RAS Amplification as a Negative Predictor of Benefit from Anti-EGFR–Containing Therapy Regimens in Metastatic Colorectal Cancer

Abstract

Background

RAS short variant (SV) mutations in colorectal cancer (CRC) are associated with lack of benefit from epidermal growth factor receptor (EGFR) monoclonal antibody (EGFRmAb). However, the clinical implications for RAS amplification (RASa) as a biomarker for anti-EGFR therapy in CRC remain ill defined.

Methods

Genomic analysis was performed using the Foundation Medicine (FM) comprehensive genomic profiling database of 37,233 CRC cases. Clinical outcomes were assessed using two independent cohorts: the City of Hope (COH) cohort of 338 patients with metastatic CRC (mCRC) and the Flatiron Health–FM real-world clinicogenomic database (CGDB) of 3,904 patients with mCRC.

Results

RASa was detected in 1.6% (614/37,233) of primarily mCRC. RASa 6–9 (n = 241, 39%), 10–19 (n = 165, 27%), and ≥ 20 (n = 209, 34%) copy number subsets had co-RAS SV/BRAF V600E in 63%/3%, 31%/0.6%, and 4.8%/0% of cases, respectively. In the COH cohort, six patients with RASa (13–54 copies) received EGFRmAb, four of six had progressive disease, two had stable disease, and median time to treatment discontinuation (TTD) was 2.5 months. Of the CGDB EGFRmAb-treated patients, those with RASa (n = 9) had median TTD of 4.7 months and overall survival (OS) of 11.4 months, those with RAS SV (n = 101) had median TTD and OS of 5.3 and 9.4 months, and those with RAS/BRAF wild-type (n = 608) had median TTD and OS of 7.6 and 13.7 months.

Conclusion

Patients with RASa without RAS mutations (1.1% of mCRC) may have poor outcomes on EGFRmAb, although numbers herein were small, and interpretation is confounded by combination chemotherapy. Larger independent studies are warranted to determine if RASa, including degree of amplification, may act similarly to RAS mutation as a resistance mechanism to EGFRmAb therapies.

Implications for Practice

Genomic data suggest that RAS amplification occurs as the sole RAS/RAF alteration in >1% of colorectal cancer cases and that degree of amplification inversely correlates with co-occurring MAPK pathway alterations. Preliminary clinical evidence suggests that RAS amplification may function similarly to RAS mutation as a negative predictor of benefit from anti-epidermal growth factor receptor therapies in colorectal cancer. More clinical data are needed, and comprehensive genomic profiling, including detection of RAS amplification, should be used in trial design to inform therapy selection.

Introduction

RAS is one of the most frequently mutated oncogenes across human cancers [1, 2]. In metastatic colorectal cancer (CRC), the prevalence of RAS mutations varies between 40% and 50% [3, 4]. Compared with RAS wild-type tumors, RAS mutations in colorectal cancer are associated with worse overall survival [5]. Activating RAS mutations, which include KRAS and NRAS exon 2, 3, and 4 mutations, have been associated with resistance to anti-epidermal growth factor receptor (EGFR) therapies, which is reflected in the National Comprehensive Cancer Network (NCCN) guidelines and approved labels for EGFR therapeutic antibodies cetuximab and panitumumab [6–8]. Specifically, NCCN recommends that all patients with metastatic CRC (mCRC) have KRAS and NRAS genotyping performed and, based on data from multiple trials, suggests that cetuximab or panitumumab may be combined with chemotherapy as first-line treatment for RAS wild-type mCRC (Colon Cancer NCCN Guidelines, version 3.2020).

