Patients With an Unclassified Genetic Variant in the BRCA1 or BRCA2 Genes Show Different Clinical Features From Those With a Mutation
http://www.100md.com
《临床肿瘤学》
the Department of Clinical Genetics, University Hospital and Research Institute Growth & Development, and Department of Methodology and Statistics, the University of Maastricht, Maastricht, the Netherlands
ABSTRACT
PURPOSE: To obtain and compare the probabilities of finding a mutation in the BRCA1 or BRCA2 genes, the clinical features, and the family history among patients with an unclassified variant (UV) and those with a pathogenic mutation.
PATIENTS AND METHODS: The study included 70 patients: 24 with a UV (BRCA1, n = 4; BRCA2, n = 19; both, n = 1), and 46 with a mutation (BRCA1, n = 32; BRCA2, n = 14). Two of the UVs were novel variants; the rest had been reported previously as UVs. Probabilities of finding a mutation were retrospectively obtained using BRCAPRO and Myriad II programs.
RESULTS: The probability to detect a mutation was significantly lower in the group of patients with a UV than in those with a mutation (BRCAPRO [mean ± standard deviation], 0.297 ± 0.312 v 0.627 ± 0.315, P = .001; and Myriad II, 0.124 ± 0.090 v 0.283 ± 0.176, P = .001, respectively). Independent predictive factors of finding either a UV or a mutation were number of affected relatives (2.9 ± 1.4 v 4.0 ± 1.9; P = .039) and number of tumors among relatives (3.3 ± 1.4 v 4.4 ± 1.8; P = .031), respectively.
CONCLUSION: The combined data about the predictive models show significant differences between both groups. Individual probabilities can be regarded as a help to guide the clinical management of patients with a UV in those genes. However, a definitive conclusion about the pathogenicity of a UV can not be obtained from the clinical features alone, but only in combination with biochemical and epidemiologic data.
INTRODUCTION
Whole gene mutation analyses of BRCA1 and BRCA2 genes have led to the identification of an increasing amount of nucleotide variants known as genetic variants of uncertain significance or unclassified variants (UVs). These genetic variants have as common denominator that their pathogenicity is unclear, and therefore their clinical relevance is uncertain. Approximately one third of the genetic variants in BRCA1 and half of those in BRCA2 reported in the Breast Cancer Information Core (BIC), are UVs.1
Several biochemical and epidemiologic criteria are used to establish whether a genetic variant is either pathogenic, ie, a true mutation or a normal variant (a polymorphism).2-9 Biochemical criteria include alteration in the size, expression, conformation, or properties of the protein; conservation of the amino acid among species; and function of the region where the variant has occurred. Epidemiologic criteria are whether it has been previously reported in the population, and the prevalence among 100 control alleles.
In addition, cosegregation studies can also be helpful: the absence of cosegregation can exclude pathogenicity when inheritance has an autosomal dominant pattern, as is the case for hereditary breast and ovarian cancer, and provided that the history of cancer is limited to one side of the family.
Co-occurrence of a genetic variant with a deleterious mutation in the same gene (BRCA1 or BRCA2) argues against causality, given that it would be rare statistically. An additional argument against causality applies for when the genetic variant occurs on a different allele that the mutation (ie, in trans), based on the assumption that compound heterozygotes are embryonically lethal.8
Analysis of loss of heterozygosity in the tumor9 and functional studies in cell lines can provide a tailored test for each variant, but generally are not available and are not always conclusive.
A polymorphism (ie, a neutral variant) is defined as a non–disease-causing change and/or a change found in at least 1% of the alleles in the population.10 Genetic variants are considered UVs when it is unclear whether they have a functional effect in the protein. These variants usually appear in less than 1% of the alleles in the population. The vast majority of the UVs described so far are missense mutations, but also mutations leading to stop codons at the end of the gene or variants in intronic regions (potentially involved in RNA splicing) have been reported. Presymptomatic testing cannot be offered in those situations and genetic counseling can only be based on the evaluation of the clinical features and the family history.
Whether patients with a UV in the BRCA1 or BRCA2 genes have different clinical features than those with a mutation had not been investigated. We retrospectively obtained the estimated probability of finding a mutation in the BRCA1 or BRCA2 genes in those patients in whom a UV in either one of these genes had been found. Second, we analyzed their clinical features and family history and compared them with those from patients with an established pathogenic mutation.
PATIENTS AND METHODS
Patients
Eligibility criteria. All probands originate from the Southeast region in the Netherlands. Eligible candidates for BRCA mutation analysis were selected when they met one of the inclusion criteria, based on the number of affected relatives (for detailed criteria description, see Table 1). We included in this study all eligible candidates who underwent complete BRCA1 and BRCA2 mutation analysis at our center and who were found to have a mutation or a UV in those genes. The study included 70 patients: 24 with a UV, and 46 with a mutation in the BRCA1 or BRCA2 genes. Seven individuals did not have a tumor but underwent DNA analysis because all affected relatives in the family were already deceased: six of them were found to have a mutation in BRCA1, and one had two UVs (one in BRCA1 and one in BRCA2).
