Colorectal Cancer and Genetic Alterations in the Wnt Pathway

S Segditsas and I Tomlinson

Molecular and Population Genetics Laboratory, London Research Institute, Cancer Research UK, London, UK

In colorectal tumours, Wnt pathway genetics continues to other Wnt pathway components in benign and malig- be dominated by mutations in the adenomatous polyposis nant colorectal tumours. The functional consequences coli (APC) gene. Germline mutations cause familial of these changes are mentioned briefly, but many are adenomatous polyposis and at least two-thirds of sporadic covered more comprehensively by other papers in this colorectal tumours also acquire APC mutations, quite volume. In order to minimize repetition in this issue, we possibly as the initiating events in tumorigenesis. These have not included a figure outlining the Wnt pathway. mutations almost always cause loss of the C-terminal Please refer to other papers if needed. functions of the APC protein – probably involved in microtubule binding, cell polarity and chromosome segregation – and deletion of the SAMP repeats that are important for binding to axin and formation of the APC mutation spectrum beta-catenin phosphorylation complex. The truncated APC APC proteins are, in general, stable and almost certainly Although germline mutations only account for 1% retain some activity in beta-catenin binding. The ‘two hits’ of all colorectal cancer cases, they almost invariably at APC are coselected so as to produce an optimal cause a FAP phenotype. This condition is typified by activation of Wnt signalling (just-right hypothesis). In a hundreds or thousands of colorectal adenomas, one or minority of colorectal tumours, Wnt activation can occur more of which usually progresses to carcinoma by the through mutations that affect phosphorylation sites within fourth decade if the patient is untreated (Galiatsatos and exon 3 of beta-catenin, causing protein stabilization. In Foulkes, 2006). The great majority of FAP-associated APC other tumours, epigenetic transcriptional silencing or mutations lead to a truncated protein. Frameshift mutation of the secreted frizzled-related proteins may mutations outnumber nonsense changes by about two- modulate Wnt levels. Mutations in the Wnt components to-one, and the most common nonsense changes are AXIN1, AXIN2 and TCF4 have been found in micro- C>T mutations, probably resulting from spontaneous et al satellite-unstable colon cancers, but it is not clear in every deamination of 5-methylcytosine (Laurent-Puig ., case whether these changes are functional. Therapeutic 1998). In addition to protein-truncating mutations, modulation of the Wnt pathway remains an attractive deletions of the whole gene or exon(s) have been et al therapeutic possibility for colorectal carcinomas. reported in about 5% of FAP cases (Sieber ., 2002). Oncogene (2006) 25, 7531–7537. doi:10.1038/sj.onc.1210059 The protein-truncating mutations can occur anywhere in the APC gene, but the classical phenotype of more Keywords: APC; beta-catenin; ‘just right’ than 100 adenomas is associated with mutations between codons 178 and 309 and between codons 409 and 1580, corresponding to exons 5–8, 9–14 and the first half of the large final exon 15 (Fearnhead et al., 2001). Frameshift mutations at codons 1061 and 1309 are Introduction particularly common in western European populations; both of these regions contain short repeats, and a degree It is over 15 years since the adenomatous polyposis coli of hypermutation, plus genetic drift, probably accounts (APC) gene was found to be mutated in familial for their relatively high frequency. From the persepec- adenomatous polyposis (FAP) and, subsequently, in a tive of APC mutations, the most important functional very high proportion of sporadic colorectal tumours. Of domains of the APC gene appear to be the first serine course, the connection with the Wnt pathway was not alanine methionine proline (SAMP) (axin binding) known at that time, but it is arguable that without the repeat at codon 1580 (Smits et al., 1999) and the first, cloning of APC, this issue of Oncogene would not exist. second and third 20-amino acid repeats (20AARs) In this paper, we shall focus on the genetic and involved in beta-catenin binding and degradation. The epigenetic changes that occur in APC, beta-catenin and great majority of pathogenic APC mutations truncate the protein before the first SAMP repeat and leave a stable, truncated protein that encodes 0–3 20AARs Correspondence: Professor I Tomlinson, Molecular and Population Genetics Laboratory, London Research Institute, Cancer Research (Figure 1). UK, 44, Lincoln’s Inn Fields, London WC2A3PX, UK. In colorectal tumours from FAP patients, the E-mail: [email protected] germline wild-type allele either undergoes loss of Colorectal cancer and genetic alterations in the Wnt pathway S Segditsas and I Tomlinson 7532 Heptad repeats 15aa repeats SAMP repeats Basic domain (dimerisation) (β-catenin binding) (axin binding) (microtubule binding)

