Genomic alterations associated with early-stage disease and early recurrence in patients with colorectal cancer

Abstract

Background

The molecular characterization of early-stage (1-3) colorectal cancer (CRC) remains incomplete, as opposed to metastatic disease, where comprehensive genomic profiling (CGP) is routinely performed. This study aimed to characterize the genomics of stages 1-3 versus IV CRC, and the genomics of patients recurring within 1 year of diagnosis.

Patients and Methods

Patients from a de-identified CRC clinico-genomic database who received Foundation Medicine testing (FoundationOne/FoundationOne CDx) during routine clinical care at approximately 280 US cancer clinics between March 2014 and June 2023 were included. Genomic alterations (GA) were compared by Fisher’s exact test.

Results

A total of 4702 patients were included; 1902 with stages 1-3 and 2800 with stage 4 disease. Among patients with stages 1-3 disease, 546 recurred within 1 year. Patients staged 1-3 had higher prevalence of microsatellite instability (MSI-H, 11.4% vs 4.5%, P < .001), tumor mutational burden (TMB) ≥ 10 Mut/Mb (14.6% vs 6.8%, P < .001), GA in RNF43 (11.2% vs 5.7%, P < .001), MSH6 (3.9% vs 1.7%, P < .001), MLH1 (2.3% vs 0.7%, P < .001), and MSH2 (1.5% vs 0.6%, P < .01) compared to those with stage 4 disease. Patients who recurred within 1 year had higher prevalence of MSI-H (13.2% vs 4.4%, P < .001), TMB ≥ 10 Mut/Mb (16.2% vs 6.9%, P < .001), BRAF V600E (17.2% vs 7.9%, P < .003), GA in RNF43 (13.7% vs 5.3%, P < .001), MSH6 (4.2% vs 1.6%, P = .035), and BRCA1/2 (6.2% vs 3.0%, P = .030). On recurrence, more patients received targeted therapy when CGP was performed before versus after first-line therapy (43% vs 19%, P < .001).

Conclusions

Early-stage CRC patients can have distinct genomic profiles and CGP in this population can help expand access to targeted therapies.

Introduction

Colorectal cancer (CRC) remains a disease with high incidence and mortality worldwide, with an estimated 153 020 new cases and 52 550 deaths in the United States in 2023, representing the third most common cause of cancer-related death.¹ The molecular characterization of early-stage CRC remains incomplete, as opposed to metastatic disease where comprehensive genomic profiling (CGP) is performed as part of standard of care and where it has resulted in significant improvements in outcomes. Given that over 80% of disease at initial presentation is localized,² an improved understanding of the unique molecular characteristics of this population is paramount. Early-stage CRC disease is heterogeneous, and to date, chemotherapy is the only therapy offered in higher-risk patients based on clinical and pathological criteria. There may be an opportunity with genomic profiling to both identify which patients may benefit most from intensification of systemic therapy^3,4 and to identify patients with targetable alterations, as targeted therapies are transforming the management of early-stage disease across malignancies, albeit not yet in CRC.^5-9 Indeed, in lung cancer molecular characterization of early-stage disease has become increasingly part of standard clinical practice, given increasing evidence of the presence of actionable alterations in localized disease whose therapeutic targeting has either improved clinical outcomes or is under clinical investigation.^8,10-12