Other MAPK pathway activating alterations have also been associated with resistance to anti-EGFR therapies, including BRAF V600E mutation, most notably, as well as RAS amplification and overexpression and MAP2K1 mutation in in vitro models and small series of patients with CRC [9, 10]. However, the clinical implications of RAS amplification on CRC outcomes, and particularly its impact on clinical benefit from anti-EGFR therapy, remain ill defined. This is partly due to the relative rarity of RAS amplifications in comparison with RAS mutations. When analyzed by fluorescent in situ hybridization (FISH), only 0.7% of CRC samples were found to harbor KRAS amplification [9]. On the other hand, when using a clinically validated next-generation sequencing (NGS) platform, the frequency of RAS amplification (primarily KRAS amplification) was reported for 1.7% of patients [11]. The discordance in frequency between FISH and NGS is likely related to the capture of low levels of amplifications via NGS, which may not be captured via FISH, although other factors such as cohort selection bias may also contribute. For example, Foundation Medicine (FM) NGS-based comprehensive genomic profiling assays typically report copy number alterations of six or more as amplifications.

The quantitative relevance of RAS copy number on the presence of other RAS coalterations and on response to anti-EGFR therapy has not been adequately studied. In this report, we investigated the relevance of RAS copy number on the presence of additional comutations, especially RAS short variant (SV) and other MAPK activating alterations. Finally, we report on the clinical outcome of 15 total patients with RAS-amplified but RAS SV/BRAF V600E wild-type mCRC tested by FM using a biopsy collected prior to treatment with anti-EGFR therapies. Patients were treated at City of Hope or within the Flatiron Health (FH) network and captured in the Flatiron Health–Foundation Medicine (FH-FM) real-world clinicogenomic database (CGDB).

Materials and Methods

Comprehensive Genomic Profiling and Foundation Medicine Genomic Database

Hybrid capture–based comprehensive genomic profiling (CGP; FoundationOne CDx or FoundationOne) was performed prospectively for 37,233 unique patients with CRC on formalin-fixed paraffin-embedded (FFPE) tumor tissue submitted during routine clinical care in a Clinical Laboratory Improvement Amendments–certified, College of American Pathologists–accredited, New York State–regulated reference laboratory (Foundation Medicine Inc., Cambridge, MA). DNA (>50 ng) was extracted from FFPE CRC specimens; NGS was performed on hybridization-capture, adaptor ligation–based libraries to high, uniform coverage (>500×) for coding exons of 236–324 cancer-related genes plus selected introns [12]. To determine microsatellite instability (MSI) status using sequencing data generated via a hybrid capture–based NGS protocol, 95 or 114 intronic homopolymer repeat loci with adequate coverage were analyzed for length variability and compiled into an overall MSI score via principal components analysis [13]. Tumor mutational burden (TMB) was calculated by counting the number of synonymous and nonsynonymous mutations across a 0.8–1.2 megabase (Mb) region, with computational germline status filtering, and reporting the result as mutations/Mb. This method has been previously validated for accuracy against whole exome sequencing [14].

As previously described [12], a statistical copy number model was generated and fitted to each sample to determine gene copy numbers. Normalized coverage data for exonic, intronic, and single-nucleotide polymorphism (SNP) targets accounting for stromal mixture were plotted on a logarithmic scale, and minor allele SNP frequencies were concordantly plotted across the genome. Clustering of targets and minor allele SNPs were used to define upper and lower bounds of genomic segments. Empirical Bayesian algorithms used a distribution of parameters including base ploidy and purity to fit these data and generate copy number alteration variant calls. RAS amplification as described here includes amplifications at least four copies above the overall ploidy of the specimen. In the majority of cases, this represents amplifications at least six copies. We analyzed the FM tissue genomic database of CRC samples from 37,233 unique patients to determine the impact of RAS copy number on the coexpression of other MAPK activating alterations, including RAS SV and BRAF V600E mutations. RAS amplifications were categorized into three different categories: minimally amplified (copy number 6–9), moderately amplified (copy number of 10–19), and highly amplified (copy number of 20 and above). Approval for this study, including a waiver of informed consent and a Health Insurance Portability and Accountability Act waiver of authorization, was obtained from the Western Institutional Review Board (protocol no. 20152817).

City of Hope Cohort

We reviewed a molecular database from a single center (City of Hope [COH]) to identify 338 patients with mCRC treated between years 2015 and 2018. Only cases with mCRC with tissue CGP (FoundationOne or FoundationOne CDx) were included in the analysis. Tumors with RAS amplification were identified and were linked to deidentified patient characteristics and patient clinical response to anti-EGFR therapy (n = 6; see CONSORT diagram in supplemental online Fig. 2). The study was approved by the institutional review board of the COH National Comprehensive Cancer Center (14361).

