The BRCA1 and BRCA2 genes are responsible for a high proportion of hereditary breast cancer; germline mutations in these genes confer a lifetime risk of about 70% for developing breast cancer.1 Most of these mutations are small deletions or insertions that originate a truncated protein (BIC) whose function is lost. However, in many cases, the identified germline alterations produce amino acid changes that occasionally destroy a functional domain of the protein but usually remain as unclassified variants since their role in the function of the protein is unknown; in these cases, there are several indirect tests that can help to elucidate the meaning of the variant (type of change, location, cosegregation with the disease, frequency in control population), but in most cases it is impossible to obtain a conclusive result. Any method that allowed us to distinguish if a variant of unknown significance was a deleterious mutation or a common polymorphism not associated to the disease would be extremely useful, especially for genetic counseling.
The BRCA genes act as classical tumor suppressors, being nonfunctional in cancer cells as a result of a germlime mutation following inactivation of the second allele in the tumor (the Knudson 2-hit hypothesis). Given that somatic mutations in the BRCA genes have been rarely identified,2–4 it is believed that somatic loss of the wild-type allele is a common mechanism of tumor inactivation, though other mechanisms, such as hypermethylation of the promoter, are not excluded.5–7 In addition, the BRCA genes are hypothesized to have an important role in the development of sporadic breast cancer as expression of BRCA1 is reduced in at least 30% of the cases,8 mainly by loss of heterozygosity (LOH) and, in some cases, by hypermethylation of the promoter.9–11 In the case of BRCA2, although there are fewer studies, similar behavior has been suggested.12, 13
In the present study, we analyzed 47 breast tumor samples from patients with a high-risk familial history of the disease who had been completely screened for the BRCA1 and BRCA2 genes and had shown either a germline mutation, unclassified variant or polymorphism in either of the 2 genes. Our aims were (i) to determine whether complete loss of the wild-type allele is the main mechanism of inactivation of BRCA1 and BRCA2 in tumors harboring a deleterious mutation in these genes, (ii) to analyze the indicence of LOH at the BRCA loci in familial tumors not associated with mutations in these genes and (iii) to determine whether LOH analysis could be used to analyze the pathogenicity of missense mutations, which are usually classified as variants of unknown significance.
MATERIAL AND METHODS
Paired blood and tumor samples were obtained from 47 patients who had been surgered in Fundación Jiménez Díaz and Hospital Clínico San Carlos (Madrid, Spain). These 47 patients belonged to a group of 100 high-risk breast/ovarian cancer families previously analyzed for the BRCA1 and BRCA2 genes and shown to be heterozygous for a germinal variant (mutation, polymorphism or unclassified variant) in either of the 2 genes. Criteria for selection of families were previously described.14, 15
Two hundred unrelated healthy individuals from the Blood Transfusion National Service (Madrid, Spain) were used as controls.
Previous screening for BRCA mutations in the high-risk families was performed in an index case from each family by either single-stranded conformation polymorphism (SSCP), conformation sensitive gel electrophoresis (CSGE) or denaturing gradient gel electrophoresis (DGGE) analysis and direct sequencing. Fragments were amplified using primers previously described16, 17 (M. Stratton et al., personal communication) under standard PCR conditions. Mutational screening has been described previously.14, 15
Tumor selection and DNA extraction
In 2 cases, frozen tumor samples were available, but in the remaining 45 cases material was obtained from paraffin-embedded tissues selected from the files of the Department of Pathology of Fundación Jiménez Díaz and San Carlos University Hospital. Selection of an appropriate block was made after a complete optical and immunophenotypic study. Morphologic diagnosis was made in 4 μ hematoxylin and eosin (HE)–stained sections of tumors, after 10% formaldehyde fixative (24 hr) and paraffin embedding (<70°C). Cases were classified and graded with a modified Bloom-Richardson score.18 HE staining was used to determine the proportion of tumor cells in the tissue sample. Only samples with >95% tumor cells were selected.
We used standard procedures for extracting DNA from frozen tumors and in the case of paraffin-embedded tissue, DNA was extracted using either the DNA easy tissue kit (Qiagen, Chatsworth, CA) or a modification of the protocol of Sarkar et al.19
For each tumor sample and its corresponding normal tissue, we performed PCR amplification of the fragment in which the previously identified variant in either BRCA1 or BRCA2 had been located. After amplification, the fragment was analyzed by SSCP or DGGE.20 These 2 techniques were chosen because they allowed us not only to detect the variants but also to separate the 2 alleles and determine which was the lost one in the tumor samples showing LOH.
