Transcriptional Profiling and Consensus Molecular Subtype Assignment to Understand Response and Resistance to Anti–Epidermal Growth Factor Receptor Therapy in Colorectal Cancer

Abstract

PURPOSE

Activating mutations in KRAS, NRAS, and BRAF are known to cause resistance to anti–epidermal growth factor receptor (EGFR) therapy; however, only approximately 40% of patients with colorectal cancer (CRC) with RAS^WT tumors respond to anti-EGFR treatment. We sought to discover novel biomarkers to predict response to anti-EGFR antibody treatment in CRC and to understand mechanisms of resistance to anti-EGFR therapy.

MATERIALS AND METHODS

Transcriptomic profiles from three clinical and two preclinical cohorts treated with cetuximab were used to assign consensus molecular subtypes (CMS) to each sample and correlated with outcomes.

RESULTS

Restricting to RAS^WT patients, we observed that CMS2 tumors (canonical subtype) had significantly higher response rates relative to other CMS when treated with cetuximab combination with doublet chemotherapy (Okita et al cohort: 92% disease control rate (DCR) for CMS2, chi-square P = .04; CALGB/SWOG 80405 cohort: 90% objective response rate (ORR) for CMS2, chi-square P < .001) and with single-agent cetuximab (68%, chi-square P = .01). CMS2 tumors showed best response among right-sided (ORR = 80%) and left-sided (ORR = 92%) tumors in the CALGB/SWOG 80405 cohort. CMS2 cells lines were most likely to be sensitive to cetuximab (60%) and CMS2 patient-derived xenograft had the highest DCR (84%). We found Myc, E2F, and mammalian target of rapamycin pathways were consistently upregulated in resistant samples (enrichment score >1, false discovery rate <0.25). Inhibitors of these pathways in resistant cell lines exhibited additive effects with cetuximab.

CONCLUSION

These data suggest that CRC transcriptional profiles, when used to assign CMS, provide additional ability to predict response to anti-EGFR therapy relative to using tumor sidedness alone. Notably both right-sided and left-sided CMS2 tumors had excellent response, suggesting that anti-EGFR therapy be included as a treatment option for right-sided CMS2 tumors.

Transcriptomic profiling predicts response to anti-EGFR therapy in RAS-WT colorectal cancer.

INTRODUCTION

The anti-epidermal growth factor receptor (EGFR) monoclonal antibodies cetuximab and panitumumab inhibit the mitogen-activated protein (MAP) kinase pathway by blocking EGFR at the cell membrane.¹ It is now well known that activating mutations in KRAS, NRAS, and BRAF—all downstream of EGFR in the MAP kinase pathway—confer resistance to anti-EGFR therapy.² However, among RAS^WT (ie, KRAS^WT/NRAS^WT/BRAF^WT) tumors, the response rate to anti-EGFR therapy is only about 40%.^2,3 In addition to RAS mutation status, primary tumor location (left v right side of colon) is predictor of anti-EGFR antibody response.^4-6 Given that there is not a clear anatomic explanation as to why right-sided tumors would be resistant to anti-EGFR inhibition, it has been suggested that the observed differences in left-sided and right-sided tumors are due to differences in somatic mutation, gene expression, and/or epigenetic modifications.^7,8 However, unlike RAS mutation, where the response rate is zero in RAS^mut patients, a significant fraction of RAS^WT right-sided tumors do respond to anti-EGFR therapy.^9-11 Additional biomarkers are needed to more precisely predict treatment response.

CONTEXT

• Key Objective

• Approximately 60% of patients with colorectal cancer (CRC) with wild-type KRAS, NRAS, and BRAF genes (RAS^WT) do not respond to cetuximab treatment. There is a need for biomarkers to more accurately identify the subgroup of patients who are sensitive to anti–epidermal growth factor receptor (EGFR) antibody therapy. Previous research has shown that dividing CRC into four transcriptomic consensus molecular subtypes (CMS) is a reliable prognostic biomarker.

• Knowledge Generated

• Here, we find that CMS can also predict response to anti-EGFR therapy. For RAS^WT tumors on either left or right side of the colon, CMS2 tumors had the best response rate and progression-free survival with cetuximab treatment.