Foundation Medicine–Flatiron Health CGDB Cohort

The FH-FM CGDB included 5,453 patients who had a diagnosis of mCRC, received care within the FH network between January 2011 and December 2019, and underwent tissue CGP (FoundationOne or FoundationOne CDx) performed between August 2012 and December 2019. CGDB is a nationwide (U.S.-based) deidentified electronic health record (EHR)–derived database and includes patients in FM who also received care within the FH network. Patients who were diagnosed with metastatic disease more than 90 days prior to their first visit within the FH network or who received their FM report more than 60 days after their last FH visit date were excluded to ensure all therapies received prior to CGP were captured, which left 3,904 unique patients eligible for this study. Clinical outcomes were assessed for nine patients with RAS amplification (see CONSORT diagram in supplemental online Fig. 2). Clinical characteristics and treatment information were obtained via technology-enabled abstraction of clinical notes and radiology/pathology reports and linked to CGP data. Institutional review board approval with waiver of informed consent was obtained prior to study conduct.

Statistical Analysis and Real-World Endpoints

Fisher's exact test was used to assess significance of RAS alteration co-occurrences. For analysis of clinical outcomes, time to treatment discontinuation (TTD) was defined as time from first to last administration or noncanceled order of EGFR monoclonal antibody (EGFRmAb) or until death. TTD was computed for the first EGFRmAb line received and also for consecutive lines of EGFRmAb-containing therapy. EGFRmAb was considered discontinued for patients with evidence of a change in therapy or with a structured activity date 120 days or more after last administration date and was used as an alternative endpoint to progression-free survival (PFS) for the CGDB cohort [15]. Patients whose EGFRmAb treatment continued until the data cutoff or until their structured medication administration records were censored at the date of last EGFRmAb administration. Overall survival (OS) was defined as time from first administration of EGFRmAb to date of death. Patients without a death event were censored at their last known activity date. To account for left truncation, to reflect the process of cohort eligibility, a patient's entry date into the CGDB was considered the later of the date of a patient's second visit within the FH network or their first eligible report for FM CGP. Risk set adjustment was used to ensure patients treated prior to entry date were not included in the at-risk population in OS analysis until they reached their entry date.

Results

Foundation Medicine Genomic Database

Analysis of the FM genomic database of samples from 37,233 patients with CRC showed an overall prevalence of RAS amplification (RASa) of 1.6% (n = 614). The RASa population was distributed almost evenly between the three categories of low amplification (6–9 copies, n = 241, 39%), moderate amplification (10–19 copies, n = 165, 27%), and high amplification (≥20 copies, n = 209, 34%). Distinct differences were noted in the co-occurrence of RAS and BRAF V600E mutations within the three subgroups (Fig. 1A). RAS SV mutations co-occurred more commonly in the low and moderate amplification cohorts (63% and 31%, respectively) but were nearly absent in the high-amplification cohort (4.8%) (p < .001 for all three comparisons). BRAF V600E occurred only in the low-amplification (3%) and moderate-amplification (0.6%) cohorts and was absent in the high-amplification cohort (p = .02 and .44, respectively). According to a univariate logistic regression model, RAS copy number is negatively correlated with the likelihood of a RAS mutation or BRAF V600E (odds ratio = 0.95).

We also observed a near-absence of high MSI in the RASa subset (1/540, 0.19%), whereas high MSI was observed in 4.1% (1,518/37,233) of all CRC cases, 2.8% (479/17,077) of cases with RAS SV, and 28% (592/2,125) of BRAF V600E CRC. MSI-high status differences in comparison with RASa were significant for all subsets (p < .001). The distribution and overlap of RAS and BRAF alterations are illustrated in Figure 1B. Comparing frequent coalterations in CRC cases with RAS SV mutations versus RASa suggests that the genomics of these subsets is similar and that overall, RASa tumors do not have a higher incidence of coalterations that may contribute to resistance to anti-EGFR therapy (median TMB 3.8 vs. 3.5 mutations/Mb in RAS SV vs. RASa cases, p = .20). We did note that in general RASa cases had more copy number alterations (including amplifications of CCND2, FGF23, FGF6, KDM5A, MYC, FLT3, and others); PIK3CA alterations were notably more frequent in the RAS SV subset versus the RASa subset (23% vs. 6.6%, p < .001), whereas HER2 (ERBB2) alterations were more frequent in the RASa versus RAS SV subset (8.6% vs. 3.3%, p < .001) (supplemental online Fig. 1).

