HRAS1 Rare Minisatellite Alleles and Breast Cancer in Australian Women Under Age Forty Years

Transcription

1 HRAS1 Rare Minisatellite Alleles and Breast Cancer in Australian Women Under Age Forty Years Frank A. Firgaira, Ram Seshadri, Christopher R. E. McEvoy, Gillian S. Dite, Graham G. Giles, Margaret R. E. McCredie, Melissa C. Southey, Deon J. Venter, John L. Hopper Background: A recent meta-analysis of 23 studies supported the empirically derived hypothesis that women who lack one of the four common minisatellite alleles at the HRAS1 locus are at increased risk of breast cancer. These studies relied on visual sizing of alleles on electrophoretic gels and may have underreported rare alleles. We determined whether this hypothesis applied to early-onset breast cancer by using a new method to size minisatellite alleles. Methods: We conducted a population-based, case control-family study of 249 Australian women under 40 years old at diagnosis of a first primary breast cancer and 234 randomly selected women, frequency matched for age. We sized HRAS1 minisatellite alleles with an Applied Biosystems model 373 automated DNA sequencer and GENESCAN software. All P values are two-sided. Results: We found no association of rare alleles with breast cancer, before or after adjustment for risk factors and irrespective of how their effects were modeled (crude odds ratio = 1.04; 95% confidence interval [CI] = ; P =.8). The rare allele frequency was (95% CI = ), three times the pooled estimate of (95% CI = ) from previous studies (P<.001), and was similar for case, (95% CI = ), and control, (95% CI = ) (P =.7). Conclusion: There was no support for an association between rare HRAS1 alleles and the risk of early-onset breast cancer, despite 80% power to detect effects of the magnitude of those associations (1.7-fold) previously suggested. Implications: The question of whether cancer risk is associated with rare minisatellite HRAS1 alleles needs to be revisited with the use of new methods that have a greater ability to distinguish rare alleles from similarly sized common alleles. [J Natl Cancer Inst 1999;91: ] A meta-analysis of 23 case control studies supported the empirically derived hypothesis (1) that the 11% of individuals who inherit one or two rare minisatellite alleles at the HRAS1 locus are at increased risk of cancer (2). In particular, it was estimated that the increased risk of female breast cancer is 1.68-fold (95% confidence interval [CI] ) (2). If true, the locus would explain 7.5% (95% CI 2.5% 14%) of all cases of breast cancer, considerably more than is currently attributed to mutations in the genes BRCA1 and BRCA2 (3). The HRAS1 minisatellite is derived from a tandemly repeated 28-base-pair sequence. The four common alleles, designated a1 to a4, have approximate sizes of 1.0 kilobase (kb), 1.5 kb, 2.1 kb, and 2.5 kb, respectively, and have been reported to account for approximately 94% of HRAS1 alleles in Caucasians (1,2,4). About 40 other allelic variants have been described, and these constitute what are termed the rare alleles. Many of these variants, however, differ from a common allele by only one or a few repeat units. Previous studies, from which the putative association of the rare alleles with breast cancer was derived, have relied on poorly resolving electrophoretic systems and visual sizing of the HRAS1 minisatellite alleles. This procedure may have led to the underreporting of rare alleles, and the consequent misclassifications may have influenced their results and contributed to the conflicting results in many studies (2). In this study, we have used an Applied Biosystems model 373 automated DNA sequencer and GENESCAN technology to more precisely size HRAS1 minisatellite alleles (5) in a population-based sample of Australian women with breast cancer diagnosed before the age of 40 years and in a randomly selected sample of women without breast cancer, who were frequency matched for age. The frequencies of HRAS1 minisatellite alleles in case and control have been compared to determine whether the rare alleles are associated with an increased risk of breast cancer. If inheriting at least one rare allele does increase the risk of breast cancer, the frequency of rare alleles should be greater in women with a family history of the disease (see Appendix section). HRAS1 allele association analyses are also presented separately for women with and women without a family history of breast cancer. METHODS Subjects As