Introduction

Simple tandem repeat arrays (STRs), such as mono-, di-, tri- and tetranucleotide repeat arrays, are known to mutate by intra-allelic mechanisms and the role of the mismatch repair (MMR) pathway in the instability of STRs has been investigated extensively. MMR is fundamental for the maintenance and stability of the genome and is required for the removal of base mismatches, single-nucleotide insertion/deletion loops and larger insertion/deletion loops that can arise from DNA damage or errors in DNA replication. MMR is also required for the correct resolution of recombination events (Harfe and Jinks-Robertson, 2000).

The MMR proteins are highly conserved from Escherichia coli to humans and are named in relation to their homology to the E. coli proteins MutS, MutL and MutH. In humans, there are several MutS homologues (hMSH2–6) and base mismatches and insertion/deletion loops are recognized principally by heterodimer proteins of hMSH2 with hMSH6 (MutS) or hMSH3p (MutS; Harfe and Jinks-Robertson, 2000; Jiricny and Marra, 2003). There is some overlap in the roles of the hMSH2–hMSH6 and hMSH2–hMSH3 heterodimers, although the hMSH2–hMSH3 heterodimer primarily binds to larger insertion /deletion loops (Kolodner and Marsischky, 1999). Following the initial binding of a MutS protein to a base mismatch or insertion/deletion loop, a MutL protein binds to the complex. The principal MutL homologue in humans, hMLH1, forms heterodimers with hPMS2 (MutL), hPMS1 (MutL) and hMLH3, and these heterodimers are involved in the subsequent signalling to downstream parts of the repair pathway. The interaction of MMR proteins with mismatch-containing recombination intermediates can be resolved, either by simple mismatch correction, resulting in gene conversion, or by complete abortion of the recombination event (Harfe and Jinks-Robertson, 2000). Consequently, MMR defects may also be responsible for an increase in DNA rearrangements, resulting from recombination between nonidentical or homeologous DNA strands.

Through its role in repair of errors that arise during DNA replication, the MMR pathway ensures the high fidelity of DNA replication. The crucial role of MMR in humans is highlighted when functional MMR proteins are absent, by gene mutation, by gene loss or by epigenetic silencing, resulting in the accumulation of DNA mutations (Liu et al., 1996; Jiricny and Marra, 2003). DNA errors can occur throughout the genome, but are prevalent at STR loci, which are particularly susceptible to mutations arising from replication slippage events. Failure to correct such slippage events, as a result of defective MMR, causes dramatic increases in instability of these loci (so-called microsatellite instability or MSI), resulting in a phenotype known as the microsatellite mutator phenotype (MMP) (Yamamoto et al., 1999). Of course, the MMP also generates mutations in STRs within cancer genes, for example the tumour suppressor gene transforming growth factor receptor II (TGFRII) (Jiricny and Marra, 2003). Indeed, frameshift mutations in the A₁₀ mononucleotide tract from nucleotides 709 to 718 of TGFRII have been reported to occur in approximately 90% of colorectal cancers displaying MSI (Parsons et al., 1995).

The colon contains rapidly dividing cells and therefore mutations at STR loci, which occur through error-prone intra-allelic mechanisms, accumulate over a lifetime. Detection of low-frequency, random STR mutations could be achieved by small-pool PCR across the STR locus, but this is not possible at telomeres. However, random STR or telomere mutations that have accumulated in a rapidly dividing tissue may, by chance, be detected in a clonally derived tumour sample. The study of hereditary cancer has identified a number of key cancer suppression pathways implicated in the occurrence of specific types of cancer. One such example is hereditary nonpolyposis colorectal cancer (HNPCC), which accounts for 4–13% of all colorectal cancers and is associated with inherited defects in proteins involved in the MMR pathway (Aaltonen et al., 1993; Aaltonen et al., 1994; Liu et al., 1996). Moreover, approximately 15% of sporadic colon cancers are also associated with MMR defects (Thibodeau et al., 1993).

