Molecular Biomarkers in the Personalized Treatment of Colorectal Cancer

Predictive and Prognostic Biomarkers

A prognostic factor is a clinical or biologic characteristic that is objectively measurable and that provides information on the likely outcome of the cancer in an untreated individual. Prognostic markers are helpful for identifying patients with cancer who are at high risk of metastatic relapse and therefore, potential candidates for adjuvant systemic therapy. In contrast, a predictive biomarker is defined as a marker which can be used to identify subgroups of patients who are most likely to respond (either in terms of tumor shrinkage or survival) to a given therapy. Importantly, prognostic factors define the effects of patient or tumor characteristics on patient outcome, whereas predictive factors define the effect of treatment on the tumor.

The personalized therapy of CRC has seen major advances in the setting of metastatic disease (stage IV) whereby monoclonal antibodies (mAbs) targeting tumor angiogenesis or the epidermal growth factor receptor (EGFR) signaling pathway have significantly improved patient survival when given in combination with cytotoxic chemotherapy ^16–19. Antibodies targeting vascular endothelial growth factor (VEGF) [bevacizumab] or its receptor (VEGFR; aflibercept) or EGFR (panitumumab/cetuximab) are now standard of care when combined with FOLFOX [5-fluorouracil (5-FU), leucovorin (LV) and oxaliplatin) or FOLFIRI (5-FU/LV and irinotecan) in stage IV patients ²⁰. To date, the identification of a predictive biomarker for anti-VEGF therapy remains elusive. In contrast to their benefit in the metastatic setting, the addition of antibodies against VEGF or EGFR to standard adjuvant FOLFOX chemotherapy in patients with lymph node-positive, i.e., stage III, colon cancer has failed to improve outcome in randomized trials ^21–23. Accordingly, there is no role for targeted therapy at this time in the treatment of non-metastatic CRC. In patients with stage III disease, micrometastases are responsible for disease relapse and the reasons for failure of targeted agents to improve their eradication awaits further study.

KRAS and BRAF oncogenes

KRAS and NRAS mutation status predict the efficacy of anti-EGFR antibodies with clinical benefit restricted to patient tumors with non-mutated KRAS or NRAS genes ^16–18. More than one-third of CRCs carry mutations in exon 2 of KRAS, and an additional 15% of tumors were found to carry mutations at exons 3 and 4 of KRAS and exons 2, 3 and 4 at NRAS that predict resistance to anti-EGFR antibody therapy ^{16, 17, 24, 25}. It is now recommended that expanded RAS mutation testing be performed for all patients being considered for anti-EGFR mAb treatment. In addition to RAS, mutations in the phosphoinositide 3-kinase catalytic subunit alpha (PIK3CA; exon 20 vs exon 9), which is part of the EGFR signaling pathway, may confer resistance to anti-EGFR therapies although further study is needed ²⁵. The association of mutations in KRAS with prognosis in patients with CRC has produced conflicting data ^26–28, yet recent data from adjuvant studies in node-positive, i.e., stage III, colon cancer have shown that KRAS exon 2 mutations are associated with poor clinical outcome ^{29, 30}. Furthermore, the adverse impact of KRAS mutations on prognosis appears to be stronger in the distal compared to the proximal colon cancers ³¹.

A subset (~8%) of CRCs carry a point mutation (V600E) in the BRAF oncogene that is mutually exclusive with mutation in KRAS ³². Patients whose tumors carry BRAF^V600E mutations have been consistently shown to have a poor prognosis in the metastatic setting ^33–36. In addition, data from clinical trials of adjuvant therapy in stage III disease also reveal an association between mutant BRAF^V600E and adverse outcome ^{14, 26, 29}. The adverse prognostic impact of mutant BRAF^V600E is particularly evident for overall survival and this finding may be explained, in part, by poor survival after recurrence that was observed for BRAF^V600E mutant colon cancers ^{37, 38}. In contrast to patients with BRAF^V600 mutant melanoma ³⁹, CRCs that harbor BRAF^V600E mutations were found to be resistant to inhibition of the BRAF/MEK/ERK signaling pathway by vemurafenib ⁴⁰. Resistance to vemurafenib was later found to be due to feedback activation of EGFR when BRAF is inhibited⁴¹. This finding has led to ongoing clinical trials that investigate combinations of inhibitors of BRAF, EGFR, and MEK with or without chemotherapy⁴². In a recent report, the combination of a BRAF inhibitor plus a MEK/ERK inhibitor demonstrated modest activity in a subset of patients with BRAF^V600E mutant metastatic CRC ⁴³.

DNA Mismatch Repair/Microsatellite Instability (MMR/MSI)

BRAF^V600E mutated vs non mutated tumors are associated with poorer clinical outcome in dMMR and proficient (p) MMR tumors, with its effect being somewhat attenuated in dMMR cancers³⁷.