described in the protocol outlined by Hopper et al. (7), a population-based, case control-family study of early-onset breast cancer was carried out in Melbourne and Sydney, Australia, from 1992 through 1995 (8,9). were women under the age of 40 years at diagnosis of a first primary breast cancer, identified through the Victorian and New South Wales cancer registries. were women without breast cancer, selected from the electoral roll (adult registration for voting is compulsory in Australia) by use of stratified random sampling and frequency matched for age., control, and relatives were administered the same risk factor questionnaire (8). For each case subject and control subject, a detailed family history was systematically recorded for all first-degree and second-degree relatives and subsequently checked with their living relatives at the time of their interview. Unless otherwise stated, women who reported having at least one first-degree or second-degree female relative with breast cancer were considered to have a family history of breast cancer. Verification of family cancers reported by case or relatives was obtained through cancer registries, pathology reports, hospital records, treating clinicians, and death certificates (8). Blood samples were collected from case and control at the time of interview. Of 644 eligible case, 467 (72.5%) case were interviewed. Attrition was due to death (1.7%), refusal (surgeon 8.4% or patient 11.8%), nonresponse (surgeon 0.6% or patient 1.4%), or a change in place of residence (3.6%). Of Affiliations of authors: F. A. Firgaira, R. Seshadri, C. R. E. McEvoy, Department of Haematology and Genetic Pathology, Flinders University and Flinders Medical Centre, Bedford Park, South Australia; G. S. Dite, J. L. Hopper, The University of Melbourne, Centre for Genetic Epidemiology, Carlton, Victoria, Australia; G. G. Giles, Cancer Epidemiology Centre, Anti-Cancer Council of Victoria, Carlton, Australia; M. R. E. McCredie, Cancer and Epidemiology Research Unit, New South Wales Cancer Council, Kings Cross, Australia, and Department of Preventative and Social Medicine, University of Otago, New Zealand; M. C. Southey, D. J. Venter, Department of Pathology and Research, Peter MacCallum Cancer Institute, Melbourne, Victoria, Australia, and Department of Pathology, The University of Melbourne, Parkville, Australia. Correspondence to: John L. Hopper, Ph.D., The University of Melbourne, Centre for Genetic Epidemiology, 200 Berkeley St., Carlton, Victoria 3053, Australia ( j.hopper@gpph.unimelb.edu.au). See Notes following References. Oxford University Press Journal of the National Cancer Institute, Vol. 91, No. 24, December 15, 1999 REPORTS 2107

2 the 632 eligible control, refusals (25.8%) and nonresponse (9.8%) resulted in 408 control being interviewed (64.4%). Blood samples were available from 393 case (84.2% of participating and 61.0% of eligible case ) and 294 control (72.1% of participating and 46.5% of eligible control ). Analyses of HRAS1 genotypes were performed for a subset of 249 case (53.3% of participating and 38.7% of eligible case ) and 234 control (57.4% of participating and 37.0% of eligible control ). Selection of case and control for these analyses was not made on the basis of measured risk factor information. For case and control, there were no differences between those included and those not included in the study for the following factors associated with breast cancer in the full sample of case and control (8): age, marital status, level of education, parity, height, weight, age at menarche, and country of birth. Genetic analyses were performed for 60.3% of case and 63.6% of control who had a first-degree relative with breast cancer and for 52.3% of case and 57.0% of control who did not have a first-degree relative with breast cancer. Written informed consent was obtained from all case and control, and the study was approved by institutional review boards of The University of Melbourne, The Anti-Cancer Council of Victoria, the New South Wales Cancer Council, and Flinders University. Molecular Analysis DNA was prepared as previously described (9). The HRAS1 minisatellite region was amplified by polymerase chain reaction as described (5). The fluorescently labeled products were precisely sized for genotyping of HRAS1 alleles by use of native polyacrylamide gels on a Perkin-Elmer Applied Biosystems model 373 DNA sequencer and GENESCAN software (Perkin-Elmer Corp., Foster City, CA) as