Colorectal cancer risk in individuals with biallelic or monoallelic mutations of MYH

In a recent study published in the International Journal of Cancer, Fleischmann et al.1 described the results of screening for germline mutations in the human homolog of the E. coli mutY gene (MYH) in a series of 358 individuals affected with early-onset colorectal cancer (CRC), arbitrarily defined as a diagnosis before age 56, and 354 control individuals. In Caucasians, the 2 MYH mutations Y165C and G382D, identified in a subset of familial adenomatous polyposis (FAP) patients who lack mutations in the APC gene, are by far the most frequent known disease-causing mutations.2, 3 With respect to these mutations characterized previously, one CRC case carried biallelic MYH mutations (0.3%), and 5 CRC cases (1.4%) and one control (0.3%) carried monoallelic MYH mutations. Even though the frequency of carriers of monoallelic MYH mutations in CRC cases was not statistically different from the frequency in controls, the authors suggested that MYH heterozygotes might be at increased risk of developing CRC. In the present study, we have similarly assessed the frequencies of disease-causing MYH mutations in 555 CRC cases and 918 controls. A consecutive series of individuals affected with CRC was identified, consisting of 555 individuals who visited the oncology clinic at Memorial Hospital in New York City.4 Blood specimens were obtained regardless of ethnic origin and religion from 2002 to present. Before permanent anonymization of these samples, primary CRC diagnosis was confirmed by inspection of the electronic pathology record, and gender, religion, ethnic origin, history of chemotherapy, date of birth, age at diagnosis and date of diagnosis were recorded. DNA samples were prepared from blood as described previously.5 Collection of these samples was performed under the auspices of an IRB-approved protocol. Controls (918) were obtained from a cohort study of over 17,000 volunteers of varying ethnic backgrounds, ages 30–69, unaffected by cancer, and ascertained from 2000–2002 at hospitals and blood donation centers that are geographically close to Memorial Hospital.6 To ensure comparability, controls were chosen so that numbers of persons in each age, ethnic origin and religion group were approximately equal. Subjects provided informed consent for use of their de-identified DNAs for cancer genetics research. The Y165C mutation was detected in sequences amplified by polymerase chain reaction by pyrosequencing and the G382D mutation by restriction enzyme digestion with BglII. Because the 466delE mutation was recently described as disease-causing in some Italian familial FAP and attenuated FAP families,7 we also screened for this mutation by direct sequencing. All mutation carriers were confirmed by direct sequencing. Odds ratios and exact confidence intervals were calculated using StatXact (Cytel software corporation, Cambridge, MA). Exact tests for Hardy-Weinberg equilibrium were performed for each mutation and for both mutations together using SAS version 9.0 (SAS Institute Inc., Cary, NC). Power calculations for minimum detectable odds ratios were performed using StatXact. In the 555 CRC cases, we detected 4 individuals with monoallelic MYH mutations and 2 individuals with biallelic MYH mutations; in the 918 healthy controls, 7 individuals carried monoallelic MYH mutations (Table I). In the 4 CRC cases who carried a monoallelic MYH mutation, the presence of a second MYH mutation was excluded by direct sequencing of the entire MYH gene. The 466delE mutation was not detected in any of the 555 CRC cases nor in 621 controls tested (297 Jewish controls were not tested). The Y165C and G382D mutations were not detected in the 266 Jewish cases nor in 450 Jewish controls. These data indicated that the frequency of the 466delE mutation in control individuals is <0.6% and that the frequency of both the Y165C and G382D mutations in Jewish control individuals is <0.8% at a 0.05 significance level. The Y165C and G382D mutations were detected only in the non-Jewish Caucasian and Caucasian/Hispanic groups (Table I). We detected 7 carriers of monoallelic MYH mutations in 392 healthy non-Jewish Caucasian and Caucasian/Hispanic controls combined for a population frequency of 1.8%; thus, the allele frequency of the Y165C and G382D mutations in this population is 0.9%. If the MYH Y165C and G382D mutations were in Hardy-Weinberg equilibrium in the CRC cases, the expected frequency of carriers with biallelic MYH mutations in cases would