Mutational profiles of metastatic colorectal cancer treated with FOLFIRI plus cetuximab or bevacizumab before and after secondary resection (AIO KRK 0306; FIRE‐3)

Secondary resection of metastases is recommended in metastatic colorectal cancer (mCRC). Data describing changes in mutational profiles of corresponding primary tumor and metastatic tissue after conversion treatment are limited. Next generation sequencing was performed in formalin‐fixed mCRC samples from patients of the FIRE‐3 trial (FOLFIRI plus cetuximab or bevacizumab) before treatment start (baseline) and after secondary resection of metastases (post baseline). Changes of mutational profiles and tumor mutational burden (TMB) were assessed within a post‐hoc analysis. Median overall survival (OS), progression‐free survival (PFS) and objective response rate (ORR) were compared between treatment arms. Paired tumor samples were obtained from 25 patients (19 RAS wild‐type, 6 RAS mutant by pyrosequencing). ORR (92.0% vs 58.0%) and OS (60.8 vs 35.4 months, hazard ratio = 0.39 [95% CI 0.14‐1.12], P = .08) were higher for patients receiving cetuximab. After conversion therapy, 56 alterations (42 in the cetuximab and 14 in the bevacizumab arm) were newly observed in 18 patients (9 each treated with cetuximab or bevacizumab). Gains (n = 21) and losses (n = 21) of alterations occurred during cetuximab‐based treatment, while mainly gains of alterations occurred during bevacizumab (n = 10). Three of nine patients treated with cetuximab that presented a change of mutational profiles, developed resistance to cetuximab. Mutational profiles were largely comparable before and after treatment with anti‐VEGF or anti‐EGFR directed monoclonal antibodies after secondary resection. Mutations associated with resistance to anti‐EGFR antibodies were observed in only one‐third of patients.

Mutational profiles were largely comparable before and after treatment with anti-VEGF or anti-EGFR directed monoclonal antibodies after secondary resection. Mutations associated with resistance to anti-EGFR antibodies were observed in only onethird of patients.
Secondary resection for initially unresectable metastatic colorectal cancer (mCRC) is associated with improved prognosis. Predicting opportunities for secondary resection, however, depends on the discovery of molecular changes in primary tumor and corresponding metastatic tumor tissue. Here, mutational profiles were investigated for mCRC patients in the FIRE-3 trial, a study of FOLFIRI plus cetuximab or bevacizumab as first-line therapy for irresectable mCRC. Of nine mCRC patients undergoing cetuximab therapy who experienced changes in tumor mutational profile, one-third became resistant to cetuximab. For patients treated with anti-VEGF and anti-EGFR antibodies, mutational profiles were similar before and after treatment and following secondary resection.

| INTRODUCTION
Prognosis of patients with primarily unresectable metastatic colorectal cancer (mCRC) has markedly improved over the past decades by introduction of monoclonal antibody treatment according to molecular tumor characteristics. 1,2 After cytotoxic conversion therapy of initially unresectable lesions, multidisciplinary treatment approaches (eg, surgery and local-ablative treatment) are recommended to improve long-term overall survival. 3,4 Although previous investigations mainly focused on R0 resection rate, even R1 resection of metastatic lesions has been associated with improved outcome in both liver-limited and non-liver-limited disease. [5][6][7][8][9][10] It is, therefore, essential to identify patients that benefit from multidisciplinary treatment and to evaluate biomarkers for increasing their frequency. Various factors influence secondary resectability, such as surgeons' experience, location of metastases, but also surrogate parameters of response like early tumor shrinkage or depth of response. The presence of BRAF V600E mutations (MUT) was associated with a lower likelihood of secondary resectability. 11 However, little is known about dynamics of molecular tumor characteristics in paired samples of primary tumors and corresponding metastases after conversion treatment.

FIRE-3 was an open-label multicenter randomized controlled phase
III trial that evaluated the combination of FOLFIRI plus cetuximab or bevacizumab as first-line regimen in irresectable KRAS wild-type (WT) mCRC patients. 12 Of these, 29% of patients underwent secondary resection after conversion. 5 Additionally, a subgroup of 373 patients provided formalin-fixed paraffin embedded (FFPE) samples for targeted next generation sequencing (NGS) analysis of 315 genes (FoundationOne, Roche). 13 We were able to re-perform NGS in FFPE specimens of metastases from patients who underwent secondary resection and to correlate this analysis with data from corresponding primary tumors. Our aim was to assess dynamic changes in molecular characteristics of paired specimen (primary tumor and metastases) after conversion treatment with cytotoxic agents (5-FU, LV, Irinotecan) and biologicals (cetuximab or bevacizumab). Penzberg, Germany), quality assessment and type of data were reported previously, 2,12-14 and are briefly summarized in the Appendix S1, Material and Methods section. The 315 genes that were investigated by the above-mentioned assay are listed in the Table S1. The sequencing coverage and quality statistics for each sample are summarized in Table S2.

