Diagnostic Strategy for Analytical Scanning of BRCA1 Gene by Fluorescence-assisted Mismatch Analysis Using Large, Bifluorescently Labeled Amplicons¹

The identification and characterization of large and structurally complex genes involved in cancer predisposition, such as the breast and ovarian cancer predisposing genes BRCA1 and BRCA2 (1, 2), have stimulated efforts aimed at improving mutation detection in large DNA regions to enable the recognition of high-risk individuals. The great allelic heterogeneity of inactivating changes in these genes increases the complexity of the task (3).

BRCA1 is a tumor suppressor gene spanning a region of 81 kb of genomic DNA (4) and is characterized by 22 coding exons (open reading frame, 5592 nucleotides) including exon 11, which represents 61.3% of the coding sequence (3426 nucleotides). Introns contain a large number of repetitive sequences of the Alu family, amounting to 41.5% of the total gene sequence. Whereas linkage analysis indicates an 81% contribution of the BRCA1 gene to the pathogenesis of breast/ovarian cancers at high risk of genetic predisposition (i.e., families with at least four breast cancer cases and one case of ovarian cancer), BRCA1 mutations were detected in 54.8% of these families, using direct sequencing or different mutation scanning methods (5). This result indicates a sensitivity of detection of 68%. Several studies using a variety of methods also suggest a 30% discrepancy between the expected and observed contributions of BRCA1 to breast/ovarian cancer predisposition (3, 6, 7). The difference may be attributable to suboptimal sensitivity of the methods used or to the presence of defects not detectable by these methods, particularly exon deletions or duplications (8).

Current diagnostic strategies of mutation detection are based on PCR amplification of target DNA. Nucleotide sequencing using the dideoxy method represents the ultimate proof of the nature of the nucleotide change. However, it usually requires large numbers of redundant reactions to reach optimal diagnostic accuracy. To reduce the costs and efforts involved in the analysis of sequence data, several scanning methods have been developed to select PCR products before sequencing (9). Chemical cleavage of mismatch is able to identify all nucleotide substitutions and microdeletions/insertions (10) by cleaving the DNA at mismatches present in heteroduplex molecules at positions corresponding to mutations or sequence variants. FAMA⁵ (11, 12) is a mutation scanning procedure that uses bifluorescent DNA heteroduplexes as substrates to detect and simultaneously localize mutations thus maximizing the reliability, sensitivity, and accuracy of mutation detection. Few diagnostic methods are able to scan large gene regions for unknown mutations; the protein truncation test scans sequence segments in the 1–2 kb range, but it is limited to coding regions or large open reading frames and detects only truncating mutations (13). The aim of the present study was to develop a reliable and cost-effective diagnostic strategy, based on FAMA, to detect all unknown mutations in large and complex genes such as BRCA1 by scanning long PCR amplicons (>1 kb) fluorescently labeled using strand-specific universal primers.

Genomic DNA was extracted from the peripheral blood leukocytes of patients affected by breast or ovarian cancers in the context of familial breast/ovarian cancer kindreds. Twelve different DNA samples with mutations in exon 11 (14), were tested blindly over the entire length of exon 11 to assess the diagnostic accuracy of FAMA. Another set of seven DNAs from unrelated breast or ovarian cancer patients with a high probability of carrying BRCA1 mutations (15) were screened along the entire coding frame, including all exon-intron junctions and the promoter region.

