The research interests of the Genetic Epidemiology Laboratory involve the contribution of germline variation in cancer-predisposing genes to population cancer incidence and the role of gene-environment interactions. These interests involve both the development of statistical methodology and the application of these techniques to bowel cancer, melanoma and testis cancer.
Bowel Cancer (Head of Group: Tim Bishop)
In terms of contribution of genes to cancer incidence, we have been continuing our investigation of the contribution of mismatch repair genes to the population incidence of colorectal cancer focusing on a case-population control-based colorectal cancer study and also a hospital-based colorectal cancer case series. Immunohistochemical analysis of colorectal tumours suggests that loss of expression (LOE) of the two predisposing genes (hMSH2 and hMLH1) differs. hMSH2 LOE is only found in 2% of tumours which are predominantly of early onset, suggestive of germline mutations as being the likely cause while hMLH1 LOE increases with the age of onset, is more common in females than males and is more common in tumours derived from the right side of the bowel. Overall, these results are consistent with epigenetic phenomena which are thought to involve the silencing of hMLH1 by methylation.
We have analysed two case-control studies of colorectal neoplasia focussing on gene-environment interaction. The first is a case-control study of adenomas in cases and distal adenoma free controls as identified in the MRC-ICRF FLEXISCOPE trial (co-ordinated by Dr Wendy Atkin, St Mark's). In total, 370 cases and a similar number of controls were interviewed about their diet and provided a DNA sample. The second study is a case-control study of colorectal cancer (in collaboration with Professor Roland Wolf (ICRF, Dundee), Professor David Forman (NYCRIS, Leeds), Professor Colin Garner (University of York) and Professor Alan Boobis and Dr Nigel Gooderham (University of London), 500 cases and GP matched controls were recruited in Dundee, Leeds and York. There is evidence for the previously reported risk factors of red meat (in the carcinoma study) and smoking (in both the adenoma and carcinoma study). We have explored two genes involved in N-acetylation (NAT1 and NAT2) and do not find that they are associated with an increased risk of having an adenoma or carcinoma. Furthermore, there is no evidence of an interaction between diet and genotype at these loci (Barrett, Smith et al., Carcinogenesis 2003; 24: 275). Investigations of other p450 genes have also been conducted. (Sachse, Smith et al., Carcinogenesis 2002; 23: 1839).
Melanoma (Head of Group: Julia Newton Bishop)
Research continues directed towards understanding the genes predisposing to melanoma, and their interaction with the environment. In families at high risk of melanoma, mutations in CDKN2A (p16) underlies susceptibility. We have identified a mutation deep in an intron that is predicted to affect splicing of p16 (Harland, Mistry et al., Hum Mol Genet 2001; 10: 2679). This mutation has been found in 6 UK families, making it the most common CDKN2A mutation found in the UK to date and has since been found in Australia and Canada.
Figure 1 shows the estimated penetrance (cumulative risk by age) for melanoma by geographical residence of the mutation carriers. These estimates are significantly different (p < 0.03).
The Leeds group works with the Melanoma Genetics Consortium, which Dr Julia Newton Bishop chairs. We have published with this group an analysis of families with CDKN2A germline mutations and at least two cases of melanoma estimating the penetrance of such mutations by the location of residence of the mutation carriers (Bishop, Demenais et al., J Natl Cancer Inst 2002; 94: 894). This work has involved data from the Melanoma Genetics Consortium and analysis involving Dr Florence Demenais (Paris) and Dr Alisa Goldstein (Washington DC). Interestingly, the estimated penetrance differs between Europe, USA and Australia with the lifetime penetrance being highest in Australia. Broadly, the estimated penetrances reflect the baseline rates across these populations and suggest that factors which influence risk of melanoma in the general population also affect risk of melanoma in those with germline mutations.
