THE CHROMOSOMAL INSTABILITY PATHWAY IN COLON CANCER

Abstract

The acquisition of genomic instability is a crucial feature in tumor development and there are at least 3 distinct pathways in colorectal cancer pathogenesis: the chromosomal instability (CIN), microsatellite instability (MSI), and CpG island methylator phenotype (CIMP) pathways. Most cases of colorectal cancer arise through the CIN pathway, which is characterized by widespread imbalances in chromosome number (aneuploidy) and loss of heterozygosity (LOH). It can result from defects in chromosomal segregation, telomere stability, and the DNA damage response, although the full complement of genes underlying CIN remains incompletely described. Coupled with the karyotypic abnormalities observed in CIN tumors are the accumulation of a characteristic set of mutations in specific tumor suppressor genes and oncogenes that activate pathways critical for colorectal cancer initiation and progression. Whether CIN creates the appropriate milieu for the accumulation of these mutations or vice versa remains a provocative and unanswered question. The goal of this review is to provide an updated perspective on the mechanisms that lead to CIN and the key mutations that are acquired in this pathway.

INTRODUCTION

Colorectal cancers (CRCs) develop through an ordered series of events beginning with the transformation of normal colonic epithelium to an adenomatous intermediate and then ultimately adenocarcinoma, the so-called “adenoma-carcinoma sequence”.¹ Multiple genetic events are required for tumor progression, and genomic instability is now recognized as an essential cellular feature that accompanies the acquisition of these mutations. In colon cancer, at least 3 distinct pathways of genomic instability have been described, the chromosomal instability (CIN), microsatellite instability (MSI), and CpG island methylator phenotype (CIMP) pathways. This review focuses on the CIN pathway that underlies the majority of all colon cancers.

In 1990, Fearon and Vogelstein first proposed a multistep genetic model of colorectal carcinogenesis that has come to serve as a paradigm for solid tumor progression. Inactivation of the Adenomatous Polyposis Coli (APC) tumor suppressor gene occurs first, followed by activating mutations of KRAS (Figure 1).² Subsequent malignant transformation is driven by additional mutations in the TGF-β, PIK3CA, and TP53 pathways.^2–6 This model predicts that at least 7 distinct mutations are required. Recent genome-wide sequencing efforts have calculated as many as 80 mutated genes per colorectal tumor, but a smaller group of mutations (< 15) were considered to be the true “drivers” of tumorigenesis.^{7, 8} Although there continue to be refinements to the original model, several key principles have been established: multiple genetic hits are required, there are discrete intermediates in the progression to cancer, and the temporal acquisition of these genetic changes matters. For example, APC mutations serve as the initiating event in adenoma formation in human and mouse models. In contrast, mutational activation of KRAS cannot initiate cancer in vivo, and only when combined with a mutation in APC does mutant KRAS promote tumor progression.⁹

Figure 1. Multistep genetic model of colorectal carcinogenesis.

The initial step in colorectal tumorigenesis is the formation of aberrant crypt foci (ACF). Activation of the Wnt signaling pathway can occur at this stage as a result of mutations in the Adenomatous polyposis coli (APC) gene. Progression to larger adenomas and early carcinomas requires activating mutations of the proto-oncogene KRAS, mutations in TP53, and loss of heterozygosity at chromosome 18q. Mutational activation of PIK3CA occurs late in the adenoma–carcinoma sequence in a small proportion of colorectal cancers. Chromosomal instability is observed in benign adenomas and increases in tandem with tumor progression.

Genomic instability and CRC

Baseline mutation rates are insufficient to account for the multiple mutations that are required for cancer to develop. The rate of mutations per nucleotide base pair is estimated to be as low as 10⁻⁹ per cellular generation.¹⁰ As proposed by Loeb et al., cancer cells must acquire some form of intrinsic genomic instability, a “mutator phenotype”, that increases their rate of new mutations.¹¹ Chromosomal instability (CIN) is observed in 65%–70% of sporadic colorectal cancers; the term refers to an accelerated rate of gains or losses of whole or large portions of chromosomes that results in karyotypic variability from cell to cell.¹² The consequence of CIN is an imbalance in chromosome number (aneuploidy), sub-chromosomal genomic amplifications, and a high frequency of loss of heterozygosity (LOH). Challenges to the field have been the methodological approaches to measure chromosomal instability and standardizing the precise quantitative criteria that define a “CIN-positive” tumor. Approaches to measure CIN have included cytometry, karyotyping, loss of heterozygosity analysis, fluorescent in situ hybridization (FISH), and most recently, comparative genomic hybridization (CGH). Newly developed CGH microarrays (array CGH), in which the metaphase chromosomes used in the conventional CGH are replaced by cloned DNA fragments with known genomic locations, have advanced the field due to their ability to identify chromosomal and segmental amplifications and deletions with higher resolution. Due to the use of different methods and criteria, it is not always straightforward to designate a tumor as CIN-positive vs. CIN-negative. In addition, some have proposed sub-categories of CIN-high and CIN-low for CIN-positive tumors.^13–15

A defect in the DNA mismatch repair (MMR) system leads to a second pathway characterized by instability in stretches of DNA microsatellites. The 3^rd pathway is designated the CpG island methylator phenotype (CIMP), which exhibits gene silencing due to hypermethylation of CpG islands.¹⁶ Because the definitions of these 3 pathways are not mutually exclusive, a tumor can occasionally exhibit features of multiple pathways. For example, up to 25% of MSI colorectal cancers can exhibit chromosomal abnormalities.¹⁷ In addition, whereas the CIMP phenotype can account for most of the MSI-positive/CIN-negative CRCs, up to 33% of CIMP-positive tumors can exhibit a high degree of chromosomal aberrations.¹⁵ Conversely, as many as 12% of CIN-positive tumors exhibit high levels of MSI.¹⁸ The significance and implications of these overlapping features are not yet fully defined.