Clinical data. Probabilities of finding a mutation in the probands were retrospectively obtained using BRCAPRO11 and Myriad II12 models (Cancergene software program; Southwestern Medical School at Dallas, Dallas, TX). Breast cancer (in probands and in relatives) refers to invasive breast cancer, both for probability estimations and for the clinical evaluation. Ductal carcinoma in situ (DCIS) was recorded in the category "other tumors."
After excluding the eight patients who harbored a UV that later was reported as a polymorphism, we recorded the clinical data and family history of 62 patients: 16 patients with a UV and 46 patients with a mutation in the BRCA1 or BRCA2 genes (Table 1).
Methods
Laboratory diagnosis. BRCA1 and BRCA2 were analyzed by denaturing high performance liquid chromatography (DHPLC), as described previously.13 Additional technical details, primers, and DHPLC elution profiles are available on request. Changes in DHPLC elution profiles were verified by standard sequence analysis. In the past, a protein truncation test was used to analyze exon 11 of BRCA1 and exons 10 and 11 of BRCA2. In those cases, the rest of the gene was fully analyzed by DHPLC later. In addition, multiplex ligation-dependent probe amplification analysis14 was performed for BRCA1 to detect large duplications or deletions.
Definition of UV. A number of epidemiologic and biochemical criteria were recorded for each UV (summarized in Table 2). Genetic variants that were found together with a mutation in the same gene were not included because it was assumed that coexistence of two pathogenic mutations would be likely to be incompatible with life and/or extremely rare. A missense variant was considered a UV based on its frequency: less than 1% in our control and patient population or when it had previously been reported as UV in the literature and/or databases. Cosegregation studies could only be performed in three of the patients; all three showed cosegregation with the disease.
Four of the genetic variants that were initially classified as UV (Ser1040Asn, Asp1420Tyr, Val2728Ile, and Lys3326stop) were reported later to be polymorphisms2,3,15; therefore, we excluded the patients having one of those variants from additional clinical analysis. After excluding those four variants, 14 different UVs remained (Table 2). All but one of the UVs are missense variants. The variant in intron 19 of the BRCA1 5313 v 25 A > C, which may result in altered splicing in the BRCA1 gene, and Lys467Arg in the BRCA2 gene, are novel genetic variants.
The UV Glu2856Ala in exon 20 of BRCA2 has a borderline frequency of 1%. We still included it among the UVs because it has been reported previously as UV in the Breast Cancer Information Core database1 and because all of the other criteria (it causes polarity change and lies on a conserved region) suggest that it may be pathogenic.
Among the 46 patients with a mutation, there were 24 different mutations identified: 12 in BRCA1 and 12 in BRCA2. The vast majority (19 mutations) are nonsense, four have an effect on splicing, and one is a missense mutation.
Statistical Analysis
Descriptive analyses were performed with the whole group: 16 UVs, 46 mutations, and subgroup analysis with BRCA2 patients alone. The tests used were independent two-sample t test for the continuous variables and 2 test for the categoric variables. Logistic regression was used as a predictive model. The mutation probability was obtained as a function of one independent variable. Receiver operating characteristic curves were used to compare these predictive models with BRCAPRO and Myriad II.16
RESULTS
All patients included in the study fulfilled one of our indication criteria to perform DNA analysis. However, highly significant differences were found when we retrospectively obtained the chance of finding a mutation in a BRCA1 or BCRA2 gene, both with BRCAPRO and Myriad II software programs, between the group with a mutation (n = 46) and those with a UV (n = 16). The mutation probability (mean ± standard deviation [SD]) was 0.297 ± 0.312 for the group with UVs versus 0.627 ± 0.315 in the group with mutations (P = .001) with BRCAPRO, and 0.124 ± 0.090 versus 0.283 ± 0.176, respectively (P = .001), with Myriad II (Table 1). From the eight patients with a polymorphism, probabilities were obtained in seven of them. In one patient, probabilities could not be estimated because of the lack of information about the family. The mean ± SD chance of a mutation was 0.175 ± 0.293 with BRCAPRO and 0.173 ± 0.165 with Myriad II. No significant differences were observed between this group and the group with UVs, whereas a statistically significant difference was found with BRCAPRO between this group and the group with a mutation (P = .001; data not shown).
After excluding the eight patients with a polymorphism, the clinical study included 16 patients with a UV: two in BRCA1, 13 in BRCA2, and one in both genes, and 46 patients with an established mutation in BRCA1 (n = 32) or in BRCA2 (n = 14). Table 1 lists the clinical data of the patients and their families from both groups.
The presence of tumors other than invasive breast and ovarian cancer among the probands was significantly more frequent in the group with UVs (four patients; 25%), compared with that among the patients with a mutation (three patients; 7%; P = .044). These tumors include DCIS (n = 2), colon (n = 1), and melanoma (n = 1) among the probands with a UV, and DCIS (n = 2) and colon (n = 1) among those with a mutation.
In both groups, the most frequent indication to perform DNA analysis was the presence of three or more affected relatives. One affected individual in the family was the indication criteria to perform DNA analysis in 31% of the patients with a UV, as opposed to only 11% in the group with a mutation.
In agreement with those results, we also found that the group with mutations showed a significantly higher number of affected relatives (mean ± SD, 4.0 ± 1.9 v 2.9 ± 1.4; P = .039), as well as higher number of tumors among relatives (4.4 ± 1.8 v 3.3 ± 1.4; P = .031). These differences were not due to a different family size, given that the total mean number of relatives was similar.