6 57 453 767 1020 1169 1266 1495 2033 2200 2400 2560 2843 1379 1581 Armadillo repeats 20aa repeats EB1 and HDLG (β-catening binding/degradation) binding sites a Gl mut allele

Gl wt allele

b Gl mut allele

Gl wt allele

c Gl mut allele

Gl wt allele

Figure 1 The ‘just-right’ model. The figure shows the multiple domains of the APC protein and the correlation between the position of the germline (gl) mutation and that of the somatic mutation. (a) Germline mutations between the first and the second 20AAR are associated with LOH. (b) Germline mutations before the first 20AAR are associated with somatic mutations between the second and third 20AAR. (c) Germline mutations after the second 20AAR are associated with somatic mutations before the first 20AAR.

heterozygosity (LOH) or acquires a protein-truncating one-third to one-half of tumours, a frequency lower mutation. In general, these are thought to be sponta- than that at many other bona fide tumour suppressor neous events and APC is often cited as a classical loci. tumour-suppressor gene, both alleles of which must be The reason for the MCR and relatively low frequency inactivated for tumorigenesis to begin. It has been of LOH at APC came from studies of FAP (Lamlum argued, however, that the number of adenomas in the et al., 1999). It was found that LOH is strongly FAP colon is too great for each one to have acquired a associated with germline mutations between the first ‘second hit’, especially if FAP adenomas are frequently and second 20AAR (codons 1285–1379). Germline polyclonal. The validity of such calculations is inhe- mutations before codon 1280 are associated with rently difficult to confirm or refute, since they depend on somatic mutations between the second and third the one hand on estimations of the susceptible cell 20AAR (codons 1400 and 1495); and germline muta- population and, on the other, on finding ‘second hits’ in tions after codon 1400 are associated with somatic almost all early tumours, a technically very demanding mutations before codon 1280 (Lamlum et al., 1999; task. For the purposes of this paper, it will be assumed Albuquerque et al., 2002; Crabtree et al., 2003) that FAP adenomas do have (at least) two hits, although (Figure 1). Most tumours end up with APC alleles that the possibility that very early lesions can grow without a encode a total of two 20AARs. Similar associations exist somatic APC mutation must be borne in mind. for sporadic colorectal cancers. This association has The somatic mutations that occur in sporadic colo- been proposed to cause an optimal level of Wnt rectal tumours are similar in spectrum and location to signalling/beta-catenin activation (Lamlum et al., 1999; the somatic mutations in FAP polyps. Albuquerque et al., 2002). Whether this level is the maximum level is unclear, since APC protein truncated at codon 1309 has been shown to have increased Selective constraints on APC mutations in inherited and stability compared with other mutants (Dihlmann sporadic tumours et al., 1999) and many truncated APC proteins retain some ability to bind and, perhaps, degrade beta-catenin In recent years, it has become established that the (Rubinfeld et al., 1997). Absence of APC appears to be somatic APC mutation in FAP tumours – and sporadic selectively disadvantageous (Schneikert and Behrens, colorectal carcinomas – is not simply an event that 2006). Whatever the case, it is clear that selective removes function of the protein encoded by the germline constraints act on colorectal tumours such that some wild-type allele. It had been known for some years that combinations of APC mutations provide a superior somatic APC mutations tend to occur in a restricted growth advantage for the tumour cell. region of the gene (mutation cluster region (MCR)) that The ‘1st hit-2nd hit’ association at APC has been lay approximately between codons 1250 and 1450 dubbed the ‘just right’ hypothesis (Albuquerque et al., (Miyoshi et al., 1992). LOH occurred in about 2002). This is a convenient term that encapsulates the

