|Home | About | Journals | Submit | Contact Us | Français|
Every year in the United States, 160,000 cases of colorectal cancer are diagnosed, and 57,000 patients die of the disease, making it the second leading cause of death from cancer among adults.1 The disease begins as a benign adenomatous polyp, which develops into an advanced adenoma with high-grade dysplasia and then progresses to an invasive cancer.2 Invasive cancers that are confined within the wall of the colon (tumor–node–metastasis stages I and II) are curable, but if untreated, they spread to regional lymph nodes (stage III) and then metastasize to distant sites (stage IV).3-5 Stage I and II tumors are curable by surgical excision, and up to 73% of cases of stage III disease are curable by surgery combined with adjuvant chemotherapy.3,4,6 Recent advances in chemotherapy have improved survival, but stage IV disease is usually incurable.3,4
The clinical behavior of a colorectal cancer results from interactions at many levels (Fig. 1). The challenges are to understand the molecular basis of individual susceptibility to colorectal cancer and to determine factors that initiate the development of the tumor, drive its progression, and determine its responsiveness or resistance to antitumor agents. This review summarizes areas of current knowledge, recognizing that the topics presented are only fragments of the total picture.
The loss of genomic stability can drive the development of colorectal cancer by facilitating the acquisition of multiple tumor-associated mutations. In this disease, genomic instability takes several forms, each with a different cause (Table 1).7-26
The most common type of genomic instability in colorectal cancer is chromosomal instability, which causes numerous changes in chromosomal copy number and structure.7 Chromosomal instability is an efficient mechanism for causing the physical loss of a wild-type copy of a tumor-suppressor gene, such as APC, P53, and SMAD family member 4 (SMAD4), whose normal activities oppose the malignant phenotype.2,27,28 In colorectal cancer, there are numerous rare inactivating mutations of genes whose normal function is to maintain chromosomal stability during replication, and in the aggregate, these mutations account for most of the chromosomal instability in such tumors.8 In contrast to some other cancers, colorectal cancer does not commonly involve amplification of gene copy number29 or gene rearrangement.
In a subgroup of patients with colorectal cancer, there is inactivation of genes required for repair of base–base mismatches in DNA, collectively referred to as mismatch-repair genes (Fig. 2 and and3).3). The inactivation can be inherited, as in hereditary nonpolyposis colon cancer (HNPCC), also known as the Lynch syndrome, or acquired, as in tumors with methylation-associated silencing of a gene that encodes a DNA mismatch-repair protein.
In patients with HNPCC, germ-line defects in mismatch-repair genes (primarily MLH1 and MSH2) confer a lifetime risk of colorectal cancer of about 80%, with cancers evident by the age of 45 years, on average.10-13,30,31 The loss of mismatch-repair function in patients with HNPCC is due not only to the mutant germ-line mismatch-repair gene but also to somatic inactivation of the wild-type parental allele.31 Genomic instability arising from mismatch-repair deficiency dramatically accelerates the development of cancer in patients with HNPCC — some cancers arise within 36 months after normal results on colonos-copy.32 For this reason, yearly colonoscopy is recommended for carriers of an HNPCC mutation,30,32 and prophylactic colectomy should be considered for patients with high-grade lesions. Germ-line mutations of another mismatch-repair gene, MSH6, attenuates the predisposition to familial cancer.9,33,34 Somatic inactivation of mismatch-repair genes occurs in approximately 15% of patients with nonfamilial colorectal cancer. In these patients, biallelic silencing of the promoter region of the MLH1 gene by promoter methylation inactivates mismatch repair15-17 (Fig. 2 and and33).