detailed previously (5). Genotyping was performed in a blinded fashion for source (case subject or control subject) of DNA. Statistical Methods Under the assumptions of Hardy Weinberg equilibrium, the maximum likelihood estimator of the frequency of rare alleles is f (2n 11 + n 01 )/2n, where n n 11 + n 01 + n 00 and n ij is the observed number of with the ij genotype (i,j 0,1), where 1 represents the presence of a rare allele and 0 represents the absence of a rare allele (i.e., presence of one of the four common alleles) and has asymptotic standard error (SE) [(f [1 f ])/2n] 1/2 and approximate 95% CI (f 1.96SE f SE). Estimates of allele frequency for different groups were compared by assuming that they were each normally distributed with standard deviation equal to SE derived from that group alone. The Hardy Weinberg equilibrium assumption was assessed by comparing the observed numbers of individuals with different genotypes with those expected under Hardy Weinberg equilibrium for the estimated allele frequency and then comparing the Pearson goodness-of-fit statistic with a 2 distribution with 1 df. Given no evidence of departure from Hardy Weinberg equilibrium, we analyzed and modeled the frequency of rare alleles as a function of potential covariates by use of linear logistic regression, by assuming that the number of rare alleles was the sum of two independent binomial variables. The influence of HRAS1 genotype on the risk of breast cancer was assessed by standard case subject/ control subject analyses by use of multiple linear logistic regression, with and without adjustment for the risk factors identified in this study (8). Genotype was modeled the following three ways: 1) by number of rare alleles (two parameters), 2) by a linear effect per number of rare alleles (one parameter), and 3) by the presence or absence of any rare allele (one parameter). Logistic regression analyses were performed with STATA (10). All statistical tests and P values are two-sided. For logistic regression analyses, P values were calculated by use of the likelihood ratio test. RESULTS Table 1. HRAS1 genotype distributions for case and control Table 1 shows the complete genotype distributions for the case and control. The four common alleles, a1 to a4, are labeled 1 4, respectively. The rare alleles are labeled in terms of the difference in number of repeat units from that of the common allele closest in size. For example, refers to the allele with one repeat unit more than a1, and 4 2 has two fewer repeat units than a4. For case and control combined, the allele frequencies for a1, a2, a3, and a4 were (95% CI ), (95% CI ), (95% CI ), and (95% CI ), respectively. Except for a1, each of these allele frequencies was less than the corresponding values of (95% CI ), (95% CI ), (95% CI ), and (95% CI 1 1, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Total *Genotype ij, where i and j represent the two alleles of a given genotype. n i,j is the number of with alleles i and j. The four common alleles, a1 to a4, are labeled 1 to 4, respectively. The rare alleles are labeled in terms of the difference in number of repeat units from that of the common allele closest in size. For example, refers to the allele with one repeat unit more than a1, and 4 2 has two fewer repeat units than a REPORTS Journal of the National Cancer Institute, Vol. 91, No. 24, December 15, 1999

3 ) found by pooling control in the meta-analysis (2). More over, the deficit increased as the allele size increased, being most pronounced for a4. Table 2 shows that, overall, the frequency of the rare alleles was (95% CI ), so that 32% of individuals carried at least one rare allele. This allele frequency was almost three times greater than the (95% CI ) observed in the pooled control of the meta-analysis (2) (P<.001). There was no difference in the frequency of rare alleles between case and control overall (P.7), among those with a reported family history of breast cancer (P.2), or among those without a reported family history of breast cancer (P.2). Comparing women with and without a reported family history of breast cancer, there was no difference overall (P.5), among case (P.1), or among control (P.4). The rare allele frequencies for case and control, respectively, were (95% CI ) and (95% CI ) for women born in Australia (P.7) and (95% CI ) and (95% CI ) for the subset of these women for whom both parents were also born in Australia (P.7). There was no evidence for deviation