Single Center Experience: COH Patient Characteristics and Response to Anti-EGFR Therapy for RAS-Amplified Subset

Between January 2015 and December 2018, 338 patients with mCRC underwent tissue CGP. Twelve (3.6%) patients had tumors harboring RASa (supplemental online Fig. 2). Among patients with RASa tumors, 11 of 12 had a left-sided primary tumor. All 12 tumors were wild type for BRAF alterations, and 9 of 12 (75%) were wild type for RAS SV mutations. Three patients had concurrent RAS SV mutations (KRAS G12S [n = 2] and G12V [n = 1]), which occurred in the settings of low (seven and eight copies) and intermediate (13 copies) KRAS amplifications. Seven of nine patients had high levels of RASa (20–79 copies); six of these seven had left-sided tumors. The characteristics of patients segregated by RAS/BRAF alteration status are summarized in Table 1.

Among the nine patients with tumors wild type for RAS SV/BRAF V600E as well as and positive for RASa, seven patients received anti-EGFR therapy. Six of these patients (KRAS amplification range, 13–54 copies) received anti-EGFR after CGP biopsy: three in the third line; one each in the first, second, and fourth lines, including treatment with folinic acid, fluorouracil and irinotecan (FOLFIRI)/panitumumab (n = 3); and one each with irinotecan/cetuximab, capecitabine/cetuximab, and irinotecan/panitumumab (supplemental online Fig. 3). Four patients experienced progressive disease (PD) as a best response, two had stable disease (SD), and none had an objective response to anti-EGFR therapy. All six patients were wild type for HER2 amplification and PIK3CA mutation in addition to RAS/BRAF mutation. SD occurred in a patient receiving first-line treatment (FOLFIRI/panitumumab), lasting for 4 months, and in another patient with second-line treatment (irinotecan/cetuximab), lasting for 7 months. Patients with PD had a median PFS, which was equivalent to TTD, of 2.3 months, and all patients including those with SD had a median PFS of 2.5 months. Tumor characteristics, prior therapy, anti-EGFR regimen, and associated outcome are summarized in Table 2.

CGDB Patient Characteristics and Response to Anti-EGFR Therapy for RAS-Amplified Subset

At FM, 3,904 patients underwent tissue CGP, were diagnosed with metastatic CRC less than 90 days prior to their first visit within the FH network, and received their FM report within 60 days of a FH visit. Sixty-six (1.7%) patients had tumors harboring RASa (supplemental online Fig. 2). Among patients with RASa tumors, 39/66 (59%) were wild type for RAS SV/BRAF V600E. In this cohort, RASa was associated with a significantly lower age at diagnosis compared with patients with RAS/BRAF V600E wild-type disease (p = .03). The characteristics of patients segregated by RAS/BRAF alteration status are summarized in Table 1.

Among the 39 patients with tumors wild type for RAS SV/BRAF V600E and positive for RASa, 18 patients received anti-EGFR therapy. Nine of these patients, all with tumors harboring RASa (range 9–62 copies), and wild type for HER2 amplification and PIK3CA mutation in addition to RAS/BRAF mutation, received anti-EGFR post-CGP biopsy across multiple lines, with the majority (eight of nine) receiving EGFRmAb in combination with chemotherapy (Table 2 and supplemental online Fig. 3). Tumor characteristics, prior therapy, anti-EGFR regimen,and associated TTD are summarized for patients with RASa tumors in Table 2.

