clinical translation. Nevertheless, rational, tissue-specific combination therapies are necessary to provide precise and effective disease control, in light of tissue-specific resistance mechanisms.

Among the most clinically studied upstream targets for combination therapy is EGFR. Currently, small-molecule inhibitors targeting mutationally activated EGFR are standard-of-care and have had long-standing clinical success against EGFR-mutant lung adenocarcinoma (33, 34). Similarly, anti-EGFR mAbs (e.g., cetuximab; ref. 35) have been highly effective against colorectal cancer in combination with fluorouracil-based chemotherapy. In addition, following the determination that EGFR mediates resistance to BRAF inhibition in BRAF-mutant colorectal cancer, cetuximab was found to synergize with BRAF inhibitors to prolong survival relative to standard therapy, and is now standard second-line therapy (31, 36). Consequently, combining EGFR-targeted agents (i.e., gefitinib, afatinib) with both adagrasib and ARS-1620 was found to reduce downstream MAPK signaling and tumor volume in two mouse xenograft models of KRAS^G12C (8, 18). Similarly, addition of cetuximab to sotorasib blocked adaptive EGFR-driven reactivation of MAPK signaling specifically in colorectal cancer cell lines and mouse xenograft models, and led to sustained suppression of pERK and pMEK, decreased cell viability, and near-complete tumor regression.

On the other hand, there are indications that SHP2 inhibitors, which have very limited activity as single agents against KRAS-mutated cell lines, can restore the sensitivity of KRAS-mutant NSCLC to MEK inhibition, and thereby inhibit tumor growth (37). Indeed, coadministration of SHP2 inhibitors with ARS-1620 was found to diminish adaptive reactivation of GTP-bound KRAS^G12C in mouse xenografts, an effect further augmented in a triplet combination of KRAS^G12C, EGFR, and SHP2 inhibitors (8). As SHP2 mediates signaling between RTKs and RAS, coadministration of SHP2 inhibitors with ARS-1620 was also found to decrease RTK-mediated MAPK reactivation independent of the RTK (29). Finally, SHP2 inhibitors were found to increase inactive GDP-bound KRAS and, in combination with ARS-1620, to induce suppression of pERK and increase T-cell infiltration, eliciting tumor regression in mouse models of PDAC and NSCLC (38). Finally, a combination of SHP2, PI3K, and KRAS^G12C inhibitors was found to suppress both pAKT and pERK and induce durable tumor regression in EMT-induced mouse xenografts, which exhibit hallmarks of FGFR- and IGF1R-induced MAPK and PI3K reactivation, respectively (32). Accordingly, early-phase clinical trials are ongoing to test KRAS^G12C inhibitors in combination with either EGFR inhibitors, EGFR mAbs, or SHP2 inhibitors (Fig. 1; Table 2). Finally, BI 1701963, developed by Boehringer Ingelheim, is a distinct therapeutic class that acts as a pan-KRAS inhibitor by preventing SOS1 from binding inactive GDP-bound KRAS, thus inhibiting exchange of GDP to GTP (39, 40) and indirectly inactivating all forms of KRAS. Early-phase clinical trials of this drug are ongoing, alone or in combination with trametinib, in patients with any KRAS mutation (Fig. 1; Table 2).

In addition, combining KRAS^G12C inhibitors with inhibitors of multiple downstream mediators has been tested preclinically, considering that downstream effectors of both the MAPK (RAF–MEK–ERK) and PI3K–AKT–mTOR pathways are clearly reactivated following exposure to KRAS^G12C drugs. For example, MEK inhibitors, although of limited utility alone, have been combined with BRAF inhibitors with great success against melanoma, and are now standard-of-care for these tumors (41). In addition, MEK inhibitors notably enhanced the potency of chemotherapy in mouse models of lung cancer, particularly of tumors with KRAS^G12C (42). Finally, Canon and colleagues noted synergy between sotorasib and MEK inhibitors to reduce tumor volume in H358 mouse xenografts. On the basis of these findings, a clinical trial of sotorasib in combination with a MEK inhibitor is ongoing (26). Similarly, combining KRAS^G12C inhibitors with either PI3K (27, 32) or mTOR (18) inhibitors overcame the adaptive increase in PI3K signaling, increased inhibition of MAPK/PI3K signaling, and reduced tumor volume in mouse xenografts. Early-phase clinical trials of KRAS^G12C inhibitors and mTOR inhibitors are ongoing (Fig. 1; Table 2). On the other hand, there are no clinical trials combining KRAS^G12C inhibitors with AURKA inhibitors at this time, although AURKA has been shown to mediate resistance to KRAS^G12C inhibitors and remains a valid target for future combination therapy.