Early attempts at molecular subtyping of CRC have demonstrated the predictive and prognostic potential of genomic signatures in localized disease. One of the most notable molecular categorization schemes is the consensus molecular subtypes (CMS) classification, which subdivides CRC into 4 subtypes based on gene expression data from over 3000 patients.¹³ The CMS classification has been found to enhance the prognostic ability of the TNM staging system in localized disease^13,14 as well as the predictive ability of adjuvant chemotherapy.^14,15 However, the CMS classification was never clinically validated to inform clinical decision-making. Studies in smaller groups of early-stage patients have shown that ZNF217, MET, and PKHD1 alterations are enriched in early-stage patients compared to the metastatic setting.¹⁶KRAS mutations have been found to occur between 51% and 67% in metastatic disease^17,18 as opposed to 26%-33%^19,20 in localized disease, with an opposite trend for BRAF V600E mutations that occur in 6%-7% of metastatic cases^21,22 as opposed to 8%-14%^23-25 of patients with localized disease. HER2 overexpression is a relatively rare event in CRC, although with important therapeutic implications, and has been found to occur in 1%-2% of early-stage and advanced disease.²⁶ More recently, in-depth multiomic sequencing efforts of CRC including whole exome, whole-genome, and RNA sequencing data have demonstrated a far greater complexity in the molecular diversity of CRC.^27,28

Patients with stages 2 and 3 CRC still have a moderate risk of cancer recurrence. Ten-year follow-up data from the pivotal MOSAIC study comparing adjuvant FOLFOX and 5-FU demonstrated a 25% and 38% recurrence risk for stages 2 and 3 CRC, respectively, following definitive therapy with surgery and adjuvant chemotherapy.²⁵ The addition of oxaliplatin improved overall survival (OS) for patients with BRAF V600E mutations and microsatellite stable (MSS) disease. Similarly, a recent meta-analysis of 25 studies found a 5-year disease-free survival (DFS) of 79% and 63.6% for patients with stages 2 and 3 disease treated with adjuvant chemotherapy.²⁵ The greater risk for recurrence is in the first 2 years, with recurrence rates decreasing steadily thereafter with as low as 1.5% annual risk of recurrence after 5 years of completion of therapy.²⁹ Though high-risk genetic aberrations such as BRAF, KRAS, and MSI-H disease and familial syndromes have been well described across stages in CRC and have been shown to significantly impact DFS and OS,^30,31 the genomics landscape of localized disease presenting with early as opposed to late recurrence is still not well characterized.

By analyzing a cohort of 4702 patients with CRC, we comprehensively characterized the genomics of early-stage CRC (stages 1, 2, and 3) patients to evaluate potentially targetable alterations seen in earlier-stage disease. We also specifically studied patients who recurred within a year of diagnosis to evaluate the genomic landscape of this higher-risk population which still lacks comprehensive characterization. Finally, we evaluated the utility of CGP performed for patients with stages 1-3 disease, to inform our understanding of the use of targeted therapy on recurrence in a real-world cohort. To our knowledge, this study constitutes one of the most comprehensive real-world studies of genomic profiling of early-stage CRC and demonstrates the utility of tumor profiling in localized disease.

Methods

Study population

All patients included in this study had a confirmed diagnosis of primary CRC, underwent tissue genomic testing using Foundation Medicine CGP assays, and were included in the US-wide Flatiron Health and Foundation Medicine clinico-genomic CRC database (FH-FMI CGDB) between March 2014 and June 2023. Retrospective de-identified longitudinal clinical data were derived from electronic health records from approximately 280 US cancer clinics (~800 sites of care), comprising patient-level structured and unstructured data, curated via technology-enabled abstraction of clinical notes—including oncology clinician-defined, rule-based lines of therapy—and radiology/pathology reports. These were linked to genomic data derived from Foundation Medicine testing by de-identified, deterministic matching.³²

Patients with tissue biopsy collected from the primary CRC site within 3 months from the initial diagnosis were separated into 2 groups for genomic comparisons; patients with stages 1-3 disease and stage 4 disease at diagnosis. Patients with stages 1-3 were further separated into 2 other groups for genomic comparisons; patients who recurred (metastasis) within ≤1 year from the initial diagnosis (“early recurrence”) versus >1 year from the initial diagnosis (“late recurrence”; Figure 1).

Institutional Review Board approval of the study protocol was obtained prior to the study conduct and included a waiver of informed consent based on the observational, noninterventional nature of the study (WCG IRB, Protocol No. 420180044).