from Hardy Weinberg equilibrium among case, among control, among total (all P>.8), or among any group or subgroup defined by case subject/control subject status and/or family history status. Table 3 shows that the frequency of Table 2. Frequency of HRAS1 rare alleles, stratified by family history of breast cancer*, Yes No Total,,,,, No. of rare alleles 0 61 (75) 42 (65) 103 (71) 108 (64) 119 (70) 227 (67) 169 (68) 161 (69) 330 (68) 1 18 (22) 21 (32) 39 (27) 54 (32) 46 (27) 100 (30) 72 (29) 67 (29) 139 (29) 2 2 (2) 2 (3) 4 (3) 6 (4) 4 (2) 10 (3) 8 (3) 6 (3) 14 (3) Total Allele frequency (95% confidence interval) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Two-sided P *Family history of breast cancer is defined as having a reported first- or second-degree female relative with breast cancer. Test of allele frequencies: family history (yes versus no) case only, P.1; control only, P.4; case and control combined, P.5. Columns may not add to 100% because of rounding. See Statistical Methods section for more details. Family history definition Table 3. Frequency of HRAS1 rare alleles by reported or verified family history of breast cancer* Allele frequency (95% confidence interval) Total Reported first-degree relative with breast cancer Yes ( ) ( ) ( ) No ( ) ( ) ( ) Two-sided P Verified first-degree relative with breast cancer Yes ( ) ( ) ( ) No ( ) ( ) ( ) Two-sided P Reported first-degree or second-degree relative with breast cancer Yes ( ) ( ) ( ) No ( ) ( ) ( ) Two-sided P Verified first-degree or second-degree relative with breast cancer Yes ( ) ( ) ( ) No ( ) ( ) ( ) Two-sided P *P values determined by use of likelihood ratio test. Journal of the National Cancer Institute, Vol. 91, No. 24, December 15, 1999 REPORTS 2109

4 rare alleles did not differ by the presence of a reported or verified family history of breast cancer overall, among case, or among control, irrespective of the definition of family history. The frequency of rare alleles was actually lower among case with a verified first-degree relative with breast cancer (P.05). The frequency of rare alleles did not differ according to age, country of birth, state, highest level of education, marital status, or any of the other variables and risk factors for breast cancer measured by questionnaire (8). After adjustment for these factors, there was no difference in the frequency of rare alleles between case and control (odds ratio [OR] 1.05; 95% CI ; P.8) or between women with and without a family history of breast cancer, whether combining case and control (OR 0.86; 95% CI ; P.5) or analyzing only case (OR 0.59; 95% CI ; P.1) or only control (OR 1.29; 95% CI ; P.4). Table 4 shows that, irrespective of how the putative effect of rare alleles was modeled, there was no association of rare HRAS1 allele status with risk of breast cancer, either before or after adjustment for the risk factors identified in the full dataset (8). Although the point estimates for the effect of having two rare alleles were greater than for having one rare allele, as was also found by the metaanalysis for all cancers (2), none of the estimates was statistically significant. The differences in effects were also not statistically significant (all P>.5). When modeled as a linear effect on the logarithmic OR scale, the crude effect per number of rare alleles was 0.06 (95% CI 0.28 to 0.40; P.7), equivalent to predicted ORs of 1.06 and 1.12 for one and two rare alleles, respectively. After adjustment, the effect on the logarithmic OR was 0.09 (95% CI 0.29 to 0.47 per allele [P.6]), equivalent to 1.09 and 1.19 for one and two alleles, respectively. On the logarithmic odds scale, the estimated effect for having any rare allele was 0.04 (95% CI 0.36 to 0.44) with no adjustment and 0.05 (95% CI 0.37 to 0.47) after adjustment (P.8). These two estimates were less than the corresponding estimate for breast cancer of 0.52 (95% CI ) found by the meta-analysis (2) (P.06 and P.08, respectively). Given that the 95% CIs of these logarithmic OR estimates were about 0.8, effects equivalent to a logarithmic OR of 0.5 (i.e., OR 1.65) or more would have been detectable at the.05 level of statistical significance with more than 80% power. For women with a reported family history of breast cancer, the crude OR for presence of at least one rare allele on breast cancer risk was 0.60 (95% CI ); after adjustment for the covariates as in Table 4, it remained at 0.60 (95% CI ). For women without a reported family history