Time to Check CHEK2 in Families With Breast Cancer?

CHEK2 (cell cycle checkpoint kinase 2) is an important player in the team of all-stars implicated in DNA repair and human cancer predisposition, including BRCA1, BRCA2/FANCD2, TP53, and ATM. CHEK2 is a serine threonine kinase that is activated by ATM protein in response to DNA double-strand breaks. CHEK2 not only regulates the function of BRCA1 protein in DNA repair but also exerts a number of critical roles in cell cycle control and apoptosis. The patterns of cancer risk associated with germline mutations in the critical genes in BRCA-associated DNA repair are remarkably variable. For example, PALB2, a recently identified BRCA2interacting protein and a key player in the BRCA1-BRCA2 biochemical complex, appears to be rarely mutated in families with multiple patient cases of early-onset breast cancer. For ATM, however, the paucity of high-penetrance mutations has been a continuing puzzle. One particular CHEK2 mutation, 1100delC in exon 10, which abolishes its kinase function, has been evaluated in numerous reports because of its association with breast and other cancers in both unselected and familial breast cancer cohorts. Other CHEK2 mutations have been evaluated in vitro and in vivo and have shown diminished CHEK2 function and population specificity but no clear relationship to breast cancer risk. In this issue of the Journal of Clinical Oncology, Weischer et al attempt to provide more stable estimates of the breast cancer risk associated with CHEK2*1100C, and, on the basis of these estimates, raise a challenge: Are we ready to consider screening for this mutation as part of the clinical evaluation of families with breast cancers? To address this challenge, it is instructive to examine the insights and the limitations afforded by the many studies included in the meta-analysis of this particular cancer-predisposing mutation. Both family-based (genetic) and population-based epidemiologic methodologies have been employed to determine the cancer spectrum (phenotype), risk (penetrance), number of individual mutations (allelic heterogeneity), and distribution of mutations in different geographic groups (population heterogeneity) associated with candidate cancer predisposition genes, such as CHEK2. Turning first to the issue of phenotype, what is the cancer spectrum associated with CHEK2*1100C? Weischer et al focus appropriately on breast cancer. However, CHEK2*1100delC was initially described in a kindred with classic Li-Fraumeni syndrome; two carriers had breast cancer, but the third had a histiocytoma. In a multicenter study of cancer incidence in 67 families with at least one CHEK2*1100C mutation carrier, the risk of cancers other than breast was not significantly elevated, except for colorectal and lung cancers in women aged 20 to 50 years. No excess of cancer was observed in male relatives, though underreporting was demonstrated. Others have observed increases in breast cancer in men and in colorectal and prostate cancers, but these have not been characterized with the same level of detail as the breast cancer risk in women. The next critical question is whether CHEK2*1100delC is really a low penetrance mutation. In their meta-analysis, Weischer et al calculated a cumulative breast cancer risk by age 70 years in families with breast cancer of 37% (95% CI, 26% to 56%) for the 1100delC allele. They consider this comparable to the recently reported 57% (95% CI, 47% to 66%) penetrance for BRCA1 and 49% (95% CI, 40% to 57%) for BRCA2. Several caveats should be considered in this context. First, the upper band of the risk estimate in the meta-analysis by Weischer et al is derived from nine studies of patients with familial breast cancer. The methods of ascertainment for these studies varied widely, and at least one was population-based. Several of these studies utilized ascertainments that selected for familial clustering of breast cancer, and a range of criteria was used to define “familial.” Of note, the BRCA meta-analysis cited by Weischer et al explicitly sought to limit its scope to exclude familial patient cases and to limit to populationbased ascertainments. Thus, comparison of the BRCA populationbased risk estimates and the CHEK2*1100delC familial-based penetrance estimates runs the risk of what Weischer et al call the apples and oranges dilemma. Attention to the type of studies included in these meta-analyses is important because of the now well-established appreciation that selected ascertainments appear to create a bias toward high penetrance estimates. The degree to which penetrance estimates vary for BRCA1 and BRCA2 remains a topic of intense controversy among cancer epidemiologists; explanations for the differences in risk estimates among studies include the biases in patient case (and control) ascertainment resulting from selective participation or other methodologic factors, the differing environmental or other modifying genetic factors, and genotype-phenotype associations that affect kindreds segregating BRCA mutations. In the subset analysis by Weischer et al, using study designs least prone to ascertainment bias (ie, unselected breast cancer cases), the aggregated odds ratio for breast cancer was 2.7, and this dropped to 1.8 (95% CI, 1.2 to 2.5) when the two largest studies were excluded. One of these large studies was published earlier this year in JCO by this same group. In that study, only 13 1100delC mutation carriers were identified in more than 1,000 breast cancer patient cases, which gives a broad confidence for the observed risk of 1.3 to 5.4. Another consideration in all genetic studies of this type is population heterogeneity, which is the result of different mutations having JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 26 NUMBER 4 FEBRUARY 1 2008