No significant differences were observed between the probands of both groups with regard to sex, prevalence of breast cancer or ovarian cancer, and age of diagnosis of the first tumor. No difference was found in either the mean age of diagnosis of breast or ovarian cancer, or in the frequency of bilateral breast cancer or of cancer in male patients in the families.
A subgroup analysis was performed for the BRCA2 patients only, which showed the same results (see P values in Table 1). Logistic regression analysis showed that the number of affected relatives (P = .024) and the number of tumors among relatives (P = .023) were significant independent univariate predictive factors of finding either a mutation or a UV. The probability of finding a mutation in BRCA1 or BRCA2 is given by the following functions: P = 1/[1 + exp(0.586 – 0.430 x number of tumors among relatives)], or P = 1/[1 + exp(0.455 – 0.442 x number of affected relatives)]. Receiver operating characteristic curves were used to compare these two models with BRCAPRO and Myriad II models. The BRCAPRO and Myriad II models seemed to be better predictive models because they had a larger area under the curve (0.787 and 0.816, respectively) than the above-mentioned models based on one variable (area under the curve, 0.686 and 0.694, respectively).
Table 2 shows the individual BRCAPRO and Myriad II mutation probabilities as well as the biochemical and epidemiologic features of each UV. BRCAPRO mutation probabilities were exceptionally high (ie, above the mean value of the mutation probability in the group with mutations) in three probands with a UV (the one with Asp1739Gly in BRCA1 and those with the UVs Ser384Thr and Ser1750Phe in BRCA2).
DISCUSSION
In this study we provide evidence that patients with a UV in BRCA1 or BRCA2 have different clinical features compared with those with a pathogenic mutation. First, when we retrospectively obtained the probabilities of finding a mutation using model analysis software, these were significantly lower in probands with a UV than in those with a mutation. An additional detailed clinical analysis revealed that the frequency of tumors other than invasive breast and ovarian cancer in the proband, the number of affected relatives, and the number of tumors among first- and second-degree relatives correlated with that difference.
Families with a mutation in BRCA2 have overlapping but different clinical features from those with a mutation in BRCA1.17 Given that most of the UVs were localized in BRCA2, whereas the majority of the mutations were localized in BRCA1 in our study, the differences observed between both groups could actually have been caused by the localization either in BRCA1 or in BRCA2, and not because of the type of genetic variant. Therefore, we performed a subanalysis only for the patients with a genetic abnormality in BRCA2, and observed similar results as those found in the whole group.
Fewer affected relatives and fewer tumors in the families of patients with a UV as compared with those with a mutation suggest that if some of the UVs are pathogenic, they may have less penetrance than the established pathogenic mutations, or need the co-occurrence of a mutation in another as yet unidentified genetic risk factor. The lower penetrance might be explained by the different nature of these genetic variants (ie, UVs are mostly missense variants, whereas most of the mutations are causing protein truncation products).
Both BRCAPRO and Myriad II are widely used programs to estimate the chance of finding a mutation in BRCA1 or BRCA2 genes.18-21 So far only BRCAPRO has been validated and seems to be more accurate than Myriad II.18 According to our results, the group of UVs significantly differ in their estimated probabilities of finding a mutation compared with the group with established pathogenic mutations with both programs.
Petersen et al5 have described a Bayesian approach that uses clinical information about families to assess causality of missense mutations in hereditary disorders. Although it is a quantitative method, when used for hereditary breast or ovarian cancer it does not allow the inclusion of all clinical data; for example, type of tumor (ie, breast or ovarian cancer) or age of diagnosis. In addition, it assumes that the penetrance does not change with age, which is not the case for hereditary breast and ovarian cancer.
One has to take into account, however, that our results are valid for the group as a whole but it is still premature to use the results for individual UVs. For this purpose, analysis of several families with the same UV would be needed to ascertain if the probability results are similar. In this respect, it is also possible that the associated probabilities of a mutation are not related to the identified UV but to a nonidentified mutation in another region of the gene, such as intronic sequences or regulatory regions, or in other as yet unidentified genes. So far, individual probabilities can help categorize these UVs into high or low probability of being pathogenic to be used as a chief guidance for medical management.
Our study shows that, in the majority of the patients, it is reasonable to ignore variants clinically and base the counseling on the family history. However, our results also show that three probands with a UV had exceptionally high BRCAPRO mutation probabilities (ie, above the mean mutation probability of the group with mutations). One of those UVs is Asp1739Gly in BRCA1. Using biochemical criteria and conservation among species, Abkevich et al7 analyzed the likelihood that a number of missense variations in BRCA1 are pathogenic; among them, one localized in the same codon (Asp1739Glu) as the UV included in our study. They also come to the conclusion that the change of Asp1739 is likely to be deleterious.
After our article was submitted, Goldgar et al8 provided strong evidence for five UVs to be either deleterious or neutral using a multifactorial likelihood-ratio model. That model includes the following parameters: co-occurrence with deleterious mutations, cosegregation data, sequence conservation, severity of substitution, and functional characteristics. Three of the UVs included in our study (Arg841Trp in BRCA1 and Tyr42Cys and Pro655Arg in BRCA2) have also been analyzed by Goldgar et al8 and are considered neutral (ie, they do not cause disease). Evidence against causality was overwhelming only for the first two UVs and came mainly from the cosegregation and co-occurrence criteria.8 Our results support these conclusions. The patients with those UVs had a low probability of mutation that was much more evident for the first two UVs. In addition, a tight probability range was observed among the patients showing Tyr42Cys, which further supports the hypothesis against this UV being pathogenic.