Oncogene Colorectal cancer and genetic alterations in the Wnt pathway S Segditsas and I Tomlinson 7533 notion well – and it will be used below – although we did (i) Loss of germline wt + truncating mutation not use such a term originally, because the selective constraints are evidently not so rigid that deviation from the total of two 20AARs cannot be tolerated in a Phenotype substantial minority of lesions (Lamlum et al., 1999; Crabtree et al., 2003). Further studies have shown that a LOH as first somatic event the ‘just right’ model applies to extracolonic lesions in FAP, but that the combination of mutations is different X (Lamlum et al., 1999; Groves et al., 2002). In desmoids LOH Allelotype + + or and uppergastrointestinal tumours, LOH is associated with germline APC mutations between the second and third 20AARs, and the optimum mutant proteins X appear to encode a total of three or four 20AARs. Another manifestation of the requirement for optimal b LOH as second somatic event Wnt signalling in colorectal tumorigenesis is the LOH existence of attenuated disease (AFAP) (Spirio et al., Allelotype + + 1993). Although the division between classical FAP and XXX AFAP is indistinct (Knudsen et al., 2003), in the latter, patients typically have fewer than 100 colorectal adenomas at presentation. AFAP cases generally have germline APC mutations in distinct regions of the gene (ii) Loss of germline mut+ truncating mutation (glwt) (exons 1–4, exon 9 and distal to codon 1580 in exon 15). LOH It is now clear that at least some AFAP tumours have to + Allelotype + acquire three hits at APC (Spirio et al., 1998; Su et al., X XX 2000; Sieber et al., 2006). This has been particularly Figure 2 Main pathways of APC mutation in AFAP. Two main reported in patients with exon 9 mutations, a region that pathways of APC mutation have been identified in AFAP polyps: undergoes physiological alternative splicing. Two main (i) loss of the germline wild-type (wt) allele (LOH) (which can occur genetic pathways exist: (i) an additional somatic as a first (a) or second (b) event), together with acquisition of mutation, typically at codon 1556, affecting the germline a truncating mutation on the germline mutant allele; or (ii) acquisition of a truncating mutation on the germline wild-type mutant allele together with loss of the germline wild- (gl wt) allele and loss of the germline mutant. LOH typically occurs type; and (ii) mutation, typically between the first and by mitotic recombination, rather than deletion, thus resulting in second 20AAR, of the germline wild-type allele plus loss a duplication of the alternative allele. (’) Germline mutation, of the germline mutant (Figure 2). (x) somatic mutation(s), LOH is loss of heterozygosity. In this way, it seems that the AFAP colorectal adenoma can acquire a ‘just right’ level of Wnt signalling. The unanswered question is why AFAP cases more colorectal adenomas, and hence an earlier onset of develop more adenomas than the general population. colorectal carcinoma (Nugent et al., 1994). We have Pathway (ii) is no different from the sporadic situation. proposed (Lamlum et al., 1999) that this severe For pathway (i), it has been proposed that codon 1556 phenotype occurs because under the ‘just right’ model mutations are rarely selected in colorectal tumours, but these patients tend to acquire their ‘second hits’ by occur frequently – because this is a ‘slippery’ oligonu- LOH, a relative frequent event compared with a cleotide repeat tract – and are selected in AFAP. Other truncating APC mutation (although direct functional data show that mutations close to codon 1556 also occur effects of the truncated APC protein may also play a specifically in AFAP, suggesting that the driving force is role). Afurther finding is that FAPpatients with selection for an allele that encodes some protein germline mutations between the second and third truncated in exon 9 and some between the second and 20AARs have more severe disease than those with third 20AAR. Perhaps a combination of these factors is mutations before the first AAR (excluding AFAP- to blame (Su et al., 2000; Sieber et al., 2006). associated mutations); again, this may be because the former patients can acquire ‘just right’ ‘second hits’ in the large region of the gene before the first AAR, Genotype–phenotype associations in FAP and whereas mutations in the latter are more constrained to modifier loci the region between the second and third AARs. For desmoids, the most severe disease is associated with There exist well-established genotype–phenotype asso- germline APC mutations after codon 1400, again the ciations for FAP. The best defined is the association LOH-associated part of the gene (Sturt et al., 2004). between the number of colorectal adenomas (classical Although the relative probabilities of a ‘just right’ FAP versus AFAP) and the location of the germline somatic mutation may explain much of the variation in mutation (see above). Afurther association is that disease severity in FAP, there remains considerable, patients with germline APC mutations between the first unexplained phenotypic heterogeneity. There is evidence and second 20AARs (with a particularly high frequency that variants at genetically distinct, disease-modifying of mutations occurring at codon 1309) tend to have loci can explain at least some of the residual variation.