MECHANISMS LEADING TO CHROMOSOMAL INSTABILITY

More than 100 genes can cause chromosomal instability in the yeast Saccharomyces cerevisiae. Although many of these genes have human homologues, only a limited number have been implicated in human tumors (see Table 1).

Table 1.

Selected Genes That Regulate Chromosomal Instability

Pathway	Deregulated gene	Gene alteration
Chromosome segregation	BUB1; BUBR1	Somatic mutations
	hZw10; hZwilch; hRod	Somatic mutation
	AURKA	Gene amplification and overexpression
	PLK1	Gene overexpression
	PTTG	Gene overexpression
	CENP-A; CENP-H; HEC1; INCENP	Gene overexpression
	APC	Germline and somatic mutations
Telomere regulation	TERC	Overexpression
DNA damage response	ATM; ATR	Germline mutation
	BRCA 1/2	Germline mutation
	TP53	Germline mutation and somatic mutations
	MRE11	Germline mutation

Chromosome segregation defects and CIN

The CIN phenotype could result from defects in pathways that ensure accurate chromosome segregation. The mitotic checkpoint (also known as the spindle assembly checkpoint) is the major cell cycle control mechanism that ensures high fidelity of chromosome segregation by delaying the onset of anaphase until all pairs of duplicated chromatids are properly aligned on the metaphase plate. Defects in checkpoint signaling lead to chromosome mis-segregation and subsequent aneuploidy with abnormal numbers of chromosomes being distributed to daughter cells. Nearly 100 years ago, Theodor Boveri suggested that malignant tumors may arise from a single cell with an abnormal genetic constitution acquired as a result of defects in the mitotic spindle apparatus.¹⁹

Products of the genes MAD (Mitotic Arrest Deficient) and BUB (Budding Uninhibited by Benzimidazoles) operate as checkpoint sensors and signal transducers that control sister chromatid separation (Figure 2). Their activation causes inhibition of the anaphase promoting complex/C (APC/C), a large ubiquitin-protein ligase, and cell cycle arrest. MAD3/BUBR1, MAD2, and BUB3 associate with the APC/C activating molecule CDC20 to form the mitotic checkpoint complex and induce a conformational change in the APC/C, which prevents binding and ubiquitination of its substrates.²⁰ Activation of the APC/C leads to degradation of securin and activation of separases, a class of caspase-related proteases. Separase in turn regulates a multiprotein complex termed cohesin, which creates physical links between sister chromatids that are maintained until late mitosis. Errors in this complex and ordered process result in unequal chromosomal segregation. Cahill et al. demonstrated that cancer cell lines with mutations in hBUB1 and hBUBR1 failed to arrest when spindles were depolymerized.²¹ Mutations in hBUB1 were dominant since exogenous expression of the mutant gene conferred an abnormal spindle checkpoint and CIN to chromosomally stable cell lines.²² To date, hMAD2 and hMAD1 mutations have been identified in breast cancer and leukemia, respectively, but not in CRC.^{23, 24} However, mutations in the hZw10, hZwilch/FLJ10036, and hRod/KNTC kinetochore proteins that function at the spindle checkpoint, and in Ding, which is essential for proper chromosome disjunction, have been reported in CRC.²⁵ Another protein involved in the mitotic checkpoint is the centromere-associated protein E (CENP-E), a kinetochore-associated kinesin-like protein. CENP-E stabilizes the interaction between microtubules and kinetochores and stimulates the recruitment and kinase activity of BubR1. Fibroblasts derived from mice heterozygous for CENP-E show significantly higher rates of aneuploidy than wildtype fibroblasts due to chromosomal mis-segregation.²⁶

Figure 2. Regulation of sister chromatid separation at the metaphase-anaphase transition.

In prometaphase, highly condensed chromosomes establish bipolar attachments to the mitotic spindle. Unattached or malorientated chromosomes generate a signal to delay the onset of anaphase until all pairs of sister chromatids are properly aligned on the metaphase plate. This signal is transduced by a relay of spindle-checkpoint proteins, including MAD1, MAD2, BUB1, BUBR1, BUB3 and centrosome protein E (CENP-E), which inhibits cell division cycle 20 (CDC20)-mediated activation of an E3 ubiquitin ligase, the anaphase promoting complex/cyclosome (APC/C). Following attachment and alignment of all the chromosomes at metaphase, the checkpoint signal is silenced and APC/C initiates the ubiquitin-dependent degradation of securin and activation of separase. Separase in turn cleaves a multiprotein complex termed cohesin, which creates physical links between sister chromatids to initiate anaphase.

