Chips, candidate genes, and CLL

Comment on Rudd et al, page 638

Rudd and colleagues have conducted a large-scale association study using nonsynonymous SNPs on a population sample of patients with CLL and healthy controls and identified a group of plausible candidate SNPs, including ATM, CHEK2, and BRCA2.

The search for the specific genes that account for susceptibility to most human disease in general and hematologic cancer in particular is a pivotal goal in human genetics and a major anticipated benefit of the sequencing of the human genome. The handful of major (ie, high-penetrance) genes identified for a few human cancers (eg, BRCAs in breast cancer) account for only a small component of the genetic predisposition that exists, and the minor (low-penetrance) genes that likely account for the balance of susceptibility are not known with certainty for any human malignancy, despite more than a decade of candidate testing. The situation with chronic lymphocytic leukemia (CLL) is typical. Striking kindreds with multiple patients with CLL, population registry, and twin studies clearly implicate heredity as important in CLL. Family history predicts risk of CLL far better than any known extrinsic risk factor. Yet to date, the genes that account for elevated risk in relatives of patients with CLL are unknown. Separate linkage studies using the largest existing collections of high-risk kindreds in the United States¹  and the United Kingdom²  have failed to provide strong evidence for specific chromosomal regions that may harbor these genes. Linkage is a powerful technique, but among its limitations is the requirement for many families, and this approach cannot detect the multiple weaker genes (those that convey less than a 4-fold risk) that are precisely the ones thought to contribute most to hereditary susceptibility. Thus, so-called “association studies,” based in the general population, are an essential complement to family studies. To date, however, population-based studies that test one gene or a few genes at a time have not proven fruitful. While critical observers have emphasized the need for attention to study design (need for adequate power, careful control selection, attention to population structure and false positives), there is an increasing consensus that new genomic technologies that rapidly assess multiple genomic variants will provide the breakthrough. Here, 2 general approaches are possible. The first, or indirect approach, uses chips with tens to hundreds of thousands of single nucleotide polymorphisms (SNPs) and tests for statistical associations of each SNP with disease. Since the SNP detected on the chip that exhibits the association may not be the one that accounts for the critical disease-causing alteration, but only one nearby (ie, in linkage disequilibrium), as in linkage analysis, a follow-up round of genetic studies must be conducted to verify the locus and find the specific gene. A second approach is the “direct” method selected by Rudd and colleagues in this issue of Blood. Using diverse genomic tools, they selected only SNPs calculated to cause an amino acid change (nonsynonymous SNPs [nsSNPs]), thus limiting selection to a group much more likely to have functional significance. The ability to select both genes with relevance to cancer biology and the specific SNPs within those genes that cause deleterious protein changes combines the strength of the candidate approach (genes are selected because of a priori interest) with the power of chip-based approaches to study thousands of candidates. Accordingly, Rudd and colleagues find associations with genes that are highly plausible (ATM, BRCA2, and CHEK2).

The critical and complementary role of population studies can be appreciated by the ATM finding. ATM has a critical role in DNA repair, and its location near a commonly observed cytogenetic abnormality in CLL at 11q13 makes it a highly plausible candidate gene in CLL. Nevertheless, an earlier linkage study excluded the gene.³  Since Rudd and colleagues observed estimated odds ratios in the 1.7 to 2.1 range, the signal would have “passed below the linkage radar”; that is, occurred below the limit of detection for linkage but comfortably within the range detectable in case-control design using large population samples.

Much remains to be done. Future linkage studies should involve consortia that combine precious high-risk CLL kindreds to provide the best chance to detect signals from the sofar elusive high-penetrance genes. It is axiomatic that the specific candidates implicated require independent validation in well-designed studies from diverse populations. In spite of the reasonable sample size and conservative threshold for significance, the general concern for false positives mandates a very cautious view of the specific candidates identified. Since the authors selected both the genes and the SNPs, it is likely that many possible pathways and genes are under- or unrepresented, so alternate whole-genome approaches (eg, the indirect approach) that represent the whole genome in an anonymous but more balanced manner will be eagerly awaited. Chip-based approaches like that of Rudd and colleagues should see broader application as investigators move to apply advanced genomic approaches to the problem of genetic susceptibility in common hematologic malignancy. ▪