- © 2005 by American Society of Clinical Oncology
Role of Genomic Markers in Colorectal Cancer Treatment
- From the Drug Resistance Group, Centre for Cancer Research and Cell Biology, Queen's University Belfast, University Floor, Belfast City Hospital, United Kingdom
- Address reprint requests to Patrick G. Johnston, MD, PhD, Department of Oncology, Queen's University Belfast, University Floor, Lisburn Road, Belfast Lisburn Road, Belfast, BT9 7AB, UK; e-mail: oncology{at}qub.ac.uk.
Abstract
For the last four decades, fluorouracil (FU) has been the main treatment of choice in colorectal cancer (CRC) in both the advanced and adjuvant settings. In the advanced setting, FU monotherapy produces response rates of only 10% to 20%. Furthermore, in resected stage III CRC, FU monotherapy has increased overall survival by only 20%. The combination of FU with newer therapies such as oxaliplatin and irinotecan has significantly improved response rates to 40% to 50%. Despite these improvements, more than half of advanced CRC patients derive no benefit from treatment; this is due to either acquired or inherent drug resistance. This review aims to highlight the current prognostic and predictive markers that have been identified for CRC to date. The limited use of these predictive markers underscores the importance of and need for multiple marker testing in order to improve response rates and decrease toxicity. This review will also focus on high throughput methods to identify panels of predictive markers for CRC, which ultimately aim to tailor treatment according to an individual patient and tumor profile.
INTRODUCTION
Colorectal cancer (CRC) is the second leading cause of cancer-related deaths in the Western world. Approximately 75% of patients with CRC present with localized disease; however, despite curative surgery, approximately 40% of patients still experience disease relapse leading to morbidity and eventual mortality. In patients with resectable stage III CRC, adjuvant therapy has been demonstrated to improve disease-free survival and overall survival by 35% and 22%, respectively. Yet the role of adjuvant therapy in stage II CRC remains controversial. The 5-year survival for patients with stage II CRC is 75%, which demonstrates that the majority of patients are cured by surgery alone. However, 40% of these patients will develop recurrent disease within their lifetime; hence, there is a need to identify which of these patients would benefit from adjuvant therapy. Molecular profiling of tumors may identify patients who are more likely to benefit from adjuvant therapy. This would enable the clinician to tailor treatment according to an individual patient and tumor profile. In CRC, a limited number of predictive markers have been identified to date (Table 1).1-5 However, the use of these predictive markers individually has led to somewhat inconclusive results. This limited use of individual predictive markers highlights the importance of and need for multiple marker testing in order to improve response rates and decrease toxicity in CRC patients. The purpose of this review is to highlight the current prognostic and predictive markers that have been identified to date. It will also focus on the recent advances in high throughput technology and the impact that this will have for CRC.
Current Predictive Markers for FU, Oxaliplatin, and Irinotecan in Colorectal Cancer
Loss of Heterozygozity
It has been reported that 70% of patients with CRC have lost a portion of chromosome 17p or 18q or both.17 The 17p chromosome contains p53, which is an important tumor suppressor, and is reported to be mutated in 40% to 60% of patients with CRC p53 has been described as the universal sensor of genotoxic stress.18 p53 status has been studied as a prognostic factor, and more recently as a predictor of response to cancer chemotherapy. Two methods have been employed to assess p53 status; DNA analysis, to detect a variety of mutations, and immunohistochemistry, to detect abnormal nuclear accumulation of the p53 protein. Immunohistochemically detected p53 overexpression has been used as a surrogate marker for p53 mutation; however, this assumption is not always correct. Many genetic changes do not result in p53 overexpression, and positive immunohistochemical analysis of p53 may occur in the absence of p53 mutation. A number of studies19-21 have demonstrated that p53 overexpression correlated with poor survival when measured by immunohistochemistry in stage II disease. However, a number of other studies22-25 assessed p53 by both immunohistochemistry and polymerase chain reaction (PCR) single-stranded conformational polymorphism (SSCP) and found that p53 did not display any independent prognostic role in early stage CRC.
A recent study by Tang et al12 found that p53 mutation was associated with a poorer prognosis in stage II and III CRC patients who received surgery alone, whereas p53 was not a prognostic factor among those patients who had received fluorouracil (FU) -based adjuvant chemotherapy. Yet, Ahnen et al13 found that patients with stage III CRC whose tumors overexpressed p53 did not derive significant survival benefit from adjuvant FU-based treatment, whereas those without p53 over-expression did. These conflicting findings may be due to the techniques used as well as the various antibodies used to detect p53. Due to these conflicting results the use of p53 as a predictive marker for FU remains controversial.
