Biomarker-Based Prediction of Response to Therapy for Colorectal Cancer: Current Perspective
Abstract
The diagnosis and management of colorectal cancer (CRC) has been impacted by the discovery and validation of a wide variety of biomarkers designed to facilitate a personalized approach for the treatment of the disease. Recently, CRC has been reclassified based on molecular analyses of various genes and proteins capable of separating morphologic types of tumors into molecular categories. At the same time, a number of new prognostic and predictive single genes and proteins have been discovered that are designed to reflect sensitivity and/or resistance to existing therapies. Multigene predictors have also been developed to predict the risk of relapse for intermediate-stage CRC after completion of surgical extirpation. More recently, a number of biomarkers tested by a variety of methods have been proposed as specific predictors of chemotherapy and radiotherapy response. Other markers have been successfully used to predict toxic effects of standard therapies. In this review, a series of novel biomarkers are considered and compared with standard-of-care markers for their potential use as pharmacogenomic and pharmacogenetic predictors of disease outcome.
Sessile serrated adenoma with invasive adenocarcinoma. Low-power magnification of a sessile serrated polyp/adenoma with an area of invasive carcinoma arising in the center of the polyp. Invasive colorectal cancer (CRC) associated with sessile serrated precursor lesions appears to be a distinct molecular subclass of sporadic CRC. Approximately 10% to 15% of sporadic CRCs arising in the proximal colon may feature a serrated precursor lesion with diploid microsatellite instability, frequent DNA methylation, KRAS mutation, loss of chromosome 1p, and overexpression of genes such as Ephrin receptor B2 and hypoxia-inducible factor 1-α12–15 (H&E, ×20).
Molecular Classification of CRC
Colorectal cancer (CRC) is the second leading cause of cancer mortality in the United States.1 There are 160,000 new cases of CRC diagnosed each year and 57,000 CRC-related deaths in the United States.1 Despite a slow decline during the last 20 years, for men, the age-adjusted incidence is 61.2 cases per 100,000 and for women, it is 44.8 per 100,000.1
Pathologists have traditionally had a major role in the care of patients with newly diagnosed CRC in the initial establishment of the diagnosis of invasive disease and in the morphologic classification and final staging of surgically resected specimens.2,3 In the last 30 years, there has been significant advancement in the understanding of the molecular origins of CRC and characteristics of tumor aggressiveness.4–7 This major expansion of genomic and proteomic data has simultaneously contributed to the discovery of novel targeted therapies for patients with CRC in whom recurrent and metastatic disease have developed.8–10 Moreover, the use of gene sequencing, expression profiling, and other molecular techniques has introduced putative biomarkers that have been reported to predict the clinical response and toxic effects for the nontargeted traditional antineoplastic drugs that have been used to the treat the disease.11 These new techniques have led to the discovery of a number of biologically interesting genes and pathways associated with CRC development and progression, as well as new biomarkers that are emerging as bedside clinical tests for the prediction of response to therapy for the disease. As a result, practicing pathologists have now become responsible for overseeing the performance of these emerging tests designed to personalize the selection of postsurgical treatment for CRC.
This review is focused primarily on the discovery and development of predictive biomarkers designed to predict therapy-specific disease outcome parameters such as overall and recurrence-free survival. Prognostic markers are also considered, and their interactions with predictive markers and indirect impact on therapy selection and disease outcome are also reviewed.
In the past several years, several groups have attempted to produce a molecular classification of CRC based on a number of features, including chromosomal instability (tendency for chromosome breakage [CIN]), microsatellite instability (defective DNA repair capability [MSI]), and frequent CpG island hypermethylation (gene silencing due to methylation of the promoter gene sequence [CIMP]).12–15 In contrast with the breast cancer molecular “portraits” system that has achieved wide acceptance by pathologists and oncologists,16 the CRC molecular classification continues to undergo refinement in the number of classes and their clinical practice applicability.
Prediction of Response to Primary Treatment for CRC
Pathologic Staging
The molecular classifications of CRC have uniformly attempted to link DNA stability and gene expression status with 7 factors: (1) tumor type, (2) tumor grade, (3) extent and distribution of associated inflammatory changes, (4) presence of “serrated” architecture, (5) hereditary basis, (6) pathologic stage, and (7) general prognosis.12–15 However, it should be noted that the molecular profiles that characterize the discrete pathogenic pathways (CIN, serrated, and CIMP) differ significantly from the gene expression patterns that reflect the predictive and prognostic gene expression profiles that are based primarily on clinical outcome associations.
