Rare genetic variants and the risk of cancer

There are good reasons to expect that common genetic variants do not explain all of the inherited risk of the common cancers, not least of these being the relatively low proportion of familial relative risk that common cancer SNPs currently explain. One promising source of the unexplained risk is rare, low-penetrance genetic variants, a class that ranges from low-frequency polymorphisms (allele frequency < 5%) through subpolymorphic variants (frequency 0.1–1.0%) to very low frequency or ‘private’ variants with frequencies of 0.1% or less. Examples of rare cancer variants include breast cancer susceptibility loci CHEK2, BRIP1 and PALB2. There are considerable challenges associated with the discovery and testing of rare predisposition alleles, many of which are illustrated by the issues associated with variants of unknown significance in the Mendelian cancer predisposition genes. However, whilst cost constraints remain, the technological barriers to rare variant discovery and large-scale genotyping no longer exist. If each individual carries many disease-causing rare variants, the so-called missing heritability of cancer might largely be explained. Whether or not rare variants do end up filling the heritability gap, it is imperative to look for them alongside common variants.

Introduction

The first and most clear-cut achievement in the study of inherited cancer risk that has come from the revolution in molecular genetics is the successful discovery of genes which cause the Mendelian cancer predisposition syndromes. We use the term “Mendelian” here to refer to patterns of familial inheritance that at least closely approximate to Mendelian expectations, implying penetrances of the associated variants that are usually at least of the order of 0.7. To date, about 100 such genes are known. Most of these have been discovered by a combination of positional cloning by linkage analysis, and choice of candidate genes. A variety of indirect, sometimes serendipitous, approaches, such as seeking rare, large deletions or finding a signature of genomic instability, have also made significant contributions.

In all cases, the disease-causing variants are, at least for dominant effects, very rare, being maintained in the population by mutation-selection balance. Whilst some founder mutations exist and other mutations are recurrent, possibly because of unusually high germline mutation rates, there is considerable within-locus variant heterogeneity. Thus, over 500 different adenomatous polyposis coli (APC) mutations have been reported in patients with familial adenomatous polyposis (FAP). Even for recessive cancer predisposing conditions, such as Xeroderma pigmentosum, the individual allele frequencies rarely exceed 0.005.

Although the relative risks associated with the Mendelian cancer genes are, by definition, very large and so should be readily detectable, there remains speculation that there may exist undiscovered cancer loci with high-penetrance variants. The evidence for this is largely based on comparisons with the clinical features of existing syndromes. Large cancer families of unknown cause, unusual cancers in consanguineous families, individuals with multiple tumours of one type and, occasionally, co-occurrence of specific cancers all suggest Mendelian inheritance. In some cases, such as hyperplastic polyposis syndrome, disease may be recessive and sometimes occult or of variable severity, and hence the condition might rarely present in families that would allow linkage analysis. In other cases, there might be considerable locus heterogeneity, confounding attempts at gene mapping and identification. However, it seems most unlikely that such new discoveries will significantly increase the overall population frequency of Mendelian cancer syndromes.

For the common cancers, a maximum of about 5% of cases are associated with known Mendelian susceptibility. For colorectal cancer (CRC), for example, Mendelian syndromes include FAP, Lynch syndrome (caused by mutations in DNA mismatch repair genes), the recessive MUTYH-associated polyposis (MAP) and a small number of rarer polyposis syndromes. However, overall, about 30% of the variation in CRC risk is thought to be due to inherited susceptibility, a proportion far too large to be explained by the Mendelian syndromes. Breast cancer has a similar gap between Mendelian and overall genetic risk and for prostate cancer, the risk is even higher, as very few cases are attributable to high-risk alleles. It is that gap which must be filled by studies to identify cancer predisposition alleles in the general population.

A priori, the missing cancer heritability must be derived from genetic variants of lower risk. In principle, these variants could exist at any frequency in the population. The first approach to looking for the basis of this heritability, exemplified by association studies on HLA and disease, was to search for associations between common polymorphic variants and a specific disease. Initially, this was based on candidate gene approaches, testing the hypothesis that either the discovery variant itself was disease-causing or that it was in linkage disequilibrium with the disease-causing variant. Candidate genes often produced conflicting results [1, 2, 3]. However, with the advent of high-throughput technology, it became possible to do genome-wide association (GWA) searches, with no a priori assumptions about mechanisms, on very large numbers of polymorphic variants and on very large cohorts of cases and controls. It became apparent, moreover, that large sets of thousands of cases and controls were needed when the observed odds ratios (ORs) for even the most significantly associated variants were quite low and rarely exceeded even 1.4 or 1.5. At this level of OR, only very large studies can give significance when allowance is made for the number of comparisons being tested.

GWA studies, based on what is now sometimes called the tagging SNPs model, have been used successfully by many groups, where success is measured by the significance of the association rather than its magnitude. However, the common cancer alleles detected to date typically account for only ∼10% of the familial relative risk of disease, still leaving open the question of what can explain the remaining genetic risk of cancer. The strong possibility is that rare variants with ‘subpolymorphic’ allele frequencies account for most of this remaining inherited risk [4•, 5•]. Direct evidence for this has come, for example, from the study of genetic variation in the level of plasma HDL-type cholesterol [6].