In addition to adaptive reactivation of MAPK signaling, increased proliferative signaling via disinhibition of the cell-cycle transition is another source of KRAS^G12C therapy resistance, particularly in NSCLC. Indeed, up to 20% of KRAS-mutant NSCLCs have concurrent loss-of-function mutations in CDKN2A, a cell-cycle regulator and tumor suppressor, which, in turn, leads to constitutive CDK4/6-associated RB phosphorylation and cell proliferation (43–45). Furthermore, previous reports by Puyol and colleagues suggest that interphase CDKs, particularly CDK4, are necessary for lung tumor development in conditional KRAS^G12V mouse models, such that CDK4 inactivation, by either conditional knockout or a null allele, led to reduced tumor development and induction of senescence, as defined by expression of β-galactosidase (19). Notably, CDK4 inactivation induced senescence only in the lung, as β-galactosidase was not detected in other tissues, including colon, pancreas, and stomach, suggesting a tissue-specific role. Finally, treatment with a CDK4 inhibitor led to decreased pRB and tumor volume.

In the context of KRAS^G12C, the combination of adagrasib and palbociclib, a CDK4/6 inhibitor, showed significant synergy as evidenced by p27 accumulation, decreased pRB, and marked decrease in tumor volume in CDKN2A-deficient xenograft models (18). Similarly, combining sotorasib with carboplatin, a commonly used frontline agent in NSCLC, shrank tumors in a mouse xenograft model (26). Accordingly, there exists significant translational potential for combining KRAS^G12C inhibitors with either cytotoxic chemotherapy (particularly in NSCLC) or inhibitors of interphase CDKs, as in ongoing clinical trials (Fig. 1; Table 2).

Finally, a third mechanism of KRAS^G12C therapy resistance is impaired antitumor immunity. In light of the growing use of immune checkpoint therapy across the cancer landscape, Canon and colleagues explored the impact of KRAS^G12C inhibitor therapy on antitumor immunity (26). Interestingly, sotorasib was able to induce durable cures against CT26 KRAS^G12C cells injected into immunocompetent mice. In sharp contrast, sotorasib induced only short-lived tumor regression followed by recurrence in nearly all immunodeficient BALB/c mice xenografted with the same cells, suggesting that an impaired host immune system may confer resistance independent of MAPK reactivation or proliferative signaling. Furthermore, treatment with KRAS^G12C inhibitors appeared to induce immune response to tumorigenic tissue, with sotorasib inducing marked infiltration of CD8⁺ T cells, macrophages, and dendritic cells into CT26 KRAS^G12C tumors after five days of treatment. Gene expression analysis also revealed increased expression of genetic signatures of IFN signaling, chemokine production, and antigen processing, suggesting that KRAS^G12C inhibition boosts T-cell priming. Moreover, combining sotorasib with anti–PD-1 therapy augmented T-cell infiltration and led to complete and durable remissions. Finally, mice treated with this combination and cured of xenografted CT26 KRAS^G12C rejected a subsequent rechallenge with CT26 KRAS^G12C and parental CT26, but not 4T1 mouse breast cancer cells. Analysis of splenocytes from rechallenged mice demonstrated marked increase in IFNγ, a marker of T-cell priming, in the presence of CT26 tumor cells but not in the presence of 4T1 tumor cells (26).