The proximal ends of human telomeres consist of interspersion patterns of telomere (TTAGGG) and variant repeat types (for example, TGAGGG and TCAGGG), which extend approximately 2 kb into the telomere repeat array (Allshire et al., 1989; Baird et al., 1995). Beyond this, the telomere consists of a homogeneous array of TTAGGG repeats and the total telomere length can range from 2 to 20 kb (Harley et al., 1990; Hastie et al., 1990). Telomere maintenance is essential for cell immortality and in tumour cells it is achieved by de novo synthesis of telomere repeats at the terminus by the enzyme telomerase (Kim et al., 1994) or by a recombination-based mechanism known as alternative lengthening of telomeres (ALT) (Lundblad and Blackburn, 1993; Bryan et al., 1995). Comparison of the proximal regions of single telomeres, by telomere variant repeat mapping by PCR (TVR–PCR), in populations of unrelated individuals has shown that there must be a high frequency of germ line mutations probably dominated by simple intra-allelic events (Baird et al., 1995; Coleman et al., 1999; Baird et al., 2000). The frequency of germline telomere mutations has been measured directly in families as 0.006 (0.6%) per kb per gamete (95% confidence limits: 3.25 10^-4 to 0.036; HAP and NJR, unpublished). Mechanisms that can give rise to intra-allelic mutations include unequal sister-chromatid exchange, slippage during replication and errors of MMR following DNA damage (Sinden, 2001). They are dependent on DNA replication and therefore more likely to occur in rapidly dividing cells or tissues. In contrast, complex mutations have been identified in cells that use the ALT pathway for telomere maintenance (Varley et al., 2002). These complex mutations involve the acquisition of novel telomere repeats and likely arise from the recombination-based process responsible for telomere maintenance in ALT⁺ cells.

Telomere length maintenance is essential for the progression to malignancy, and yet telomeres are exposed to all the mutation processes that underlie variation at STR loci. Indeed the accumulation of mutations at STR loci, as a result of failure by MMR, predicts that there will be instability in telomeric sequences. To determine whether telomere mutations occur in somatic cells in vivo and to examine the role of MMR on telomeres during tumorigenesis, we have compared the proximal ends of human telomeres in normal/tumour samples in relation to the presence of instability at the polyA tract of the TGFRII gene and MMR defects. We observed considerable telomere instability and this was greatest in tumours with compromised MMR.

Results

To determine whether human telomere repeat arrays mutate in vivo, two panels of colon cancers have been screened for mutations using TVR–PCR (Baird et al., 1995). The first panel comprised pairs of blood DNA and colorectal adenocarcinoma DNA samples from patients with sporadic colon cancer. The DNA samples had been characterized previously for instability at four microsatellites and for instability at the minisatellite MS1 (D1S7; Lothe et al., 1993; Hoff-Olsen et al., 1995). The selected 52 pairs of samples in panel 1 included 27 that showed instability at D1S7 and 25 that did not. In addition, a subset of the tumours in this panel had been screened for somatic mutations in the MMR gene, hMSH2 (Borresen et al., 1995). The second panel comprised 37 randomly selected pairs of normal colon and colon tumour tissue samples from patients (aged 37–88 years) with suspected sporadic colon cancer. Pathological examination of the tissue samples revealed that two were from benign adenomas, three from well-differentiated carcinomas, 24 from moderately differentiated carcinomas and eight from poorly differentiated carcinomas.

Genotype analysis of the telomere-adjacent sequences

Analysis of single alleles at the Xp/Yp and 12q telomeres can be conducted by PCR amplification from single-nucleotide polymorphisms (SNPs) in the DNA sequence immediately adjacent to the telomere. Therefore, each DNA sample was genotyped at SNPs in the Xp/Yp telomere-adjacent DNA that define the common haplotypes in this region (Baird et al., 1995). In panels 1 and 2, 30 and 14 samples (respectively) were heterozygous for T/A at the SNP 30 bases proximal to the telomere, allowing mutation detection in 60 and 28 Xp/Yp single alleles, respectively (Tables 1 and 2). The Xp/Yp telomeres in the tumour DNA samples that are homozygous at the -30 base were also screened for mutations through the analysis of diploid telomere maps. Thus, 104 and 74 Xp/Yp telomeres were screened for mutations in panels 1 and 2, respectively.

Table 1 - Summary of telomere mutation detection in colon tumour DNA samples in panel 1.

Full table

Table 2 - Summary of telomere mutation detection in colon tumour DNA samples in panel 2.