Previously, we reported a study of the genetic heritability of melanocytic naevi. In this study, adolescent UK twins pairs have been examined for the number and size of their naevi (Wachsmuth, Gaut ,et al., J Invest Dermatol 2001; 117: 348). Naevi are of considerable interest because of their association with susceptibility to melanoma and so understanding the causes of naevi would give further information on the causes of melanoma. This study was comparative with an Australian study conducted on a population of largely UK ancestry (Zhu, Duffy et al., Am J Hum Genet 1999; 65: 483). Interestingly, the UK and Australian studies showed similar results for heritability and population variation in the number of naevi but twins in Queensland had thirty more naevi, on average. This study suggests therefore that while UV exposure affects the mean number of naevi, variation between individuals is largely determined by genetic factors. We have subsequently shown that the genes which influence naevi in sun exposed anatomical sites are likely to be the same as those that influence the number of naevi in sun protected sites. However, the precise genes involved in determining such variation are unknown to date. We have found that there is no evidence of linkage to CDKN2A unlike the evidence from the Australian study (Barrett et al., to appear Br J Cancer).
The group is extending its area of interest towards understanding the control of relapse from melanoma, and the effect of the environment. In December 2000, we started recruiting a large cohort study of melanoma patients, which will accrue for the next 5 years. A nested case control study of late relapsing melanoma is running concurrently which will accrue until 2004.
Testis Cancer (Head of Group: Tim Bishop)
Our collaborative research work into identifying predisposing genes involving Dr Mike Stratton and Dr Liz Rapley (ICR, Sutton) and Dr Doug Easton (CRUK, Cambridge) is focusing on the X chromosome (Rapley, Crockford et al., Nature Genet 2000; 24: 197). In this study, we showed significant evidence for a gene in the region of the fragile X gene. Evidence was strongest for regions containing a gene that had a predominant effect on families with either bilateral cancer or families in which testis cancer cases had previously been diagnosed with an undescended testis. We have been attempting to identify further families to provide better delineation of the critical region on the X chromosome, to determine the complete sequence across the region and to identify and screen candidate genes in this region. We now know that the region is of the order of 4 Mb following the completion of the sequencing by the MRC Sanger Centre in Cambridge and is relatively gene poor although annotation is still in progress. In fact, only three genes have been identified in the region, and to date, no mutations have been identified. We have examined the fragile X gene (FRAXA) in the most detail and cannot show any coding mutations. The critical region has not been refined to date as too few informative families have been collected but the overall family set for further analysis is approximately a third greater than for previous analyses.
Statistics Group (Head of Group: Jenny Barrett)
The statistics group in the Genetic Epidemiology Division is involved in three main activities:
- Statistical input to the epidemiological and clinical studies carried out
in the Division, including the design of new studies and data analysis. This
includes providing training and support for clinical researchers;
- Methodological research into improved methods of design and analysis motivated
by the research interests of the Division. Our methodological interests include
gene-environment interaction, segregation analysis and the analysis of haplotype
data;
- Development and management of the Division's databases.
Since a major focus of our bowel cancer and melanoma studies is investigating the joint action of genes and environmental risk factors, we have been evaluating the properties of various study designs for the identification of gene-environment and gene-gene interaction. Numerous variations on the basic case-control design have been suggested, including using family members as controls, recruiting cases with a family history of cancer or with a second primary cancer, matching controls to cases according to various sampling schemes, or using cases only. The optimal design depends on numerous factors, and we have provided software which enables researchers to quickly evaluate the power of a range of designs for any particular problem (http://cruk.leeds.ac.uk/katie; Saunders et al, The Stata Journal 2003; 3: 47). For example, two-stage sampling schemes may be considered where a crude measure of exposure (such as latitude as a surrogate for sun exposure) is available on a large cohort, and controls are sampled according to frequency of this surrogate measure. The graph compares (for one set of parameter values) the required number of cases to detect interaction for various such sampling schemes as a function of the specificity of the surrogate exposure measure (Figure 1).
Segregation analysis has been performed with a view to understanding the likely contribution of genes to colorectal cancer susceptibility after taking into account germline mutations in mismatch repair genes. Similar techniques have also been applied in collaboration with other researchers at the University of Leeds to estimate the heritability of various quantitative traits associated with cardiovascular disease. We have investigated the properties and limitations of current methods of segregation analysis for qualitative and quantitative traits as part of the Genetic Analysis Workshop. (Crockford et al., to appear in BMC Genetics.)
Fig 1. The required number of cases to detect interactions by sampling scheme.
|