Oncogene Colorectal cancer and genetic alterations in the Wnt pathway S Segditsas and I Tomlinson 7534 The identities of these genes – which are good candidate MYH associated Position of somatic APC mutations colorectal cancer susceptibility loci for the general FAP and sporadics population – are as yet unknown (Crabtree et al., 25 2001, 2002). 20

15 MYH-associated polyposis

It was something of a surprise when a recessive ‘version’ Frequency 10 of FAP was identified in 2002 (Al-Tassan et al., 2002). There was a prevalent belief that >100 colorectal 5 adenomas was pathognomonic of FAP and, hence, a germline APC mutation. Indeed, the identification of 0 MYH/MUTYH mutations in an AFAP family arose as 0-100 700- 1000- 1250- 1400- 1800- 2700- 800 1100 1300 1500 1900 2800 an unexpected consequence of the investigation of a Codon missense APC variant, E1317Q (see below). MYH itself Figure 3 APC mutations in MAP. MYH-associated lesions tend has no known involvement in the Wnt pathway, but to accumulate somatic APC mutations in accordance with the ‘just- appears to act by increasing the rate of G:C>T:A right’ model, with one hit occurring before the first 20AAR (codon transversions in susceptible cells in the gastrointestinal 1266) and the other hit usually after the second 20AAR (codon tract. The predominant pairs of biallelic APC mutations 1400). FAP and sporadic lesions tend to accumulate somatic in MAP polyps are one located before the first 20AAR mutations around the second 20AAR. and the other, after the second 20AAR, thus following the ‘just right’ model (Figure 3). The predominance of European population, and has been associated with a G:C>T:Achanges means that LOH is rare, as are risk of colorectal tumours in some studies. Most of these mutations between the first and second 20AARs studies, including our own, have lacked statistical power compared with sporadic colorectal tumours. MYH- and any role for E1317Q in tumour risk is currently associated polyposis (MAP) accounts for about 30% highly uncertain. More common non-synonymous SNPs of cases who present with 5–100 adenomas and about in the APC gene include D1822 V (MAF ¼ 0.18), 10–20% of APC mutation-negative classical FAP. E1120G (MAF unknown), Q1370K (MAF unknown), S1496T (MAF ¼ 0.01), T1973T (MAF ¼ 0.003), S2502G (MAF ¼ 0.038), S2523A(MAFunknown) and D2655E Low-penetrance APC variants (MAF unknown). None of these variants has been shown convincingly to increase the risk of colorectal The missense variant APC I1307K acts as an unusual tumours, although there have been suggestions of effects tumour predisposition allele, because it acts as a ‘pre- of D1822V in some subgroups (Slattery et al., 2001). mutation’ (Laken et al., 1997). The change, which is Afurther possibility is that polymorphisms or rare largely present in those of Ashkenazi Jewish origin, variants in other Wnt pathway components contribute creates a hypermutable A8 tract. This tract is prone to to the risk of colorectal tumours. Such potential disease- replication-mediated slippage and the resulting single- associated variants include N287S of beta-catenin and base insertion or deletion creates a truncated APC P312T, R398H, L445M, D545E, G700S and R891Q of protein. In accordance with the ‘just right’ model, the AXIN1 (Fearnhead et al., 2004). However, any associa- other allele undergoes LOH and a tumour results. The tions of these variants with disease currently remain tumour risk associated with I1307K has been contro- tantalizing, but unproven. versial, but most recent reports suggest that it has a relatively small effect (perhaps only 1.5-fold risk of colorectal cancer), suggesting that the A8 tract is only APC promoter methylation modestly hypermutable (Gryfe et al., 1999). Moreover, most tumours from I1307K patients do not show Transcriptional silencing of APC by promoter methyla- frameshift mutations within the A8 tract (Sieber et al., tion has been suggested as an alternative to somatic 2003). The most likely explanation is that the patients mutation. Methylation of one of the two main under study have been highly selected and many have promoters, 1A, appears to occur densely in the core a separate, additional predisposition to colorectal promoter region and can occur in adenomas, suggesting tumours. It is also possible that I1307K has some that it is an early event in tumorigenesis (Esteller et al., direct functional effect. Nevertheless, I1307K – with a 2000). Few studies, however, have considered both the population frequency of 5–10% in Ashkenazim – relationship of methylation to mutation of APC, and the remains an important predisposition allele for colorectal evidence that does exist suggests that methylation of tumours. mutated alleles may occur, thus raising some doubts as Another missense APC variant that has been asso- to the functional importance of methylation, and other ciated with disease is E1317Q (Frayling et al., 1998). studies have not yet shown convincingly that methyla- This allele has a frequency of about 1% in the northern tion leads to a reduction in or loss of APC expression.