Comprehensive genomic profiling

Hybrid capture-based NGS assays (FoundationOne or FoundationOne CDx) were performed on patient tumor specimens in Clinical Laboratory Improvement Amendments (CLIA)–certified, College of American Pathologists (CAP)-accredited laboratory (Foundation Medicine, Inc.). Samples were evaluated for genomic alterations (GA) as previously described.³³ Tumor mutational burden (TMB) was determined on up to 1.1 Mb of sequenced DNA with the clinical and analytic validation as previously described.³⁴ To determine microsatellite instability (MSI) status via NGS, we used a fraction-based MSI algorithm to categorize a tumor as MSI-H, MSI-intermediate (MSI-L), or MSS. This algorithm calculates the fraction of unstable microsatellite loci based on an analysis across >2000 loci. PD-L1 22C3 was run according to manufacturer instructions in a CLIA-certified and CAP-accredited laboratory (Foundation Medicine, Inc) and scored with a tumor proportion score = number of PD-L1-positive tumor cells/(total number of PD-L1-positive + PD-L1-negative tumor cells). All patient cases were tested with manufacturer-recommended system level controls, H&E stained slide, negative reagent control slide, and PD-L1 22C3 IHC slide.

Statistical analysis

Chi-square tests and Wilcoxon rank-sum tests were used to assess differences between groups of categorical and continuous variables, respectively. We first evaluated the frequency of known or likely pathogenic actionable GAs in CGP specimens for all eligible CRC cases and compared the frequencies between stages 1-3 versus stage 4 disease and between early and late recurrences. Actionable GAs evaluated included KRAS mutations, BRAF V600E mutation, NRAS mutations, BRCA1/2 alterations, ERBB2 amplification, MSH2 mutations or deletion, MSH6 mutations or deletion, MLH1 mutations or deletion, PMS2 mutations or deletion, NTRK1/2/3 rearrangements, RET rearrangements, ALK rearrangements, ROS1 rearrangements, RNF43 mutations, and POLE/POLD1 mutations. We then compared the frequency of alterations in all 324 genes included in the FoundationOne or FoundationOne CDx assays. Because the associations can be influenced by the different prevalence of MSI-H and TMB ≥ 10 mut/Mb among groups, those genes that were differentially altered in a statistically significant manner in the latter analysis were then compared between the groups in patients with MSS disease and no POLE/POLD1 mutations. Additional subgroup analyses were performed, including only MSI-H patients, and stratified by tumor side (right and left), sex, age (<50 years and 50+ years), tumor location (rectum and colon), and by metastatic systemic treatment (received vs did not receive treatment). All comparisons were made using Fisher’s exact test, adjusted for multiple comparisons. R version 4.1.3 software was used for all statistical analyses.

Results

Genomic comparison of stages 1-3 and stage 4 CRC

We included a total of 12 639 CRC tissue CGPs, of which 6196 were collected from the primary CRC site. After excluding those specimens with an unknown collection or diagnosis date, unknown stage at diagnosis, or with specimens collected after 3 months of the diagnosis date, 1902 patients with stages 1-3 and 2800 patients with stage 4 disease at diagnosis were included in the analysis (Figure 1).

Patients with stages 1-3 disease were older (62 vs 60 years, P < .001) and more frequently had tumors with high TMB (P < .001) and microsatellite instability (P < .001; Table 1). Among actionable alterations, RNF43 mutations, MSH6, MLH1, and MSH2 mutations/deletions were more prevalent in patients with stages 1-3 compared to stage 4 disease (Figure 2A). However, when restricting the analysis to MSS and non-POLE/POLD1 mutated CRC, these differences were no longer observed (Supplementary Figure S1A). No differences were observed for specific KRAS mutations (Figure 2B). A list with the numerical frequencies of the top 50 gene alterations by stage is presented in Supplementary Table S1.

