Towards Multiomics-Based Dissection of Anti-EGFR Sensitivity in Colorectal Cancer

Summary

Overexpression of the EGFR ligands amphiregulin (AREG)/epiregulin (EREG) may be a surrogate of EGFR dependency regardless of sidedness in metastatic colorectal cancer. High AREG/EREG may be coupled with negative hyper-selection (i.e., lack of genomic drivers of primary resistance beyond RAS and BRAF) to identify patients with right-sided tumors and potential sensitivity to EGFR blockade.

See related article by Williams et al., p. 4153

In this issue of Clinical Cancer Research, Williams and colleagues have externally validated the prognostic role of the two EGFR ligands amphiregulin (AREG) and epiregulin (EREG) expression levels in patients with RAS wild-type metastatic colorectal cancer (mCRC) receiving anti–EGFR-based upfront therapy (1). This research group should be congratulated for its continuous effort towards the clinical implementation of transcriptional biomarkers of resistance/sensitivity to EGFR inhibition in mCRC.

The efficacy of cetuximab and panitumumab in colorectal cancer is sustained by the blockade of the functional activation of EGFR by its ligands, irrespective of EGFR expression and in the context of a permissive genomic background that does not lead to the constitutive activation of EGFR downstream pathways (2). At the beginning of this research field, a post hoc analysis of the CO.17 trial with cetuximab versus best supportive care in chemo-refractory disease suggested the potential predictive role of EREG expression as assessed by RT-PCR (3). However, this study was limited by the availability of KRAS exon 2 status alone and the lack of knowledge on the predictive role of primary tumor location. Then, Seligmann and colleagues showed for the first time the predictive role of AREG and/or EREG overexpression by RT-PCR in the all-RAS wild-type subgroup of patients of the PICCOLO trial, that has compared irinotecan plus panitumumab versus irinotecan alone in the second-line setting (4). However, the large scale investigation of this promising biomarkers has been limited by: (i) the potential lack of reproducibility and clinical applicability of measuring and categorizing the mRNA expression levels of EGFR ligands; (ii) the increasing use of anti-EGFR therapy in the first-line setting thanks to the results of bevacizumab versus anti-EGFR head-to-head phase III trials (5, 6), in parallel with improved patients selection by means of all-RAS/BRAF status and primary tumor location (7, 8); (iii) evidence of significant association of low AREG and EREG expression with right-sidedness and BRAF mutations (4, 9), thus not allowing to clearly assess the individual impact of these biomarkers in a properly selected or stratified patient population. To overcome these limitations, the research group led by Prof. Seligmann continued to investigate and expand the knowledge on EGFR ligands as biomarkers and showed several crucial results. First, an artificial intelligence-assisted AREG/EREG IHC has showed internal consistency with the initially reported predictive analyses of the PICCOLO trial and external validity in this observational prospective study (1, 10). This assay has a clear impact on the potential large-scale use of AREG/EREG to guide patient selection. In addition, here the Authors were able to confirm the prognostic and likely predictive role of AREG and EREG in the first-line setting, where anti-EGFR agents are combined with chemotherapy. Interestingly, the results were significant and clinically meaningful also in the RAS and BRAF wild-type subgroup and, above all, regardless of primary tumor location.

A different line of research has been carried out by our group in parallel with the widespread use of comprehensive genomic profiling: the genomic-based negative hyper-selection. In fact, we demonstrated the negative prognostic impact of a panel of uncommon genomic alterations of resistance to EGFR inhibition beyond RAS and BRAF mutations (including HER2 amplification/mutations, MET amplification, gene fusions, PIK3CA exon 20/PTEN/AKT1 mutations), first by a case–control prospective study and then by a pre-specified analysis of the VALENTINO first-line trial (11, 12). Of note, these resistance alterations—except for HER2 overexpression/amplification—are significantly enriched in right-sided tumors (11, 12, 13). The predictive impact of negative molecular hyper-selection has been recently demonstrated by a translational analysis of the PARADIGM trial, a head-to-head, first-line trial of panitumumab- versus bevacizumab-FOLFOX therapy in patients with RAS wild-type mCRC (13). Interestingly, a significant interaction in terms of overall survival between genomic hyper-selection and the type of monoclonal antibody has been reported, with this trend retained irrespective of primary tumor location. As expected, a greater benefit from panitumumab-FOLFOX was seen in patients with left-sided and hyper-selected tumors, but it was also observed in those in the molecularly hyper-selected and right-sided subgroup.