of breast cancer, the crude OR was 1.32 (95% CI ); after adjustment, it was 1.29 (95% CI ). Because all of these 95% CIs included unity, there was no evidence of an effect of rare alleles on risk of breast cancer in women with a family history of breast cancer or in women without a family history of breast cancer. Furthermore, the effect was no different between women with and women without a family history of breast cancer in terms of either a crude or an adjusted OR (P.2 and P.4, respectively). Finally, after adjustment for both family history of breast cancer and presence of a rare allele, there was no evidence for an additional statistical interaction in their effects on risk of breast cancer, irrespective of whether family history was defined in terms of at least one reported or verified affected first-degree relative or one reported or verified affected first-degree or second-degree relative or whether analysis was in terms of crude or adjusted OR. There was little power, however, to detect nonadditivity on the logarithmic odds scale. DISCUSSION In this study, we have found that, by sizing HRAS1 minisatellite alleles with automated and more precise methodology, the frequency of rare minisatellite alleles at the HRAS1 locus was 0.17, almost three times greater than that found in previous studies that have used visual sizing (2). Because, to our knowledge, this is the first study of HRAS1 alleles in the Australian population and because the studies of the meta-analysis were carried out in the United States or Western Europe, the difference may be due to an underlying real difference in allele frequency. This is unlikely, however, because the great majority of our case and control were of British or Western European descent (8,9). The difference, however, is consistent with the automated sizing method that can distinguish more clearly between a common allele and the rare alleles that differ by one or a few repeats, especially for the larger alleles. For example, from Table 1, we can calculate that the allele frequency for the combined alleles of 2 1, 2, and was (0.108 in case and in control ; compare with for the a2 allele derived from the meta-analysis). For 3 2, 3 1, 3, Table 4. Association of HRAS1 allele status with breast cancer* Crude odds ratio (95% CI) Two-sided P Adjusted odds ratio (95% CI) Two-sided P No. of rare alleles 0 Reference Reference ( ) 1.01 ( ) ( ) ( ).4 No. of rare alleles (linear) 1.06 ( ) ( ).6 Any rare allele No Reference Reference Yes 1.04 ( ) ( ).8 *CI confidence interval. Data were adjusted for family history (first- or second-degree relative) of breast cancer, state, country of birth, reference age, amount of education, marital status, number of live births, height, weight, age at menarche, and duration of oral contraceptive use. P values determined by use of likelihood ratio test REPORTS Journal of the National Cancer Institute, Vol. 91, No. 24, December 15, 1999

5 3 + 1,and 3 + 2,it was (0.104 in case and in control ; compare with for the a3 allele derived from the meta-analysis). For 4 3,...,4,...,and4+3,itwas0.089 (0.098 in case and in control ; compare with for the a4 allele derived from the meta-analysis). (In none of these instances was the frequency different between case and control.) We suggest, therefore, that a major difficulty as well as a possibly contributory cause of conflicting association results in past studies has been the inability of the methods, based on Southern blotting, to distinguish the larger rare alleles from similarly sized common alleles. We failed to detect an association between the presence of the rare alleles and an increased risk of breast cancer before the age of 40 years by comparing case with control or by comparing individuals with and without a family history of breast cancer. Given the estimated frequency of rare alleles and that we studied more than 200 case and 200 control, our study had more than 80% power to detect an effect equivalent to an OR of the size found by the meta-analysis. A smaller effect, however, cannot be dismissed. Because typically less than 10% of cases of breast cancer in Western societies occur before the age of 40 years, it is possible that the effect of the rare HRAS1 alleles may be confined to or may be stronger in cancers occurring at a later age. If the rare alleles were associated with an increased risk of breast cancer, carriers would be more likely to have a family history of breast cancer. For example, if the