Oncogene Colorectal cancer and genetic alterations in the Wnt pathway S Segditsas and I Tomlinson 7535 Alternatives to APC mutation for Wnt activation MSI þ colorectal cancers (Liu et al., 2000), primarily involving insertions or deletions within short, coding It is arguable whether every colorectal tumour develops, oligonucleotide repeats. Although most of these cancers at least in part, as a result of Wnt pathway activation. showed nuclear beta-catenin expression, taken as Any exceptions to this rule will be interesting if shown in indicative of Wnt activation, and no mutation in beta- the future, but whatever the case, it is clear that not all catenin or APC, almost all changes were present as colorectal tumours activate Wnt through mutation (or heteroygotes with the wild-type AXIN2 allele, and epimutation) of APC. functional consequences of these AXIN2 changes remain The best characterized alternatives to APC mutations uncertain. Intriguingly, germline mutations in AXIN2 are activating mutations of beta-catenin (Ilyas et al., have been suggested as predisposing to colorectal cancer 1997; Morin et al., 1997). These either delete the whole (Lammi et al., 2004). However, the principal phenotype of exon 3 or target individual serine or threonine in these families was tooth agenesis. The colorectal residues encoded by this exon. These serines/threonines tumour phenotype in the single family studied was (codons 45, 41, 33 and 37) are phosphorylated by the highly variable, ranging from isolated carcinoma at an degradation complex that contains APC, and hence unremarkable age to nearly 70 adenomas at 57 years their mutation causes beta-catenin to escape from and 10 hyperplastic polyps at 35 years. Additionally, the proteasomal degradation. Mutations at these sites are normal expression of AXIN2 in the intestine is unclear not found together with APC mutations, showing that and no analyses of the colorectal tumours – for example, changes in the two genes are alternatives in some way, for loss of the wild-type allele or nuclear beta-catenin although their functional effects are unlikely to be expression – were undertaken. In the absence of this identical given that the C-terminal functions of APC corroborative evidence and the lack of similar families in that are removed by almost all pathogenic mutations do the literature, it remains possible that the colorectal not appear to be replicated in the beta-catenin protein. tumour phenotype was coincidental in this kindred. Indeed, this difference may explain why, despite only Recently, it has even been proposed that Wnt signalling monoallelic mutations being necessary, beta-catenin leads to chromosomal instability through upregulation mutations only occur in o5% of colorectal cancers, of conductin, which is itself a Wnt target. since additional events may be necessary for functional TCF4 contains an oligonucleotide repeat tract in the parity with mutations in APC. 30 region of the gene that frequently undergoes slippage Recent data suggest that beta-catenin mutations in microsatellite-unstable colorectal cancers. It has been largely occur in the context of the hereditary non- proposed that this change creates a truncated, more active polyposis colon cancer syndrome, which is caused by form of the transcription factor, although TCF4 has germline mutations in DNAmismatch repair genes multiple splice variants that make functional assessment MSH2 and MLH1 (Johnson et al., 2005). In addition, problematic (Duval et al., 2000; Shimizu et al., 2002). these mutations appear to be more common in carcinomas than adenomas, in contrast to the generally held view that Wnt-activating (that is, APC) mutations Epigenetic changes in Wnt pathway components initiate colorectal tumorigenesis. Axin (AXIN1) is an important part of the complex APC promoter methylation has been discussed above and that phosphorylates beta-catenin and hence tags it for several other studies have reported methylation of the degradation. Axin probably acts as a scaffold protein in promoters of other Wnt pathway members, including this context and its loss would be predicted to cause DKK1 (Aguilera et al., 2006) and WIF1 (Batra et al., effectively increased Wnt signalling. The following 2006). The best evidence that such changes have AXIN1 mutations, presumably somatic, have been functional consequences comes from study of the secreted found in colorectal cancer cell lines: S215L; and frizzled-related proteins (SFRPs) (Suzuki et al., 2004). L396M (Webster et al., 2000). The latter was shown to SFRPs are structurally similar to the Wnt-receptor affect binding of Axin to GSK3-beta. Further somatic frizzled proteins and can inhibit signalling by binding AXIN1 mutations have been found in sporadic colo- Wnts. Restoring SFRP expression in colorectal cancer rectal cancers at codons 395, 411 and 418 (GSK-3beta- cells can upregulate autocrine or paracrine Wnt signal- binding domain) and codons 101, 103 and 122 (RGS ling. Subsequently, somatic mutations in SFRP1 have domain), including one double mutant (Shimizu et al., been reported in colorectal cancers (Caldwell et al., 2004) 2002). The absence of protein-truncating mutations and and it has been suggested that these contribute to low frequency of biallelic changes means that there are increased non-canonical Wnt signaling, given the pre- unresolved issues regarding the functional importance sence of APC mutations in the same tumours. about the AXIN1 changes that have been found. In addition, at least some of the colorectal cancers with AXIN1 mutations also have APC mutations, showing Other genetic changes that affect Wnt signalling that AXIN1 changes cannot be the sole cause of Wnt activation in these tumours. Given the crosstalk and overlap between molecular In some tissues, AXIN2 (conductin) may substitute pathways, especially in cell signalling, it is entirely for AXIN1 in the beta-catenin degradation complex. expected that Wnt activation in colorectal tumours will Mutations in the AXIN2 gene have been reported in be inﬂuenced by (epi)genetic changes in genes that act