Differences in the prevalence of some GA were found when evaluating all genes baited on the CGP assay (Figure 2C). However, when restricting the analysis to MSS and non-POLE/POLD1 mutated CRC, these differences were no longer observed (Figure 2D; Supplementary Figure S1C). No significant differences were found when analyzing patients with MSI-H and POLE/POLD1 mutated disease (Supplementary Figure S2). Patients with right-sided and early-stage disease had higher frequency of RNF43, BRCA1/2, MLH1, MSH6, MSH2, and PMS2 alterations compared to patients with right-sided stage 4 disease (Supplementary Figure S3). Female and male patients with early-stage disease were more likely to have BRAF V600E, RNF43, MLH1, MSH6, and BRCA1/2 alterations as opposed to those with stage 4 disease (Supplementary Figure S4), and older patients with early-stage disease were more likely to have RNF43, BRCA1/2, MLH1, MSH6, and MSH2 alterations compared to their counterparts with stage 4 disease (Supplementary Figure S5). Furthermore, when analyzing separately patients with rectal cancer from patients with colon cancer, we found no significant differences in GA across patients with rectal cancer but found that patients with early-stage colon cancer had a significantly higher frequency of RNF43, BRCA1/2, MSH6, MLH1, and MSH2 alterations compared to those with stage 4 disease (Supplementary Figure S6).

Genomic comparison of early and late recurrence

Among the 1902 patients with stages 1-3 disease at diagnosis, 546 patients recurred within 1 year from the initial diagnosis with a median recurrence time of 203 days (early recurrence) and 758 patients recurred after 1 year from initial diagnosis with a median recurrence time of 651 days (late recurrence). Patients with early recurrences were older (63 vs 61 years) and more frequently had tumors that were MSI-H and TMB ≥ 10 (Table 2).

Among potentially actionable alterations, BRAF V600E, RNF43 mutations, BRCA1/2 alterations, and MSH6 mutations/deletions were more prevalent in patients with early recurrences compared to late recurrences (Figure 3A). However, when restricting the analysis to MSS and non-POLE/POLD1 mutated CRC, these differences were no longer observed (Supplementary Figure S7A). Similarly, when evaluating patients with MSI-H and POLE/POLD1 mutated disease, we found no statistically significant differences between those with early versus late recurrence; however, there was a trend toward more RNF43 and BRAF V600E alterations in patients with early recurrences (Supplementary Figure S8). No differences were observed for specific KRAS mutations (Figure 3B). A list with the numerical frequencies of the top 50 gene alterations stratified by timing of recurrence is presented in Supplementary Table S2.

Differences in the prevalence of multiple GA were found when evaluating all genes baited in the NGS panel including actionable and non-actionable alterations (Figure 3C). However, when again restricting the analysis to MSS and non-POLE/POLD1 mutated CRC, these differences were no longer observed (Figure 3D). We next evaluated genomic differences between patients stratified on receipt of systemic therapy prior to diagnosis of metastatic disease. Patients with an early recurrence pattern were more likely to have BRAF V600E and RNF43 mutations irrespective of receipt of prior systemic therapy; however, the magnitude of the difference in the frequency of these alterations was higher in patients who had not received prior systemic therapy for localized disease. Those patients were also more likely to have MLL2 alterations. In contrast, patients who had received prior systemic therapy and had an early recurrence were more likely to have BRCA1/2 alterations (Supplementary Figure S9).

Patients who recurred and had a CGP report provided before the start of therapy tended to receive more precision therapy in the first line

We found that in our patient cohort, CGP reports for early stages of CRC increased numerically each year between 2014 and 2023 (P < .001; Figure 4A). Among patients with left-sided, RAS/RAF wild-type tumors that recurred, for those that had a CGP report available before first-line therapy, 43.0% received anti-EGFR therapy (in accordance with NCCN guideline recommendations), while 19.0% of patients with a GCP report available after first-line therapy initiation received anti-EGFR therapy in the first line (P = .0002; Figure 4B).

