We have confirmed that SNP close to the TERC
locus are associated with leucocyte telomere length in humans, using methods that were different from those in the previous studies of Codd et al13
and Levy et al
We also found independently that the same TERC
SNP are associated with susceptibility to CRC. However, the SNP data show that, contrary to expectations, individuals with longer telomeres are predicted to be at increased risk of CRC.
The causal variation close to TERC and its mode of action remain to be identified, although it is most likely to lie within the region 170.90 Mb–171.05 Mb. We investigated the candidate SNP rs2293607 as a result of its genetic association with rs10936599 and rs12696304, together with its predicted effects on mRNA structure and/or expression. Our evidence suggests that rs2293607 might indeed have functional effects on TERC expression and thus on telomere length. If increased CRC risk does directly result from long telomeres, one plausible model is that telomere lengthening increases a normal cell's stem-like properties and thus its chances of acquiring tumour-causing mutations, consistent with the increased risk of both adenomas and carcinomas associated with rs10936599.
Despite the genetic evidence predicting that longer telomeres increase CRC risk, evidence from retrospective studies, such as our own, has shown that telomeres in peripheral blood cells are shorter in CRC cases than controls. However, associations between telomere length and CRC risk have not been shown in prospective studies, despite such studies probably including individuals with occult cancers. Although the prospective studies may have been too small to detect associations between telomere length and cancer reliably, the data as a whole suggest that the association between short telomeres and CRC in retrospective studies is in some way the result of disease or differential survival rather than the cause. Our samples taken closer to diagnosis (COGS study) had longer telomeres than those taken months or years after diagnosis (CORGI or VICTOR). These data are consistent with an association between longer telomeres and poorer survival, or with telomere attrition in the aftermath of treatment. However, our three sets of CRC cases differed in several ways other than sampling time, such as the therapeutic regimens used, inclusion criteria and ethnicity. We therefore also searched for evidence that telomere length depended on time between surgery and sampling or recurrence-free survival in VICTOR, because it was a clinical trial with excellent follow-up data. However, we found no association in either case. We also found no evidence that 5-fluorouracil-based chemotherapy or radiotherapy was associated with shorter telomeres in VICTOR. Given the long time between surgery and sampling in VICTOR and consequent sampling bias against early recurrence, we suggest that a requirement is an analysis of telomere length and survival based on a set of CRC cases sampled close to diagnosis, ideally before chemotherapy, and with excellent follow-up data.
The rival models linking telomere length and colorectal tumorigenesis are shown in . However, several further issues remain regarding telomeres, TERC and CRC risk.
Figure 3 Models of associations between telomere length and colorectal cancer (CRC) risk. Arrows indicate causality. (1) The current model is based on observation of shorter leucocytes telomeres in CRC cases than controls, assumes that this is mirrored in colorectal (more ...)
- Telomerase may repair DNA other than at telomeres,38–40 because de novo synthesis of telomere repeats may occur at double-stranded breaks. Despite our evidence to the contrary, it may be these extra-telomeric functions that are causally associated with increased CRC risk.
- Telomere length is most conveniently measured in peripheral blood: this can lead to non-genetic variation between samples (eg, if the composition of the leucocyte count varies) and it may not reflect the situation in the target tissue for cancer development. In particular, we believe it most unlikely that circulating DNA shed from cancer cells that have short telomeres contributed anything other than a tiny proportion of the total blood DNA in which we measured telomere length in CRC cases. We attempted to minimise sampling issues by collecting in a similar fashion from cases and controls when possible (notably in CORGI and COGS). However, this had to be balanced against the practicalities of empowering the study with enough samples (eg, we used NBS controls as a comparison with VICTOR as a result of the absence of unaffected controls from those clinical trial sample sets). Data were analysed throughout having stratified by sample set.
- There was overlap between the NBS samples analysed both in this study and that of Codd et al.13 However, even excluding these samples, the association between rs10936599 and telomere length remained (OR 0.673, 95% CI 0.511 to 0.886, p=5×10−3). In fact, the consistency between our work and that of Codd et al13 provides good evidence of the technical validity of our assays.32
- It is possible that the association between telomere length and cancer risk is non-montonic, allowing shorter telomeres to be associated with CRC overall, yet the minor TERC allele to be protective.
- It is a further conundrum that TERT tagSNP have been reported to be associated with an increased risks of lung, breast and other cancers, yet common genetic variants at TERT are not associated with telomere length in the studies undertaken to date. The reason underlying the organ or site-specific associations of cancers with TERC or TERT polymorphisms remains unclear.
Telomere length is a trait that is highly likely to be under stabilising selection. If telomeres are too short, they predispose to chromosomal aberrations and the diseases that result. However, they may also lead to earlier cell death, especially if DNA damage checkpoints are active. Very long-lived humans tend to have long telomeres, so what is the downside of having very long telomeres? Our work suggests that at least part of this downside is increased CRC risk, and there is evidence to support this contention in that chronically increased TERT expression promotes tumorigenesis in mouse models.41
Although some of the data are paradoxical, there is a clear link between rs10936599 and longer telomeres in leucocytes and increased risk of CRC. This finding may overcome some of the problems inherent in directly measuring telomere length in retrospectively collected cancer cases and controls. We propose that the previously reported associations between cancer risk and short telomeres in the general population wrongly emphasised the chromosomal instability aspects of cancer risk over the aspects relating to cell lifespan and stem-like properties.