be approximately 8/100,000 (the square of the allele frequency in controls). We detected, however, 2 carriers of biallelic mutations in the 255 non-Jewish Caucasian and Caucasian/Hispanic CRC cases tested (0.8%), which is a 100-fold difference. This deviation from Hardy-Weinberg equilibrium in the cases for both mutations together was statistically significant (p < 0.001), and the deviation was also significant for the G382D mutation alone (p < 0.001). We note that the reported carrier frequencies of MYH mutations in control populations vary over a wide range even within the same population (Table II). Al-Tassan et al.8 and Sieber et al.3 estimated the carrier frequency in the British population to be 2% (2/100) and 1.9% (2/107), respectively, whereas Fleischmann et al.1 found the frequency to be 0.3% (1/354). The wide range of population frequencies may reflect real differences between populations, as we suspect is the case in comparing the Jewish and non-Jewish Caucasian populations. Differences in estimates of carrier frequencies could occur due to sampling from populations that consist of different population mixtures. These considerations emphasize the importance of appropriate matching of controls with cases. In the entire sample set tested here, the frequencies of carriers of monoallelic MYH mutations was 0.7% in CRC cases and 0.8% in healthy controls; in the non-Jewish Caucasian and Caucasian/Hispanic groups, the frequencies were 1.6% in cases and 1.8% in controls. Hence, in our study, there is not a statistically significant association between monoallelic MYH mutations and CRC (odds ratio [OR] = 0.95; 95% confidence interval [CI] = 0.20–3.75). In fact, assuming the population carrier frequency of MYH mutations is 1%, our study had 80% power to detect an association with a significance level of 5% only if the genotype risk ratio is 3.3 or greater. If there is a risk associated with monoallelic MYH mutations, the results from our study and previously published studies have been vastly under-powered to detect it. In a naive analysis of pooled data from all 5 published case-control series combined with data from this study (Table II), the frequency of carriers in CRC cases was 1.10% (29/2,632) and in controls 0.61% (13/2,114) for an OR = 1.80 (exact 95% CI = 0.90–3.78); this association was not statistically significant (p = 0.10). We note that even with the sample size of the combined data analysis, we would have only been able to detect an OR of 2.4 or greater with 80% power and a significance level of 5%. Power calculations show that to detect an OR of 1.5, the minimum sample size required would be approximately 16,500 cases and 16,500 controls. It is important to note that this naïve, combined analysis does not account for the differences in allele frequencies in the control populations or differences in the ascertainment of cases among these studies. Accounting for these differences would likely reduce the power of the combined analyses even further. We have found, as others have before us, that individuals who carry biallelic MYH mutations have a high risk of developing CRC. However, the question of whether individuals who carry known disease-causing, monoallelic MYH mutations are at increased risk of CRC remains unanswered. Analysis of the data from all studies performed to date suggests that there might be an association between CRC development and monoallelic mutations in MYH; however, even the combined data are underpowered to detect an association (i.e., to detect an OR below 2.4). With the allele frequency of the MYH mutations being low in all populations studied to date, either much larger sample sizes (e.g., n > 10,000) or analysis of selected case groups (e.g., persons with familial CRC) will be required to address this important question. This work was supported by the National Institutes of Health grants CA103500 (to N.A.E.) and CA103394 (to N.M.), a fellowship from the Cancer Research and Prevention Foundation (to P.P.), a fellowship from the Frankel Foundation (to T.K.), the Byrne Foundation, and the Academic Medicine Development Company. Dr. P. Gregersen provided DNAs from controls through the New York Cancer Project. Dr. J. Shia was instrumental in the collection of Jewish cases who underwent surgery at Memorial Hospital. The authors thank Mr. M. Ilzarbe and Ms. D. Wong for their help with DNA sequencing. Yours sincerely, Paolo Peterlongo, Nandita Mitra, Shaokun Chuai, Tomas Kirchhoff, Crystal Palmer, Helen Huang, Khedoudja Nafa, Kenneth Offit, Nathan A. Ellis.