The WECARE Study of Gene-Radiation Exposure Interactions as a Cause of Contralateral Breast Cancer: Extending the study of interactions

Jonine Bernstein, Memorial Sloan-Kettering Cancer Center 
*Dan Stram, Department of Preventive Medicine, Keck School of Medicine, University of Southern California 
Duncan Thomas, Department of Preventive Medicine, Keck School of Medicine, University of Southern California 

Keywords: Gene x Environment Interactions; Contralateral Breast Cancer; GWAS

The WECARE Collaborative Study Group

Broadening GWAS studies to encompass alleles that are involved in gene-environment interactions is important for the understanding of the full range of genetic influence on cancer risk and for adding to our basic knowledge of cancer etiology. In radiation epidemiology the discovery of alleles that modify individual genetic sensitivity to the carcinogenic effects of radiation exposure would have potential consequences for radiation protection policy.

The Women's Environment Cancer and Radiation Epidemiology (WECARE) Study includes a two stage genome-wide association study (GWAS) in which women with second primary breast cancers in the contralateral breast are compared to matched controls for the purpose of finding genetic variants that either increase susceptibility to contralateral breast cancer on their own or interact with estimated radiation exposure (constructed from treatment records for the first cancer) to increase contralateral cancer risk. At present the WECARE Study has completed stage 1 of the GWAS study, in which approximately 1 million SNPs were genotyped with the Illumina OMNI platform. After removing genotyping failures a total of 641 contralateral breast cancer cases and 1247 matched cases without contralateral breast cancer are currently available. The second stage will involve genotyping another 1000 women with contralateral breast cancer and 1000 controls.

As reported at this meeting in 2010 based on case-control analyses the first stage results themselves show little evidence for novel gene x radiation interactions, and Teraoka et al [1] reported that there were no significant gene by radiation interactions when 21 known breast cancer variants identified through other GWAS studies were considered. There is a considerable amount of research concerning the development of more powerful tests for gene x environment (GxE) interactions. These include the empirical Bayes-modified case-only analyses of Mukhejee and Chatterjee [2] the two-stage approach of Murcray et al [3] which tests, in the first stage, for gene exposure associations while ignoring case-control status, followed by traditional case-control testing in the second stage, and others [4] [5] [6] [7]. These methods generally seek to exploit gene x environment independence underlying the powerful case-only method of Piegorsch et al [8] while retaining protection against the effects of occasional failures of this assumption for individual SNPs. These additional approaches are currently being implemented in analyses of the WECARE GWAS Study and will be discussed at the meeting.

Nearly all variants identified by GWAS in breast cancer (and many other diseases) have quite modest effects on risk of initial cancers with odds ratios in the range from 1.1 to 1.3 for identified variants globally significant after genome-wide correction for multiple comparisons. At this time it appears that it may take hundreds of variants to explain, for example, the increase in cancer risk that a women with a sister or mother diagnosed with breast cancer experiences, relative to other women without first degree affected relatives. Even if a given variant does interact with radiation exposure, this is likely to be only of limited interest for risk prediction unless the interaction odds ratios are much larger than the main effects of common variants seen to date.

Of ultimate interest will be the combined behavior of all variants that are involved in breast cancer risk and the interaction of the combination with radiation exposure. Assuming that a polygenic model for breast cancer susceptibility is reasonable and that ultimately a large enough set of variants will be discovered that the polygene will be well characterized (perhaps as a weighted sum of hundreds of variants), it will be very important to evaluate the full polygene in relation to radiation exposure to determine whether individuals who are especially likely to have a first cancer (because of their polygenic load) are more likely to experience a second cancer if they receive radiotherapy than not. While conceived as a discovery project the WECARE Study may play its most important role in evaluation of the sum total of polygenic effects in relation to radiation exposure as the contributors to the polygene are gradually discovered. The ability of the WECARE Study to play this role in the future will depend upon two principal factors: 1) the true state of nature regarding the role of polygenic inherited variation relative to the role that complex interactions between genes (or genes and other environments) may play, and 2) The ability of the 1 million SNPs genotyped in the WECARE Study to recover the components of the polygene.

With respect to (1), there is an ongoing debate about the role of polygenic components in risk (c.f. [9]) relative to epistatic components (i.e. the influence of complex gene by gene interactions). This is a crucial issue for risk prediction in the WECARE Study or any other study, since epistatic components of disease risk are inherently far more difficult to identify than are single genes contributing additively to risk.

There are strongly suggestive results from GWAS based variance components analyses (c.f. [10]) for a number of traits ([10] [11] [12] [13]) that polygenic components do indeed play an important role in the overall heritability of traits or risk of disease. However, in most instances to date, only a very small fraction of this evident polygenic component has been directly attributable to known variants. At this point it time it appears likely that much of the heritability of many traits or diseases, can be attributed to polygenic components.