Tumor response and PFS were not available for patients in the CGDB cohort, but median TTD for all lines of anti-EGFR therapy was 4.7 months for patients with RASa in the absence of RAS SV and BRAF V600E mutations (n = 9). In the same data set, patients with RAS SV mutations (n = 101) and patients wild type for both RAS alterations and BRAF V600E (n = 608) had a median TTD on anti-EGFR therapy of 5.3 months and 7.6 months, respectively. Limiting to the first exposure to anti-EGFR therapy, median TTD was 4.7 months for patients with RASa, 3.9 months for patients with RAS SV mutations, and 6.5 months for patients who were wild type for RAS/BRAF alterations. Similarly, median OS was 11.4 months for patients with RASa, 9.4 months for those with RAS SV mutations, and 13.7 months for patients with RAS/BRAF V600E wild-type disease (Fig. 2). Because of the low number of patients with RASa in our treatment cohort, direct statistical comparisons between these groups were not performed, and larger studies are needed to confirm the differences in outcomes seen in this study.

Discussion

KRAS and NRAS mutations are well-established predictors of lack of response to anti-EGFR therapies in CRC, and data also suggest that other MAPK pathway activating alterations, such as BRAF V600E, negatively predict EGFR antibody response. In a set of previously published studies, KRAS amplification was shown to mediate resistance to cetuximab in cell line models; silencing of KRAS expression restored sensitivity and has also been reported as a driver of acquired resistance after response and subsequent progression on anti-EGFR therapy in two patient cases [10, 16]. Furthermore, among CRC cases wild type for KRAS mutations treated with cetuximab or panitumumab (where the authors note that response could be attributed to anti-EGFR therapy), KRAS amplification was detected in none of 44 cases with responses but in 4 of 53 cases without response [9].

To assess the prevalence and genomic characteristics of RAS-amplified CRC, we interrogated the FM genomic database of 37,233 patients with CRC and CGP results. Overall, RAS amplification was observed in 1.6% of cases (n = 614), including 64% without co-occurring RAS or BRAF V600E mutations. Notably, co-occurrence of RAS amplification with RAS mutations was significantly more common in tumors with low-level RAS amplification compared with those with intermediate- or high-level amplification. BRAF V600E co-occurrence was infrequent and exclusive to the low- and moderate-amplification subset. These findings suggest that a high level of amplification (≥20 copies) may mitigate selective pressure for emergence of other alternate pathways for MAPK activation.

We looked at patients with CRC whose tumors were negative for RAS and BRAF V600E mutations as well as for HER2 amplification and PIK3CA mutation, but positive for RAS amplification by CGP prior to treatment with anti-EGFR therapy at a single institution (n = 6) or captured in the FH-FM joint CGDB (n = 9). Clinical data from the CGDB suggest that patients with RAS amplification as the sole RAS alteration present before initiation of anti-EGFR therapy have TTD on anti-EGFR therapy similar to those with RAS mutations treated with anti-EGFR (median 4.7 vs. 5.3 months, p = .57). Similarly, in the COH cohort, none of six patients without RAS SV mutations but with RAS amplification present before initiation of anti-EGFR therapy had an objective response, and the median PFS was 2.5 months. In addition, all four patients within the COH cohort who received anti-EGFR therapy after progression on prior standard cytotoxic therapy had PD on their first imaging study, with a median TTD because of progression of 2.3 months. Only 2 of 15 patients (cases 7 and 9) in our overall RAS-amplified set had TTD >7 months. Of these, both had treatment with cetuximab/FOLFIRI including a period of maintenance therapy, and case 7 had the lowest RAS copy number (nine copies) in the amplified cohort. Further analysis of a larger treatment group is needed to determine the predictive value of RAS amplification for various anti-EGFR treatment regimens. We did not observe a trend for longer TTD or PFS for patients with low-level versus higher-level RAS amplification treated with anti-EGFR (data not shown), but this analysis should be performed in larger cohorts powered for statistical comparison.