The loss of mismatch-repair function is easy to recognize by the associated epiphenomenon of microsatellite instability, in which the inability to repair strand slippage within repetitive DNA sequence elements changes the size of the mononucleotide or dinucleotide repeats (micro-satellites) that are scattered throughout the genome. Mismatch-repair deficiency can also be detected by immunohistochemical analysis, which can identify the loss of one of the mismatch-repair proteins.14,35-37 Cancers characterized by mismatch-repair deficiency arise primarily in the proximal colon, and in sporadic cases, they are associated with older age and female sex.30 In mismatch-repair deficiency, tumor-suppressor genes, such as those encoding transforming growth factor β (TGF-β) receptor type II (TGFBR2) and BCL2-associated X protein (BAX), which have functional regions that contain mononucleo-tide or dinucleotide repeat sequences, can be inactivated.2,27,28
An alternative route to colorectal cancer involves germ-line inactivation of a base excision repair gene, mutY homologue (MUTYH, also called MYH).25,33 The MYH protein excises from DNA the 8-oxoguanine product of oxidative damage to guanine.24,25,33 In persons who carry two inactive germ-line MYH alleles, a polyposis phenotype develops, with a risk of colorectal cancer of nearly 100% by the age of 60 years.33 MYH-associated polyposis is increasingly recognized: one third of all persons in whom 15 or more colorectal adenomas develop have MYH-associated polyposis.33 The diagnosis requires genetic testing, which is facilitated by two mutations, Y165C and G382D, that together account for 85% of cases.33 Thus far, somatic inactivation of MYH has not been detected in colorectal cancer.
Epigenetic silencing of genes, mostly mediated by aberrant DNA methylation, is another mechanism of gene inactivation in patients with colorectal cancer.18,20 A methylated form of cytosine in which a methyl group is attached to carbon 5 (5-methylcytosine) defines a fifth DNA base, introduced by DNA methylases that modify cyto-sines within CpG dinucleotides.18 In the normal genome, cytosine methylation occurs in areas of repetitive DNA sequences outside of exons; it is largely excluded from the CpG-rich “CpG islands” in the promoter regions of approximately half of all genes.18 By comparison, in the colorectal-cancer genome, there is a modest global depletion of cytosine methylation but considerable acquisition of aberrant methylation within certain promoter-associated CpG islands.18 This aberrant promoter-associated methylation can induce epigenetic silencing of gene expression.18 In sporadic colorectal cancer with microsatellite instability, somatic epi-genetic silencing blocks the expression of MLH1.18
Among the loci that can undergo aberrant methylation in colorectal cancer, a subgroup seems to become aberrantly methylated as a group, a phenomenon called the CpG island methylator phenotype (CIMP, or CIMP-high).18,19 The molecular mechanism for CIMP remains unknown, but the phenomenon is reproducibly observed in about 15% of colorectal cancers and is present in nearly all such tumors with aberrant methylation of MLH118,19,21,38 (Fig. 2 and and3).3). The pathogenetic consequence of MLH1 silencing is well established, but the contribution of other epigenetic silencing events to colorectal carcino-genesis remains an area of ongoing study. An intermediate level of aberrant methylation in CIMP may define a subtype (i.e., CIMP2 and CIMP-low) that is thought to account for 30% of CIMP cases.22,23 A third pattern of aberrant methylation is exemplified by exon 1 of the gene encoding vimentin. Although this locus is not expressed by normal colon mucosa or colorectal cancer, it is aberrantly methylated in 53 to 83% of patients with colorectal cancer in a pattern that is independent of CIMP.39,40
Colorectal cancers acquire many genetic changes, but certain signaling pathways are clearly singled out as key factors in tumor formation (Fig. 2 and Table 2).41-62 One of these changes, the activation of the Wnt signaling pathway, is regarded as the initiating event in colorectal cancer.2,28,43 Wnt signaling occurs when the oncoprotein β-catenin binds to nuclear partners (members of the T-cell factor–lymphocyte enhancer factor family) to create a transcription factor that regulates genes involved in cellular activation.2,28,43 The β-catenin degradation complex controls levels of the β-catenin protein by proteolysis. A component of this complex, APC, not only degrades β-catenin but also inhibits its nuclear localization.