Discussion

The genomic characterization of early-stage CRC remains limited. In this study, we analyzed a large well-annotated real-world cohort of patients with early-stage and metastatic disease who underwent CGP. As opposed to their metastatic counterparts, patients with early-stage disease more often had mutations in mismatch repair genes and RNF43, although when examining patients with MSS disease, no differences were observed between patients with metastatic and early-stage disease. In our overall cohort, we found that the 10 most frequently altered genes in patients with both localized and metastatic disease were APC, TP53, FBXW7, RNF43, ARID1A, MLL2, ASXL1, PTEN, ATM and CTNNB1. (Supplementary Table S1). These results confirm that most patients with early-stage and metastatic disease in CRC occur in a continuum, with little genomic differences between them. However, significant differences were observed when stratifying patients by age of diagnosis, gender, and sidedness, indicating that these epidemiologic variables correlate with distinct tumor genomic profiles and biology.

Despite encouraging DFS rates for stages 2 and 3 colorectal cancer, anywhere from 15% to 40% of patients recur,³⁵ dependent on clinical and pathologic risk factors, with the majority of those recurrences seen in the first 2 years following completion of therapy. The genomics of patients who recur early has been poorly characterized to date. In our overall cohort, we found that the most commonly altered genes in patients who recurred early were APC, TP53, PIK3CA, BRAF, SOX9, RNF43, ARID1A, MLL2, PTEN and ATM. In patients with disease that recurred late, the 10 most frequently altered genes were the same but with differences in terms of the individual frequencies at which they were mutated compared to patients who recurred early (Supplementary Table S2). Patients with early recurrences were also relatively enriched for TMB high and MSI-H disease. However, among patients who were MSS, no significant genomic differences were seen between patients with early and late disease recurrences. Among patients with a known status of receipt of systemic therapy, patients with recurrent disease within one year from diagnosis had a relatively higher frequency of BRAF V600E and RNF43 alterations, and the magnitude of this difference was higher for patients who did not receive prior systemic therapy. The patients who did not receive prior systemic therapy were more likely to have low-risk stage 2 disease on diagnosis. These findings are not surprising, given that BRAF is a known factor of high risk for disease progression³⁶ and RNF43 is a key tumor suppressor in CRC.³⁷ Interestingly, KRAS mutations were found in similar proportions across patient subsets, likely due to enrichment of recurrent disease for BRAF mutations which are more proximal in the MAP kinase pathway.

Our study also demonstrated the utility of CGP for informing decision-making in a real-world setting. Between 2014 and June 2023 we found a consistent annual increase in CGP for patients presenting with early-stage disease. Most notably, we found that patients diagnosed with early-stage disease who had a CGP result prior to initiation of first-line therapy for recurrent/metastatic disease, more often received EGFR-targeted therapy after recurrence as opposed to patients who did not have a CGP result available at the time of first-line therapy initiation. As more targeted therapies such as KRAS inhibitors,^38,39 HER2 targeting approaches⁴⁰ and immunotherapy^41-43 enter the treatment paradigm of CRC, it will be important to have CGP available at the time of diagnosis of recurrent disease, in order to best inform future therapeutic strategies. Our study, which constitutes the most comprehensive study of the genomic landscape of early-stage and early-recurrent CRC to date, highlights the higher frequency of key alterations such as BRAF V600E, RNF43, and MSI-H disease in these patients, results that can have a significant impact on informing therapeutic decisions.

Our study is limited by the fact that our database may not reflect a generalizable cancer population, as all patients currently included in the FH-FMI CGDB need to have a CGP performed during routine practice. Out of the 1902 patients with early-stage CRC, 598 (31.4%) do not have a recurrence date documented, suggesting that the recurrence rate in our early-stage cohort is at least 68.6%. This rate is higher than what would be expected in this population, typically at 25%-26% for stage 3 disease, indicating a form of selection bias in our population which was likely skewed toward the inclusion of patients with higher-risk diseases. Similarly, our database contains alterations assessed by the Foundation Medicine genomics platform, which may slightly differ from other available genomic assays, although there is no expected significant difference between most genomic profiling platforms for the key cancer-associated genes. Finally, our study was limited to patients who had NGS testing sent from their tumor, which for early-stage patients may lead to recruitment bias from centers that often profile such patients and recruitment of more high-risk early-stage patients.