Another proposed cause of CIN is abnormal centrosome number and function. Centrosomes coordinate mitosis by serving as an anchor for the reorganization of cytoplasmic microtubules into a mitotic spindle apparatus. It has been proposed that extra centrosomes lead to the formation of multiple spindle poles during mitosis, resulting in the unequal distribution of chromosomes. Ganem et al. visualized thousands of cell divisions in a variety of cancer cell lines, including colon, and showed that when centrosomes cluster, an increased rate of merotelic chromosomal attachments to the spindle can cause chromosomal mis-segregation and ultimately CIN.²⁷ Evidence of a causal relationship between centrosome amplification and cancer has recently been provided in flies.²⁸ Transplantation of larval brain cells with extra centrosomes induced the formation of metastatic tumors, suggesting that centrosome amplification can initiate tumorigenesis.

A pathogenic role for centrosome-associated Aurora and Polo-like (Plk) kinases has been also identified. There are three human Aurora kinases with distinct subcellular localizations and roles. Aurora A (AURKA) is involved in centrosome function and duplication, mitotic entry, and bipolar spindle assembly. Aurora A overexpression is associated with centrosome amplification, arrested mitosis with incomplete cytokinesis, and multinucleation.²⁹ AURKA is amplified and positively associated with the degree of CIN in colorectal tumors.³⁰ Aurora B is the catalytic component of the chromosomal passenger complex that regulates the accurate segregation of chromatids at mitosis, histone modification, and cytokinesis. Although its involvement in tumorigenesis is less clear, Aurora B overexpression correlates with advanced stages in colorectal cancer.³¹ Plk1 to Plk4 are serine/threonine kinases, which regulate entry into mitosis, centrosome duplication, transition from metaphase to anaphase, and cytokinesis. Two recent studies have provided evidence of a cross-talk between Plk1 and Aurora A.^{32, 33} Elevated expression of Plk1 has been observed in up to 73% of primary CRCs, and this correlated with tumor invasion, lymph node involvement, and Dukes’ stage.³⁴

Telomere dysfunction and CIN

Evidence indicates that CIN can be driven by telomere dysfunction. Telomeres are DNA-protein complexes that consist of stretches of hexameric repeats (TTAGGG in humans) that protect the ends of eukaryotic chromosomes from fusing and breaking during segregation. In somatic cells, a portion of telomeric DNA is lost after each round of cell division, due to the inability of DNA polymerase to completely synthesize the 3′ end of linear chromosomes, also known as the “end-replication problem”. When a critical short telomere length is reached, DNA damage checkpoints trigger cell senescence pathways and apoptosis. Cells that fail to undergo senescence enter a crisis-like state characterized by a period of massive cell death. Cells that survive the crisis checkpoint activate telomerase, the enzyme complex that elongates telomeres, or alternative mechanisms that lead to the lengthening of telomeres (ALT).

When telomere end protection is compromised, chromosomal ends enter breakage-fusion-bridge (B/F/B) cycles that can continue for multiple cell generations and lead to dramatic genome reorganization. Genomic amplification at fragile sites can commonly result from B/F/B cycles.³⁵ The importance of telomere loss and breakage-fusion-bridge cycles has been demonstrated in vivo. In mice deficient in the RNA component of telomerase (Terc^−/−), telomere shortening led to increased rates of spontaneous tumor formation and initiation of aberrant crypt foci (ACF) and microadenomas in the gastrointestinal tract.³⁶ The authors also demonstrated that telomeres are shortened at the stage of high grade dysplasia.³⁷ Several groups have reported shorter telomeres in 77% to 90% of colon cancer samples, compared with adjacent normal tissues.^38–42 However, increased telomerase activity has also been reported in colorectal tumors.^{38, 43, 44} To reconcile these observations, it is reasonable to propose that whereas telomere shortening promotes the chromosomal instability that drives early carcinogenesis, telomerase activation during later stages confers immortality to the tumor cells. Consistent with this view, longer telomeres have been identified in Dukes’ C and D compared with Dukes’ A and B tumors, and increasing telomere length and telomerase expression has been correlated with increasing depth of tumor invasion.^{38, 42}

DNA damage response and CIN

The DNA damage response machinery protects cells from exogenous as well as endogenous genotoxic stress by initiating a cascade that culminates in cell cycle arrest to allow sufficient time to repair the damage, or in the case of irreparable damage, by inducing senescence or apoptosis. Some DNA repair proteins are involved in human cancer, such as the ataxia telangiectasia mutated (ATM) and ataxia telangiectasia and Rad3-related (ATR) protein kinases.⁴⁵ Inactivating mutations in these genes predispose to specific syndromes, all characterized by an increased susceptibility to cancer, such as ataxia-telangiectasia (ATM mutations), Seckel (ATR mutations), Li-Fraumeni (TP53 mutations), and hereditary breast-ovarian cancer (BRCA1 and BRCA2 mutations) syndromes. Of all these checkpoint genes, TP53 has been the only one directly implicated in human colon cancer, having at least a permissive role for the development of CIN. Recently, Wang et al. identified more than 1,000 genes that could cause chromosomal instability based on their homology to genes of yeast and Drosophila melanogaster.²⁵ From this list, 100 candidate genes were sequenced in a panel of colorectal cancers. Interestingly, 19 mutations in 5 genes were identified, and one of these genes, MRE11, is involved in DNA double-strand break repair.