Oncogene Colorectal cancer and genetic alterations in the Wnt pathway S Segditsas and I Tomlinson 7536 primarily in other pathways. One recent example is the signalling is constrained even in tumour cells and that effect of K-ras mutations – present in 30–40% of there may be further modifications to Wnt levels, colorectal cancers – that can induce Wnt signalling molecular targeting of this pathway is an attractive in vitro by increasing beta-catenin stability, perhaps therapeutic option. Genetic and epigenetic studies have through inhibition of the kinase activity of GSK-3beta provided the basis for these findings. Potentially, small (Li et al., 2005). In other cases, Wnt targets are disturbances in Wnt activation – up or down – can lead genetically altered in colorectal tumours. c-Myc may to a suboptimal state. Non-steroidal anti-inflammatory be amplified in some cases, and cyclinD1 in a very few. drugs, including aspirin, may partly act through Wnt EPHB2 undergoes slippage in a short repeat tract in suppression, and analogues of molecules such as sFRPs about half of microsatellite-unstable colon cancers and WIF1 may provide useful benefits. More intrigu- (Alazzouzi et al., 2005), although biallelic changes are ingly and controversially, the data suggest that Wnt apparently uncommon; other tumours show transcrip- activation may also be therapeutically beneficial under tional silencing. The functional consequences, if any, of some circumstances. these EPHB2 changes are currently unknown. The chief promise for Wnt pathway genetics in the near future lies in the possibility that low-penetrance alleles of Wnt genes, with small effects on risk, play roles Concluding remarks in predisposition to colorectal tumours. The results from such studies are expected within months. These analyses Given that nearly all colorectal tumours activate Wnt might lead eventually to a rash of genetic testing and signalling, that the ‘just-right’ model shows that Wnt novel and/or targeted methods for disease prevention.