frequency of rare alleles is 0.17 and the associated risk of breast cancer really is 1.7 as was found by the meta-analysis (2), the probability that the mother is affected is 1.32 times higher if the daughter is a carrier, compared with her not being a carrier (see Appendix section). In our study, however, the probability of having a family history of breast cancer was actually less, though not statistically significantly so, in case subject carriers and in carriers overall. On the logarithmic odds scale, the 95% CIs were about 1.2 among case and 0.8 overall. Therefore, although there was less than 50% power to detect the predicted effect of about 0.25 on the logarithmic odds scale, for both case and overall, the predicted effect lies outside the 95% CIs for the observed effects of 0.5 (95% CI 1.1 to 0.1) and 0.15 (95% CI 0.55 to 0.25), respectively. That is, examination of data on family history of breast cancer provided no support for an association between the rare alleles and risk of breast cancer. The findings presented in this report for early-onset breast cancer suggest that the putative associations of HRAS1 minisatellite alleles with cancers of the breast and other sites need to be re-evaluated with newer methods of sizing alleles. We have presented our raw data in Table 1 so that pooling with similar studies may yet reveal genotypes or alleles that discriminate between case and control. APPENDIX Let the frequency of the rare alleles be p and assume the Hardy Weinberg equilibrium and random mating. Let X 1, if an individual has at least one of the rare alleles and otherwise let X 0. Then, following the original formulation by Fisher (6), for a mother daughter pair, P(X m 1 X d 0) p and P(X m 0 X d 0) q 1 p, P(X m 1 X d 1) (1 + pq)/(2 p), P(X m 0 X d 1) q 2 /(2 p), where the subscripts m and d refer to mother and daughter, respectively. Let Y 1, if an individual has breast cancer, and otherwise Y 0, and RR represent the increase in risk of breast cancer associated with having at least one of the rare alleles. That is, P(Y i 1 X i 1) RR P(Y i 1 X i 0), where i m or d. Also, assume that if i j, P(Y i X i,x j ) P(Y i X i ), so that if the individual s genotype is known, risk is independent of the relation s genotype. Then, P(Y m 1 X d 1) P(Y m 1 X m 1) P(X m 1 X d 1) + P(Y m 1 X m 0) P(X m 0 X d 1) P(Y m 1 X m 0) [RR(1 + pq) + q 2 ]/(2 p) and P(Y m 1 X d 0) P(Y m 1 X m 1) P(X m 1 X d 0) + P(Y m 1 X m 0) P(X m 0 X d 0) P(Y m 1 X m 0) [RRp + q]. Therefore, when p 0.17 and RR 1.7, the ratio P(Y m 1 X d 1)/P(Y m 1 X d 0) [RR(1 + pq)+q 2 ]/[RRp + q](2 p) is equal to REFERENCES (1) Krontiris TG, DiMartino NA, Colb M, Parkinson DR. Unique allelic restriction fragments of the human Ha-ras locus in leukocyte and tumour DNAs of cancer patients. Nature 1985; 313: (2) Krontiris TG, Devlin B, Karp DD, Robert NJ, Risch N. An association between the risk of cancer and mutations in the HRAS1 minisatellite locus. N Engl J Med 1993;329: (3) Ford D, Easton DF, Stratton M, Narod S, Goldgar D, Devilee P, et al. Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. Am J Hum Genet 1998;62: (4) Garret PA, Hulka BS, Kim YL, Farber RA. HRAS proto-oncogene polymorphism and breast cancer. Cancer Epidemiol Biomarkers Prev 1993;2: (5) McEvoy CR, Seshadri R, Firgaira FA. Large DNA fragment sizing using native acrylamide gels on an automated DNA sequencer and GENESCAN software. BioTechniques 1998;25: (6) Fisher RA. Correlations between relatives on the supposition of Mendelian inheritance. Trans Roy Soc (Edinb) 1918;52: (7) Hopper JL, Giles GG, McCredie MRE, Boyle P. Background, rationale and protocol for a case control-family study of breast cancer. Breast 1994;3: (8) McCredie MR, Dite G, Giles GG, Hopper JL. Breast cancer in Australian women under 40. Cancer Causes 1998;9: (9) Southey MC, Batten LE, McCredie MR, Giles GG, Dite G, Hopper JL, et al. Estrogen receptor polymorphism at codon 325 and risk of early onset breast cancer. J Natl Cancer Inst 1998;90: (10) STATA Statistical Software: Release 5.0, College Station (TX): Stata Corporation; NOTES Supported by grants from the Kathleen Cuningham Foundation, the National Health and Medical Research Council of Australia, the Victorian Health Promotion Foundation, The NSW Cancer Council, and The Peter MacCallum Cancer Institute. Manuscript received June 23, 1999; revised October 6, 1999; accepted October 12, Journal of the National Cancer Institute, Vol. 91, No. 24, December 15, 1999 REPORTS 2111