• Relevance

• There was not a significant difference in response rates (93% left v 80% right) between right- and left-sided CMS2 tumors, suggesting anti-EGFR therapy should be considered for CMS2 tumor regardless of primary tumor location.

Recently, an international consortium of six independent research groups categorized colorectal cancer (CRC) into four robust consensus molecular subtypes (CMSs).¹² Although the CMS can predict survival and are associated with tumor location (eg, CMS2 enriched in left-sided tumors) and distinct mutational subtypes of CRC (eg, BRAF-mutation enriched in CMS1), CMS has not been convincingly demonstrated as a predictive clinical biomarker and its use in clinical practice is limited.^5,6,13-15 The purpose of this study was to determine whether CMS classification could predict response to anti-EGFR treatment in RAS^WT CRC patients. We hypothesize as others that response to anti-EGFR therapy is determined by the intrinsic transcriptional state of the tumor,^16,17 which is captured in the CMS. Therefore, to better identify patients with CRC who will most likely benefit from anti-EGFR therapy, we performed a comprehensive analysis of all publicly available clinical and preclinical data sets having both anti-EGFR treatment and transcriptomic profiling evaluating molecular and clinical predictors of response.

MATERIALS AND METHODS

Materials and methods used in this work are available in the Data Supplement. Written informed consents were obtained from either the participants or their legal guardians, and this study was approved by the institutional review board.

RESULTS

Implication of the CMS Framework to Predict Anti-EGFR Antibody Response

Characteristics of all four study cohorts (clinical and preclinical) are summarized in the Data Supplement. There were 983 CRC samples available from all four study cohorts.^17-21 However, after filtering patients who were missing either transcription or clinical response data, 208 (21%) RAS^WT samples with anti-EGFR treatment response and CMS calls were available for analysis (Fig 1).The disease control rate (DCR) of RAS^WT CMS2 patients was 91.7% (33/36) in the Okita et al cohort (chemotherapy plus cetuximab-treated), whereas the rates for CMS1, CMS3, and CMS4 subtypes were 50% (2/4), 65% (13/20), and 83.3% (15/18), respectively (Fig 2A). Similar results were also observed in the Khambata-Ford et al cohort (cetuximab-monotherapy), where CMS2 patients with RAS^WT showed the highest DCR (68.2% [15/22]) among the other molecular subtypes (CMS1: 0% [0/4]; CMS4: 29% [5/17]). The CMS3 subtype was not observed in this data set (Fig 2B).

FIG 1.

Flow diagrams of data-filtering steps used for all four study cohorts. CMS, consensus molecular subtypes; CRC, colorectal cancer; MSI, microsatellite instable; MSS, microsatellite stable; PDX, patient-derived xenograft.

FIG 2.

RAS^WT CRC samples of each CMS subgroup that clinically benefited from or were sensitive to cetuximab treatment. The clinical cohorts contained patients treated with (A) chemotherapy and cetuximab and (B) cetuximab monotherapy. The preclinical cohorts of the (C) CRC cell lines and (D) PDX models were treated with cetuximab monotherapy. (E) Objective response rates of RAS^WT mCRC patients of each CMS subgroup who received first-line cetuximab plus chemotherapy in the CALGB/SWOG 80405 trial. CMS, consensus molecular subtypes; CRC, colorectal cancer; PDX, patient-derived xenograft.

In the preclinical cohorts, the CMS2-subtype CRC cell lines and patient-derived xenograft (PDX) models with RAS^WT (KRAS^WT/NRAS^WT/BRAF^WT) genes also showed the highest response rate to cetuximab treatment (Figs 2C and 2D). In the RAS^WT CRC cell lines of Medico et al (n = 27) cohort, 58.8% (10/17) of microsatellite stable (MSS), RAS^WT CMS2 cell lines were sensitive to cetuximab treatment. MSS, RAS^WT CMS1 and CMS3 cell lines were not present in this cohort. By contrast, all MSS, RAS^WT CMS4 cell lines (n = 10) showed resistance to cetuximab treatment (Fig 2C). In the Bertotti et al cohort, MSS, RAS^WT PDX models (n = 60) in the CMS2 subgroup (n = 24) had 75% (18/24) response rate to cetuximab treatment, whereas the other subtypes CMS1 (n = 11), CMS3 (n = 9), and CMS4 (n = 16) had response rates of 36.4% (4/11), 55.6% (5/9), and 62.5% (10/16), respectively (Fig 2D).