We have developed a two-step PCR protocol (Fig. 1,A) for efficient and cost-effective strand-specific labeling of BRCA1 amplicons based on the design of appropriate chimeric primers. These consist of two different hexadecamer sequences added onto the 5′ side of the site-specific portions of the forward and reverse PCR primers, respectively. To ensure that such “universal” extensions are compatible with high yields and high specificity of PCR reactions for amplicons in the 1.0–1.4-kb range, we found it essential to choose sequences that are rare in the human genome by using different combinations of rare octamers (16). Similarly, rare 3′-octamer positions were selected along the target DNA sequence (BRCA1; GenBank accession no. U14680) by using the PC-Rare software.⁶ In a second step, amplicons were bifluorescently labeled using fluorescent HUPs consisting (for each strand) of the same rare 16-mers to which a “GG” dinucleotide was added on the 5′-end between the strand-specific fluorophore (FAM, sense; HEX, antisense) and the hexadecamer. In this study the forward and the reverse universal primers were: 5′-6FAM-ggACCgTTAgTAgTCgAC-3′ and 5′-HEX-ggTCggATAgCTAgTCgT-3′, respectively. The sequence of the 25 primer pairs covering all exons and exon-intron boundaries of BRCA1 is shown in Table 1. The first PCR reaction was carried out in a final volume of 25 μl with 4.5 pmol of HUP-tailed primers and at the annealing temperature of 56°C for 22 of 25 amplicons. The optimal annealing temperatures for robust amplification of amplicons 11.1, 6, and 8 were 48°C, 64°C, and 68°C, respectively. In the second PCR reaction a 1:100 dilution was reamplified using fluorescently labeled HUP primers and cycling conditions of: 95°C, 1′; 48°C, 1′; and 72°C, 1′ for 30 cycles, with a final extension at 72° for 10′. Heteroduplex formation was carried out by linking onto this PCR a denaturation step at 98°C for 10 min and a reannealing step at 65°C for 1 h. PCR-products were run onto a 1% agarose gel to determine the proper volume for the chemical cleavage reaction.

The protocol was as previously described (11, 12), with the following modifications. After heteroduplex formation, ∼0.2 pmol of bichrome PCR fragments were ethanol-precipitated in a dry ice/ethanol bath, upon addition of 60 μg of glycogen carrier (Boehringer, Mannheim), and resuspended in 18 μl of water. Six μl of DNA were treated at 37°C for 45 min with 20 μl of 7 m hydroxylamine hydrochloride, and another aliquot of 6 μl was incubated for 15 min at 15°C in 0.5% osmium tetroxide/2.5% pyridine/5 mm HEPES (pH 8.0)/0.5 mm Na₂EDTA in a total volume of 19 μl. Aliquots of 7 m hydroxylamine hydrochloride (Merck) solution, titrated to pH 6.0 by addition of diethylamine (Fluka), were stored at −80°C. Osmium tetroxide [Aldrich, 4% (w/v) in water was diluted in distilled water to give a 1% stock solution, aliquoted, and stored at −80°C. Mixes were prepared on ice for the osmium tetroxide reaction and at room temperature for the hydroxylamine reaction. Chemical reactions were terminated by transferring the samples to ice and adding 200 μl of 0.3 m sodium acetate/0.1 mm Na₂EDTA (pH 5.2), and the nucleic acids were ethanol-precipitated twice. Pellets were resuspended in 50 μl of 1 m piperidine (Aldrich) and incubated at 90°C for 20 min. Fifty microliters of 0.6 m sodium acetate (pH 6.0) were added, and the nucleic acids were ethanol-precipitated and dried. Pellets were resuspended in 2 μl of a 5:1 mixture of 100% formamide/25 mm Na₂EDTA. After the addition with 0.5 μl of fluorescent-labeled size standard (GS2500; Applied Biosystems) samples were electrophoresed in a 4% acrylamide gel in a PE Applied Biosystems 377 DNA sequencer. Data were analyzed using the Genescan software.

For DNA sequencing, dye-terminator reactions were performed according to the manufacturer’s instructions using, in most cases, HUP primers (i.e., without the fluorescent label and the GG-dinucleotide).

The new labeling strategy was first tested on exon 11 (Fig. 1). Four amplicons (1269 bp, 1021 bp, 1231 bp, and 1262 bp) were designed to span a genomic region of 3896 nucleotides comprising exon 11 (3426 nucleotides) and the flanking introns (274 and 196 nucleotides of intron 10 and 11 sequences, respectively) and also providing overlaps of at least 200 nucleotides within the exon. The same strategy was applied to scan 16140 nucleotides of the BRCA1 gene (19.1% of the total gene length) comprising the 22 coding exons (5592 nucleotides), the flanking introns (at least 100 nucleotides on either side), and the promoter region. In addition to the four amplicons covering exon 11, three additional large amplicons were used to cover the promoter region and the first two exons (two overlapping amplicons of 1334 and 1351 nucleotides, respectively) and exons 6 and 7 (one amplicon of 1201 nucleotides). The 25 amplicons used for BRCA1 range in size from 350 to 1351 bp (primer sequences are shown in Table 1).This design resulted in a robust and economical protocol for amplification and bifluorescent labeling of large amplicons using uniform PCR conditions. Efficient incorporation of the 6-FAM and the HEX fluorophore was observed consistently in the upper and the lower DNA strand, respectively (examples are discussed below).