Full table

Similar haplotype analysis in the 12q telomere-adjacent DNA (Baird et al., 2000) showed that 32 and 22 single alleles could be analysed from the SNP 197 bases from the start of the telomere in panels 1 and 2, respectively (Tables 1 and 2). Some alleles at the 12q telomere do not contain the same telomere-adjacent sequence and they do not amplify with the available PCR primers (null alleles) (Baird et al., 2000). Consequently, the remaining samples in the tumour panels are either homozygous or hemizygous at the -197 SNP. As the 12q null alleles can only be identified with certainty by pedigree analysis, the number of such alleles in each panel was estimated using the known frequency of null alleles in Caucasians. Thus an estimated 80 and 64 12q telomeres were screened for mutations in panels 1 and 2, respectively.

Telomere mutations in colon cancer samples

The screen for telomere mutations was conducted using TVR–PCR. In brief, TVR–PCR amplifies sequences between a primer located immediately adjacent to the Xp/Yp or 12q telomere and one of three primers that anneal to the TTAGGG, or TGAGGG or TCAGGG repeat types within the telomere repeat array. Size resolution of the PCR products from the three separate reactions (in adjacent tracts) reveals the distribution of the TTAGGG and variant repeat types along the array. If DNA samples from individuals are homozygous at the Xp/Yp -30 SNP or the 12q -197 SNP, then telomere maps from both alleles are superimposed, but if the DNA sample is heterozygous at the telomere-adjacent SNPs, telomere maps of single alleles are generated.

Telomere mutations were detected in both panels (Tables 1 and 2), by the comparison of the interspersion of TTAGGG and variant telomere repeats (telomere maps) in normal and colon cancer DNA sample pairs. The types of mutations include loss or gain of one or a few repeats from a block of like repeat types and the change of one repeat type to another (Figure 1). It was not possible to define all the mutations fully for a variety of reasons. Some tumour samples contained a mutation near the start of the telomere repeat array that could be defined but, in addition, other, reproducible changes were identified further into the array where individual repeats were not fully resolved (eg 1268, Table 1). In others, definition of the mutation was difficult because of the presence of a telomere map from the normal (progenitor) allele (Figure 2). Also the presence of the two superimposed progenitor telomere maps in tumours that are homozygous at SNPs in the telomere-adjacent DNA (Figure 2) made mutation detection difficult in some cases. Where possible, mutations were confirmed by sequence analysis following amplification of DNA extracted from bands in TVR gels (Tables 1 and 2). This was limited to mutations that occurred near the start of the telomere repeat array and where the purified amplicons were free of products from the progenitor or the homologous allele. Some of the mutations involved more than one repeat type. If the different repeat types were adjacent in the telomere (eg 955, panel 1, Table 1), it is likely that the changes arose in a single mutation event. If, however, the mutant telomere had changes in blocks of repeats that were not adjacent (eg 1268, Table 1), then more than one mutation event may have been involved. However, in order to compare mutation frequencies each mutant telomere was scored as having arisen from a single mutation event.

Figure 1.

Telomere maps showing the types of telomere mutations observed in panels 1 and 2. Tumour 965 shows the loss of one repeat from a block of C-type repeats (arrowed) in an Xp/Yp haplotype B-associated telomere. Tumour 7 shows the map of a mutated haplotype B-associated telomere, which includes a gain of eight C-type repeats (vertical bar), superimposed on the telomere map from the progenitor allele. Both these mutations are accompanied by shifts in the interspersion pattern of T-type and variant repeats further into the telomere array. Tumour 2 shows a change of a null (N-type) to a T-type repeat (arrowed), identified by sequence analysis as a TAGGG to TTAGGG mutation in the Xp/Yp haplotype A-associated telomere. This mutation does not alter the telomere map distally. T, T-type repeat (TTAGGG); G, G-type repeat (TGAGGG); C, C-type repeat (TCAGGG); n, normal; t, tumour

Full figure and legend (161K)

Figure 2.