If we take all of these results into account, it seems evident that a definitive answer about pathogenicity cannot be obtained from a single perspective alone. An ideal scoring system should therefore include a combination of biochemical, epidemiologic, and clinical data. Finally, an international effort, which includes keeping an up-to-date genetic variant database, is also needed to ascertain the pathogenicity of individual UVs.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank Y. Detisch and B. Dols-Caanen, genetic counselors, for running the software programs with the predictive models and drawing the family pedigrees, respectively; R. Helsdingen, medical student, for recording the clinical data; and D. Marcus-Soekarman, MD, PhD, and U. Moog, MD, PhD, clinical geneticists, for the critical discussions.
NOTES
Presented as an abstract at the Family Cancer Meeting, June 5-7, 2003, International Union Against Cancer (UICC), Oklahoma City, OK, and at the Familial Cancer Meeting, May 6-7, 2004, the European School of Oncology, Madrid, Spain.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
National Human Genome Research Institute: Breast Cancer Information Core (BIC) database: An open access on-line breast cancer mutation data base. http://research.nhgri.nih.gov/bic/
Arnold N, Peper H, Bandick K, et al: Establishing a control population to screen for the occurrence of nineteen unclassified variants in the BRCA1 gene by denaturing high-performance liquid chromatography. J Chromatogr B 782:99-104, 2002
Deffenbaugh AM, Frank TS, Hoffman M, et al: Characterization of common BRCA1 and BRCA2 variants. Genet Test 6:119-121, 2002
Fleming MA, Potter JD, Ramirez CJ, et al: Understanding missense mutations in the BRCA1 gene: An evolutionary approach. Proc Natl Acad Sci U S A 100:1151-1156, 2003
Petersen GM, Parmigiani G, Thomas D: Missense mutations in disease genes: A Bayesian approach to evaluate causality. Am J Hum Genet 62:1516-1524, 1998
Petrucelli N, Lazebnik N, Huelsman KM, et al: Clinical interpretation and recommendations for patients with a variant of uncertain significance in BRCA1 or BRCA2: A survey of genetic counseling practice. Genet Test 6:107-113, 2002
Abkevich V, Zharkikh A, Deffenbaugh AM, et al: Analysis of missense variation in human BRCA1 in the context of interspecific sequence variation. J Med Genet 41:492-507, 2004
Goldgar DE, Easton DF, Deffenbaugh AM, et al: Integrated evaluation of DNA sequence variants of unknown clinical significance: Application to BRCA1 and BRCA2. Am J Hum Genet 75:535-544, 2004
Osorio A, de la Hoya M, Rodríguez-López R, et al: Loss of heterozygosity analysis at the BRCA loci in tumor samples from patients with familial breast cancer. Int J Cancer 99:305-309, 2002
Human Genome Variation Society: Nomenclature for the description of sequence variations. http://www.genomic.unimelb.edu.au/mdi/mutnomen/recs.html
Parmigiani G, Berry DA, Aguilar O: Determining carrier probabilities for breast cancer-susceptibility genes BRCA1 and BRCA2. Am J Hum Genet 62:145-158, 1998
Frank TS, Manley SA, Olopade OI, et al: Sequence analysis of BRCA1 and BRCA2: Correlation of mutations with family history and ovarian cancer risk. J Clin Oncol 16:2417-2425, 1998
Wagner T, Stoppa-Lyonnet D, Fleischmann E, et al: Denaturing high-performance liquid chromatography detects reliably BRCA1 and BRCA2 mutations. Genomics 62:369-376, 1999
Hogervorst FB, Nederlof PM, Gille JJ, et al: Large genomic deletions and duplications in the BRCA1 gene identified by a novel quantitative method. Cancer Res 63:1449-1453, 2003
Mazoyer S, Dunning AM, Serova O, et al: A polymorphic stop codon in BRCA2. Nat Genet 14:253-254, 1996
Altman DG: Practical Statistics for Medical Research. London, UK, Chapman & Hall, 1991
Verhoog LC, Berns EMJJ, Brekelmans CTM, et al: Prognostic significance of germline BRCA2 mutations in hereditary breast cancer patients. J Clin Oncol 18:119S-124S, 2000 (suppl)
Berry DA, Iversen ES Jr, Gudbjartsson DF, et al: BRCAPRO validation, sensitivity of genetic testing of BRCA1/BRCA2, and prevalence of other breast cancer susceptibility genes. J Clin Oncol 20:2701-2712, 2002
Euhus DM, Smith KC, Robinson L, et al: Pretest prediction of BRCA1 or BRCA2 mutation by risk counselors and the computer model BRCAPRO. J Natl Cancer Inst 94:844-851, 2002
Shannon KM, Lubratovich ML, Finkelstein DM, et al: Model-based predictions of BRCA1/2 mutation status in breast carcinoma patients treated at an academic medical center. Cancer 94:305-313, 2002
De la Hoya M, Diez O, Perez-Segura P, et al: Pre-test prediction models of BRCA1 or BRCA2 mutation in breast/ovarian families attending familial cancer clinics. J Med Genet 40:503-510, 2003(Encarna B. Gómez-García, )
ABSTRACT
PURPOSE: To obtain and compare the probabilities of finding a mutation in the BRCA1 or BRCA2 genes, the clinical features, and the family history among patients with an unclassified variant (UV) and those with a pathogenic mutation.