In prospective clinical trials, patients with mCRC with RAS mutations treated with first-line anti-EGFR therapy–containing regimens have reported PFS of 7.3–7.4 months compared with 10.1–11.4 months for those wild type for RAS mutations, and OS of 15.6–16.4 months for mutated RAS versus 26.0–28.4 months for wild type RAS [6, 7]. Our data, from two independent clinical cohorts including patients treated in the first to fourth lines, suggest that RAS amplification acts similarly to RAS mutation as a negative predictor of response to anti-EGFR therapy in CRC, and additional clinical studies are warranted to inform treatment selection for this subset of patients. In this study, 101 CGDB patients with RAS SV mutations received anti-EGFR therapy in real-world clinical practice between 2014 and 2019; however, in the majority of cases (67%), the biopsy used for CGP was collected prior to anti-EGFR therapy initiation, but the CGP testing was performed and reported after therapy initiation; thus the patient was likely to have been believed to be RAS wild type at the time of treatment decision. A subset of remaining patients (n = 8) had a prior RAS wild type hotspot test.

Limitations of this study include retrospective analysis of relatively small cohorts of patients with RAS amplification treated with anti-EGFR therapy at COH or captured within the CGDB, limiting our ability to make statistical comparisons. Yet this represents the largest reported series of patients with RAS amplification treated with anti-EGFR therapy. For CGDB patients, clinical information was captured as documented in the EHR, and testing events and treatments received outside of the Flatiron Health network that were not documented in the available EHR were not captured. For CGDB patients, PFS was unavailable. TTD is a pragmatic endpoint that is often used as a proxy for progression; however, patients may also discontinue therapy for reasons other than progression, including drug toxicity. Patients treated at COH or captured in the CGDB received anti-EGFR therapy in varying lines with inconsistent prior therapy, and most patients received anti-EGFR therapy in combination with other regimens, leaving open the possibility that benefit was not specific to EGFR inhibition. When limiting our analysis to patients who progressed on prior cytotoxic therapy, PD on anti-EGFR therapy was noted in all patients in the COH cohort.

Conclusion

Overall, our study found that CGP can identify RAS amplification in a subset of patients with CRC who were wild type for known negative predictors (i.e., RAS mutations, BRAF V600E mutation, HER2 amplification, and PIK3CA mutation) of response to anti-EGFR therapy. Further study is warranted to refine and determine the utility of RAS amplification, as well as other exploratory biomarkers, to inform effective selection of patients likely to benefit from anti-EGFR therapies.

Author Contributions

Conception/design: Alexa B. Schrock, Marwan Fakih

Provision of study material or patients: Marwan Fakih

Collection and/or assembly of data: Jessica K. Jaideep Sandhu

Data analysis and interpretation: Alexa B. Schrock, Jessica K. Lee, Russell Madison, Cheryl Cho-Phan, Jeremy W. Snider, Emily Castellanos, Jeffrey M. Venstrom

Manuscript writing: Alexa B. Schrock, Jessica K. Lee, Jaideep Sandhu, Russell Madison, Cheryl Cho-Phan, Jeremy W. Snider, Emily Castellanos, Jeffrey M. Venstrom, Marwan Fakih

Final approval of manuscript: Alexa B. Schrock, Jessica K. Lee, Jaideep Sandhu, Russell Madison, Cheryl Cho-Phan, Jeremy W. Snider, Emily Castellanos, Jeffrey M. Venstrom, Marwan Fakih

Disclosures

Alexa B. Schrock: Foundation Medicine (E), Roche (OI); Jessica K. Lee: Foundation Medicine (E), Roche (OI); Russell Madison: Foundation Medicine (E), Roche (OI); Cheryl Cho-Phan: Flatiron Health (E), Roche (OI); Jeremy W. Snider: Flatiron Health (E), Flatiron Health, Roche (OI); Emily Castellanos: Flatiron Health (E), Flatiron Health, Roche (OI); Jeffrey M. Venstrom: Foundation Medicine (E), Roche (OI); Marwan Fakih: Amgen, Pfizer, Array, Bayer (C/A), Amgen (H), Amgen, Novartis, AstraZeneca (RF—institution), Amgen, Guardant360 (other—speaker's bureau). Jaideep Sandhu indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board

References

Author notes