New insights that suggest a direct relationship between the DNA damage response and CIN have been provided by in vivo studies. Haploinsufficiency of histone H2AX, an ATM and ATR substrate, can compromise genomic integrity, and in a p53-deficient background, enhance tumor susceptibility.^{46, 47} ATM and H2AX are closely linked and map to a cytogenetic region (11q23) often deleted in human cancers. Mouse embryonic fibroblasts derived from ATM and H2AX double deficient mice show severe genomic instability.⁴⁸ In addition, Peddibhotla et al. reported that in vivo deficiency of Chk1, another DNA damage checkpoint effector, causes multiple mitotic defects and displaces Aurora B from late mitosis structures, resulting in failure of cytokinesis and multinucleation.⁴⁹

LOH and CIN

Loss of heterozygosity (LOH) is considered to be a hallmark feature of CIN-positive tumors. An average 25%–30% of alleles are lost in tumors; it is not unusual to observe losses in 75% of alleles in tumor samples.¹² Surprisingly, there are relatively few insights into the mechanisms that underlie the generation of LOH. Nevertheless, several pathways have been implicated, including mitotic nondisjunction, recombination between homologous chromosomes, and chromosomal deletion. In a study of LOH in colon cancer, Thiagalingam et al. performed a detailed molecular and cytogenetic analysis of LOH in the 5 chromosomes (1,5, 8, 17, and 18) most frequently lost in human colorectal cancers.⁵⁰ The majority of losses on chromosome 18 involved the whole chromosome and appeared to be caused by mitotic nondisjunction. However, when the losses were limited to only part of a chromosome, these appeared to be a consequence of interchromosomal recombinations and deletions associated with DNA double-strand breaks rather than to mitotic recombination, break-induced replication, or gene conversion. The specific genetic defects that underlie these LOH events remain unidentified.

THE “CHROMOSOMAL INSTABILITY PATHWAY”

Coupled with the typical karyotypic abnormalities observed in CIN tumors is the accumulation of a characteristic set of mutations in specific tumor suppressor genes and oncogenes (Table 2). These mutations activate oncogenic pathways that are critical for colorectal cancer pathogenesis. It is not clear whether CIN creates the appropriate environment for the accumulation of these mutations or vice versa, although a role for APC in the establishment of the CIN phenotype has been recently proposed. Nevertheless, the combination of these mutations in colorectal cancer in the setting of CIN has been historically designated the “chromosomal instability pathway.”

Table 2.

Overall Prevalence of Genetic Mutations in CIN-positive Colorectal Cancers

Gene	Chromosomal location	Prevalence of mutations	Function of gene product
Oncogenes
KRAS	12p12	~ 30–50%	Cell proliferation, survival, and transformation
CTNNB1	3p22	~ 4–15% (~ 50%*)	Regulation of Wnt pathway target genes that promote tumor growth and invasion
PIK3CA	3q26	~20%	Cell proliferation and survival
Tumor suppressor genes
APC	5q21	~ 30–70%	Inhibition of Wingless/Wnt signaling; cytoskeletal regulation
TP53	17p13	~ 40–50%	Cell cycle arrest, apoptosis induction
SMAD4, SMAD2	18q21	~ 10–20%	Intracellular mediators of the TGF-β pathway
DCC	18q21	~ 6%	Cell surface receptor for netrin-1

^*Identified in 50% of tumors without APC mutations

APC/β-Catenin

The earliest genetic event in colorectal tumorigenesis is activation of Wnt signaling through the genetic disruption of APC (Adenomatous Polyposis Coli) on 5q21.⁵¹ Germline mutations in APC are responsible for FAP (familial adenomatous polyposis).⁵² Somatic APC mutations are observed in 5% of dysplastic ACF, 30%–70% of sporadic adenomas, and in as many as 72% of sporadic tumors, indicating that functional loss of APC is an early event in tumor initiation.^53–55 Although germline inactivating mutations are distributed throughout the entire gene, somatic mutations are clustered in the mutation cluster region (MCR) between codons 1286 and 1513.⁵⁶ Hypermethylation of the APC promoter has been also reported in 18% of primary colorectal carcinomas and adenomas, representing an alternative mechanism for APC gene inactivation.⁵⁷

The APC gene product is a large protein with multiple functional domains that regulates differentiation, adhesion, polarity, migration, development, apoptosis, and even chromosomal segregation. A critical function is its interaction with glycogen synthase kinase-3β (GSK-3β) and β-catenin, each an essential component of the Wingless/Wnt signaling pathway (Figure 3). APC binds to β-catenin, GSK-3β and CK1α/ε (casein kinase 1α/ε) on an axin-conductin scaffold; the subsequent phosphorylation of β-catenin by GSK-3β leads to proteasome-dependent degradation and suppression of the Wnt signal. ⁵⁸ Mutant APC disrupts complex formation and the increased cytoplasmic levels of β-catenin can translocate to the nucleus, where it drives the transcription of multiple genes implicated in tumor growth and invasion through its interaction with the T-cell factor (TCF)/lymphoid enhancer factor (LEF) family of transcription factors.⁵⁹

Figure 3. The Wnt signaling pathway in the “OFF” and “ON” states.