In the CALGB/SWOG 80405 cohort (first-line doublet chemo + cetuximab) of 128 RAS^WT mCRC patients, CMS2 patients (n = 52) had the highest objective response rate (ORR) 90.4% (47/52), and no patients experienced progressive disease (PD). CMS1 patients (n = 26) had the lowest ORR at 42.3% (11/26) including four patients with PD; CMS3 (n = 12) and CMS4 (n = 38) had ORRs of 66.7% (8/12) and 76.3% (29/38), respectively (Fig 2E). A statistically significant association was found between CMSs of RAS^WT patients and response to treatment in all three clinical cohorts as determined by chi-square tests (Okita et al, χ²(3) = 8.4, P value = .04; Khambata-Ford et al, χ²(3) = 9.6, P value = .01; CALGB/SWOG 80405 cohort, χ²(3) = 21.4, P value = .0005). The odds ratios (ORs; CMS2 v non-CMS2) of treatment responses for the cohorts were 6.6, 6.8, and 5.5, respectively, indicating that CMS2 patients were more likely to achieve a response to anti-EGFR treatment compared with CMS1/3/4 patients.

Having found that CMS2 tumors had the best objective response to cetuximab therapy, we further assessed the interaction using a Cox regression model for both univariate and multivariate analyses in both clinical cohorts. We also observed that mutation status of RAS genes and CMSs of patients with CRC were significantly associated with progression-free survival (PFS) in the Okita et al cohort (Data Supplement). In this cohort, Cox hazard ratios of the univariate and multivariate studies for CMS2 versus others were 0.56 (95% CI, 0.38 to 0.81; P value** = .0023) and 0.62 (95% CI, 0.41 to 0.96; P value* = .0031), respectively (Data Supplement). In the Khambata-Ford et al dataset, we found the covariates sex, RAS gene mutation, and CMS of patients with CRC were significant (P values < .05) in univariate analysis (Data Supplement). However, in multivariate analysis, CMS was the only covariate significantly (CMS2 v others, P value = .0039) associated with PFS of patients with cetuximab-treated CRC. In this cohort, CMS2 showed predictive power over other subtypes in both univariate- (HR, 0.44; 95% CI, 0.27 to 0.73; P value** = .0016) and multivariate (HR, 0.44; 95% CI, 0.25 to 0.77; P value** = .0039) Cox-regression analyses (Data Supplement). Comparing PFS of patients of all four subtypes in both cohorts, CMS2 patients showed significantly higher median PFS (approximately 7.5 months and approximately 3.1 months) than other subtypes (log-rank test; P values < .001; Data Supplement).

Comparison of CMS and Tumor Sidedness to Anti-EGFR Antibody Response

Given multiple prior reports suggesting that left-sided tumors have better response to anti-EGFR treatment,^22,23 we specifically examined anti-EGFR response in left-sided and right-sided tumors taking into account of CMS. We analyzed objective response of RAS^WT mCRC patients (n = 128) treated with first-line doublet chemotherapy plus cetuximab from the CALGB/SWOG 80405 trial classified by both CMSs and tumor sidedness. Consistent with prior report,¹² left-sided tumors were more likely to be CMS2 (approximately 32% [42/128] left-sided v approximately 8% [10/128] right-sided) and were more likely to respond than right-sided tumors (ORR, 57% v 16%). However, incorporating CMS further refined response for both left-sided (Fig 3A) and right-sided (Fig 3B) tumors. For both left-sided and right-sided tumors, CMS2 had the best response, with an ORR of 93% (complete response [CR]: 9, partial response [PR]: 30) and 80% (CR: 2, PR: 6), respectively (Fig 3C). By contrast, both left-sided and right-sided tumors of CMS1 tumors had poor response (CMS1-left: ORR = 45%, CR: 1, PR: 4; CMS1-right: ORR = 40%, CR: 1, PR: 5). Interestingly, a small subset of patients with left-sided CMS3 tumors demonstrated responses nearly as favorable as CMS2, with an objective response rate (ORR) of 87.5%. In contrast, right-sided CMS3 tumors showed the poorest response among all groups, with an ORR of only 25% (Fig. 3C).

FIG 3.