Mixed telomere maps may affect mutation detection and definition. Detection of mutations in tumour samples with two superimposed telomere maps can be more difficult. The 1314 tumour DNA sample contains a mixture of the progenitor and mutant telomere maps of a haplotype B-associated telomere. Comparison with the progenitor telomere map, seen in the 1314 normal tissue sample, shows that a mutation affects the telomere from repeat 9 onwards (star), but the mutant telomere map is obscured by the progenitor map (dashed arrow). Sequence analysis confirmed the loss of one repeat from the mutant telomere but other changes (such as change of one repeat type to another) may remain undetected. Indeed, other changes were detected further into the telomere repeat array of this allele (not shown). The 1363 samples were homozygous at -30 Xp/Yp SNP and therefore TVR–PCR amplified two haplotype B-associated alleles. The superimposed allelic telomere maps can be seen as a diploid pattern. For example, the arrows adjacent to the 1363 telomere maps show the presence of T- and C-type repeats at the same position, making mutation detection more difficult. T, T-type repeat (TTAGGG); G, G-type repeat (TGAGGG); C, C-type repeat (TCAGGG); n, normal; t, tumour

Full figure and legend (318K)

In the nonrandom collection of sporadic tumour samples, which comprise panel 1, 19/52 (36.5%) revealed a mutation in at least one telomere, whereas 6/37 (16.2%) of the random collection of sporadic tumour samples in panel 2 contained at least one telomere mutation. Comparison of the mutation frequency per allele from tumours homozygous or heterozygous at telomere-adjacent SNPs suggests that detection of mutant telomeres in diploid telomere maps was not significantly reduced (Table 3). Thus the average frequency of telomere mutations per allele was 13% in panel 1 (15.4% at Xp/Yp; 10% at 12q Table 3) and 5.8% in panel 2 (5.4% at Xp/Yp; 6.3% at 12q, Table 3).

Table 3 - Summary of telomere mutation frequencies per allele.

Full table

Telomere mutations and tumours with instability at the polyA tract of the TGFRII gene

It has been shown that 90% of colon cancers with MSI have mutations in the polyA tract of the TGFRII gene (Parsons et al., 1995). In this study, tumours with microsatellite instability, based on the analysis of the TGFRII polyA tract, contained significantly more telomere mutations than tumours without (Table 3). The telomere mutation frequency was 18.6% per allele in tumours with TGFRII polyA instability compared to the mutation frequency of 6.2% per allele in tumours without TGFRII polyA instability (²=11.6, d.f.=1, P=0.001). Panel 1 includes five tumours (Table 1), which contain characterized mutations within the hMSH2 gene (Borresen et al., 1995), and they all contain mutations in the Xp/Yp telomere and two also have mutations in the 12q telomere. If each mutant telomere represents only one mutation event, then the mutation frequency in colon tumours with a defect in MMR is 35% per allele. The significantly higher frequency of telomere mutations in tumours with instability at the TGFRII polyA tract suggests a role for MMR in the maintenance of telomere repeat arrays, and this is strongly supported by the very high telomere mutation frequency in tumours with a mutation in hMSH2.

Telomere mutations: losses versus gains

The mutant telomeres can be grouped according to whether the mutations, in the first 100–120 repeats, are likely to cause an overall gain or loss of repeats in the array. Among the 16 Xp/Yp mutant telomeres observed in panel 1, 10 cause a loss of repeats, three a gain and the effect of the three remaining mutations is uncertain. Among the eight mutant telomeres at the 12q telomere in panel 1, five cause a loss of repeats, one a gain and the effect on repeat number was not changed or could not be ascertained for two other mutants. If the mutation profile, seen in the proximal 100–120 repeats of telomeres, is present in the homogeneous array of TTAGGG repeats, the overall effect would tend to decrease telomere length.