In the absence of a Wnt signal, the destruction complex containing adenomatous polyposis coli (APC), glycogen synthase kinase 3β (GSK-3β) and casein kinase 1α/ε (CK1α/ε) on an axin-conductin scaffold targets the degradation of cytoplasmic β-catenin in a proteasome-dependent manner. In the nucleus, Wnt target genes are also kept silent by the repressor Groucho interacting with DNA-bound T cell factor (TCF). In the presence of a Wnt ligand, occupancy of the receptors Frizzled (Frz) and coreceptor low-density lipoprotein receptor-related protein (LRP) triggers the phosphorylation of the cytoplasmic tail of LRP by CK1 and GSK-3β as well as the disheveled (Dsh)-dependent recruitment of axin on phosphorylated LRP. Phosphorylation of β-catenin no longer occurs, and the increased cytoplasmic levels of β-catenin translocate to the nucleus, where the transcription of multiple genes is initiated through displacement of Groucho and the interaction of β-catenin with the T-cell factor (TCF)/lymphoid enhancer factor (LEF) family of transcription factors.

Gain-of-function mutations in β-catenin (CTNNB1) have been identified in as many as 50% of colon tumors with intact APC, reflecting the importance of the Wnt pathway.⁶⁰ Interestingly, Samowitz et al. found the frequency of CTNNB1 mutations in small adenomas (12.5%) to be significantly higher than in large adenomas (2.4%) and invasive cancers (1.4%), indicating that adenomas with alterations in β-catenin are less likely to progress to malignancy and that APC and CTNNB1 mutations are not functionally equivalent.⁶¹ Mutations in the AXIN and AXIN2/conductin genes have also been reported, but only in CRCs with MSI.^{62, 63}

K-RAS

KRAS is mutated in 30%–50% of CRCs.⁶⁴ Single nucleotide point mutations that occur in codons 12 and 13 of exon 2, and to a lesser extent in codon 61 of exon 3, lock the enzyme in the GTP-bound, activated form, leading to constitutive activation of RAS downstream signaling. Interestingly, KRAS mutations are observed in 60% to 95% of nondysplastic or hyperplastic ACF, indicating that these microscopic lesions are unlikely to be precursors of adenomas and cancer.^{65, 66}

Activated RAS regulates multiple cellular functions through well-described effectors (Figure 4). The best characterized effector is the Raf-MEK-ERK pathway. The Raf family includes 3 serine/threonine kinases (ARAF, BRAF, and RAF1) that activate MEK1 and MEK2, which in turn phosphorylate ERK1 and ERK2. ERK then phosphorylates cytosolic and nuclear substrates, including JUN and ELK1, that regulate enzymes such as Cyclin D1, which is involved in the control of cell cycle progression.⁶⁷ The significance of ERK activation in colon cancer pathogenesis is not clear. Whereas MEK is frequently activated in human colorectal tumors, recent studies have indicated that inhibitors of MEK are ineffective therapeutics.⁹

Figure 4. The RAS signaling pathway.

Growth factors binding to their cell surface receptors activate guanine exchange factors (GEF) such as SOS (son of sevenless) that are attached by the adaptor protein GRB2 (growth-factor-receptor bound protein 2). SOS stimulates the release of bound GDP from RAS, and it is exchanged for GTP, leading to the active RAS-GTP conformation. The GTPase-activating proteins (GAP) can bind to RAS-GTP and accelerate the conversion of RAS-GTP to RAS-GDP (guanosine diphosphate), which terminates signaling. Mutated RAS is constitutively active in the RAS-GTP conformation. Activated RAS regulates multiple cellular functions through effectors including the Raf-MEK-ERK pathway, PI3K, RALGDS, RALGDS-like gene (RLG) and RGL2.

RAS-GTP also binds the catalytic subunit of type I PI3Ks, which translocate to the plasma membrane to generate phosphoinositol lipids. Of note, PIK3CA is mutated in almost 20% of CRCs; mutations that occur in the hotspots located on exons 9 (E542K, E545K) and 20 (H1047R) are oncogenic in cellular models of CRC.^{5, 68} AKT/PKB is a critical downstream target of PI3K that regulates a cascade of responses including cell growth, proliferation, and survival by inactivating several pro-apoptotic proteins such as BAD and Forkhead (FKHR) transcription factors.^{69, 70} PI3K also activates Rac, a Rho family GTPase, which is important for RAS-mediated transformation in some cellular contexts.⁷¹

RAS can also activate a family of exchange factors for the Ral small GTPases, which include RALGDS, RALGDS-like gene (RLG) and RGL2. RALGDS is a key component for tumor formation in a mouse model of RAS-dependent skin carcinogenesis.⁷² RAS is also linked to NF-κ B, a transcription factor that regulates inflammatory and immune responses and cell survival. RALB stimulates TBK1, which can activate NF-κB by phosphorylating its inhibitor IκB. Barbie et al. demonstrated that RAS-mutant cells depend on TBK1 activation of NF-κB–regulated anti-apoptotic signals; Meylan et al. simultaneously showed that inhibition of NF-κB can block RAS-induced formation of lung tumors in mice.^{73, 74}