Efficacy of anti-EGFR antibody treatment to left-sided and right-sided RAS^WT tumors of the colon in patients with CRC of the CALGB/SWOG 80405 trial. Distributions of the total number of RAS^WT CRC patients who exhibited OR (OR = CR + PR), SD, and PD to chemo and anti-EGFR antibody therapy divided into four CMSs of (A) left-sided and (B) right-sided disease groups. (C) Objective response rates segregated by both CMS and sidedness. CR, complete response; CRC, colorectal cancer; EGFR, epidermal growth factor receptor; CMS, consensus molecular subtypes; OR, objective response; PD, progressive disease; SD, stable disease.

Furthermore, a logistic regression model was performed to get insights about the influence of tumor sidedness, CMSs, and their potential interaction in predicting the objective response to anti-EGFR treatment. The results of the logistic regression model indicated that CMS subtype was a significant predictor of the response to anti-EGFR treatment. Specifically, patients with non-CMS2 subtype tumors had lower odds of responding to treatment compared with CMS2 patients (OR, 0.21; 95% CI, 0.06 to 0.74; P = .02). Patients with right-sided tumors had lower odds of responding to treatment than left-sided tumors (OR, 0.31; 95% CI, 0.07 to 1.31; P = .23), although the effect was not statistically significant, and the wide CI suggests a lack of precision in this estimate. The interaction between tumor-sidedness and CMS subtypes was insignificant (P = .97), suggesting that the effect of tumor-sidedness did not vary significantly across different CMS subtypes.

Identification of Active Pathways in Cetuximab-Resistant CRC Cells

Gene set enrichment analysis (GSEA) was used to identify active pathways or gene sets in nonresponder versus responder groups of CMS2/RAS^WT/cetuximab. Enrichment analysis showed that E2F targets, MTORC1 signaling, G2M checkpoint, mitotic spindle, and Myc target gene sets were enriched (normalized enrichment score or NES >1) in nonresponders compared with responders in the Okita et al cohort (false discovery rate = 0.32). GSEA was then conducted on MSS CMS2 cell lines and PDX models with RAS^WT genes to identify gene sets associated with cetuximab-resistance mechanisms. E2F, Myc, and MTORC1 pathway gene sets were consistently enriched in both cetuximab-resistant CRC cell lines and nonresponder group of PDX models (Fig 4A). E2F gene set had a NES of 1.22 in the nonresponder group of PDX models but a nonsignificant false discovery rate value of 0.37.

FIG 4.

(A) Gene set enrichment analysis showing that E2F, mTOR, and Myc pathways are consistently active in the cetuximab-refractory CMS2 subgroup with RAS^WT genes. NES of ssGSEA performed in the cetuximab-refractory, CMS2, RAS^WT gene group of (B) chemotherapy plus cetuximab-treated patients (Okita et al), (C) cetuximab monotherapy–treated patients (Khambata-Ford et al), (D) CRC cell lines (Medico et al), and (E) PDX models (Bertotti et al). The NES values of MTORC1, E2F, and MYC gene sets from CMS2/RAS^WT samples of each cohort in the heatmaps were transformed into Z-scores. The heatmap x-axis represents the patient IDs (B, C), CRC cell-line name (D), and PDX model IDs (E) of each study cohort. CMS, consensus molecular subtypes; CRC, colorectal cancer; GSEA, gene set enrichment analysis; NES, normalized enrichment scores; PDX, patient-derived xenograft; ssGSEA, single-sample GSEA.

