Same Name, Different Game: EGFR Drives Intrinsic KRAS^G12C Inhibitor Resistance in Colorectal Cancer

Since their discovery nearly 40 years ago, oncogenic RAS mutations have been a tantalizing drug target that escaped decades of intense efforts at direct inhibition. The recent establishment of KRAS^G12C-specific inhibitors has helped to revitalize the promise of RAS inhibition, and early results of the AMG510 compound in patients with non–small cell lung cancer (NSCLC) has provided hope for efficacy, showing a 48% response rate as reported at the European Society for Medical Oncology 2019 meeting. In contrast, patients with colorectal cancer have shown a meager 3% response rate to the same drug, suggesting tissue-specific applicability of KRAS^G12C inhibition and leaving open the door for effective combination therapies.

The preliminary success of AMG510 in NSCLC comes after a long history of unsuccessful RAS regulation- or pathway-directed targeted therapy trials. One of the notable earlier efforts in the 2000s to inhibit RAS regulation was farnesyltransferase inhibitors, which were designed to prevent RAS membrane localization and subsequent activation; however, this drug class showed no benefit in overall survival versus standard supportive therapy (1). Subsequent research efforts have largely focused on targeting RAS downstream targets in the MAPK (MEK, ERK, pan-RAF, SHP2, etc.) and/or PI3K cascades (2). However, clinically tested agents in these classes have largely shown low efficacy as single agents—likely due to incomplete RAS pathway inhibition, toxicity limitations, and/or feedback pathway activation—and acute toxicity and/or insufficient efficacy in various drug combinations. Consequently, the vast majority of KRAS-mutant cancers including colorectal cancer continue to lack approved RAS pathway inhibitors, and chemotherapy or other non–RAS-related therapies largely remain the standard of care. Given the shortcomings of downstream RAS inhibitors, parallel research on direct RAS inhibition has continued to hold the promise of more comprehensive inhibition of its downstream activity. The major obstacles have been the extremely high affinity of RAS for GTP/GDP and, before 2013, the absence of known allosteric sites for small-molecule targeting.

Reflecting this early uncertainty about viable sites for inhibition, in 2011, KRAS structural data was limited to 9 structures on the Protein Data Bank, whereas 150 were available as of December 2019. Indeed, structural analysis played a key role in the discovery of KRAS^G12C inhibitors in 2013 by a group led by Kevin Shokat. This study leveraged the unique nucleophilicity of the cysteine residue in the KRAS^G12C mutant to identify cysteine-reactive small molecules that modify KRAS^G12C (3). An initial hit from a compound screen led to the generation of an acrylamide analogue that showed promising activity in G12C-mutant lung cancer cell lines. Structural analysis found that this irreversible inhibitor bound to a novel allosteric site called the switch-II pocket, located between the central β-sheet and the switch-II and α3-helices. The induced allosteric changes blocked both the binding of RAS to effector proteins and the nucleotide binding pocket. Since then, multiple KRAS^G12C inhibitors have been developed, among which AMG510 (4) was one of the first to enter clinical trials. Expectedly, these compounds have no activity against other KRAS mutants such as the more common G12V and G12D, yet the relatively high prevalence of G12C mutants in lung cancer led it to be a top target cancer type for G12C inhibitors. The surprising clinical disparity in NSCLC and colorectal cancer response rates indicates that tissue-specific molecular differences might underlie the mechanism of the intrinsic resistance in colorectal cancer.

In this issue of Cancer Discovery, Amodio and colleagues (5) identify a higher basal EGFR activity level in colorectal cancer versus NSCLC as bypassing KRAS^G12C inhibition (Fig. 1), manifesting as a rapid rebound in MAPK pathway activity. The discovery began with a non–phospho-RTK analysis in patient samples, but this did not yield a candidate resistance mechanism; instead, a phospho-RTK array in cell lines identified basal pEGFR as prominently higher in colorectal cancer than in NSCLC cells. Stimulation with the EGFR ligand EGF increased pEGFR across all cell lines, but activation of AKT, MEK, and ERK occurred in only colorectal cancer and not NSCLC cells, possibly due to higher basal activation levels in NSCLC. This suggests that colorectal cancer cells are more primed to activate downstream effectors in response to EGFR activation. Concomitant targeted inhibition of EGFR (cetuximab) and KRAS^G12C led to sustained RAS effector inactivation and significant in vitro activity. Impressively, in vivo administration led to frank tumor regression in two independent PDX models, including a 73-day sustained response in one of the models, compared to no regression by either single agent.

As noted by the team, the results broadly mirror the seminal discovery of higher basal EGFR activity as underlying the intrinsic resistance of BRAF-mutant colorectal cancer to BRAF inhibitors (3%–5% clinical response rate) when compared with sensitive BRAF-mutant melanomas (50%–70% response rate; refs. 6, 7). A casual observation might suggest a foregone conclusion of the current study in light of this knowledge, but in fact it was not obvious that the mechanisms would be similar, as BRAF- and KRAS-mutant colorectal cancer are extremely different from each other clinically, pathologically, and molecularly. BRAF-mutant colorectal cancer is nearly always characterized by microsatellite instability (MSI) and a CpG island methylator phenotype (CIMP), whereas KRAS-mutant colorectal cancer is nearly always microsatellite-stable (MSS) and CIMP-negative (Fig. 1). Transcriptome-wide gene-expression differences are almost entirely sufficient to delineate MSI and MSS colorectal cancer, to the point where they are nearly as different from each other as they are from stomach cancers (8). Thus, the potential existence of differential EGFR activity and its effect on highly differential global transcriptomic signatures would not necessarily be predicted to straightforwardly map from MSI onto MSS. The finding that the mechanism is, in fact, the same speaks to the power of EGFR signaling and perhaps lineage-specific factors in determining molecular and clinical cancer behavior.

In a recent trial, triple encorafenib (BRAF), binimetinib (MEK), and cetuximab (EGFR) administration in BRAF-mutant colorectal cancer resulted in a 26% response rate (9), which, although an encouraging improvement, is perhaps somewhat modest compared with expectations from preclinical data. This suggests that cautious optimism should be applied to KRAS^G12C plus EGFR inhibition as well, as it remains to be seen how robust the therapeutic window of the combination is in humans. Even so, this study represents a solid move forward against KRAS-mutant colorectal cancer, which has been therapeutically refractory for decades and in desperate need of novel interventions. Moreover, the recent discovery of AMG510 driving antitumor immunity hints at potential trials combining KRAS^G12C inhibition and immune checkpoint blockade (10); this study provides the rationale for adding EGFR inhibition to this combination, especially as immune checkpoint inhibitors alone are largely ineffective in MSS colorectal cancer, due in part to low neoantigen counts compared with MSI colorectal cancer. Finally, the findings raise the question of whether all KRAS-mutant colorectal cancers, not just G12C, also exhibit high EGFR activity levels, possibly explaining the particular failure of downstream-targeted drugs in this disease. Overall, the fact that decades of frustrating direct RAS inhibition research are finally starting to pay dividends in ever finer details suggests that the wider journey toward pan–mutant RAS inhibition is all the more worthwhile.

No potential conflicts of interest were disclosed.