TP53

TP53 is located on the short arm of chromosome 17; it encodes a 393 amino acid transcription factor that is a tumor suppressor and a central coordinator of cellular responses to stress, including DNA damage, aberrant proliferative signals, and oxidative stress.⁷⁵ TP53 is induced by several oncogenic proteins, such as c-Myc, RAS, and adenovirus E1A.⁷⁶ Under normal conditions, p53 is negatively regulated by MDM2, E3-ubiquitin ligase, and the related protein MDM4 (also known as MDMX), which bind to the transactivation domain of p53 and target it for degradation by ubiquitination. In cells with a high level of stress, the interactions between MDM2, MDM4, and p53 are disrupted, allowing activated p53 to exert its transcriptional activity. Defined as the “guardian of the genome”, p53 is a master regulator that controls the transcription of hundreds of genes involved in DNA metabolism, apoptosis, cell cycle regulation, senescence, angiogenesis, immune response, cell differentiation, motility, and migration. Some of the best-studied targets of p53 are CIP1/WAF1, the p21 cell cycle inhibitor; GADD45 and 14-3-3, which contribute to G2 arrest; and BAX, FAS (APO1), PIG3, and KILLER (DR5), which regulate caspase activation and apoptosis.^{77, 78}

P53 dysfunction is an almost universal hallmark of human tumors and its loss of function has been reported in 4%–26% of adenomas, 50% of adenomas with invasive foci, and in 50%–75% of CRCs, defining its role in the transition from an adenoma to carcinoma.⁷⁹ The majority (~80%) of TP53 mutations are missense mutations: GC to AT transitions that occur principally in 5 hotspot codons (175, 245, 248, 273, and 282).⁸⁰ These mutations lead to the synthesis of an inactive protein with an abnormally long half-life that is detectable by immunohistochemistry. p63 and p73 are functionally and structurally related to p53, and they may also play a role in tumor suppression. Although loss of p63 and p73 has been recently found in many tumor types, their involvement in colon tumor pathogenesis has not yet been determined.⁸¹

18q Loss

Allelic loss at chromosome 18q has been identified in as many as 70% of primary colorectal tumors, particularly in advanced stages.² The gene Deleted in Colorectal Carcinoma (DCC) was initially proposed to encode a colorectal tumor suppressor, but its product is a cell surface receptor for the neuronal protein netrin-1 and DCC mutations are rarely detected in human colorectal tumors (6%). Furthermore, DCC mutant mice do not develop malignancies, so doubts have been raised about its role in colon cancer pathogenesis.^82–84 Other tumor suppressor genes reported on 18q are SMAD2 and SMAD4, which are intracellular mediators of the transforming growth factor-β pathway that are involved in the regulation of cell growth, differentiation and apoptosis. However, SMAD4 and SMAD2 mutations have been found in fewer than 20% and 10% of colon cancers, respectively.^{85, 86}

Cables, recently mapped to chromosome 18q11.2–12.1, appears to be another important candidate.^{87, 88} Functionally, Cables acts as a linker protein or “cable” that increases tyrosine phosphorylation of cyclin-dependent kinases (cdk2, cdk3, cdk5) by non-receptor tyrosine kinases (Src, Abl, Wee1). Loss of Cables expression occurs in 60%–70% of sporadic colorectal cancers, which is probably related to hypermethylation of CpG islands in its promoter coupled with 18q LOH.⁸⁸ Exposure to the carcinogen 1,2-dimethylhydrazine increased the incidence of colorectal tumors and reduced survival rates in Cables^−/− mice, compared to Cables^+/+ mice.⁸⁹

COX-2

Aberrant overexpression of cyclooxygenase-2 (COX-2) is thought to have an important role in development of CRC. COX-2, an early-response gene induced by growth factors, cytokines, inflammatory mediators and tumor promoters, is overexpressed in as many as 43% of adenomas and 86% of carcinomas.⁹⁰ Direct genetic evidence of the role of COX-2 in colorectal cancer came from a key study by Oshima et al. ⁹¹ In APC⁷¹⁶ knockout mice, the number of intestinal polyps was reduced by 34% when one allele of COX-2 was knocked out and by 86% when both alleles were deleted.

The tumorigenic effects of COX-2 can be attributed to the production of PGE₂; increased levels of PGE₂ have been reported in colorectal adenomas as well as carcinomas.^{92, 93} COX-2 and PGE₂ regulate proliferation, survival, migration, and invasion in colorectal tumors.⁹⁴ COX-2 also regulates angiogenesis; overexpression of COX-2 induces the production of pro-angiogenic factors such as vascular endothelial growth factor and basic fibroblast growth factor.⁹⁵ Homozygous deletion of COX-2 in vivo not only impairs the growth of tumor xenografts but also reduces tumor vascularity.⁹⁶

THE TIMING OF CIN: THE CHICKEN OR THE EGG?

The question of whether CIN is a cause or a consequence of the malignant process is unanswered. The stage of tumorigenesis at which the CIN phenotype arises is controversial, with some reports suggesting that CIN initiates tumorigenesis. However, others have contended that CIN is acquired and has a role in maintaining the tumorigenic process.