We aimed to determine if the cetuximab treatment-resistant pathways are concurrently or mutually active in each sample of the CMS2/RAS^WT group that is resistant to cetuximab. Therefore, we performed single-sample GSEA (ssGSEA) on each sample in the CMS2/RAS^WT group in each of the four study cohorts (two clinical and two preclinical) to assess the activity of the MTORC1, E2F, and Myc pathways. The NESs obtained from ssGSEA for each gene set were transformed into Z-scores. The per-sample Z-scores of each gene set were extracted from the cetuximab treatment-refractory groups of all four study cohorts and plotted into heatmaps (Figs 4B-4E). Each column in the heatmap represents a CMS2/RAS^WT sample that is resistant to cetuximab (tissue/cell line/PDX model), and a higher Z-score of a pathway/geneset in a given sample indicates greater activation of that pathway in that sample. The MTORC1 (55%), E2F (60%), and MYC (55%) pathways were active (Z-score >0) in the CMS2 CRC patients with RAS^WT genes (n = 20) who had shown PD after getting treatment with cetuximab plus chemotherapy cohort (Okita et al; Fig 4B). In the cetuximab treatment/refractory group of the cetuximab-monotherapy cohort (Khambata-Ford et al), the higher enrichments of MTROC1, E2F, and two MYC gene sets were observed in 57%, 57%, 71%, and 42%, respectively, of the RAS^WT CMS2 patients (n = 7; Fig 4C). Similar activity rates for the same gene sets were observed in the preclinical cohorts. More than 50% of the cetuximab-refractory CRC cell lines (Fig 4D) and PDX models (Fig 4E) in the CMS2 group with RAS^WT genes showed higher enrichments of MTORC1, E2F, and MYC gene sets.

In Vitro Validation of Cetuximab Resistance Pathways

We hypothesized that using inhibitors of mammalian target of rapamycin (mTOR), E2F, and/or MYC pathways in combination with cetuximab could potentially reverse cetuximab resistance. We selected two CMS2/RAS^WT CRC cell lines known to be resistant to cetuximab, HT55 and SNUC1, and evaluated their viability in response to cetuximab in combination with inhibitors of the E2F (CHEK1/2 inhibitors AZD7762 and MK-8776),^24,25 MYC (BRD4 inhibitor JQ1),²⁶ and MTORC1 (everolimus)²⁷ pathways in drug-screening assays. We first performed single-drug screening assays to establish dose-response curves of individual drugs (Data Supplement), and calculated IC₅₀ and AUC values (Data Supplement). As previously reported, cetuximab alone could not suppress the proliferation of HT55 and SNUC1 cells in vitro (IC₅₀ >100 μg/mL; Data Supplement).¹⁸ We also retrieved the IC50 values of these drugs in the cetuximab-sensitive DiFi and NCI-H508 cell lines and the cetuximab-resistant HT55 and SNUC1 CRC cell lines from the Genomics of Drug Sensitivity in Cancer (GDSC2) database. We found that both the cetuximab-sensitive and cetuximab-resistant CRC cell lines are resistant to these drugs when used in single-agent drug-screening assays (Data Supplement). Therefore, none of the inhibitors had significant single-agent activity against either HT55 or SNUC1 cells (Data Supplement). Next, to test drug combinations with cetuximab, a dose of each small molecule inhibitor that was sublethal but high enough to block its target was selected and combined to a dose range of cetuximab. The addition of AZD7762 (Fig 5A), MK-8776 (Fig 5B), everolimus (Fig 5C), and JQ1 (Fig 5D) significantly decreased cell viability relative to cetuximab alone; however in each case, the combinatorial effect was additive, rather than synergistic. These additives rather than synergistic in vitro results are consistent with single-sample GSEA of clinical and preclinical samples (Fig 4), which indicated that multiple resistant pathways are active simultaneously in cetuximab-resistant CRC cells.

FIG 5.

Normalized cell viabilities of HT55 cells in single drug and drug combination assays. Average normalized cell viabilities of HT55 cell line in single-drug (cetuximab) assay and drug combination assays with fixed dose of (A) AZD7762, (B) MK-8776, (C) everolimus, and (D) JQ1 drugs and their combinations with different concentrations of cetuximab. ***P value < .01.