The diagnostic accuracy of this method was assessed in a blind test on twelve DNA samples from unrelated patients with reported exon 11 mutations. These samples were coded by one of the participating laboratories (Dominique Stoppa-Lyonnet, Paris, France; see Table 2) and subjected blindly to FAMA screening in another laboratory (Enrico Ricevuto, University of L’Aquila, Italy). Specific cleavage patterns of mismatches were observed in each sample at the positions within the relevant amplicons that are listed in Table 2. These cleavage patterns were then attributed correctly, without sequencing, to the reported mutations⁷ and to three different allelic variants (shown in italics in Table 2). Each of the four microdeletions/insertions involving more than one nucleotide (samples IC2, IC22, IC25, and IC27 in Table 2) showed doublets of cleavage in the C- and T-specific reactions occurring concomitantly at the corresponding position of the complementary strand. Some of these cleavages occurred at adjacent matched positions and were attributable to the extensive structural destabilization around microdeletions/insertions. Multiple cleavages were also observed with microdeletions involving only one nucleotide (IC23 and IC29). Single nucleotide substitutions resulting in six missense mutations or in three different allelic variants were all detected, and each yielded at least two fluorescent cleavage signals. In several samples, a double change was detected in the same amplicon (samples IC8, IC22, and IC23).

Fig. 2 shows the FAMA “signature” of the reported 4184del4 mutation. This microdeletion is the most frequent mutation in exon 11 and can be missed in the protein truncation test because it is close to the end of exon 11. As is often the case for microdeletions, quite strong cleavage signals were observed both on the upper and the lower strands (blue and green extra-bands, respectively, in panel A and in both the C- and the T-specific reactions). The T- and the C-specific profiles of the antisense strand are shown in panels C and D, respectively. These patterns reflect the sequence features of the region comprising the mismatches in the complementary heteroduplexes (panel B). In addition to the size calibration based on the internal size marker, the position of mismatches was often refined by using the baseline sequence information provided by the weak C- or T-specific cleavages at paired residues (see Fig. 2 C for an example). Because of the variable marker density and the decreasing gel resolution for large fragments, sizing is precise almost to the nucleotide level in the range up to the 536-nucleotide size marker, and it is in within 10 nucleotides up to the size of the largest amplicons studied. In several cases the estimated sizes of the observed cleavage fragment corresponded exactly to the position of the reported mutation or sequence variant within the relevant amplicon.

The results of our blind tests on exon 11 prompted us to evaluate the sensitivity of FAMA as a single scanning method for the entire BRCA1 gene and its applicability to the routine diagnosis of BRCA1 mutations. We therefore selected seven families with unknown status with regard to BRCA1 mutations, but with high probability of pathogenic changes at this locus (Table 3). Among these, two families (F245, five primary ovarian cancers; and F1028, six breast cancers and no ovarian cancer) with lod-scores for the BRCA1 locus of 0.48 and 1.47, respectively, had not been examined previously for BRCA1 mutations. The remaining five breast/ovarian cancer families shown in Table 3 had positive lod-scores for the BRCA1 locus ranging between 0.37 and 0.98. They had not revealed pathogenic mutations in a preliminary screen, in which the sequencing technology available at that time was used to screen directly most but not all exons (15). All 25 amplicons covering the BRCA1 gene were scanned and each of the nucleotide changes detected by FAMA was confirmed by targeted DNA sequencing. Three truncating mutations were detected in families F153, F326, and F322 (4184del4 in exon 11; 259 insA in exon 5; and 4302 C→T in exon 12, resulting in Gln1395stop, respectively). Three distinct missense mutations were detected in two other families. In family F338, two single nucleotide substitutions were found: 5215 G→A (Arg1699Gln) in exon 18 and 5468 G→A (Met1783Ile) in exon 22 (Table 3; Fig. 3). The former had once been reported in the Breast Cancer Information Core database. As shown in Fig. 3,D, this change cosegregates with cancer (patients F338.6 and F338.7), whereas the additional Met1783Ile substitution in exon 22, found in the proband F338.1, seems to be a nonpathogenic, previously unreported, allelic variant, because it is not present in two affected members of this family (F338.6 and F338.7; see Fig. 3, panelD). In addition, patient F338.6 reveals fluorescence signals indicating the presence of unreported variants that can readily be mapped to introns 17 and 18, respectively. Another family (F245) yielded a previously unreported Asp1739Val substitution of unknown functional significance. This family was originally characterized by five cases of ovarian cancer, but a female breast cancer was diagnosed (age, 45 years) while this study was in progress. All three missense mutations affect the highly conserved BRCT domain of BRCA1, but their functional relevance is unknown.