Discussion

Tandem repeat instability in colon cancer has been studied most extensively at microsatellites (including mono-, di- and tri-nucleotide repeats), but it has also been detected at some minisatellites that have longer repeat units, for example, the D1S7 minisatellite (MS1) with a 9 bp repeat unit (Hoff-Olsen et al., 1995) and at D7S21 (Coleman et al., 2001). Telomeres, consisting of long arrays of a 6 bp repeat unit, have the potential for MSI but in contrast to many microsatellite loci, telomeres have an essential chromosomal function that is required for cell immortality. By comparison of the first 100–120 repeats (600–720 bp) of the Xp/Yp and 12q telomeres in normal and tumour DNA samples, we have demonstrated for the first time that human telomeres mutate in colon cancers. The telomere mutation frequencies reported here are likely to be underestimates for a number of reasons, but most notably because alleles were screened for mutations across a maximum of 120 repeats, which is only a small portion of the whole telomere array. At triplet repeat arrays, for example, the myotonic dystrophy ((CAG)_n), fragile X syndrome ((CGG)_n) and Friedreich ataxia ((GAA)_n) loci, the length of the repeat array affects the mutation dynamics as expanded alleles are highly unstable. Sequence homogeneity within the repeat array also affects instability, for example, instability at the CGG repeat array of the FRAXA locus is increased as a consequence of sequence homogeneity, in men with alleles in normal length range (Crawford et al., 2000). If these general features of microsatellite instability hold true for telomere repeat arrays, then mutations in the homogeneous portion of the telomere (TTAGGG)n will be more numerous. This would further increase the telomere mutation frequency per kb.

Mixed progenitor and mutant telomere maps were only seen in the tumour-derived samples (as shown in Figures 1 and 2), indicating that the progenitor telomere is seen in the normal tissue sample. However comparison of telomere maps in a normal tissue sample and a tumour sample does not indicate when the telomere mutation arose with respect to tumour development. If telomere mutations accumulate over a lifetime, as a result of errors in the rapidly dividing cells of the colon, they will be present at a low frequency in the normal tissue (below the level of detection), but any individual mutant would only be detected if it were fortuitously present in the cell that develops into a tumour. Indeed, one of the two benign adenomas present in panel 2 (sample 7) contained a mutant Xp/Yp telomere that comprised a gain of eight TCAGGG repeats. However, the higher frequency of telomere mutations in tumours that show instability at the TGFRII polyA tract and in tumours with a mutation in hMSH2, indicates that the majority of telomere mutations detected in this study arise during tumour formation.

The instability of microsatellites has been studied extensively using a variety of repeat sequences, cell types and in different species. Analysis of germline mutations at tetranucleotide repeat arrays distributed all over the human genome has shown that more than 85% of the mutations result in the gain or loss of a single repeat unit. The rate of mutation expansions is constant for all alleles, irrespective of allele length, while the contraction mutation rate increases exponentially as the allele length increases. Interestingly, the overall rate of gain versus loss mutations is constant at these loci, thus explaining why they do not become excessively long (Xu et al., 2000), and it seems that dinucleotide repeat arrays have similar mutation dynamics in humans (Ellegren, 2000). The dynamics of disease-associated triplet repeat array instability in somatic and germline tissues have also been studied extensively. The relationship between allele length, homogeneity of repeats within the array, the GC content, cell-type factors and any bias towards gain or loss of repeats by mutation is complex and varies between loci. For example, expanded alleles of the myotonic dystrophy (CTG)_n triplet repeat array show a bias towards expansion mutations (Monckton et al., 1995; Martorell et al., 1998; Fortune et al., 2000), while in contrast expanded alleles of the Friedreich ataxia (GAA)_n repeat array show a bias towards contractions (Sharma et al., 2002). In addition, the mutation dynamics of simple tandem repeat arrays can be altered by a defect in MMR. For example, the bias of expansion mutations at the CTG repeat array in the myotonic dystrophy gene shifted towards contractions in mice that lack the Msh2 gene (Savouret et al., 2003), whereas Msh6 deficiency resulted in an increase in the frequency of somatic expansions (van den Broek et al., 2002). It is therefore difficult to predict what the effect of telomere instability will be on telomere length.