Several studies have demonstrated that CIN, measured as allelic imbalance in specific chromosomal regions, occurs at very early stages of tumorigenesis.^97–99 Shih et al. analyzed 32 sporadic colorectal adenomas averaging 2 mm in diameter for allelic imbalances using digital single nucleotide polymorphism PCR.⁹⁹ A relatively high frequency of allelic imbalances on chromosomes 5q (55%), 1p (10%), 8p (19%), 15q (28%), and 18q (28%) was identified, with over 90% of the adenomas exhibiting allelic imbalance of at least one chromosomal arm. Cardoso et al. evaluated the CIN status of polyps from patients with germline mutations in APC or MYH.¹⁰⁰ As many as 60% and 80% of the polyps exhibited aneuploid changes, respectively, with gains of chromosome 7 and 13, and losses of chromosomes 17p, 19q, and 22q, being the most prevalent aberrations. These findings support the view that chromosomal abnormalities can occur during very early stages of tumorigenesis. Whether these abnormalities truly constitute an underlying CIN and whether they occur before or after APC inactivation are questions difficult to answer. Nowak et al. devised a stochastic mathematical model to define under what conditions CIN would likely be the initial event in colorectal carcinogenesis or the second event following mutation of the first copy of the APC gene.¹⁰¹ For a wide range of realistic parameter values, CIN mutations appeared to initiate colon cancer. However, there is very little direct experimental evidence to support this model.

Interestingly, APC has been proposed as a potential initiator of CIN. In addition to its central role in regulating the Wingless/Wnt cascade, it has a role in cytoskeletal regulation through its ability to bind the plus ends of cytoplasmic and spindle microtubules and centrosomes through an EB1-binding domain. The first evidence that associated APC with CIN came in 2001, when 2 groups independently reported that mouse embryonic stem cells with mutations in APC became aneuploid in culture and accumulated multiple chromosomal abnormalities not seen in isogenic lines with wildtype APC.^{102, 103} Close inspection revealed that the mutant cells actually underwent polyploidization in whole-genome increments rather than losses and gains of one or a few individual chromosomes, the latter of which is more characteristic of the CIN phenotype. Therefore, the role of APC in the early generation of CIN remains highly provocative but incompletely defined. An additional twist is provided by studies demonstrating that inappropriate activation of Wnt signaling might contribute directly to CIN.^{104, 105} The authors defined a role for conductin in the regulation of the spindle checkpoint, and overexpression of conductin led to chromosomal losses and gains and the generation of CIN. In addition to APC, Giaretti et al. reported a strong correlation between KRAS mutations and aneuploidization.¹⁰⁶

A number of mouse models have also provided provocative new insights.¹⁰⁷ MAD2 overexpression in transgenic mice leads to structural and numerical chromosomal defects that predispose the animals to a wide range of tumors.¹⁰⁸ Rao et al. showed that reduced levels of BUBR1 in mice that carry a heterozygous mutation of the APC gene (Apc^Min/+ mice) increases not only the number of new colonic tumors but also their progression towards malignancy.¹⁰⁹ Finally, CENP-E heterozygous animals exhibit an increased incidence of spontaneous spleen lymphomas and benign lung tumors, suggesting that genetic instability could be an initiator of tumor formation.²⁶ Interestingly, CENP-E haploinsufficiency reduced the incidence of tumors in carcinogen- or genetically-induced models.²⁶ Taken together, these mouse models suggest that chromosomal instability could be an important initiating event in tumor formation. However, in certain contexts, aneuploidy also appeared to suppress tumorigenesis.

Although there are increasing reports that chromosomal instability could be an initiating event in some tumor models, it is not firmly established in colorectal cancer pathogenesis. In the same vein, there are no data that provide a direct connection between CIN and the acquisition of specific mutations in key genes required for colorectal development. This contrasts with our current understanding of DNA microsatellite instability, in which mutations in the DNA mismatch repair genes are directly linked to mutations in TGFBR2 (TGF-β receptor type II), IGF2R (insulin growth factor 2 receptor), and BAX. However, it is likely that CIN emerges during the early steps of the adenoma–carcinoma sequence, increasing the mutation rate and facilitating the progression to malignancy.

FROM STEM CELLS TO TUMOR: REFINING THE ADENOMA-CARCINOMA MODEL

The adenoma–carcinoma model requires some subtle refinement to incorporate recent insights from the study of cancer stem cells. A subpopulation of tumor cells within each CRC, approximately 0.25%–2.5% of the total number of cells, are proposed to be cancer-initiating stem cells. Cancer stem cells (CSCs) are defined by the ability to self-renew, perpetuate themselves for an extended period of time, and maintain the ability to generate a variety of differentiated cells through asymmetrical division. Upon transplantation into immunodeficient mice, CSCs can generate a phenocopy of the original tumor.¹¹⁰ Although CSCs could be derived from normal stem cells, they might also arise through dedifferentiation of mature somatic cells to reacquire stem cell characteristics.