Sixteen of the 47 tumor samples (cases 1–16) belonged to patients in whom we had previously detected a germline deleterious mutation in the BRCA genes, 10 in BRCA1 and 6 in BRCA2 (Table I). Fourteen of the 16 cases showed selective loss of the wild-type allele in the tumor when compared to normal tissue. LOH was not detected for patients 10 and 13: in case 10, the high percentage of adjacent normal tissue found in the paraffin-embedded sample could have contaminated the tumor sample, preventing detection of LOH; in case 13, this contamination was discarded. Not considering the contaminated sample, the percentage of cases showing loss of the wild-type allele was 93.3%.
Table I. LOH study in the cases harbouring a germline variant in the BRCA genes
In 13 patients (case 17–29), germline variants were classified as being of unknown significance as they represented amino acid changes whose implication in the function of the protein was unknown; 11 of them were detected in BRCA1 and 2 in BRCA2. Ten cases harboring 6 different variants did not show loss of the wild-type allele in the tumor, suggesting that these alterations were not deleterious (Fig. 1a, Table I). For variants S1040N and T1720A in BRCA1 (cases 17–20 and 27), we had previous evidence suggesting that they were not pathogenic as both were later detected in a patient harboring a deleterious mutation in BRCA2 (cases 19 and 27). Variants S1040N and N1236K in BRCA1 and S2483G and T3349A in BRCA2 were detected in 4 patients (cases 18, 22, 28 and 29) who showed retention of the wild-type allele and loss of the allele in which the variant was located, confirming that they were not deleterious. For variant S1512I in BRCA1 (case 23), as we did not have previous information that could confirm the LOH result, we performed a screening in a control population of 152 individuals in whom the percentage of heterozygotes for the variant was 1.3% (data not shown), confirming that it was a common polymorphism.
We found loss of the wild-type allele in 3 samples (cases 24–26), all harboring the same variant, G1706E in BRCA1; indeed, cases 24 and 25 were sisters, both affected with ovarian and bilateral breast cancer (Fig. 1b). This variant was highly suspicious of representing a deleterious mutation for several reasons: (i) it resulted in the substitution of a small hydrophobic amino acid by a hydrophilic one and it was localized in the COOH terminal region of the gene, which is highly conserved and supposed to be functionally important;21(ii) we had performed a study in a control population of 100 individuals in whom the variant was not detected;14 and (iii) in 1 of the 2 families in which the variant was detected, we could confirm its segregation with the disease (Fig. 2).
Eighteen cases were not associated with deleterious mutations in the BRCA genes, but 17 of them were heterozygous for a common polymorphism in the BRCA1 gene (cases 30–46) and 1 for the BRCA2 gene (case 47); the polymorphisms analyzed are described in the BIC database. Only 1 of the 18 tumors (case 30) showed LOH at the BRCA1 locus (Table I).
Sensitivity and specificity of the test
With all these results, we tried to estimate the sensitivity and specificity of the test. We took into account only the cases for which we were certain about the pathogenicity of the variant: cases 1–16, which were positive for a deleterious mutation in the BRCA genes and cases 30–48, which were not associated with BRCA and in which a common polymorphism was used for LOH analysis. For this analysis, we included 23 BRCA1-positive cases from different European and American groups, reported by Esteller et al.,22 which were not included in our series and from which 21 showed loss of the wild-type allele in the tumor. Comparing the previous information we had from these variants and the results obtained by LOH analysis, the sensitivity of the test was 92.3% (IC95% = 83.9–100%) and the specificity 94.4% (IC95% = 83.9–100%).
Although most of the deleterious alterations described in the BRCA genes are frameshift mutations that originate in a truncated protein (BIC), in many cases the detected alterations give rise to amino acid changes whose implication in the function of the protein is unknown. In these cases, it is difficult to offer genetic counseling to patients as we do not know if these familial cases are associated with the BRCA genes.
The BRCA genes act as tumor supressors, in which a germline mutation in one allele is followed by inactivation of the wild-type allele in the tumor, giving rise to a nonfunctional protein. Given that somatic mutations have been rarely described for these genes, it is beleived that a common mechanism of inactivation in the tumor is complete loss of the wild-type allele.5, 6, 23 If this is true, LOH analysis would allow us to determine if a given variant is a deleterious mutation or a common polymorphism.