Several studies26,27 have demonstrated that patients who have lost 18q or lack protein expression in their tumors have an inferior 5-year survival compared with those who retain both alleles. In addition, retention of both 18q alleles had a more favorable outcome after adjuvant FU-based chemotherapy in stage III CRC.28 At least two other tumor suppressor genes are located near to deleted in colon cancer (DCC); they are DPC4 and JV18-1.29-31 Therefore it is possible that the region that is lost may contain these two genes as well as DCC, as a result, any one of these genes may affect prognosis.
Microsatellite Instability
Genetic instability has been recognized as a central element in the genesis of malignant lesions, resulting in clonal evolution of genetic events acquired in the course of tumor progression.32 Microsatellite instability (MSI) is common to many forms of cancer and is found in half to two thirds of sporadic colon cancers. MSI is caused by mutations in the mismatch repair (MMR) genes such as hMSH2, hMLH1 and hMSH6, which results in failure of the DNA mismatch repair system to correct errors that occur during replication.33
Various studies have investigated the prognostic role of MSI in stage II CRC. The studies have confirmed a consistent and independent association between MSI-high (MSI-H) phenotype and improved survival in stage II and stage III CRC patients (76% v 54%).34 Furthermore, Lim et al35 demonstrated that patients with MSI tumors exhibited better recurrence-free survival compared with those with microsatellite-stable (MSS) tumors. Moreover, the use of adjuvant chemotherapy did not benefit these patients.
There also appears to be a link between MSI and transforming growth factor-beta (TGFβ) RII mutation. TGFβ is a multifunctional polypeptide that regulates a number of cellular processes including growth, differentiation, deposition of the extracellular matrix (ECM), immunosuppression, embryogenesis, repair of soft and hard tissue and regulation of hematopoiesis.36-39 A recent study demonstrated that 61% of stage III colorectal cancer patients with MSI-high tumors also had a TGFβ RII mutation.28 This study also found that patients who had MSI-high tumors and TGFβ RII mutations had a 5-year survival of 74% following adjuvant FU therapy compared with 46% in patients with MSI-high tumors without TGFβ RII mutations.28
Oncogenes
K-ras controls growth and differentiation by transduction of extracellular signals. K-ras belongs to the RAS family (K-ras, H-ras, and N-ras) of cellular proto-oncogenes. It has been established that approximately 30% of CRC have a mutation in the K-ras gene. K-ras abnormalities have been identified in codons 12, 13, 31, and 61, with more than 80% of cases in codons 12 and 13. Studies have shown that, in stage I and II CRC, K-ras mutations are positively associated with recurrence and poorer long-term survival.40 These studies noted that mutations within codons 12 and 13 were linked to increased risk of nodal metastasis.41
In 1998 the RASCAL (K-ras in Colorectal Cancer Collaborative group) study was initiated to assess the association between K-ras mutation and patient outcome and tumor characteristics. The results from the RASCAL studies concluded that the glycine-to-valine mutation on codon 12 (10% of the study) was an independent factor for increased risk of death.42 The second RASCAL study allowed further subgroup analysis in patients with stage II and III CRC (8.6% of the study), which demonstrated that the valine mutation on codon 12 lost its prognostic role in this particular subgroup of patients.43 The conclusions from these studies suggest that not only does the mutation lead to cancer progression, but also may lead to a tumor with more aggressive biologic behavior.
Chemotherapeutic Agents
The most active drug in CRC, the antimetabolite FU, was developed more than 40 years ago. In the last 5 years, the median survival for patients with metastatic colorectal cancer has nearly doubled from 12 to 22 months, and new agents in late phase clinical treatment may soon extend this survival benefit further. In the metastatic disease setting, single-agent FU produced response rates of only 10% to 20%44; however, combination of FU with new classes of drugs such as oxaliplatin and irinotecan, has significantly improved response rates to the 40% to 50% range in patients with metastatic colorectal cancer.45,46 Despite these improvements there is a need to identify novel panels of molecular and biochemical markers that can be used to predict response to traditional and novel therapies.
FU
A primary mechanism of action of FU is inhibition of the nucleotide synthetic enzyme thymidylate synthase (TS) by its active metabolite fluorodeoxyuridine monophosphate (FdUMP) resulting in thymidylate depletion, which if prolonged causes apoptosis via thymineless death (Fig 1).47 TS is a cytosolic enzyme that catalyses the reductive methylation of deoxyuridine monophosphate (dUMP) to yield deoxythymidine monophosphate (dTMP), a precursor of deoxythymidine triphosphate (dTTP), which is required for DNA synthesis and repair.48 FdUMP forms a stable ternary complex with TS and 5,10-methylene tetrahydrofolate (CH2THF), which blocks dTMP production thereby inhibiting DNA synthesis and repair.