Thus, KRAS^G12C inhibition appears to induce a proinflammatory transcriptional signature that primes antigen-presenting cells and cytotoxic T cells, which, in turn, have antitumor activity. This process can be prolonged and durable when PD-1 checkpoint is inhibited. However, co-occurring genetic alterations may modulate the immune response to tumors. For example, mutations of KEAP1 and STK11 appear to induce a colder immune microenvironment with decreased T-cell infiltration, and are associated with poor clinical outcome in NSCLC treated with first-line chemo-immunotherapy (46). In contrast, co-occurring mutations in TP53 are associated with increased intratumoral T-cell infiltration, PD-1 expression, and prolonged clinical benefit from anti–PD-1 immunotherapy in NSCLC (47). Hence, ongoing clinical trials are pursuing combinations of KRAS^G12C inhibitors and immunotherapy, particularly for advanced NSCLC where immunotherapy has already shown significant efficacy (Fig. 1; Table 2).

Innate and acquired resistance to KRAS^G12C inhibitors has impeded their development and remains an obstacle to their long-term success, as seen with other targeted therapies. For example, first-line osimertinib elicits nearly 80% ORR against EGFR-mutant metastatic lung adenocarcinoma, but patients invariably experience eventual recurrence (34). The mechanisms of recurrence are multifold and heterogeneous, encompassing both EGFR-dependent and EGFR-independent mechanisms. These mechanisms include compensatory MET amplification, activation of MAPK signaling, and even transformation to small-cell or squamous cell histology (20).

Similarly, possible resistance mechanisms to singular inhibition of KRAS^G12C appear to be diverse, as investigated preclinically, with primary drivers consisting of reactivation of multiple MAPK effectors both upstream and downstream of RAS, disinhibition of cell-cycle transition, and defects in immunity. Importantly, these resistance mechanisms appear to be tissue-specific, with colorectal cancer developing resistance primarily via activation of upstream EGFR, and NSCLC deploying all three mechanisms, depending upon the presence of co-occurring alterations in CDKN2A, STK11, and TP53. Thus, clinical trial design warrants an understanding of these tissue-specific differences in escape from KRAS^G12C blockade. Multiple mechanism-driven clinical trials combining KRAS^G12C inhibitors with a wide array of resistance mediators are now active and recruiting. It remains to be seen how well these combination therapies will be clinically tolerated. However, these collectively represent a leap forward to combat therapy resistance.

With the discovery of the switch II region in KRAS^G12C, targeted agents are finally being translated to the clinic, to the benefit of many patients with KRAS mutations. Unfortunately, durable response to these novel agents is yet to be achieved due to complex and diverse mechanisms of adaptive resistance. However, there is now great promise from combination therapies, which are based on an improved understanding of resistance mediators, to elicit long-term disease control or remission.

D.S. Hong reports grants and personal fees from Amgen, Adaptimmune, Genentech, Infinity, Pfizer, Seattle Genetics, and Takeda; grants from AbbVie, Aldi-Norte, Astra-Zeneca, BMS, Daiichi-Sankyo, Eisai, Fate Therapeutics, Ignyta, Kite, Kyowa, Lilly, Merck, Medimmune, Molecular Templates, Mologen, Navier, NCI-CTEP, Novartis, Numab, Turning Point Therapeutics, Vernstatem, and VM Oncology, and grants from Mirati during the conduct of the study; grants, personal fees, and other from Bayer; grants and other from Genmab and LOXO; other from miRNA Therapeutics, AACR, ASCO, SITC, Molecular Match, and OncoResponse, and other from Presagia outside the submitted work; personal fees from Alpha Insights, Acuta, Axiom, Baxter, Boxer Capital, COG, Ecor1, GLG, Group H, Guidepoint, HCW Precision, Janssen, Merrimack, Medscape, Numab, Prime Oncology, ST Cube, Tavistock, Trieza Therapeutics, and WebMD. No disclosures were reported by the other authors.

We acknowledge support from an NIH Cancer Center Support Grant (CA016672; awarded to P. Pisters).