Molecular Staging
A proposed molecular classification of CRC is summarized in Table 1, and examples of the morphologic features of specific molecular classes are shown in Image 1, Image 2, and Image 3. This classification has been developed by using a variety of molecular techniques, including fluorescence in situ hybridization, comparative genomic hybridization, DNA sequencing, and routine and methylation-specific messenger RNA (mRNA) expression profiling. To date, although of significant interest, the molecular classification of CRC has not achieved widespread clinical use, nor has it been validated or reproduced by multiple institutions on a large scale.
The Molecular Classification of CRC*
Open in new tab
The staging of resected CRC by pathologists remains the cornerstone for the prediction of future disease relapse and progression.2 After many decades of use of the staging method of Dukes et al, the current TNM staging system has now achieved near universal use.2,17 Recent attention has focused on understaging of disease caused by the evaluation of an insufficient number of retrieved lymph nodes from the resected pericolorectal mesentery or perirectal adventitia.18 The minimum number of lymph nodes recovered for accurate CRC TNM staging has varied from 10 to 18.18,19
Molecular class 1 colorectal cancer (CRC) with high microsatellite instability (MSI-H) and high CpG island hypermethylation. This image is taken from a poorly differentiated CRC arising in the proximal colon in an elderly woman. Note the high nuclear grade and intratumoral lymphocytic infiltrates. In other regions, signet-ring cell histologic features and mucinous differentiation were noted. This case of sporadic MSI-H CRC has been termed medullary carcinoma in some studies and typically features a favorable prognosis (H&E, ×200).
Molecular class 1 colorectal cancer (CRC) with high microsatellite instability (MSI-H) and high CpG island hypermethylation. This image is taken from a poorly differentiated CRC arising in the proximal colon in an elderly woman. Note the high nuclear grade and intratumoral lymphocytic infiltrates. In other regions, signet-ring cell histologic features and mucinous differentiation were noted. This case of sporadic MSI-H CRC has been termed medullary carcinoma in some studies and typically features a favorable prognosis (H&E, ×200).
Oncotype Dx Colon Cancer Test
Cytotoxic Chemotherapy for CRC: Biomarkers for the Prediction of Efficacy and Toxicity
ColoPrint
A number of recently introduced technologies have enabled the detection of circulating tumor cells (CTCs) in patients with early-stage and advanced CRC.28 In a recent meta-analysis of 36 published studies, the identification of CTCs in peripheral blood was an independent predictor of reduced recurrence-free (HR, 3.24) and overall (HR, 2.28) survival.29 However, to date, the ability of CTC determinations to guide therapy selection and the changing of therapy in CRC has not been confirmed in prospective clinical trials.30
Compiled from a series of reviews,31–36Table 2 lists a series of prognostic biomarkers studied in patients with CRC. Although a significant number of prognostic factors have achieved independent prognostic significance for disease-free and overall survival, no single routine or molecular test has achieved widespread use in daily practice.37,38 To date, the pathologic stage of disease based on primary resection specimen evaluation remains the cornerstone of prognosis assessment and likely success of the primary therapy for CRC.39,40
5-Fluorouracil (5-FU) and its oral prodrug, capecitabine (Xeloda) are the cornerstones of multiagent chemotherapy for relapsed and metastatic CRC.10,11 These drugs are part of the 2 major regimens for CRC treatment: FOLFIRI (5-FU, folinic acid [Leucovorin], and irinotecan [Campostar]) and FOLFOX (5-FU, folinic acid [Leucovorin], and oxaliplatin [Eloxatin]). Three biomarkers have been studied widely as potential predictors of 5-FU-based chemotherapy response and potential serious drug toxic effects.
Treatment of CRC with 5-FU has long been associated with wide variations in tumor response and drug toxicity.41,42 Direct inhibition of thymidylate synthase (TS, also abbreviated TYMS) is considered the major mechanism of 5-FU anti-cancer activity. Although a series of early reports indicated that increased TS expression was associated with resistance to 5-FU treatment,43 numerous recent studies designed to measure TS protein expression by slide-based immunohistochemical analysis and mRNA expression by RT-PCR as a predictor of 5-FU response in CRC have failed to achieve consensus or widespread clinical adoption.40–42 Thus, although the majority of studies have found that TS overexpression is an unfavorable prognostic and predictive factor for 5-FU–treated CRC, the test is not routinely used before the selection of 5-FU–based therapies. Germline polymorphisms in the TS gene seem to have a major role in the regulation of TS expression, but this approach has focused more on predicting 5-FU toxic effects than efficacy.42,43 TS polymorphisms associated with low TS expression are consistently associated with increased 5-FU drug toxic effects, including bone marrow suppression, mucositis, and diarrhea.