The most common mutation in colorectal cancer inactivates the gene that encodes the APC protein. In the absence of functional APC — the brake on β-catenin — Wnt signaling is inappropriately and constitutively activated. Germ-line APC mutations give rise to familial adenomatous polyposis, an inherited cancer-predisposition syndrome in which more than 100 adenomatous polyps can develop; in carriers of the mutant gene, the risk of colorectal cancer by the age of 40 years is almost 100%.2,30,43 Somatic mutations and deletions that inactivate both copies of APC are present in most sporadic colorectal adenomas and cancers.2,43 In a small subgroup of tumors with wild-type APC, mutations of β-catenin that render the protein resistant to the β-catenin degradation complex activate Wnt signaling.2,41-43
The inactivation of the p53 pathway by mutation of TP53 is the second key genetic step in colorectal cancer. In most tumors, the two TP53 alleles are inactivated, usually by a combination of a missense mutation that inactivates the transcriptional activity of p53 and a 17p chromosomal deletion that eliminates the second TP53 allele.2,27,28,44,45 Wild-type p53 mediates cell-cycle arrest and a cell-death checkpoint, which can be activated by multiple cellular stresses.63 The inactivation of TP53 often coincides with the transition of large adenomas into invasive carcinomas.64 In many colorectal cancers with mismatch-repair defects, TP53 remains wild-type, though in these cancers the activity of the p53 pathway is probably attenuated by mutations in the BAX inducer of apoptosis.2,28
The mutational inactivation of TGF-β signaling is a third step in the progression to colorectal cancer.50 In about one third of colorectal cancers, somatic mutations inactivate TGFBR2.47,49,50,65,66 In tumors with mismatch-repair defects, TGFBR2 is inactivated by distinctive frameshift mutations in a polyadenine repeat within the TGFBR2 coding sequence.47 In at least half of all colorectal cancers with wild-type mismatch repair, TGF-β signaling is abolished by inactivating missense mutations that affect the TGFBR2 kinase domain or, more commonly, mutations and deletions that inactivate the downstream TGF-β pathway component SMAD4 or its partner transcription factors, SMAD2 and SMAD3.29,47,49-51,65-68 Mutations that inactivate the TGF-β pathway coincide with the transition from adenoma to high-grade dysplasia or carcinoma.69
Several oncogenes play key roles in promoting colorectal cancer (Fig. 2 and Table 2). Oncogenic mutations of RAS and BRAF, which activate the mitogen-activated protein kinase (MAPK) signaling pathway, occur in 37% and 13% of colorectal cancers, respectively.21,55,57,70,71 RAS mutations, principally in KRAS, activate the GTPase activity that signals directly to RAF. BRAF mutations signal BRAF serine–threonine kinase activity, which further drives the MAPK signaling cascade.70,71 BRAF mutations are detectable even in small polyps,21 and as compared with RAS mutations, they are more common in hyperplastic polyps, serrated adenomas, and proximal colon cancers, particularly in those with the CIMP phenotype (Fig. 3). Patients with numerous and large hyper-plastic lesions, a condition termed the hyperplastic polyposis syndrome, have an increased risk of colorectal cancer, with disease progression occurring through an intermediate lesion with a serrated luminal border on histologic analysis.18,22,38,58,59
One third of colorectal cancers bear activating somatic mutations in PI3KCA, which encodes the catalytic subunit of phosphatidylinositol 3-kinase (PI3K).72 Less common genetic alterations that may substitute for PI3KCA mutations include loss of PTEN, an inhibitor of PI3K signaling, as well as amplification of insulin receptor substrate 2 (IRS2), an upstream activator of PI3K signaling, and coamplification of AKT and PAK4, which are downstream mediators of PI3K signaling.73
Advances in DNA sequencing technology have made it possible to sequence the entire coding genome of a human cancer. Colorectal cancer provided the first example of the power of this technology, with high-throughput sequencing of 18,000 members of the Reference Sequence (RefSeq) database of the National Center for Biotechnology Information.65,66 Cancer-associated somatic mutations were identified in 848 genes. Of these, 140 were identified as candidate cancer genes that probably contributed to the cancer phenotype because they were mutated in at least two colorectal cancers and when corrected for gene size showed more mutations than expected by chance.