Examination of the mutant telomeres detected in this study suggests that the majority of telomere mutations will result in a loss of one to three repeats from the array, although telomere mutations may have a greater effect in the tumours that are defective in MMR (panel 1). It is interesting, therefore, that colon cancers with microsatellite instability seem to have shorter telomeres than tumours without microsatellite instability, irrespective of the level of telomerase expression (Takagi et al., 2000). Furthermore, telomerase activity is expressed in a higher proportion of normal colon mucosa biopsies from patients with HNPCC compared with similar biopsies from patients without HNPCC (Cheng et al., 1998). Together these observations indicate that there is a greater requirement for telomere length maintenance (extension) in normal mucosa from HNPCC patients, which could be a consequence of increased telomere instability. Accordingly, colon cancers in HNPCC patients may arise from cells that already express telomerase or there may be a requirement to activate telomerase early and hence stabilize telomere length and possibly the karyotype. Interestingly, karyotypes of MMR-defective colon cancers are more stable, with fewer changes in chromosome structure or number, than karyotypes in colon cancers that are MMR proficient (Eshleman et al., 1998). However, while the data presented here suggest a bias towards loss of repeats as a result of telomere instability, the mutation dynamics within the homogeneous array of TTAGGG repeats may be different in the presence or absence of effective MMR and requires investigation before any firm conclusion can be drawn.

Telomerase-deficient yeast can survive cell death by the activation of a pathway(s) that allows telomere length maintenance by recombination-based mechanisms (Lundblad and Blackburn, 1993; Lundblad, 2002). One subset of the yeast survivors (type II survivors) resemble telomerase-negative immortal human cells that show ALT and maintain telomere length by a recombination-based mechanism (Bryan et al., 1995; Dunham et al., 2000; Varley et al., 2002). In telomerase-deficient yeast, the frequency at which survivors emerge from crisis is enhanced in the presence of a defect in MMR. It has been proposed that MMR may act as a barrier to homeologous telomeric recombination and hence a defect in MMR may promote survival in the absence of telomerase (Rizki and Lundblad, 2001). Further investigation of human telomere mutation dynamics in the presence or absence of MMR will allow us to elucidate the role this pathway plays in cell immortalization in the presence or absence of telomerase.

Materials and methods

Genomic DNA from tissue samples

High molecular weight DNA samples from pairs of blood and colon cancer tissues sample that comprise panel 1 were supplied by PH-O. Following informed consent from patients (Leicestershire Research Ethics Committee, Ref 6338), paired samples of normal and tumour tissue that comprise panel 2 were taken by a pathologist immediately after removal of the colon. The samples were snap frozen in liquid nitrogen and stored prior to DNA extraction from 0.5 g of tissue, using standard methods.

Polymorphism assays and TVR–PCR

Polymorphism assays in the telomere-adjacent DNA at Xp/Yp and 12q were used to deduce the haplotypes and hence the genotypes at the Xp/Yp -30 and the 12q -197 SNPs, as described previously (Baird et al., 1995; Baird et al., 2000). Telomere analysis was conducted on heterozygous or homozygous DNA samples from the Xp/Yp -30 base, using the allele-specific primers TS-30A or TS-30T (Baird et al., 1995) and from the 12q -197 base, using the allele-specific primers 12q -197A or12q -197G (Baird et al., 2000). The allele-specific primers were used in TVR–PCR with the primers TAG-TelW, TAG-TelX and TAG-TelY that detect the TTAGGG, TGAGGG or TCAGGG repeats types as described earlier (Baird et al., 1995). The PCR buffer is described elsewhere (Jeffreys et al., 1991).

Sequence confirmation of mutations detected by TVR–PCR

Bands from the TVR–PCR polyacrylamide gels were excised (Baird and Royle, 1997) and reamplified with the telomere-adjacent primer and the TAG primer (Jeffreys et al., 1991). Gel-purified PCR products were sequenced using the Big Dye Terminator Cycle Sequencing Ready Reaction Kit (PE Biosystems).

Analysis of the TGFRII polyA tract

The TGFRII polyA tract was PCR amplified using primers TGF2A 5'-AAGCTCCCCTACCATGACT and TGF2B 5'-TGCACTCATCAGAGCTACAG (Parsons et al., 1995). Genomic DNA (100 ng) was amplified in a 10 l reaction in the presence 1 M of each primer, 0.1 U/l Vent polymerase (New England Biolabs) and 1 PCR buffer (supplied with the Vent polymerase), 2 mM dNTPs, 1.5 mM MgCl₂ and 0.2 Ci [³²P]dCTP (Saiki et al., 1988). The PCRs were cycled 30 times at 96°C for 20 s, 55°C for 30 s and 70°C for 1 min in a PTC200 DNA Engine™ (MJ Research, USA). The products were resolved by acrylamide gel electrophoresis and detected by autoradiography.