Dihydropyrimidine dehydrogenase (DPD, also abbreviated DPYD) is an enzyme associated with the breakdown of administered 5-FU. Compared with TS, studies of DPD expression as a predictor of 5-FU efficacy are generally less convincing.11,44,45 In certain patients with germline polymorphisms of the DPD gene, significant reduction in enzymatic activity is present, resulting in severe toxic effects when 5-FU is administered.11,44,45
Although a number of studies have linked a reduction in folic acid pools associated with the presence of methylenetetrahydrofolate reductase (MTHFR) germline polymorphisms, large studies have not confirmed the prognostic or predictive value of MTHFR genotyping.46–48MTHFR polymorphisms have also been associated with capecitabine toxic effects.11
CRCs deficient in DNA mismatch repair (molecular classes 1 and 2 with high MSI [MSI-H] tumors) typically feature a proximal anatomic location, mucinous features, lymphocytic infiltration, and pushing margins. There is preclinical evidence that these tumors are resistant to 5-FU but are sensitive to irinotecan and mitomycin C.49
Circulating Tumor Cells
In summary, the aforementioned genetic and molecular tests designed to predict efficacy and toxic effects for 5-FU and capecitabine have shown a high percentage of false-negative results, possibly due to incomplete sequencing and pathway analyses.50 In the future, a combination of more thorough (“deep”) sequencing of germline and tumor cells combined with gene expression profiling and analysis of epigenetic events may ultimately lead to routine evaluation of these biomarkers in the personalized selection of antineoplastic therapy for CRC.
Other Prognostic Factors
In the FOLFIRI regimen, the topoisomerase I inhibitor irinotecan is combined with the 5-FU/folinic acid regimen.51–53 Irinotecan is metabolized to its active form, SN-38, which itself is conjugated and eliminated by the liver-derived enzyme UDP-glucuronosyltransferase, also known as UGT1A1. Several biomarkers have been proposed to predict efficacy and toxic effects for irinotecan-based regimens.
Biomarkers Proposed to Predict Clinical Resistance to Cetuximab/Panitumumab Therapy
Open in new tab
Measuring tumor cell topoisomerase I expression has correlated with irinotecan response in several large studies,54,55 but this procedure is not currently performed as part of the selection of therapy for CRC.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
(
):
–
.
.
. In:
, eds.
. 2nd ed.
:
;
:
–
.
.
. In:
Jr
, eds.
.
:
;
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
(
):
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
(
):
. Abstract 4000.
et al.
.
.
;
(
):
. Abstract 4036.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
. In press.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
. In press.
et al.
.
.
;
:
. doi:10.1186/1471-2407-9-339.
et al.
.
.
;
:
–
.
et al.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
. PubMed PMID: 20406168.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
(
):
. Abstract 3570.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
. doi:10.1371/journal.pone.0007287.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
.
;
:
–
.
et al.
.
.
;
:
–
.
.
.
;
:
–
. PubMed PMID: 20425074.
Polymorphisms in the UGT1A1 gene have been strongly linked to toxicity of irinotecan-based treatment for CRC.11,51–53,56,57 One particular polymorphism, homozygosity for the 7-repeat allele (also known as UGT1A1*28) is associated with severe diarrhea when irinotecan is administered.11,56,57 Although UGT1A1 genotyping is being performed on a regular basis in some facilities, widespread use has not occurred owing to the presence of conflicting negative data58 and the lack of current endorsement for testing by specialty societies (eg, American Society of Clinical Oncology [ASCO] and College of American Pathologists) or regulatory agencies.