PATIENTS AND METHODS: The study included 70 patients: 24 with a UV (BRCA1, n = 4; BRCA2, n = 19; both, n = 1), and 46 with a mutation (BRCA1, n = 32; BRCA2, n = 14). Two of the UVs were novel variants; the rest had been reported previously as UVs. Probabilities of finding a mutation were retrospectively obtained using BRCAPRO and Myriad II programs.
RESULTS: The probability to detect a mutation was significantly lower in the group of patients with a UV than in those with a mutation (BRCAPRO [mean ± standard deviation], 0.297 ± 0.312 v 0.627 ± 0.315, P = .001; and Myriad II, 0.124 ± 0.090 v 0.283 ± 0.176, P = .001, respectively). Independent predictive factors of finding either a UV or a mutation were number of affected relatives (2.9 ± 1.4 v 4.0 ± 1.9; P = .039) and number of tumors among relatives (3.3 ± 1.4 v 4.4 ± 1.8; P = .031), respectively.
CONCLUSION: The combined data about the predictive models show significant differences between both groups. Individual probabilities can be regarded as a help to guide the clinical management of patients with a UV in those genes. However, a definitive conclusion about the pathogenicity of a UV can not be obtained from the clinical features alone, but only in combination with biochemical and epidemiologic data.
INTRODUCTION
Whole gene mutation analyses of BRCA1 and BRCA2 genes have led to the identification of an increasing amount of nucleotide variants known as genetic variants of uncertain significance or unclassified variants (UVs). These genetic variants have as common denominator that their pathogenicity is unclear, and therefore their clinical relevance is uncertain. Approximately one third of the genetic variants in BRCA1 and half of those in BRCA2 reported in the Breast Cancer Information Core (BIC), are UVs.1
Several biochemical and epidemiologic criteria are used to establish whether a genetic variant is either pathogenic, ie, a true mutation or a normal variant (a polymorphism).2-9 Biochemical criteria include alteration in the size, expression, conformation, or properties of the protein; conservation of the amino acid among species; and function of the region where the variant has occurred. Epidemiologic criteria are whether it has been previously reported in the population, and the prevalence among 100 control alleles.
In addition, cosegregation studies can also be helpful: the absence of cosegregation can exclude pathogenicity when inheritance has an autosomal dominant pattern, as is the case for hereditary breast and ovarian cancer, and provided that the history of cancer is limited to one side of the family.
Co-occurrence of a genetic variant with a deleterious mutation in the same gene (BRCA1 or BRCA2) argues against causality, given that it would be rare statistically. An additional argument against causality applies for when the genetic variant occurs on a different allele that the mutation (ie, in trans), based on the assumption that compound heterozygotes are embryonically lethal.8
Analysis of loss of heterozygosity in the tumor9 and functional studies in cell lines can provide a tailored test for each variant, but generally are not available and are not always conclusive.
A polymorphism (ie, a neutral variant) is defined as a non–disease-causing change and/or a change found in at least 1% of the alleles in the population.10 Genetic variants are considered UVs when it is unclear whether they have a functional effect in the protein. These variants usually appear in less than 1% of the alleles in the population. The vast majority of the UVs described so far are missense mutations, but also mutations leading to stop codons at the end of the gene or variants in intronic regions (potentially involved in RNA splicing) have been reported. Presymptomatic testing cannot be offered in those situations and genetic counseling can only be based on the evaluation of the clinical features and the family history.
Whether patients with a UV in the BRCA1 or BRCA2 genes have different clinical features than those with a mutation had not been investigated. We retrospectively obtained the estimated probability of finding a mutation in the BRCA1 or BRCA2 genes in those patients in whom a UV in either one of these genes had been found. Second, we analyzed their clinical features and family history and compared them with those from patients with an established pathogenic mutation.
PATIENTS AND METHODS
Patients
Eligibility criteria. All probands originate from the Southeast region in the Netherlands. Eligible candidates for BRCA mutation analysis were selected when they met one of the inclusion criteria, based on the number of affected relatives (for detailed criteria description, see Table 1). We included in this study all eligible candidates who underwent complete BRCA1 and BRCA2 mutation analysis at our center and who were found to have a mutation or a UV in those genes. The study included 70 patients: 24 with a UV, and 46 with a mutation in the BRCA1 or BRCA2 genes. Seven individuals did not have a tumor but underwent DNA analysis because all affected relatives in the family were already deceased: six of them were found to have a mutation in BRCA1, and one had two UVs (one in BRCA1 and one in BRCA2).
Clinical data. Probabilities of finding a mutation in the probands were retrospectively obtained using BRCAPRO11 and Myriad II12 models (Cancergene software program; Southwestern Medical School at Dallas, Dallas, TX). Breast cancer (in probands and in relatives) refers to invasive breast cancer, both for probability estimations and for the clinical evaluation. Ductal carcinoma in situ (DCIS) was recorded in the category "other tumors."