Therefore, both AREG/EREG overexpression and molecular hyper-selection are emerging as the most promising biomarker-enrichment strategies to further refine the personalization of EGFR inhibition in mCRC. A crucial point is that AREG and EREG overexpression may be per se a surrogate of EGFR dependency, which is on the other hand predicted with relatively high accuracy by molecular hyper-selection or even ultra-selection (14). Obviously, in patients with microsatellite stable, HER2-negative, RAS/BRAF wild-type and left-sided tumors, the chance of EGFR dependency driven by AREG and/or EREG overexpression is extremely high. Therefore, the potential added value and clinical usefulness of AREG and EREG as selection biomarkers is less certain. In right-sided tumors, mesenchymal/metabolic/immune gene expression profiles are highly enriched, as well as uncommon genomic drivers of resistance, thus explaining the limited proportion of EGFR-dependent tumors (15). However, patients with right-sided and high AREG/EREG expression do exist and represent a relatively small molecular subgroup at risk of being neglected and who may still derive a benefit from the upfront use of anti–EGFR-based therapies. Considering the lack of randomization and the observational nature of the current study, the Authors have moved forward and have designed the ongoing ARIEL open-label, randomized clinical trial, that compares first-line chemotherapy alone versus chemotherapy plus an anti-EGFR agent in terms of early tumor shrinkage in patients with RAS wild-type, right-sided mCRC selected by centrally assessed AREG and EREG overexpression (16). This study will be perfect for the scientific community to provide a definitive answer to the potential use of EGFR inhibitors in patients with RAS/BRAF wild-type and right-sided tumors, also considering some discrepancies between the main international guidelines. In fact, according to the ASCO guidelines anti-EGFR therapy is not recommended in the first line setting in RAS wild-type and right-sided tumors (17). On the other hand, according to the ESMO guidelines doublet chemotherapy plus an anti-EGFR agent may still be an option when tumor shrinkage is needed, despite the strength of the recommendation of this option being lower than chemotherapy plus bevacizumab (18). However, the selection of patients with right-sided tumors based on additional biomarkers rather than the treatment intent should be the goal of ongoing research (Fig. 1). We believe that the ARIEL trial may represent the ideal translational platform to further validate the optimal cutoff of AREG and EREG overexpression in right-sided tumors and to validate the negative predictive role of molecular hyper-selection in this patients’ population.

Figure 1.

Graphical representation of the mechanisms of EGFR dependency in colorectal cancer. (Adapted from an image created with BioRender.com.)

In conclusion, future efforts should be directed towards the combined use of EGFR ligands expression by means of artificial intelligence-assisted assays and negative hyper-selection by means of comprehensive genomic profiling. In the era of precision medicine, multi-omics assessment of both transcriptional and genomic profiles, as well as the potential radiomic/radiogenomic and digital pathology features, may lead to a more accurate prediction of the individual sensitivity to EGFR inhibition. In parallel, every effort to conduct randomized clinical trials in patients with poor outcomes such as those with right-sided mCRC should be welcomed with gratitude from the scientific community.

Authors' Disclosures

F. Pietrantonio reports grants and personal fees from Amgen and BMS; personal fees from Merck-Serono, MSD, Takeda, Servier, Pierre-Fabre, Bayer, and Astellas; and grants from Agenus, AstraZeneca, and Incyte outside the submitted work. No disclosures were reported by the other author.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.