Mechanism of action of FU. Abbreviations: DPD, dihydropyrimidine dehydrogenase; FU, fluorouracil; TP, thymidine phosphorylase; FUDR, 2′-deoxy-5′-fluorouridine; CH2THF, 5, 10-methylene-tetrahydrofolate; FUMP, 5′-fluorouridine-5′-monophosphate; FdUMP, 5′-fluoro-2′-deoxyuridine-5′-monophosphate; TS, thymidylate synthase; RR, ribonucleotide reductase; FdUDP, 5′-fluoro-2′-deoxyuridine-5′-diphosphate; FUTP, 5′-fluorouridine-5′-triphosphate; FdUTP, 5′-fluoro-2′deoxyuridine-5′-triphosphate.
A number of in vitro and in vivo studies have demonstrated that TS is a prognostic and a predictive marker in fluoropyrimidine–based chemotherapy. The primary mechanism of resistance to fluoropyrimidines is an increase in TS expression.49 Several studies have demonstrated that high levels of TS predict for worse survival in stage II and III patients.3,50,51 A number of studies have also demonstrated that patients with low tumoral TS expression have higher response rates to FU than those with higher levels of TS expression.1-3 There is, however, a subset of patients who have low TS expression who do not respond to FU treatment. These nonresponders may have other mechanisms of resistance at play, such as high levels of thymidine phosphorylase (TP) and dihydropyrimidine dehydrogenase.
The TS gene promoter is polymorphic and usually has two (TSER*2) or three (TSER*3) 28–base pair tandem repeat sequences.52 These tandem repeats may affect transcriptional and/or translational efficiency of the TS gene. Therefore, these TS gene polymorphisms have the potential to predict clinical outcome and toxicity. It has been demonstrated that TSER*3/TSER*3 homozygous patients are less likely to respond to FU than TSER*2/TSER*2 homozygous, or TSER*2/TSER*3 heterozygous patients.6 This may be due to the fact that TS promoters with the TSER*3 sequence have been reported to generate approximately a three-fold higher mRNA than those with the TSER*2 sequence,7 and therefore patients with this genotype may express higher levels of TS and be less responsive to FU. The identification of the polymorphism provides another criterion for selecting patients who are likely to respond to FU-based chemotherapy and also identifies patients who will experience increased toxicity.
TP converts FU to fluorodeoxyuridine (FUDR), which can then be converted to the active metabolite FdUMP. Initial preclinical studies demonstrated that increased TP expression correlated with increased sensitivity to FU, probably due to increased synthesis of FUDR.8 However, analysis of TP mRNA expression in 38 colorectal tumors indicated that tumors with high TP were actually less likely to respond to FU.8,9 TP is identical to platelet-derived endothelial cell growth factor, which is a well-established angiogenic factor. Therefore, high TP expression may be a marker for a more invasive and malignant tumor phenotype that is less responsive to chemotherapy. However, in vitro, the angiogenic effects of TP would not appear to be a factor. Salonga et al5 examined the combined levels of TS, dihydropyrimidine dehydrogenase (DPD), and TP in a series of colorectal tumors treated with FU. Tumors that responded to FU-based therapy had expression values of all three genes (TS, DPD, and TP) that were below the nonresponsive cutoff values. This resulted in a 92% response rate in this group of patients. Those patients whose tumors did not respond had high levels of gene expression for at least one of the markers, once again highlighting the importance of multiple marker testing.
DPD catalyses the rate limiting step in the catabolism of fluoropyrimidines. More than 80% of FU administered is degraded in the liver by DPD. Thus, DPD limits the bioavailability of FU.53 DPD has variable activity in human tumors, and tumoral DPD has been reported to be an important determinant of response to FU both in vitro4 and in vivo in the metastatic setting.5 The differences in variation among patients who receive FU must be due to genetic differences in the activity of the DPD gene. Patients deficient in DPD experience profound systemic toxicity when treated with FU, which may prove fatal,54 as a result of its decreased catabolism of FU resulting in higher systemic levels of FU in patients.