After excluding the eight patients who harbored a UV that later was reported as a polymorphism, we recorded the clinical data and family history of 62 patients: 16 patients with a UV and 46 patients with a mutation in the BRCA1 or BRCA2 genes (Table 1).
Methods
Laboratory diagnosis. BRCA1 and BRCA2 were analyzed by denaturing high performance liquid chromatography (DHPLC), as described previously.13 Additional technical details, primers, and DHPLC elution profiles are available on request. Changes in DHPLC elution profiles were verified by standard sequence analysis. In the past, a protein truncation test was used to analyze exon 11 of BRCA1 and exons 10 and 11 of BRCA2. In those cases, the rest of the gene was fully analyzed by DHPLC later. In addition, multiplex ligation-dependent probe amplification analysis14 was performed for BRCA1 to detect large duplications or deletions.
Definition of UV. A number of epidemiologic and biochemical criteria were recorded for each UV (summarized in Table 2). Genetic variants that were found together with a mutation in the same gene were not included because it was assumed that coexistence of two pathogenic mutations would be likely to be incompatible with life and/or extremely rare. A missense variant was considered a UV based on its frequency: less than 1% in our control and patient population or when it had previously been reported as UV in the literature and/or databases. Cosegregation studies could only be performed in three of the patients; all three showed cosegregation with the disease.
Four of the genetic variants that were initially classified as UV (Ser1040Asn, Asp1420Tyr, Val2728Ile, and Lys3326stop) were reported later to be polymorphisms2,3,15; therefore, we excluded the patients having one of those variants from additional clinical analysis. After excluding those four variants, 14 different UVs remained (Table 2). All but one of the UVs are missense variants. The variant in intron 19 of the BRCA1 5313 v 25 A > C, which may result in altered splicing in the BRCA1 gene, and Lys467Arg in the BRCA2 gene, are novel genetic variants.
The UV Glu2856Ala in exon 20 of BRCA2 has a borderline frequency of 1%. We still included it among the UVs because it has been reported previously as UV in the Breast Cancer Information Core database1 and because all of the other criteria (it causes polarity change and lies on a conserved region) suggest that it may be pathogenic.
Among the 46 patients with a mutation, there were 24 different mutations identified: 12 in BRCA1 and 12 in BRCA2. The vast majority (19 mutations) are nonsense, four have an effect on splicing, and one is a missense mutation.
Statistical Analysis
Descriptive analyses were performed with the whole group: 16 UVs, 46 mutations, and subgroup analysis with BRCA2 patients alone. The tests used were independent two-sample t test for the continuous variables and 2 test for the categoric variables. Logistic regression was used as a predictive model. The mutation probability was obtained as a function of one independent variable. Receiver operating characteristic curves were used to compare these predictive models with BRCAPRO and Myriad II.16
RESULTS
All patients included in the study fulfilled one of our indication criteria to perform DNA analysis. However, highly significant differences were found when we retrospectively obtained the chance of finding a mutation in a BRCA1 or BCRA2 gene, both with BRCAPRO and Myriad II software programs, between the group with a mutation (n = 46) and those with a UV (n = 16). The mutation probability (mean ± standard deviation [SD]) was 0.297 ± 0.312 for the group with UVs versus 0.627 ± 0.315 in the group with mutations (P = .001) with BRCAPRO, and 0.124 ± 0.090 versus 0.283 ± 0.176, respectively (P = .001), with Myriad II (Table 1). From the eight patients with a polymorphism, probabilities were obtained in seven of them. In one patient, probabilities could not be estimated because of the lack of information about the family. The mean ± SD chance of a mutation was 0.175 ± 0.293 with BRCAPRO and 0.173 ± 0.165 with Myriad II. No significant differences were observed between this group and the group with UVs, whereas a statistically significant difference was found with BRCAPRO between this group and the group with a mutation (P = .001; data not shown).
After excluding the eight patients with a polymorphism, the clinical study included 16 patients with a UV: two in BRCA1, 13 in BRCA2, and one in both genes, and 46 patients with an established mutation in BRCA1 (n = 32) or in BRCA2 (n = 14). Table 1 lists the clinical data of the patients and their families from both groups.
The presence of tumors other than invasive breast and ovarian cancer among the probands was significantly more frequent in the group with UVs (four patients; 25%), compared with that among the patients with a mutation (three patients; 7%; P = .044). These tumors include DCIS (n = 2), colon (n = 1), and melanoma (n = 1) among the probands with a UV, and DCIS (n = 2) and colon (n = 1) among those with a mutation.
In both groups, the most frequent indication to perform DNA analysis was the presence of three or more affected relatives. One affected individual in the family was the indication criteria to perform DNA analysis in 31% of the patients with a UV, as opposed to only 11% in the group with a mutation.
In agreement with those results, we also found that the group with mutations showed a significantly higher number of affected relatives (mean ± SD, 4.0 ± 1.9 v 2.9 ± 1.4; P = .039), as well as higher number of tumors among relatives (4.4 ± 1.8 v 3.3 ± 1.4; P = .031). These differences were not due to a different family size, given that the total mean number of relatives was similar.