Studies have consistently shown that CRCs with dMMR/MSI-H have a more favorable prognosis compared to those tumors with pMMR/ MSS ^{15, 25, 37, 44, 45}. These studies include clinical trials comparing surgery alone with 5-FU-based chemotherapy ^{25, 44}, and trials comparing 5-FU-based regimens ^{15, 37, 46}. Of note, the survival benefit of dMMR/MSI-H status appears to be stronger among colon cancers with earlier stage vs later stage disease ⁴⁶. Patients with dMMR/MSI-H stage II colon cancers have an excellent prognosis and do not receive benefit from 5-FU-based adjuvant therapy ^{47, 48}. In stage III patients, a pooled analysis of two adjuvant chemotherapy trials that evaluated FOLFOX alone or combined with cetuximab showed that patients with dMMR vs pMMR tumors had significantly better patient DFS and OS that was limited to the FOLFOX arm⁴⁹. Given that dMMR/MSI-H testing evaluates patients for LS and provides prognostic information, NCCN guidelines recommend such testing for all newly diagnosed CRC cases. While conflicting reports exist for the association of CIMP-positive vs negative tumors with clinical outcome, a meta-analysis concluded that CIMP was associated with a poor prognosis in both MSI-H and MSS tumors ^{50, 51}. In a recent study using only CIMP-positive tumors, overall survival was decreased in those lacking MLH1 methylation compared to those with MLH1 hypermethylation that is consistent with the close association between MLH1 hypermethylation and MSI-H and the reduction in significance after controlling for MSI status ⁵².

Tumors with dMMR/MSI-H are hypermutated and express abundant frameshift peptides that serve as neoantigens to elicit a brisk immune response characterized by abundant tumor infiltrating lymphocytes (TILs) ⁵³. CRCs with dMMR/MSI show increased expression of checkpoint proteins (PD-1, PD-L1, CTLA-4, LAG-3, IDO) compared to pMMR/MSS tumors that help enable them to evade immune destruction ⁵⁴. In a recent phase II study, metastatic CRCs with MSI-H were treated with an anti-programmed death ligand-1 (PD-1) antibody (permbrolizumab) which achieved a 62% objective response rate compared to 0% for pMMR/MSS CRCs that did not respond to this therapy ⁵⁵. Progression-free survival has not yet been reached for the treatment arm of this small patient cohort that included sporadic dMMR/MSI-H and Lynch Syndrome cases as well as some extracolonic malignancies associated with this syndrome. While only ~4% of metastatic CRCs are dMMR/MSI-H indicating that the target population is relatively small ⁵⁶, the potential clinical benefit appears to be quite substantial and may potentially apply to earlier stage patients.

Classification by Molecular Subtype

While single gene analysis has been shown to provide valuable predictive and prognostic information, analyses from multiple studies support the feasibility of classifying CRCs by combining molecular markers or using gene expression profiling. The identification of CRC subtypes that link molecular features to clinical outcome can guide patient stratification and inform the rational design of drugs targeting specific pathways. In participants in an adjuvant chemotherapy trial, the combination of KRAS and BRAF^V600E mutations with the status of the DNA MMR system was shown to identify colon cancer subtypes with distinct clinicopathological features and prognoses. These data were then validated in an independent patient cohort (Fig. 1)¹⁴. Evaluation of a similar molecular classifier that included CIMP status yielded similar results in a large population-based cohort of individuals with varying stages of CRC ⁵¹, suggesting the clinical relevance of this classifier. Several studies have utilized gene expression profiling to develop expression signatures that can identify distinct tumor subgroups ^3–7. Using an unsupervised classification strategy involving over 1,100 individuals with colon cancer, three main molecular subtypes were identified based upon genetic, epigenetic and clinical features ⁴. These colon cancer subtypes (CCS) included CCS1 tumors that were characterized by mutations in KRAS and TP53 genes, displayed increased Wnt signaling, and showed evidence of marked CIN. CCS2 cancers were strongly enriched for CIMP-positive tumors, showed enhanced immune cell infiltration, and were frequently located in the right colon. CCS3 tumors consisted of both MSI-H and CIN tumors, but were enriched for BRAF^V600E and PIK3CA mutations and displayed a mesenchymal phenotype. A similar molecular classification developed by other investigators included at least three major subtypes that were largely based on three biological hallmarks of the tumor: epithelial-to-mesenchymal transition, deficiency in MMR genes and MSI-H, and cellular proliferation ⁶. In these classifications, tumors with dMMR/MSI-H had the best prognosis whereas tumors with a mesenchymal expression phenotype had the worst outcome. These molecular subtypes show overlap with previous classification systems that include CIN, MSI, CIMP, or KRAS/ BRAF^V600E mutation status, but cannot be identified solely based on single mutations or epigenetic features.

Current classifiers show superficial similarities, yet important differences exist that are likely related to differences in data processing and algorithms applied to diverse patient cohorts, sample preparation methods, and gene expression platforms. To address this issue, a CRC Subtyping Consortium was established to perform a cross-comparison of subtype assignments obtained by the various approaches on a common set of tumor samples ⁵⁷. This effort identified four molecular subtypes (CMS) with distinguishing features: CMS1 [MSI immune, 14%] showing hypermutation; CMS2 (canonical, 37%) that is epithelial with marked WNT and MYC signaling activation; CMS3 (metabolic, 13%) that is epithelial with metabolic dysregulation; and CMS4 (mesenchymal, 23%) showing prominent transforming growth factor (TGF)-β activation, stromal invasion and angiogenesis. While validation in independent cohorts preferably from clinical trials is needed, the CMS groups can be considered to be the most robust classification system currently available for CRC with a clear biological basis that may inform future clinical stratification and subtype-based interventions.