The average stage IV colorectal-cancer genome bears 15 mutated candidate cancer genes and 61 mutated passenger genes (very-low-frequency mutational events). The predominance of low-frequency mutations in candidate cancer genes implies enormous genetic heterogeneity among colorectal cancers, which mirrors the heterogeneity of the clinical behavior of colorectal cancers. it difficult to determine the clinical effect of individual mutational events. Moreover, these initial results are probably conservative, because some mutations, which were initially labeled as rare “passengers” in colorectal cancer, have subsequently emerged as common and are probably pathogenetic in other cancer types (e.g., an IDH1 mutation noted initially in one colorectal cancer but subsequently in many gliomas).65,66,74
High-throughput sequencing of the colorectal-cancer genome has identified new common mutational targets. These include the ephrin receptors EPHA3 and EPHB6 (receptor tyrosine kinases), which together are mutated in 20% of colorectal cancers, and FBXW7, which functions in a protein degradation pathway that regulates levels of cyclin E and is mutated in 14% of colorectal cancers.65,66,75 An important challenge is to reduce the complexity of the 140 candidate cancer genes by identifying the biologic pathways and processes that are common downstream targets of multiple mutational events.
The sequence of transformation from adenoma to carcinoma, as initially formulated,2,28,43 was a model of the development of colorectal cancer in which specific tumor-promoting mutations are progressively acquired. This model predicts the presence of mutations that dictate specific tumor characteristics, such as the presence of regional or distant metastases (Fig. 2). Unexpectedly, the examination of results of full-genome sequencing from primary colorectal cancers and distant metastases in the same patient showed no new mutations in the metastases,76 implying that new mutations are not required to enable a tumor cell to leave the primary tumor and seed a distant site. Because the ongoing generation of mutations serves as a molecular clock, the finding that all the mutations in a metastasis are also present in the primary tumor implies that metastatic seeding is rapid, requiring less than 24 months from the appearance of the final mutation in the primary tumor.76
The activation of growth factor pathways is common in colorectal cancer (Fig. 2). An early and critical step in the development of an adenoma is the activation of prostaglandin signaling.77,78 This abnormal response can be induced by inflammation or mitogen-associated up-regulation of COX-2, an inducible enzyme that mediates the synthesis of prostaglandin E2, an agent strongly associated with colorectal cancer.78 Prostaglandin E2 activity can also be increased by the loss of 15-prostaglandin dehydrogenase (15-PGDH), the rate-limiting enzyme in catalyzing degradation of prostaglandin.79-81 Increased levels of COX-2 are found in approximately two thirds of colorectal cancers,78,82 and there is loss of 15-PGDH in 80% of colorectal adenomas and cancers.79 Clinical trials have shown that the inhibition of COX-2 by nonsteroidal antiinflammatory drugs prevents the development of new adenomas83-86 and mediates regression of established adenomas.87
Epidermal growth factor (EGF) is a soluble protein that has trophic effects on intestinal cells. Clinical studies have supported an important role of signaling through the EGF receptor (EGFR) in a subgroup of colorectal cancers.88-91 EGFR mediates signaling by activating the MAPK and PI3K signaling cascades. Recent clinical data have shown that advanced colorectal cancer with tumor-promoting mutations of these pathways — including activating mutations in KRAS,92-94 BRAF,95,96 and the p110 subunit of PI3K97 — do not respond to anti-EGFR therapy.
Vascular endothelial growth factor (VEGF) that is produced in states of injury or during the growth of normal tissue drives the production of new stromal blood vessels (angiogenesis). Clinical studies have suggested a role for angiogenic pathways in the growth and lethal potential of colorectal cancer. Treatment with the anti-VEGF antibody bevacizumab added an average of 4.7 months to the overall survival of patients with advanced colorectal cancer (15.6 months with standard therapy).98 The identification of molecular distinctions between cancers that benefit from this treatment and those that do not remains a challenge.