No significant differences were observed between the probands of both groups with regard to sex, prevalence of breast cancer or ovarian cancer, and age of diagnosis of the first tumor. No difference was found in either the mean age of diagnosis of breast or ovarian cancer, or in the frequency of bilateral breast cancer or of cancer in male patients in the families.
A subgroup analysis was performed for the BRCA2 patients only, which showed the same results (see P values in Table 1). Logistic regression analysis showed that the number of affected relatives (P = .024) and the number of tumors among relatives (P = .023) were significant independent univariate predictive factors of finding either a mutation or a UV. The probability of finding a mutation in BRCA1 or BRCA2 is given by the following functions: P = 1/[1 + exp(0.586 – 0.430 x number of tumors among relatives)], or P = 1/[1 + exp(0.455 – 0.442 x number of affected relatives)]. Receiver operating characteristic curves were used to compare these two models with BRCAPRO and Myriad II models. The BRCAPRO and Myriad II models seemed to be better predictive models because they had a larger area under the curve (0.787 and 0.816, respectively) than the above-mentioned models based on one variable (area under the curve, 0.686 and 0.694, respectively).
Table 2 shows the individual BRCAPRO and Myriad II mutation probabilities as well as the biochemical and epidemiologic features of each UV. BRCAPRO mutation probabilities were exceptionally high (ie, above the mean value of the mutation probability in the group with mutations) in three probands with a UV (the one with Asp1739Gly in BRCA1 and those with the UVs Ser384Thr and Ser1750Phe in BRCA2).
DISCUSSION
In this study we provide evidence that patients with a UV in BRCA1 or BRCA2 have different clinical features compared with those with a pathogenic mutation. First, when we retrospectively obtained the probabilities of finding a mutation using model analysis software, these were significantly lower in probands with a UV than in those with a mutation. An additional detailed clinical analysis revealed that the frequency of tumors other than invasive breast and ovarian cancer in the proband, the number of affected relatives, and the number of tumors among first- and second-degree relatives correlated with that difference.
Families with a mutation in BRCA2 have overlapping but different clinical features from those with a mutation in BRCA1.17 Given that most of the UVs were localized in BRCA2, whereas the majority of the mutations were localized in BRCA1 in our study, the differences observed between both groups could actually have been caused by the localization either in BRCA1 or in BRCA2, and not because of the type of genetic variant. Therefore, we performed a subanalysis only for the patients with a genetic abnormality in BRCA2, and observed similar results as those found in the whole group.
Fewer affected relatives and fewer tumors in the families of patients with a UV as compared with those with a mutation suggest that if some of the UVs are pathogenic, they may have less penetrance than the established pathogenic mutations, or need the co-occurrence of a mutation in another as yet unidentified genetic risk factor. The lower penetrance might be explained by the different nature of these genetic variants (ie, UVs are mostly missense variants, whereas most of the mutations are causing protein truncation products).
Both BRCAPRO and Myriad II are widely used programs to estimate the chance of finding a mutation in BRCA1 or BRCA2 genes.18-21 So far only BRCAPRO has been validated and seems to be more accurate than Myriad II.18 According to our results, the group of UVs significantly differ in their estimated probabilities of finding a mutation compared with the group with established pathogenic mutations with both programs.
Petersen et al5 have described a Bayesian approach that uses clinical information about families to assess causality of missense mutations in hereditary disorders. Although it is a quantitative method, when used for hereditary breast or ovarian cancer it does not allow the inclusion of all clinical data; for example, type of tumor (ie, breast or ovarian cancer) or age of diagnosis. In addition, it assumes that the penetrance does not change with age, which is not the case for hereditary breast and ovarian cancer.
One has to take into account, however, that our results are valid for the group as a whole but it is still premature to use the results for individual UVs. For this purpose, analysis of several families with the same UV would be needed to ascertain if the probability results are similar. In this respect, it is also possible that the associated probabilities of a mutation are not related to the identified UV but to a nonidentified mutation in another region of the gene, such as intronic sequences or regulatory regions, or in other as yet unidentified genes. So far, individual probabilities can help categorize these UVs into high or low probability of being pathogenic to be used as a chief guidance for medical management.
Our study shows that, in the majority of the patients, it is reasonable to ignore variants clinically and base the counseling on the family history. However, our results also show that three probands with a UV had exceptionally high BRCAPRO mutation probabilities (ie, above the mean mutation probability of the group with mutations). One of those UVs is Asp1739Gly in BRCA1. Using biochemical criteria and conservation among species, Abkevich et al7 analyzed the likelihood that a number of missense variations in BRCA1 are pathogenic; among them, one localized in the same codon (Asp1739Glu) as the UV included in our study. They also come to the conclusion that the change of Asp1739 is likely to be deleterious.
After our article was submitted, Goldgar et al8 provided strong evidence for five UVs to be either deleterious or neutral using a multifactorial likelihood-ratio model. That model includes the following parameters: co-occurrence with deleterious mutations, cosegregation data, sequence conservation, severity of substitution, and functional characteristics. Three of the UVs included in our study (Arg841Trp in BRCA1 and Tyr42Cys and Pro655Arg in BRCA2) have also been analyzed by Goldgar et al8 and are considered neutral (ie, they do not cause disease). Evidence against causality was overwhelming only for the first two UVs and came mainly from the cosegregation and co-occurrence criteria.8 Our results support these conclusions. The patients with those UVs had a low probability of mutation that was much more evident for the first two UVs. In addition, a tight probability range was observed among the patients showing Tyr42Cys, which further supports the hypothesis against this UV being pathogenic.