References

Aguilera O, Fraga MF, Ballestar E, Paz MF, Herranz M, Dihlmann S, Gebert J, Siermann A, Herfarth C, von Knebel Espada J et al. (2006). Epigenetic inactivation of the Wnt Doeberitz M. (1999). Dominant negative effect of the antagonist DICKKOPF-1 (DKK-1) gene in human colo- APC1309 mutation: a possible explanation for genotype– rectal cancer. Oncogene 25: 4116–4121. phenotype correlations in familial adenomatous polyposis. Al-Tassan N, Chmiel NH, Maynard J, Fleming N, Livingston AL, Cancer Res 59: 1857–1860. Williams GT et al. (2002). Inherited variants of MYH Duval A, Rolland S, Tubacher E, Bui H, Thomas G, associated with somatic G:C-T:Amutations in colorectal Hamelin R. (2000). The human T-cell transcription factor-4 tumors. Nat Genet 30: 227–232. gene: structure, extensive characterization of alternative Alazzouzi H, Davalos V, Kokko A, Domingo E, Woerner SM, splicings, and mutational analysis in colorectal cancer cell Wilson AJ et al. (2005). Mechanisms of inactivation of the lines. Cancer Res 60: 3872–3879. receptor tyrosine kinase EPHB2 in colorectal tumors. Esteller M, Sparks A, Toyota M, Sanchez-Cespedes M, Cancer Res 65: 10170–10173. Capella G, Peinado MA et al. (2000). Analysis of Albuquerque C, Breukel C, van der Luijt R, Fidalgo P, Lage P, adenomatous polyposis coli promoter hypermethylation in Slors FJ et al. (2002). The ‘just-right’ signaling model: APC human cancer. Cancer Res 60: 4366–4371. somatic mutations are selected based on a specific level of Fearnhead NS, Britton MP, Bodmer WF. (2001). The ABC of activation of the beta-catenin signaling cascade. Hum Mol APC. Hum Mol Genet 10: 721–733. Genet 11: 1549–1560. Fearnhead NS, Wilding JL, Winney B, Tonks S, Bartlett S, Batra S, Shi Y, Kuchenbecker KM, He B, Reguart N, Mikami Bicknell DC et al. (2004). Multiple rare variants in different I et al. (2006). Wnt inhibitory factor-1, a Wnt antagonist, genes account for multifactorial inherited susceptibility is silenced by promoter hypermethylation in malignant to colorectal adenomas. Proc Natl Acad Sci USA 101: pleural mesothelioma. Biochem Biophys Res Commun 342: 15992–15997. 1228–1232. Frayling IM, Beck NE, Ilyas M, Dove-Edwin I, Goodman P, Caldwell GM, Jones C, Gensberg K, Jan S, Hardy RG, Byrd P Pack K et al. (1998). The APC variants I1307K and E1317Q et al. (2004). The Wnt antagonist sFRP1 in colorectal are associated with colorectal tumors, but not always with a tumorigenesis. Cancer Res 64: 883–888. family history. Proc Natl Acad Sci USA 95: 10722–10727. Crabtree M, Sieber OM, Lipton L, Hodgson SV, Lamlum H, Galiatsatos P, Foulkes WD. (2006). Familial adenomatous Thomas HJ et al. (2003). Refining the relation between ‘first polyposis. Am J Gastroenterol 101: 385–398. hits’ and ‘second hits’ at the APC locus: the ‘loose fit’ model Groves C, Lamlum H, Crabtree M, Williamson J, Taylor C, and evidence for differences in somatic mutation spectra Bass S et al. (2002). Mutation cluster region, association among patients. Oncogene 22: 4257–4265. between germline and somatic mutations and genotype– Crabtree MD, Tomlinson IP, Hodgson SV, Neale K, Phillips RK, phenotype correlation in upper gastrointestinal familial Houlston RS. (2002). Explaining variation in familial adeno- adenomatous polyposis. Am J Pathol 160: 2055–2061. matous polyposis: relationship between genotype and phenotype Gryfe R, Di Nicola N, Lal G, Gallinger S, Redston M. (1999). and evidence for modifier genes. Gut 51: 420–423. Inherited colorectal polyposis and cancer risk of the APC Crabtree MD, Tomlinson IP, Talbot IC, Phillips RK. (2001). I1307K polymorphism. Am J Hum Genet 64: 378–384. Variability in the severity of colonic disease in familial Ilyas M, Tomlinson IP, Rowan A, Pignatelli M, Bodmer WF. adenomatous polyposis results from differences in tumour (1997). Beta-catenin mutations in cell lines established from initiation rather than progression and depends relatively human colorectal cancers. Proc Natl Acad Sci USA 94: little on patient age. Gut 49: 540–543. 10330–10334.