Activation of Cetuximab-Resistant Pathways in Single Cells of a CRC Tissue

To evaluate the influence of intratumor heterogeneity on anti-EGFR resistance, scRNA-seq was performed on a CRC tumor with poor response to FOLFOX/panitumumab. After neoadjuvant treatment, 70% of residual tumor was viable and the patient experienced progression <5 months after resection (Data Supplement). scRNA-seq of 6,160 viable cells (epithelial: 3,929; immune: 2,082; stromal: 149) from this RAS^WT, stage IV tumor were analyzed (Fig 6A). The cluster of epithelial cells was associated with scCMS2 and scCMS3 phenotypes (Fig 6B). We observed a cluster of cells with highly active MTORC1 (mean NES = 10.3), E2F (mean NES = 8.5), MYC_Targets_V1 (mean NES = 15.3), and MYC_Targets_V2 (mean NES = 5.8) gene sets/pathways in a subset of epithelial cells that were specifically aligned with scCMS2 phenotype (Figs 6C-6F). By contrast, activations of these resistant pathways were rare in immune cells (Wilcoxon rank-sum test; P value < .01), suggesting de novo cetuximab-resistance mechanisms are intrinsic to tumor epithelial cells. Additionally, enrichment of the MTORC1 gene set was positively correlated with E2F (Pearson's correlation, r = 0.81; P < .01), MYC_Targets_V1 (r = 0.93; P < .01), and MYC_Targets_V2 (r = 0.87; P < .01). Similar strong positive correlations were also observed between E2F versus MYC_Targets_V1 (r = 0.86; P < .01), E2F versus MYC_Targets_V2 (r = 0.80; P < .01), and MYC_Targets_V1 versus MYC_Targets_V2 (r = 0.80; P < .01) pathways. The strong positive correlations among these resistant pathways in epithelial cells suggest that multiple resistance pathways are simultaneously active in the same cell.

FIG 6.

Single-cell RNA sequencing profile of a patient with CRC after being treated with chemotherapy plus panitumumab. (A) UMAP plot of the distributions of different cell types (epithelial, immune, and stromal) present in tumor tissues. (B) scCMS assignment to individual single cells in the epithelial cells compartment. Distributions of active or significantly enriched tumor cells with (C) MTORC1, (D) E2F, (E) MYC targets (V1), and (F) MYC targets (V2) gene sets/pathways in the same tumor tissue. NES of a given gene set/pathway obtained from ssGSEA were Z-score transformed using the mean and standard deviation of NES across all individual cells (n = 6,160) of the tumor tissue. CRC, colorectal cancer; NES, normalized enrichment scores; scCMS, single cell consensus molecular subtypes; UMAP, uniform manifold approximation and projection.

DISCUSSION

Anti-EGFR antibodies are a highly effective treatment for a subset of patients with CRC, but are ineffective for others and can cause significant toxicity. Therefore, proper identification of the patients most likely to benefit is a critical clinical need. Previous clinical studies, such as NCIC CO.17, PRIME, CRYSTAL, and CALGB/SWOG 80405 trials, reported that individuals with left-sided CRC experience better outcomes with anti-EGFR treatment than those with right-sided CRC.²⁸ On the basis of these data, current national treatment guidelines recommended that only left-sided, RAS^WT metastatic CRC patients should be offered anti-EGFR antibodies in the first line of treatment.^5,6 Given the different embryologic origins of the right and left colon,²⁹ it has been suggested that sidedness (left v right) is a surrogate marker of the tumor transcriptional state and that sensitivity to anti-EGFR treatment is driven by a transcriptomic rather than anatomic mechanism.^29,30 Therefore, it was our hypothesis that using CMS, a transcriptomic classifier designed to group CRC into more homogenous subtypes, would better predict response to anti-EGFR antibody treatment. We find that in frontline treatment, subsequent treatment lines, PDX models, and cell lines, CMS2 tumors (canonical subtype) consistently have the best response to anti-EGFR antibodies (Fig 2). This result is consistent with prior analysis from the Prospect-C cohort where CMS2/TA subtype had prolonged benefit from anti-EGFR therapy including two right-sided patients.³ Importantly, we find that response to anti-EGFR therapy was not limited to left-sided tumors; right-sided CMS2 tumors had a response rate nearly as good as left-sided CMS2, and superior to left-sided CMS1 tumors (Fig 3). Additionally, meta-analyses pooling data from multiple randomized clinical trials have also reported that a subset of patients with right-sided CRC respond to anti-EGFR treatment.^9,10,31 Although these retrospective analyses require prospective validation, these findings have important clinical implications for the approximately 25% of right-sided CRC tumors that are CMS2 as these patients appear nearly as likely to respond to anti-EGFR therapy as patients with left-sided CMS2 tumors.¹² Given the current lack of targeted therapy options in CRC, access to anti-EGFR therapy could substantially benefit these patients, highlighting the potential utility of transcriptomic profiling of CRC tumors to guide initial therapy.