References

1. Aaltonen LA, Peltomaki P, Leach FS, Sistonen P, Pylkkanen L, Mecklin JP, Jarvinen H, Powell SM, Jen J, Hamilton SR, Petersen GM, Kinzler KW, Vogelstein B and Delachapelle A. (1993). Science, 260, 812–816. | Article | PubMed | ISI | ChemPort |
2. Aaltonen LA, Peltomaki P, Mecklin JP, Jarvinen H, Jass JR, Green JS, Lynch HT, Watson P, Tallqvist G and Juhola M. (1994). Cancer Res., 54, 1645–1648. | PubMed | ISI | ChemPort |
3. Allshire RC, Dempster M and Hastie ND. (1989). Nucleic Acids Res., 17, 4611–4627. | PubMed | ISI | ChemPort |
4. Baird DM, Coleman J, Rosser ZH and Royle NJ. (2000). Am. J. Hum. Genet., 66, 235–250. | Article | PubMed | ISI | ChemPort |
5. Baird DM, Jeffreys AJ and Royle NJ. (1995). EMBO J., 14, 5433–5443. | PubMed | ISI | ChemPort |
6. Baird DM and Royle NJ. (1997). Hum. Mol. Genet., 6, 2291–2299. | Article | PubMed | ISI | ChemPort |
7. Borresen AL, Lothe RA, Meling GI, Lystad S, Morrison P, Lipford J, Kane MF, Rognum TO and Kolodner RD. (1995). Hum. Mol. Genet., 4, 2065–2072. | PubMed | ISI | ChemPort |
8. Bryan TM, Englezou A, Gupta J, Bacchetti S and Reddel RR. (1995). EMBO J., 14, 4240–4248. | PubMed | ISI | ChemPort |
9. Cheng AJ, Tang RP, Wang JY, See LC and Wang TCV. (1998). J. Natl. Cancer Inst., 90, 316–321. | Article | PubMed |
10. Coleman J, Baird DM and Royle NJ. (1999). Hum. Mol. Genet., 8, 1637–1646. | Article | PubMed | ISI | ChemPort |
11. Coleman MG, Gough AC, Bunyan DJ, Braham D, Eccles DM and Primrose JN. (2001). Br. J. Cancer, 85, 1486–1491. | Article |
12. Crawford DC, Wilson B and Sherman SL. (2000). Hum. Mol. Genet., 9, 2909–2918. | Article | PubMed |
13. Dunham MA, Neumann AA, Fasching CL and Reddel RR. (2000). Nat. Genet., 26, 447–450. | Article | PubMed | ISI | ChemPort |
14. Ellegren H. (2000). Nat. Genet., 24, 400–402. | Article | PubMed | ISI | ChemPort |
15. Eshleman JR, Casey G, Kochera ME, Sedwick WD, Swinler SE, Veigl ML, Willson JKV, Schwartz S and Markowitz SD. (1998). Oncogene, 17, 719–725. | Article | PubMed | ISI | ChemPort |
16. Fortune MT, Vassilopoulos C, Coolbaugh MI, Siciliano MJ and Monckton DG. (2000). Hum. Mol. Genet., 9, 439–445. | Article | PubMed | ISI | ChemPort |
17. Harfe BD and Jinks-Robertson S. (2000). Annu. Rev. Genet., 34, 359–399. | Article | PubMed | ISI | ChemPort |
18. Harley CB, Futcher AB and Greider CW. (1990). Nature (London), 345, 458–460. | Article | PubMed |
19. Hastie ND, Dempster M, Dunlop MG, Thompson AM, Green DK and Allshire RC. (1990). Nature, 346, 866–868. | Article | PubMed | ISI | ChemPort |
20. Hoff-Olsen P, Meling GI and Olaisen B. (1995). Hum. Mutat., 5, 329–332. | PubMed |
21. Jeffreys AJ, MacLeod A, Tamaki K, Neil DL and Monckton DG. (1991). Nature, 354, 204–209. | Article | PubMed | ISI | ChemPort |
22. Jiricny J and Marra G. (2003). Curr. Opin. Genet. Dev., 13, 61–69. | Article | PubMed |