Stem cells in colorectal cancers are believed to be uniquely endowed with the capacity to renew themselves.99-102 Single colorectal-cancer stem cells, by definition, can lodge in a permissive site, such as the liver, and produce a metastasis. Currently, it is not possible to isolate individual colorectal-cancer stem cells, although certain cell-surface proteins (e.g., CD133, CD44, CD166, and aldehyde dehydrogenase 1) are promising markers. Normal stem cells that reside in the colonic crypt rely on adhesive and soluble stromal–epithelial interactions to maintain division and differentiation. The extent of alterations in these regulatory mechanisms in colorectal-cancer stem cells is a promising area of investigation, since agents that control the growth of colorectal-cancer stem cells could theoretically be used for cancer prevention and treatment.
One ongoing challenge is to translate the wealth of knowledge regarding colorectal-cancer genomics into clinically applicable predictive or prognostic tests (Table 3). The relation between mutations in EGFR signaling components RAS and BRAF and anti-EGFR therapy is currently the only application of colorectal-cancer genomics to treatment.92-96 A few genomic markers are useful for prognosis. For example, germ-line mutations in tumor-suppressor genes, such as APC, MLH1, and MSH2, indicate a very high risk of colorectal cancer and guide the frequency of colorectal-cancer surveillance and recommendations for prophylactic surgery. Other somatic markers have modest or unconfirmed prognostic value and are not currently used to direct care. Sporadic colorectal cancers with a mismatch-repair deficiency generally have a favorable prognosis35,103,105,108; poor survival in stage II and III colon cancers is associated with the loss of p27 (a proapoptotic regulator of the cell cycle109) or the loss of heterozygosity at chromosomal location 18q.105
The development of molecular diagnostics for the early detection of colorectal cancer is an important translation of colon-cancer genetics into clinical practice. One example is the development of assays to detect mutations that are specific to colorectal cancer and cancer-associated aberrant DNA methylation in fecal DNA from patients with colorectal cancer or advanced adenomas. These assays have a sensitivity of 46 to 77% for detecting early-stage colorectal cancer, which is superior to the sensitivity of testing for fecal occult blood although their superiority in preventing death from cancer has not been shown.39,110-113 Stool DNA testing for colorectal cancer has been added to the cancer-screening guidelines of the American Cancer Society114 and appears to be equally sensitive for detecting advanced adenomas.115 Although still in the developmental stage, assays for detecting plasma cell-free DNA may also be clinically useful,115 and assays for tumor-specific plasma protein or RNA profiles also remain targets of research. Questions that remain to be resolved are the optimal interval between serial tests and the performance and cost-effectiveness of stool DNA testing as compared with those of newer immuno-chemical fecal occult-blood tests.116
Genetic epidemiology and twin studies indicate that 35 to 100% of colorectal cancers and adenomas develop in persons with an inherited susceptibility to the disease.117-119 In addition, an HNPCC-like syndrome occurs in some families without any evidence of defects in mismatch repair.120 Several genomic loci that could harbor highly penetrant susceptibility genes have been identified with the use of linkage approaches,121-123 but the underlying mutations are unknown. Genomewide association studies have also identified germ-line DNA variants that are strongly associated with susceptibility, but the clinical use of these results is probably limited, since the relative risk associated with these variants is low.124-129
Studies that aid in the understanding of colorectal cancer on a molecular level have provided important tools for genetic testing for high-risk familial forms of the disease, predictive markers for selecting patients for certain classes of drug therapies, and molecular diagnostics for the noninvasive detection of early cancers. In addition, biologic pathways that could form the basis of new therapeutic agents have been identified. Although some high-frequency mutations are attractive targets for drug development, common signaling pathways downstream from these mutations may also be tractable as therapeutic targets. Recent progress in molecular assays for the early detection of colorectal cancer indicates that understanding the genes and pathways that control the earliest steps of the disease and individual susceptibility can contribute to clinical management in the near term.
An understanding of the signals that dictate the metastatic phenotype will provide the information necessary to develop drugs to control or prevent advanced disease. The considerable recent advances encourage us to believe that improvements in our knowledge of the molecular basis of colorectal cancer will continue to reduce the burden of this disease.
No other potential conflict of interest relevant to this article was reported.