CIN has been described in the transformation of normal stem cell to cancer stem cell. Miura et al. demonstrated that a small population of murine bone marrow-derived mesenchymal stem cells accumulated structural and numerical chromosomal aberrations after long-term culture, and when implanted in vivo generated fibrosarcomas.¹¹¹ Similarly, Shiras et al. observed the emergence of a spontaneously immortalized neural stem cell clone that had features of CSCs. These exhibited a high level of genomic instability due to defective checkpoint responses and abnormal checkpoint control.¹¹²

The normal gastrointestinal tract consists of 10⁷ crypts, each of which contains a small number of normal stem cells. They are protected at the base of the crypts, within the stem-cell niche, where they divide slowly and asymmetrically, generating a population of transit-amplifying cells that migrate up the crypt, proliferate, and progressively differentiate into enterocytes, goblet cells, enteroendocrine and Paneth cells. It has been proposed that the first mutation in APC occurs in a gastrointestinal stem cell, to generate an APC^+/− mutant stem cell. The mutant clone can then colonize the base of the crypt, taking over and replacing the non-mutant cells in the stem cell niche. Progressively, the entire niche will be colonized with APC-mutant cells and the crypt filled with their progeny, an event termed monoclonal conversion. Once the second APC allele is lost, niche succession occurs again and the crypt fills with APC^−/− cells, the monocryptal adenoma. The subsequent acquisition of mutations in genes such as KRAS and TP53 would then also occur within this cancer stem cell population.

Support of the view that a single mutation in normal intestinal stem cells is necessary and sufficient to induce tumor formation comes from studies by Barker et al. They propose that the leucine-rich repeat–containing G protein-coupled receptor 5 (Lgr5) is a biomarker of gastrointetinal stem cells.¹¹³ Deletion of APC in the Lgr5⁺ stem cells resulted in transformation with microadenomas within 8 days and large multi-villous adenomas in 14 days.¹¹⁴ Moreover, when APC was deleted in the transit-amplifying cell population, the growth of the microadenomas rapidly stalled and after 30 weeks, large adenomas were rarely seen. Thus, the accumulation of key mutations appears to begin specifically within the cancer stem cell compartment.

CLINICAL IMPLICATIONS OF CIN

The insights into the genetic basis of colon cancer have allowed the identification of new prognostic and predictive molecular markers as well as novel treatments for patients with CRC. For example, many have attempted to determine whether KRAS, TP53, or 18q alterations could serve as prognostic markers. Some data have indicated an increased risk of relapse and death among patients with a codon 12 mutation in KRAS, but other studies have failed to confirm this association.^115–118 Similarly, some have suggested that a TP53 mutation is associated with a higher risk of death, but this appears to be confined to those who already have the worst prognosis (Dukes’ stage D).^{119, 120} A number of compelling studies associated allelic deletion of chromosome 18q with poor outcome.^{121, 122} However, two recent studies have failed to validate these earlier observations.^{123, 124} Consequently, none of individual markers are used as molecular prognostic factors.¹²⁵ However, data indicate that the overall CIN phenotype is associated with a less favorable outcome in patients than those with tumors that exhibit MSI. Patients with CIN consistently exhibited low rates of overall and progression-free survival, irrespective of ethnic background, anatomic location, or treatment with 5-fluorouracil, compared to patients with MSI tumors. Prognosis was not influenced by adjuvant therapy in patients with stage II–III CRC.¹²⁶

Therapeutic targeting of pathways that directly initiate and perpetuate CIN has now reached the clinical arena. Small-molecule inhibitors of Aurora kinases, Plks, the spindle motor protein Eg5, and CENP-E have shown antitumor activity in preclinical models and are currently being evaluated in phase I and II clinical trials for the treatment of solid tumors. Of particular interest is the observation by Swanton et al. that CIN-positive tumors are intrinsically resistant to taxanes due to the similarity between pathways that regulate the separation of chromosomes at mitosis and those implicated in the response to taxanes.¹²⁷ This serves as the basis for the CINATRA (Chromosomal Instability and Anti-Tubulin Response Assessment) trial to assess whether patients with near diploid MSI-positive colorectal cancers derive benefit from EPO906, a new microtubule stabilizer, relative to patients with CIN-positive cancers.

The time course of progression from adenoma to cancer provides a window of opportunity for chemoprevention. COX-2 inhibitors successfully prevent polyp recurrence. Three prospective, randomized, placebo-controlled, multicenter trials of secondary prevention of colorectal adenomas (the Adenomatous Polyp Prevention on Vioxx (APPROVe), Adenoma Prevention with Celecoxib (APC), and Prevention of Sporadic Adenomatous Polyps (PreSAP) trials) demonstrated that selective COX-2 inhibitors reduced polyp recurrence, with a greater effect on formation of advanced adenomas.^128–130 However, the studies uncovered an increased risk of cardiovascular events, dampening some of the enthusiasm for this approach.

In the future, attempts to better define the mechanisms that initiate CIN, the relationship between CIN and tumor progression, and the feasibility of targeting chromosomally unstable cells will be critical to advance our understanding of the most common form of genetic instability in colon cancer.

Abbreviations used in this paper

CRC

colorectal cancer

CIN

chromosomal instability

MSI

microsatellite instability

CIMP

CpG island methylator phenotype

LOH

loss of heterozygosity

ACF

aberrant crypt foci