23. Kim NW, Piatyszek MA, Prowse KR, Harley CB, West MD, Ho P, coviello GM, Wright WE, Weinrich SL and Shay JW. (1994). Science, 266, 2011–2015. | Article | PubMed | ISI | ChemPort |
24. Kolodner RD and Marsischky GT. (1999). Curr. Opin. Genet. Dev., 9, 89–96. | Article | PubMed | ISI | ChemPort |
25. Liu B, Parsons R, Papadopoulos N, Nicolaides NC, Lynch HT, Watson P, Jass JR, Dunlop M, Wyllie A, Peltomaki P, de la Chapelle A, Hamilton SR, Vogelstein B and Kinzler KW. (1996). Nat. Med., 2, 169–174. | Article | PubMed | ISI | ChemPort |
26. Lothe RA, Peltomaki P, Meling GI, Aaltonen LA, Nystrom-Lahti M, Pylkkanen L, Heimdal K, Andersen TI, Moller P and Rognum TO. (1993). Cancer Res., 53, 5849–5852. | PubMed | ISI | ChemPort |
27. Lundblad V. (2002). Oncogene, 21, 522–531. | Article | PubMed | ISI | ChemPort |
28. Lundblad V and Blackburn EH. (1993). Cell, 73, 347–360. | Article | PubMed | ISI | ChemPort |
29. Martorell L, Monckton DG, Gamez J, Johnson KJ, Gich I, de Munain AL and Baiget M. (1998). Hum. Mol. Genet., 7, 307–312. | Article | PubMed | ISI | ChemPort |
30. Monckton DG, Wong LJ, Ashizawa T and Caskey CT. (1995). Hum. Mol. Genet., 4, 1–8. | PubMed | ISI | ChemPort |
31. Parsons R, Myeroff LL, Liu B, Willson JK, Markowitz SD, Kinzler KW and Vogelstein B. (1995). Cancer Res., 55, 5548–5550. | PubMed | ISI | ChemPort |
32. Rizki A and Lundblad V. (2001). Nature, 411, 713–716. | Article | PubMed | ISI | ChemPort |
33. Saiki RK, Gelfand DH, Stoffel S, Scharf SJ, Higuchi R, Horn GT, Mullis KB and Erlich HA. (1988). Science, 239, 487–491. | Article | PubMed | ISI | ChemPort |
34. Savouret C, Brisson E, Essers J, Kanaar R, Pastink A, te Riele H, Junien C and Gourdon G. (2003). EMBO J., 22, 2264–2273. | Article | PubMed | ISI | ChemPort |
35. Sharma R, Bhatti S, Gomez M, Clark RM, Murray C, Ashizawa T and Bidichandani SI. (2002). Hum. Mol. Genet., 11, 2175–2187. | Article | PubMed |
36. Sinden RR. (2001). Nature, 411, 757–758. | Article | PubMed | ChemPort |
37. Takagi S, Kinouchi Y, Hiwatashi N, Nagashima F, Chida M, Takahashi S, Negoro K, Shimosegawa T and Toyota T. (2000). Dis. Colon Rectum, 43, S12–S17. | PubMed |
38. Thibodeau SN, Bren G and Schaid D. (1993). Science, 260, 816–819. | Article | PubMed | ISI | ChemPort |
39. van den Broek WJ, Nelen MR, Wansink DG, Coerwinkel MM, te Riele H, Groenen PJ and Wieringa B. (2002). Hum. Mol. Genet., 11, 191–198. | Article | PubMed | ISI | ChemPort |
40. Varley H, Pickett HA, Foxon JL, Reddel RR and Royle NJ. (2002). Nat. Genet., 30, 301–305. | Article | PubMed | ISI |
41. Xu X, Peng M and Fang Z. (2000). Nat. Genet., 24, 396–399. | Article | PubMed | ISI | ChemPort |
42. Yamamoto H, Perez-Piteira J, Yoshida T, Terada M, Itoh F, Imai K and Perucho M. (1999). Gastroenterology, 116, 1348–1357. | Article | PubMed | ISI | ChemPort |

Acknowledgements

We thank Jenny Foxon for technical support, and Celia May, Jenny Jeypalan and Mark Hills for helpful discussion. We gratefully acknowledge the support of the MRC (UK) that funded the work.