If we take all of these results into account, it seems evident that a definitive answer about pathogenicity cannot be obtained from a single perspective alone. An ideal scoring system should therefore include a combination of biochemical, epidemiologic, and clinical data. Finally, an international effort, which includes keeping an up-to-date genetic variant database, is also needed to ascertain the pathogenicity of individual UVs.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank Y. Detisch and B. Dols-Caanen, genetic counselors, for running the software programs with the predictive models and drawing the family pedigrees, respectively; R. Helsdingen, medical student, for recording the clinical data; and D. Marcus-Soekarman, MD, PhD, and U. Moog, MD, PhD, clinical geneticists, for the critical discussions.
NOTES
Presented as an abstract at the Family Cancer Meeting, June 5-7, 2003, International Union Against Cancer (UICC), Oklahoma City, OK, and at the Familial Cancer Meeting, May 6-7, 2004, the European School of Oncology, Madrid, Spain.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
National Human Genome Research Institute: Breast Cancer Information Core (BIC) database: An open access on-line breast cancer mutation data base. http://research.nhgri.nih.gov/bic/
Arnold N, Peper H, Bandick K, et al: Establishing a control population to screen for the occurrence of nineteen unclassified variants in the BRCA1 gene by denaturing high-performance liquid chromatography. J Chromatogr B 782:99-104, 2002
Deffenbaugh AM, Frank TS, Hoffman M, et al: Characterization of common BRCA1 and BRCA2 variants. Genet Test 6:119-121, 2002
Fleming MA, Potter JD, Ramirez CJ, et al: Understanding missense mutations in the BRCA1 gene: An evolutionary approach. Proc Natl Acad Sci U S A 100:1151-1156, 2003
Petersen GM, Parmigiani G, Thomas D: Missense mutations in disease genes: A Bayesian approach to evaluate causality. Am J Hum Genet 62:1516-1524, 1998
Petrucelli N, Lazebnik N, Huelsman KM, et al: Clinical interpretation and recommendations for patients with a variant of uncertain significance in BRCA1 or BRCA2: A survey of genetic counseling practice. Genet Test 6:107-113, 2002
Abkevich V, Zharkikh A, Deffenbaugh AM, et al: Analysis of missense variation in human BRCA1 in the context of interspecific sequence variation. J Med Genet 41:492-507, 2004
Goldgar DE, Easton DF, Deffenbaugh AM, et al: Integrated evaluation of DNA sequence variants of unknown clinical significance: Application to BRCA1 and BRCA2. Am J Hum Genet 75:535-544, 2004
Osorio A, de la Hoya M, Rodríguez-López R, et al: Loss of heterozygosity analysis at the BRCA loci in tumor samples from patients with familial breast cancer. Int J Cancer 99:305-309, 2002
Human Genome Variation Society: Nomenclature for the description of sequence variations. http://www.genomic.unimelb.edu.au/mdi/mutnomen/recs.html
Parmigiani G, Berry DA, Aguilar O: Determining carrier probabilities for breast cancer-susceptibility genes BRCA1 and BRCA2. Am J Hum Genet 62:145-158, 1998
Frank TS, Manley SA, Olopade OI, et al: Sequence analysis of BRCA1 and BRCA2: Correlation of mutations with family history and ovarian cancer risk. J Clin Oncol 16:2417-2425, 1998
Wagner T, Stoppa-Lyonnet D, Fleischmann E, et al: Denaturing high-performance liquid chromatography detects reliably BRCA1 and BRCA2 mutations. Genomics 62:369-376, 1999
Hogervorst FB, Nederlof PM, Gille JJ, et al: Large genomic deletions and duplications in the BRCA1 gene identified by a novel quantitative method. Cancer Res 63:1449-1453, 2003
Mazoyer S, Dunning AM, Serova O, et al: A polymorphic stop codon in BRCA2. Nat Genet 14:253-254, 1996
Altman DG: Practical Statistics for Medical Research. London, UK, Chapman & Hall, 1991
Verhoog LC, Berns EMJJ, Brekelmans CTM, et al: Prognostic significance of germline BRCA2 mutations in hereditary breast cancer patients. J Clin Oncol 18:119S-124S, 2000 (suppl)
Berry DA, Iversen ES Jr, Gudbjartsson DF, et al: BRCAPRO validation, sensitivity of genetic testing of BRCA1/BRCA2, and prevalence of other breast cancer susceptibility genes. J Clin Oncol 20:2701-2712, 2002
Euhus DM, Smith KC, Robinson L, et al: Pretest prediction of BRCA1 or BRCA2 mutation by risk counselors and the computer model BRCAPRO. J Natl Cancer Inst 94:844-851, 2002
Shannon KM, Lubratovich ML, Finkelstein DM, et al: Model-based predictions of BRCA1/2 mutation status in breast carcinoma patients treated at an academic medical center. Cancer 94:305-313, 2002
De la Hoya M, Diez O, Perez-Segura P, et al: Pre-test prediction models of BRCA1 or BRCA2 mutation in breast/ovarian families attending familial cancer clinics. J Med Genet 40:503-510, 2003(Encarna B. Gómez-García, )