Therapeutic Antibodies for CRC: Biomarkers for the Prediction of Efficacy and Toxicity
5-Fluorouracil and Capecitabine
Although long popular in Europe, oxaliplatin-based multiagent chemotherapy (FOLFOX) for CRC was introduced and became more widely adopted in the United States during the last 10 years.61–63 Given that platin-based drugs are DNA-damaging agents, the focus on discovery of predictive biomarkers has been on genes and pathways associated with DNA.64
Irinotecan (CPT-11; Campostar)
The DNA excision repair protein, excision repair cross-complementing C1 (ERCC1), has been the focus of study for the prediction of non–small cell lung cancer response to the standard first-line platin-based drugs plus paclitaxel chemotherapy regimen.65 Recent studies have also shown that ERCC1 expression may also predict for resistance to FOLFOX treatment of CRC.66–68 However, there is no current consensus as to the best way to test CRC specimens for ERCC1 expression, with advocates for protein detection by immunohistochemical analysis and mRNA detection by RT-PCR. In addition, polymorphisms of the ERCC1 gene have been associated with decreased expression of ERCC1 protein and improved survival after FOLFOX chemotherapy.11,66 Given the established success in predicting resistance to platin-based drugs in lung cancer and the initial results of similar studies in CRC, it is possible the ERCC1 testing will, at some point in the future, become a routine procedure for patients with high-risk primary and metastatic CRC.
Excision repair cross-complementing C2 (also known as ERCC2 and XPD) is another nucleotide excision repair enzyme that may be a predictive biomarker for CRC treatment efficacy.69–71
Ribonucleotide reductase M1 (RRM1) expression has been linked to response of lung cancer to DNA-damaging agents such as platin-based drugs and gemcitabine. In CRC, only limited studies have been performed, and it is not currently known whether RRM1 expression can guide selection of oxaliplatin-based therapy for the disease.72
Oxaliplatin
X-ray repair complementing defective repair in Chinese hamster cells 1 (XRCC1) is a DNA base repair enzyme that has been linked to prognosis in CRC.66,69,73,74XRCC1 polymorphisms have been associated with reduced XRCC1 expression and enhanced response to FOLFOX.74–76
Glutathione S-transferase π 1 (GSTP1) is an enzyme that facilitates glutathione conjugation and detoxification of a number of drugs, including oxaliplatin.11 GSTP1 expression has been linked to resistance to FOLFOX chemotherapy in CRC in some studies,66,77 but not in others.78 Loss of GSTP1 expression associated with germline polymorphisms has frequently been associated with oxaliplatin-related adverse events, especially neurotoxicity.79
DNA polymerase β (POLB) is another gene associated with the excision repair pathway. In a recent study, overexpression of POLB was an unfavorable prognostic factor for CRC, but the authors concluded that the marker was suited to predicting response only to cisplatin.80
Novel Cytotoxic Agent and Antibody Therapeutics in Clinical Trials
In summary, a number of biomarkers have been established as predictive factors for response of recurrent and metastatic CRC to both of the competing FOLFIRI and FOLFOX regimens. Although these tests are not currently recommended for daily clinical use by regulatory authorities or specialty societies, there is great potential for a personalized approach to chemotherapy selection for patients with CRC to ultimately reach mainstream clinical status. In the adjuvant setting, current opinion based on clinical trial results now holds that the FOLFOX regimen is superior to the FOLFIRI regimen.81 However, for an individual patient with high-risk primary CRC, there may be germline and tumoral genomic information whose eventual clinical validation can lead to a specific selection of agents, rather than using the “one-size-fits-all” approach.
In the last decade, regulatory agencies have approved the use of 2 monoclonal antibodies, cetuximab (Erbitux) and panitumumab (Vectibix) that target the human epidermal growth factor receptor 1 (EGFR) for the treatment of EGFR-overexpressing CRC.82,83 Both antibody therapeutics were developed with an EGFR expression test (immunohistochemical analysis) to confirm patient eligibility for treatment. More recently, a panel of biomarkers has been proposed to direct treatment with these agents Table 3. These biomarkers are evaluated in Table 3 for their level of evidence using College of American Pathologists evaluation criteria.84
Summary
In the last decade, regulatory agencies have approved the use of 2 monoclonal antibodies, cetuximab (Erbitux) and panitumumab (Vectibix) that target the human epidermal growth factor receptor 1 (EGFR) for the treatment of EGFR-overexpressing CRC.82,83 Both antibody therapeutics were developed with an EGFR expression test (immunohistochemical analysis) to confirm patient eligibility for treatment. More recently, a panel of biomarkers has been proposed to direct treatment with these agents Table 3. These biomarkers are evaluated in Table 3 for their level of evidence using College of American Pathologists evaluation criteria.84