The ability (2) of the WECARE Study to recover a polygene related to risk of breast cancer is dependent upon the frequency spectrum of the variants that underlie risk. If the most important contributions to trait heritability are due to relatively common risk alleles then it should be possible to use imputation methods [14] to reasonably well estimate individual polygene values based upon coverage statistics, i.e. the Illumina OMNI chip was constructed to include enough informative SNPs to predict most common alleles on the basis of linkage disequilibrium. Imputed polygenes can then be tested to see if they interact with radiation exposure. On the other hand, if the allele spectrum of the variants making up the polygene is quite different than the GWAS chip, with many rare variants making important contributions, then it may take considerably deeper information (e.g. whole genome sequencing) in order to capture the polygene. Some have argued that the same type of variance components analyses that appear to implicate a strong polygenic component in disease risk also implies that the components of the polygene are likely to be primarily common variants. This statement however remains somewhat controversial [15].

To give an illustration of the type of work that will be of importance using the WECARE Study data we performed two analyses. First, we investigated whether the combined effect of a polygene made up of a sum of 19 SNPs (identified as risk alleles in other GWAS studies and directly genotyped on the OMNI array) was different according to whether or not a women was exposed to radiation during treatment of the initial primary. Second, we are carrying out a variance components analysis using the WECARE Study cases only to see if there are more subtle indications of an additive interaction component in the WECARE Study GWAS data. Here we use the methods of Yang et al. [10] to determine whether there is a polygenic component in relation to response to radiation.

The results of the polygene analysis for the 19 variants showed no indications of a deviation from a multiplicative model for the joint effects of genes and radiation on the risk of a second contralateral breast cancer. The variance components analysis is in progress and will be presented at the meeting.

REFERENCES 1. Teraoka, S.N., et al., Single nucleotide polymorphisms associated with risk for contralateral breast cancer in the Women's Environment, Cancer, and Radiation Epidemiology (WECARE) Study. Breast Cancer Research, 2011. 13(6): p. R114. 2. Mukherjee, B. and N. Chatterjee, Exploiting Gene-Environment Independence for Analysis of Case-Control Studies: An Empirical Bayes-Type Shrinkage Estimator to Trade-Off between Bias and Efficiency. Biometrics, 2008. 64(3): p. 685-94. 3. Murcray, C.E., J.P. Lewinger, and W.J. Gauderman, Gene-environment interaction in genome-wide association studies. Am J Epidemiol, 2009. 169(2): p. 219-26. 4. Mukherjee, B., et al., Testing gene-environment interaction in large-scale case-control association studies: possible choices and comparisons. Am J Epidemiol, 2012. 175(3): p. 177-90. 5. Chatterjee, N., Z. Kalaylioglu, and R.J. Carroll, Exploiting gene-environment independence in family-based case-control studies: increased power for detecting associations, interactions and joint effects. Genet Epidemiol, 2005. 28(2): p. 138-56. 6. Thomas, D., Gene–environment-wide association studies: emerging approaches. Nature Reviews Genetics, 2010. 11(4): p. 259-272. 7. Cornelis, M.C., et al., Gene-environment interactions in genome-wide association studies: a comparative study of tests applied to empirical studies of type 2 diabetes. Am J Epidemiol, 2012. 175(3): p. 191-202. 8. Piegorsch, W.W., C.R. Weinberg, and J.A. Taylor, Non-hierarchical logistic models and case-only designs for assessing susceptibility in population-based case-control studies. Stat Med, 1994. 13(2): p. 153-62. 9. Zuk, O., et al., The mystery of missing heritability: Genetic interactions create phantom heritability. PNAS, 2012. epub(epub): p. 1-6. 10. Yang, J., et al., Common SNPs explain a large proportion of the heritability for human height. Nature Genetics, 2010. 42(7): p. 565-569. 11. Yang, J., et al., Genomic inflation factors under polygenic inheritance. European Journal of Human Genetics, 2011. 12. Davies, G., et al., Genome-wide association studies establish that human intelligence is highly heritable and polygenic. Molecular Psychiatry, 2011. 16(10): p. 996-1005. 13. Yang, J., et al., Genome partitioning of genetic variation for complex traits using common SNPs. Nat Genet, 2011. 43(6): p. 519-25. 14. Li, Y., et al., MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genetic Epidemiology, 2010. 34(8): p. 816-834. 15. Browning, Sharon R. and Brian L. Browning, Population Structure Can Inflate SNP-Based Heritability Estimates. The American Journal of Human Genetics, 2011. 89(1): p. 191-193.