To address this matter, we evaluated the rate and significance of LOH at the BRCA1 and BRCA2 loci in familial breast tumors with different germline variants in these genes. We analyzed tumor samples from 47 patients who belonged to high-risk breast and/or ovarian cancer families and who had been previously screened for the BRCA genes; 16 of them carried a deleterious mutation in either of the genes, 13 showed an amino acid change whose implication was unknown and 18 were negative for mutations but heterozygous for any common polymorphism in these genes.
Most of the tumors (93.3%) from patients harboring a germline deleterious mutation in any of the BRCA genes presented LOH in the corresponding locus, showing selective retention of the mutated allele. These results confirm that complete loss of the wild-type allele is the most frequent mechanism of inactivation in familial tumors associated with the BRCA genes. The number of cases associated with BRCA1 and BRCA2 was approximately the same (10 and 6, respectively), suggesting that these results are valid for both genes. In 2 cases, LOH was not detected; these 2 cases were revised and in one of them, high contamination with adjacent normal tissue was detected, which could explain the absence of LOH. In the second case, which carried a mutation in the BRCA2 gene, this contamination was discarded, confirming that the wild-type allele remained intact in the tumor. This suggests that in a low percentage of cases there are alternative mechanisms of inactivation, such as hypermethylation of the promoter, that should be the subject of further study. Our results are confirmed by those of Esteller et al.,22 who performed a similar LOH analysis on 23 BRCA1-positive cases, with 21 showing loss of the wild-type allele in the tumor and 1 showing hypermethylation of the promotor as a second mechanism of tumor inactivation.
With respect to the 13 cases carrying unclassified variants, we observed different results. Ten of the 13 tumors which harbored 6 different variants did not show loss of the wild-type allele, suggesting that these alterations were probably not pathogenic as they did not show the behavior observed for the majority of deleterious mutations. For the 6 variants, we had additional evidence suggesting that they were common polymorphisms, which confirmed the LOH results.
Concerning the 3 cases in which we detected positive LOH in the tumors (2 of them members of the same family and both affeceted by ovarian and bilateral breast cancer), all were carriers of the same variant, G1706E in BRCA1, which was highly suspicious of being deleterious because of the type of amino acid change, its segregation with the disease in one of the families, its absence in a control population of 100 individuals14 and its location in the BRCA1trans-activating domain.20 Loss of the wild-type allele detected for this variant, showing the same behavior as a deleterious mutation, confirmed our suspicions and was easily interpreted.
However, not all cases showing loss of the wild-type allele for a variant of unknown significance can be directly classified as a deleterious mutation because we do not know the rate of spontaneous LOH at the BRCA loci in the familial cases not associated with these genes. Establishment of this rate would allow us to know the exact probability of having a deleterious mutation when we detect loss of the wild-type allele in the tumor for a given variant. To address this issue, we analyzed 18 tumors not associated with the BRCA genes but heterozygous for a polymorphism in BRCA1 (17 cases) or BRCA2 (1 case). Only 1 of them (5.5%) showed LOH at the BRCA1 locus, which could be explained either by the presence of an undetected mutation in the gene or by the occurrence of spontaneous LOH for the BRCA1 gene in a case not associated with this gene. Our results suggest that, unlike the sporadic cases, in which the rate of LOH for the BRCA1 locus is at least 30%,8, 12, 13, 24, 25 in the familial cases not associated with the BRCA genes, the occurrence of LOH was very uncommon, confirming the idea of different mechanisms of tumorigenesis.
In summary, we have found that in familial breast cancer associated with germline mutations in either BRCA1 or BRCA2 the most common mechanism of inactivation in the tumor is complete loss of the wild-type allele. In contrast, in the familial cases not associated with these genes, the rate of spontaneous LOH, at least for BRCA1, appears to be very low. If our results are confirmed and given the high sensitivity (92.3%) and specificity (94.4%) of this test, we could use tumor LOH analysis to determine with a low margin of error if a germline variant of unknown significance in the BRCA genes in a case of familial breast/ovarian cancer is either a deleterious mutation or a common polymorphism.
We thank Ms. A. Barroso for excellent technical assistance and Mr. M. Morente from the Tumour Bank of the Spanish National Cancer Center. AO is a fellow of Comunidad de Madrid, RR-L is a fellow of Asociación Española Contra el Cáncer and AM-R is a fellow of Instituto de Salud Carlos III.