Several studies have demonstrated that patients with low DPD expression had longer disease-free recurrence and increased survival than those with high levels of DPD.10 Tsuji11 carried out a further study to analyze the prognostic significance of tumor DPD expression in curatively resected CRC patients who received adjuvant FU or not. They found that high DPD expression in the surgery alone group was associated with a better survival (P = .02), whereas high tumor DPD expression in the adjuvant chemotherapy group was associated with poor survival (P = .03). These results would suggest that estimating tumor DPD expression will provide useful information and help assess which treatment, if any, stage II and III CRC patients should be given after curative surgery.
Oxaliplatin
Oxaliplatin is a third-generation platinum compound first synthesized in Japan in 1969. It was developed as one of several 1, 2-diaminocyclohexane (1, 2-DACH) platinum compounds in an attempt to generate compounds with a more favorable therapeutic index than cisplatin and carboplatin.55 Oxaliplatin is thought to form a positively charged species that cross-links DNA and eventually leads to cytotoxicity.56 Several important mechanisms have been identified to play a role in resistance to oxaliplatin. These include decreased drug accumulation, drug inactivation, enhanced tolerance to platinum-DNA adducts and enhanced DNA repair.57 One of the major DNA repair systems in mammalian cells is the nucleotide excision repair (NER) pathway. NER removes bulky helix-distorting adducts produced by oxaliplatin.
ERCC1 is a highly conserved protein and is an essential member of the NER pathway.58 The ERCC1-XPF complex is involved in the cleavage of damaged DNA 5′ to the DNA lesion. It has been demonstrated that ERCC1 gene expression levels had significant independent correlation with overall survival after FU/oxaliplatin therapy in patients with advanced colorectal cancer refractory to first-line chemotherapy.14 Furthermore, an independent study demonstrated that both TS and ERCC1 mRNA expression had statistically significant association with survival in patients treated with FU/oxaliplatin.56 The XPD gene, which is also known as ERCC2, encodes a helicase that is a component of the transcription factor TFIIH and also an essential member of the NER pathway.59 The importance of XPD in platinum-drug resistance has not been clearly established, but may be multifaceted. This may include an impact on NER pathways or cross talk with other factors involved in DNA repair, which are independent of NER.
Irinotecan
Irinotecan is a DNA topoisomerase I (topo-1) inhibitor which was first introduced into the clinic in the late 1980s.60 DNA topo-1 is involved in the relaxation and recombination of torsionally strained supercoiled duplex DNA during replication and transcription.61 Inhibition of topo-1 induces DNA damage, resulting in cell death.62 Irinotecan is converted in vivo to SN-38 (7-ethyl-10-hydroxy-camptothecan) by carboxylesterases, which are abundant in the liver and also found in other tissues.63 SN-38 is 100- to 1,000-fold more biologically active than irinotecan and exerts its cytotoxicity by trapping the complexes formed by topo-1 with DNA. The single strand DNA breaks generated by SN-38 are not toxic, as they are highly reversible and rapidly repaired once the drug is removed. Lethal irreversible DNA damage occurs when DNA synthesis is ongoing and the replication fork encounters the DNA-topo-1 complex, causing a double-strand break that can lead to cell death.64 Studies have indicated a positive relationship between the topo-1 activity and the cellular sensitivity to irinotecan,16 but it has not been proven.
Novel Therapies
Epidermal growth factor receptor (EGFR; c-erbB-1) is a type 1 receptor tyrosine kinase, which signals through PI3K/PKC, MAPK, and STAT3 to induce proliferation, cell cycle progression and inhibition of apoptosis.65 Overexpression and/or mutation of EGFR has been detected in a number of human cancers and is often associated with aggressive disease and poor prognosis.66 More importantly EGFR expression has been detected in 60% to 75% of colorectal cancer.67,68 A study by McKay et al demonstrated that EGFR overexpression occurred in approximately 50% of in CRC patients.69 Several tyrosine kinase inhibitors and monoclonal antibodies have been developed to target EGFR. Iressa (ZD-1839), a tyrosine kinase inhibitor, is the most advanced in terms of clinical trials; IMC-C225, a monoclonal antibody targeted to EGFR, has shown antitumor activity in a wide variety of tumor types. Panitumumab (ABX-EGF) is a humanised IgG2 monoclonal antibody that binds with high affinity to EGFR and blocks binding of EGF and TGFα. The efficacy of panitumumab was assessed in a phase II study of patients with metastatic CRC whose tumors overexpressed EGFR. The study found that panitumumab monotherapy resulted in 13% of patients with a partial response and 39% of patients with stable disease.70 Although EGFR inhibitors have shown encouraging activity their response rates have been limited, therefore it is logical to combine these inhibitors with chemotherapy in an attempt to improve response rates.