| Objectives
Main objective of this analysis was the exploratory comparison of DNA mutational profiles in paired samples of patients with metastatic colorectal cancer at baseline (eg before treatment start) and after secondary resection of metastases (post baseline). Further objectives were the assessment of mutated allele frequency changes of mutations, copy number alterations (CNA), tumor mutational burden (TMB) and RAS status (WT to MUT and vice versa). Objective response rate (ORR), progressionfree survival (PFS) and overall survival (OS) were defined as previously published, starting from randomization. 12 Furthermore, two additional exploratory survival endpoints were defined in this analysis. Disease-free survival (DFS) involved the time from secondary resection of metastases to the subsequent progression of disease. The postsurgical survival (PSS) was defined as the period from secondary resection of metastases to death by any cause. Patients alive were censored at the last patient contact, the last update on patient survival was performed in February 2021. 15

| Statistical analysis
Statistical analyses were pre-specified in a protocol analysis plan before start of the analysis. Logistic regression was applied to estimate the relative treatment benefit on ORR and the odds ratio (OR) of cetuximab versus bevacizumab and was calculated together with the 95% confidence intervals (CI). The Kaplan-Meier method and Cox proportional hazard models were used to estimate the relative treatment benefit on OS, PFS, DFS and PSS. Median survival as well as hazard ratios (HR) together with the 95% confidence intervals (CI) was provided. The t-test for paired samples compared median exon coverage and tumor mutational burden of matching samples analyzed at two timepoints. All P-values <.05 (two-sided) were considered significant. However, none of the analyses was powered for the comparisons made, as the sample size in this post-hoc analysis resulted from available tumor samples. Due to the exploratory nature of our study, no adjustment for multiple testing was applied. SAS 9.4 (SAS Institute, Cary, North Carolina) and R version 3.2.3 software were used for statistical analyses.    Table S3.) Date of secondary resection was not recorded in one patient (Table 1).

| Outcome of patients undergoing secondary resection
No differences in PFS and DFS were observed with regard to treatment arms (    (Table 3).
All changes are graphically displayed in Figure 2C. In summary, alterations associated with intrinsic resistance to cetuximab other than RAS mutations were observed in two patients after resection of

| Change of TMB status from baseline to post-baseline
Mean tumor mutational burden was lower post-baseline compared to baseline (5.56 vs 6.48 mutations per Mb). However, this difference was not significant (P = .23) ( Figure 2D).

| DISCUSSION
In this retrospective exploratory analysis of the randomized phase III FIRE-3 trial, we evaluated the response and outcome of patients who underwent secondary resection, and compared the mutational profiles of paired samples before treatment start (primary tumor) and after secondary resection of metastases.
Secondary resection of metastases after conversion of initially irresectable disease during cytotoxic treatment is recommended by current treatment guidelines 3,4 and has been observed in clinical trials at rates ranging between 15% and 80%, depending on whether the trial was designed for the evaluation of resectability or not and whether disease was liver-limited or not. [6][7][8][16][17][18] Resectability was retrospectively assessed at baseline and at best response in FIRE-3 and only 29% of all patients actually underwent resection, while the proportion of potential resectability was significantly higher. 5 In this subgroup analysis, a strong, but nonsignificant trend towards higher ORR and longer OS and PSS was observed in patients treated with cetuximab compared to bevacizumab. The neoadjuvant addition of cetuximab to FOLFIRI was previously associated with numerically higher resection rates. 6,7 Nevertheless, the addition of cetuximab in a perioperative setting (ie, neoadjuvant and adjuvant treatment) in KRAS exon 2 WT mCRC patients with (suboptimal) resectable liver metastases was associated with unfavorable outcome compared to chemotherapy alone in the NewEPOC trial. 18  Comprehensive molecular profiling confirmed the presence of RAS mutations, providing a higher sensitivity compared to pyrosequencing. 13 During cetuximab treatment, one patient with initial RAS WT mCRC developed a new KRAS G12D mutation, but eight patients did not. Additional RAS mutations with low mutated allele frequency were described as a mechanism of resistance by clonal selection. 19,20 Beyond RAS, inactivation of NF1 and gain of CNA in GNAS and SRC, respectively, were observed after resection of metastases in two patients, for which intrinsic resistance towards cetuximab has been reported previously. 21 In contrast, less alterations were detected in metastatic tissue of patients treated with bevacizumab, which is known to be an inhibitor of angiogenesis by blocking the vascular endothelial growth factor (VEGF). 27,28 Few biomarkers have been reported for efficacy of bevacizumab treatment such as chromosomal instability or angiogenesis activity, which has, however, not been considered in our analysis. 29,30 We can therefore not conclude if patients of FIRE-3 who underwent resection are particularly susceptible to anti-VEGF treatment or not.
Clonal evolution of tumor lesions must be acknowledged when analyzing more than one lesion by comprehensive genomic profiling.
We did not observe significant differences in terms of synchronous or metachronous, liver-limited or nonlimited metastatic disease in the mutational profiles of FIRE-3 patients. Some data suggest that mutational profiles of primary tumors remained consistent during systemic treatment, but that metastases are more heterogenous with a higher rate of private mutations. 31 Here, the timepoint of metastatic disease played a crucial role: while synchronous metastases showed a rate of concordance to primary tumors of 14%-84%, Our data rather support the option to biopsy a progressive lesion in case of rapid and unexpected disease progression or delayed metachronous disease. As a compromise, liquid biopsies could additionally assess the current status of known mutations in case of divergent results between primary tumor and metastases.
In this analysis, we were able to obtain paired DNA sequencing results of 25 well-described mCRC patients evaluated within a randomized controlled trial. Prior cohorts comparing DNA sequencing results between primary tumors and metastases were of comparable or lower sample size. However, the explanatory power remains limited, and the number of patients is too small to draw definite conclusions. The missing documentation for the underlying metastatic site for NGS testing, missing analyses of patients with irresectable disease and clonal heterogeneity within tumor lesions could have additionally biased our results.

| CONCLUSION
In conclusion, we observed largely comparable mutational profiles in patients with initially unresectable metastatic colorectal cancer before treatment start and after conversion and secondary resection of the