Despite the many methodological options available for mutation detection, the choice is limited, by the necessity to ensure the highest sensitivity of detection for all types of changes while keeping costs and time of analysis within a reasonable limit. The present study extends to large and complex genes, like BRCA1, the applications of the FAMA method using large PCR amplicons (up to 1.4 kb). A key feature of this diagnostic strategy is the robust and economical two-step procedure for strand-specific labeling of PCR products using chimeric oligonucleotides and universal fluorescent primers (Fig. 1). This labeling method is cost-effective, because it requires only one set of universal fluorescent primers, and allows reproducible amplification of all exons and splice sites of BRCA1 using homogeneous PCR conditions.

In the blind-test screening for changes in exon 11 (Table 2) different kinds of mutations and polymorphisms (various microdeletions/insertions involving 1 to 11 nucleotides as well as several point substitutions) were all correctly assigned, without sequencing, based on the precise sizing of cleavage products and the information obtained from the C- and T-specific cleavage profiles. Although sequence analysis represents the ultimate proof of the mutation, strand-specific fluorescent labeling combined with chemical mismatch cleavage focuses the analysis on the positions that are changed. This is in contrast with all commonly used mutation scanning methods (9), including denaturing high precision liquid chromatography (17), that use a scanning window of a few hundred bases (<500 bp) and require complete and redundant double-strand sequencing to identify mutations and/or sequence variants. High density oligonucleotide array technology, while promising (18), has not gained wide acceptance for the routine molecular diagnosis of breast cancer predispositions.

The ability to recognize mutation-specific signatures (Figs. 2 and 3) is particularly useful for the selective discrimination of multiple changes occurring within the same large amplicon (Fig. 3). Moreover, the time spent in analyzing data can be shortened by using panels of specific cleavage signatures for known mutations and polymorphisms in the DNA regions of interest.

The panel of seven high-risk breast or breast/ovarian cancer families described in Table 3 supports the conclusion that FAMA is an accurate mutation scanning method. Among the six families with breast/ovarian cancers listed in Table 3 only one (F519) did not reveal any mutation or allelic variant. One should note that this family had the lowest lod score (0.37) for linkage to BRCA1, although two ovarian cancers and four breast cancers were observed. Conversely, the other family that did not reveal BRCA1 mutations (F1028) is characterized by five female and one male breast cancer cases and no case of ovarian cancer, but displays the highest lod score for the BRCA1 locus of the seven families listed in Table 3. The lack of BRCA1 mutations or allelic variants in these families suggests the presence of large deletions, or exon duplications, as found in ∼10–15% of BRCA1 inactivations (19, 20). The presence of such rearrangements can now be tested by using multiplex fluorescent PCR (21, 22). Alternatively, other breast cancer predisposing genes, such as BRCA2, may be involved despite the positive but moderate lod-score for the BRCA1 locus.

The method described here is particularly advantageous when screening large open reading frames or genes where multiple exons and splice junctions can be encompassed by few large PCR amplicons. For example, exon 11 of BRCA1 requires only four amplicons that overlap generously (see Fig. 1). Moreover, scanning open reading frames using large amplicons also reduces the risk of one allele not being amplified because a mutation or, more likely, a polymorphic site, falls within one of the primer sequences.

Recent protocols for solid-phase chemical cleavage (23, 24, 25) allow the automation of the diagnostic strategy described here to achieve high throughput. Moreover the recent replacement of osmium tetroxide with potassium permanganate has removed the main concern about toxicity of chemical cleavage methods (25, 26). Therefore, one can envisage a variety of applications in the clinical context and in other areas of genetic analysis. For example, efficient screening of lower-risk families for breast cancer-predisposing mutations (27) can be envisioned, and analytical scanning for mutations and polymorphisms can be applied to other large and complex loci involved in the pathogenesis of genetic diseases.