Vascular endothelial growth factor (VEGF) is the ligand for two tyrosine kinase receptors expressed on vascular endothelial cells, VEGF receptor 1 (VEGF R1) and VEGF R2. It has been shown that increased VEGF expression correlates with tumor stage and poorer prognosis in colorectal cancer.71 Furthermore, Ishigami et al have shown that the recurrence rate of resected stage III VEGF-positive colon cancer is 4.5 times higher than VEGF negative colon cancer.72 Bevacizumab, which is a recombinant humanized monoclonal antibody against VEGF, is currently in clinical trials and has demonstrated encouraging clinical activity. Furthermore, Vatalanib (PTK787), which is a novel oral angiogenic inhibitor that targets all known VEGF receptor tyrosine kinases, has demonstrated biologic activity on the tumor microvasculature in hepatic metastasis from CRC.73
FUTURE DIRECTION
The advent of high-throughput methodologies such as microarray–based gene expression profiling, proteomic profiling, comparative genomic hybridization analysis and the newly developed metabolomics enables tumor samples to be profiled on a global scale. This has major implications for the diagnostic capability and prognostic classification of tumors, where we can ultimately predict response of each individual tumor to chemotherapy.
The most frequently used genomewide approach is DNA microarray profiling, this has been utilized to either diagnose/stage cancer or to predict outcome (eg, recurrence). Mariadason et al carried out gene expression profiling on 30 colorectal cancer cell lines. They were able to identify panels of genes that correlated with drug sensitivity and in addition used leave-one-out cross validation to demonstrate that these genes were predictive for either FU, irinotecan or oxaliplatin response. They noted that the FU gene set had a greater power to predict response than four “classical” determinants of FU response: TS, TP, p53, and MMR status.74,75
Several groups have used gene expression profiling for prognosis in several different tumor types including, leukemia,76 breast,77 early-stage lung adenocarcinoma,78 mesothelioma,79 and inflammatory breast cancer.80 A number of gene expression studies have been carried out in CRC.81,82 A study by Bertucci et al, demonstrated that the gene expression profiling was able to separate stage IV from stage I-III disease in an unsupervised and supervised manner using global hierarchical clustering. From their clustering they were able to predict the likelihood of metastasis, and suggested that this group of patients may benefit from a more aggressive treatment regimen.82
Wang et al also used gene expression profiling to identify markers for stage II colorectal cancer. The study contained 74 patients with stage II colorectal cancer. They used two supervised class prediction approaches to select markers from the 17,616 informative genes from the microarray. In the first approach, the patients were divided equally into a training set and a test set. This approach yielded 60 genes from 38 patients. The patients were divided into one of two groups based on unsupervised clustering results. From that each subgroup was further divided into a training set and a test set and again were analyzed to select markers. This approach yielded 23 markers from the training set, which were then analyzed in the test set. The investigators then compared the predictive power of the 23-gene set and the 60-gene set and found that only the 23-gene set was predictive. The 23-gene set was then further validated in 36 independent patients and demonstrated an overall accuracy of 78%.83 This study highlights the power of predictive marker testing but also highlights the need to carefully select the correct analysis for the purpose of the test.
CONCLUSION
To date, there have been significant limitations to the studies carried out on individual predictive and prognostic markers, such as the choice of cutoff used, the statistical methods employed to assess importance of the genes and the size of the study population. It has been previously demonstrated how a small number of genes can exert major effects on drug response, but it will be the case that the identification of key polymorphic genes and environmental factors will ultimately lead to the ability to predict enhanced response to chemotherapy while minimizing drug toxicity. In the future, molecular profiling of tumors may identify individuals more likely to benefit from chemotherapy and tailor individual treatment in the future. The area of pharmacogenomics will ultimately lead to a time of a more rationalized and molecular approach to cancer treatment and will undoubtedly make a huge contribution in the field of oncology. The next step in the field of pharmacogenomics will be to develop clinical trials that will assess prospectively the benefits of profiling a patient's particular tumor, which should translate into improvements in both overall response and toxicity. If the aim of pharmacogenomics is realized, a new age of individualized treatment will become a reality with a superior overall response rate, less toxicity and enhanced survival benefit for patients.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Footnotes
-
Authors' disclosures of potential conflicts of interest are found at the end of this article.
- Received March 11, 2005.
- Accepted April 12, 2005.
References
- 1.↵
- 2.
- 3.↵
- 4.↵
- 5.↵
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
- 12.
- 13.
- 14.
- 15.
- 16.
- 17.↵
- 18.↵
- 19.↵
- 20.
- 21.↵
- 22.↵
- 23.
- 24.
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.
- 38.
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