Oncogene Colorectal cancer and genetic alterations in the Wnt pathway S Segditsas and I Tomlinson 7537 Johnson V, Volikos E, Halford SE, Eftekhar Sadat ET, Shimizu Y, Ikeda S, Fujimori M, Kodama S, Nakahara M, Popat S, Talbot I et al. (2005). Exon 3 beta-catenin Okajima M et al. (2002). Frequent alterations in the Wnt mutations are speciﬁcally associated with colorectal signaling pathway in colorectal cancer with microsatellite carcinomas in hereditary non-polyposis colorectal cancer instability. Genes Chromosomes Cancer 33: 73–81. syndrome. Gut 54: 264–267. Sieber O, Lipton L, Heinimann K, Tomlinson I. (2003). Knudsen AL, Bisgaard ML, Bulow S. (2003). Attenuated Colorectal tumourigenesis in carriers of the APC I1307K familial adenomatous polyposis (AFAP). A review of the variant: lone gunman or conspiracy? J Pathol 199: 137–139. literature. Fam Cancer 2: 43–55. Sieber O, Segditsas S, Knudsen A, Zhang J, Luz J, Rowan A Laken SJ, Petersen GM, Gruber SB, Oddoux C, Ostrer H, et al. (2006). Disease severity and genetic pathways Giardiello FM et al. (1997). Familial colorectal cancer in in attenuated familial adenomatous polyposis vary greatly, Ashkenazim due to a hypermutable tract in APC. Nat Genet but depend on the site of the germline mutation. Gut 55: 17: 79–83. 1440–1448. Lamlum H, Ilyas M, Rowan A, Clark S, Johnson V, Bell J Sieber OM, Lamlum H, Crabtree MD, Rowan AJ, Barclay E, et al. (1999). The type of somatic mutation at APC in Lipton L et al. (2002). Whole-gene APC deletions cause familial adenomatous polyposis is determined by the site of classical familial adenomatous polyposis, but not attenuated the germline mutation: a new facet to Knudson’s ‘two-hit’ polyposis or ‘multiple’ colorectal adenomas. Proc Natl Acad hypothesis. Nat Med 5: 1071–1075. Sci USA 99: 2954–2958. Lammi L, Arte S, Somer M, Jarvinen H, Lahermo P, Thesleff I Slattery ML, Samowitz W, Ballard L, Schaffer D, Leppert M, et al. (2004). Mutations in AXIN2 cause familial tooth Potter JD. (2001). Amolecular variant of the APCgene at agenesis and predispose to colorectal cancer. Am J Hum codon 1822: its association with diet, lifestyle, and risk of Genet 74: 1043–1050. colon cancer. Cancer Res 61: 1000–1004. Laurent-Puig P, Beroud C, Soussi T. (1998). APC gene: Smits R, Kielman MF, Breukel C, Zurcher C, Neufeld K, database of germline and somatic mutations in human Jagmohan-Changur S et al. (1999). Apc1638T: a mouse tumors and cell lines. Nucleic Acids Res 26: 269–270. model delineating critical domains of the adenomatous Li J, Mizukami Y, Zhang X, Jo WS, Chung DC. (2005). polyposis coli protein involved in tumorigenesis and Oncogenic K-ras stimulates Wnt signaling in colon cancer development. Genes Dev 13: 1309–1321. through inhibition of GSK-3beta. Gastroenterology 128: Spirio L, Olschwang S, Groden J, Robertson M, Samowitz W, 1907–1918. Joslyn G et al. (1993). Alleles of the APC gene: an Liu W, Dong X, Mai M, Seelan RS, Taniguchi K, attenuated form of familial polyposis. Cell 75: 951–957. Krishnadath KK et al. (2000). Mutations in AXIN2 cause Spirio LN, Samowitz W, Robertson J, Robertson M, colorectal cancer with defective mismatch repair by activat- Burt RW, Leppert M et al. (1998). Alleles of APC modulate ing beta-catenin/TCF signalling. Nat Genet 26: 146–147. the frequency and classes of mutations that lead to Miyoshi Y, Nagase H, Ando H, Horii A, Ichii S, Nakatsuru S colon polyps. Nat Genet 20: 385–388. et al. (1992). Somatic mutations of the APC gene in Sturt NJ, Gallagher MC, Bassett P, Philp CR, Neale KF, colorectal tumors: mutation cluster region in the APC gene. Tomlinson IP et al. (2004). Evidence for genetic predisposi- Hum Mol Genet 1: 229–233. tion to desmoid tumours in familial adenomatous poly- Morin PJ, Sparks AB, Korinek V, Barker N, Clevers H, posis independent of the germline APC mutation. Gut 53: Vogelstein B et al. (1997). Activation of beta-catenin-Tcf 1832–1836. signaling in colon cancer by mutations in beta-catenin or Su LK, Barnes CJ, Yao W, Qi Y, Lynch PM, Steinbach G. APC. Science 275: 1787–1790. (2000). Inactivation of germline mutant APC alleles by Nugent KP, Phillips RK, Hodgson SV, Cottrell S, Smith- attenuated somatic mutations: a molecular genetic mechan- Ravin J, Pack K et al. (1994). Phenotypic expression in ism for attenuated familial adenomatous polyposis. Am J familial adenomatous polyposis: partial prediction by Hum Genet 67: 582–590. mutation analysis. Gut 35: 1622–1623. Suzuki H, Watkins DN, Jair KW, Schuebel KE, Markowitz SD, Rubinfeld B, Albert I, Porﬁri E, Munemitsu S, Polakis P. Chen WD et al. (2004). Epigenetic inactivation of SFRP genes (1997). Loss of beta-catenin regulation by the APC tumor allows constitutive WNT signaling in colorectal cancer. Nat suppressor protein correlates with loss of structure due to Genet 36: 417–422. common somatic mutations of the gene. Cancer Res 57: WebsterMT,RozyckaM,SaraE,DavisE,SmalleyM,YoungN 4624–4630. et al. (2000). Sequence variants of the axin gene in breast, Schneikert J, Behrens J. (2006). Truncated APC is required colon, and other cancers: an analysis of mutations that for cell proliferation and DNAreplication. Int J Cancer 119: interfere with GSK3 binding. Genes Chromosomes Cancer 74–79. 28: 443–453.

Oncogene