We also observed that RAS^WT CMS1 tumors display poor response to anti-EGFR therapy. This could be due to overexpression of IL-1R1 receptor protein and activation of IL-1–mediated pathway in CMS1 tumors, which is associated with resistance to EGFR pathway blockade and poor prognosis of patients with CRC.³² Likewise, activation of vascular endothelial growth factor (VEGF) pathway is also implicated as one of the resistance mechanisms to cetuximab, and CMS1 tumors have shown better responses to anti–vascular endothelial growth factor receptor (VEGFR) plus chemotherapy than anti-EGFR treatment arm in the CALGB/SWOG 80405 clinical trial.³³ Upon further analysis of the CALGB/SWOG 80405 data, we found that only 45% of left-sided and 40% of right-sided CMS1 patients had objective responses to frontline doublet plus anti-EGFR therapy (Fig 3). These poor responses suggest that for the approximately 10% of patients with left-sided CRC with CMS1 tumors, the potential toxicity of anti-EGFR therapy may not be warranted and anti-VEGFR may be the preferred biologic therapy for these patients.¹² Interestingly, the response rate of left-sided CMS3 tumors was remarkably greater than right-sided CMS3 (87.5% v 25%; Fig 3), suggesting that the CMS classification alone does not capture the complete range of transcriptomic differences mediating anti-EGFR response in CRC.

Prior studies have reported that RAS-MAPK–mediated activation is the most common mechanism in acquired resistance to anti-EGFR treatment.^34-38 However, in our investigation, we revealed that primary resistance to anti-EGFR inhibition in CMS2 tumors is mainly driven by simultaneous activations of MTORC1, E2F, and MYC pathways and originates from tumor epithelial cells (Fig 4). Simultaneous activation of these resistant pathways also resulted in additive but not synergistic activity in in vitro drug combination assays of cetuximab-refractory CMS2 CRC cell lines (Fig 5). This supports the idea that synergy between anticancer drugs is rare and highly dependent on cellular context.³⁹ The existence of multiple anti–EGFR-resistant pathways in cetuximab refractory CRC cells also suggests the potential benefits of using higher-order combinations of drug targets that may help overcome primary resistance against anti-EGFR monoclonal antibody treatment. Furthermore, scRNA-seq profiling of a CRC tumor treated with anti-EGFR antibody revealed that activations of MTORC1, E2F, and MYC pathways are strongly correlated with epithelial cells with CMS2 phenotype, suggesting anti-EGFR resistance is regulated by tumor intrinsic epithelial cell signaling pathways (Fig 6). A recent study has also divided the CMS into two intrinsic subtypes, i2 and i3, on the basis of the intrinsic transcriptomic signatures of colorectal tumors. The gene expression signature of the i2 subtype, which primarily included conventional CMS2 and CMS4 tissues, was also strongly correlated with the cetuximab drug response signatures of cancer cell lines from the CTRPv2 data set.¹⁶ This finding suggests that treatment outcomes for anti-EGFR therapy appear to be correlated with both intrinsic subtypes of CRC and bulk transcriptomic–based conventional CMS types.

This study has limitations, including retrospective data sets with gene expression data collected using different platforms.^17-21 Also, because of limited number of samples available for each CMS group in all four study cohorts (clinical and preclinical), we could not identify conserved cetuximab resistance pathways in CMS1, CMS3, and CMS4 tumor subtypes. In summary, our study reports a comprehensive analysis of preclinical and clinical studies of anti-EGFR antibody response of CRC to identify novel biomarkers and understand drug resistance mechanisms. We found that RAS^WT/MSS/CMS2 tumors from both left and right colon have favorable response to cetuximab treatment. We propose the CMS framework has predictive power to stratify RAS^WT/MSS CRC patients into cetuximab-sensitive and cetuximab-refractory subgroups.

AUTHOR CONTRIBUTIONS

Conception and design: Saikat Chowdhury, Scott Kopetz, John Paul Shen

Financial support: Scott Kopetz, John Paul Shen

Administrative support: Heinz-Josef Lenz, Scott Kopetz, John Paul Shen

Provision of study materials or patients: Heinz-Josef Lenz, Scott Kopetz, John Paul Shen

Collection and assembly of data: Saikat Chowdhury, Ria Gupta, Kangyu Lin, Joshua Millstein, Heinz-Josef Lenz, John Paul Shen

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).