Conclusions

Treatment paradigms for metastatic CRC have recently become more nuanced, with first-line immune checkpoint blockade now considered standard of care for MSI-H disease and BRAF V600E and KRAS G12C targeted therapies considered valid second-line options. The introduction of targeted therapies to early-stage disease has been more delayed, although recent landmark studies have demonstrated the superior benefit of neoadjuvant immune checkpoint blockade in locally advanced MSI-H colon⁴² and rectal cancer.⁴¹ Studies examining the utility of BRAF-targeted therapy in the adjuvant setting, whether via a minimal-residual disease-guided approach [NCT03803553] or a consolidative adjuvant approach [NCT05710406] are ongoing. Our study indicates that knowledge of the CGP in early-stage disease can affect treatment decisions for recurrent disease and that potentially targetable alterations can occur with clinically meaningful frequency in patients with localized and early recurrent disease. This work also highlights that further studies are needed to understand which patients may benefit most from molecular profiling. The increased adoption of CGP in early-stage CRC and more studies that can expand the breadth and depth of genomic information available for these patients will be important in expanding access to targeted therapies for patients with recurrent colorectal cancer.

Supplementary material

Supplementary material is available at The Oncologist online.

Acknowledgments

We thank the patients whose data made this research possible, the clinical and laboratory staff at Foundation Medicine, and the team at Flatiron Health.

Author contributions

Leon Pappas: data analysis and interpretation, manuscript writing, and final approval of manuscript. Julia C. F. Quintanilha: provision of study material, data analysis and interpretation, manuscript writing, and final approval of manuscript. Richard S. P. Huang: conception/design, provision of study material, data analysis and interpretation, and final approval of manuscript. Aparna Parikh: data analysis and interpretation and final approval of manuscript.

Conflicts of interest

J.C.F.Q. and R.S.P.H. are employees of Foundation Medicine, a wholly owned subsidiary of Roche, and have equity interest in Roche. A.R.P. has equity in C2i Genomics, Khora, XGenomes, Cadex, and Parithera. In the last 36 months, she has served as an advisor/consultant for Amgen, Eli Lilly, Mirati, Pfizer, Inivata, Biofidelity, Checkmate Pharmaceuticals, FMI, Guardant, Abbvie, Bayer, Delcath, Taiho ,CVS, Value Analytics Lab, Seagen, Saga, AZ, Scare Inc., Illumina, Taiho, Hookipa, Kahar Medical, Xilio Therapeutics, Sirtex, Takeda, Merck, MPM Capital, and Science For America. She receives fees from Up to Date. She has received travel fees from Karkinos Healthcare. She has been on the DSMC for a Roche study and on the Steering Committee for Exilixis. She has received research funding to the Institution from PureTech, PMV Pharmaceuticals, Plexxicon, Takeda, BMS, Mirati, Novartis, Erasca, Genentech, Daiichi Sankyo, Syndax, Revolution Medicine, and Parthenon. L.P. has served as an advisor/consultant for Astellas Pharma and CARIS Life Sciences and has received fees from Curio Science.

Funding

Foundation Medicine, a wholly owned subsidiary of Roche, is a for-profit company and producer of FDA-regulated molecular diagnostics. Authors employed by Foundation Medicine were involved in the design and conduct of the study, analysis, interpretation of the data, preparation, review, and approval of the manuscript.

Data Availability

The data that support the findings of this study originated by Flatiron Health, Inc. and Foundation Medicine, Inc. Requests for data sharing by license or by permission for the specific purpose of replicating results in this manuscript can be submitted to [email protected] and [email protected].

References

Author notes