The Δ3-LR described here has not been reported previously and is the third reported mutation that leads to total exon 3 deletion of the BRCA2 gene. The c.277.317-726delinsCCAT (putative p.Ser98Glu106delinsPro) has been detected in a Swedish breast/ovarian cancer family . The c.156_157insAlu has been described in a Portuguese family as an Alu insertion at codon 52 of BRCA2 [14, 15] and is now considered to be a frequent founder mutation that is detected in nearly one third of breast/ovarian cancer families from northern/central Portugal [16, 17]. In our large series of 2058 cases that are non informative for BRCA1-2 point mutation, the prevalence of BRCA2 large rearrangements (LR) is estimated to be 3% of all BRCA2 deleterious mutations. This value is lower than the 7% to 11% reported previously [3, 7, 18]. In our six pedigrees, there were no male breast cancer cases, which contrasts with the literature . Moreover, the low frequency of BRCA2 LR could be explicable by the lower frequency of Alu sequences in the genomic locus of BRCA2 in comparison to BRCA1. Screening for BRCA2 large constitutive rearrangements should be recommended in comprehensive genetic tests, especially for families with multiple breast and/or ovarian cancer cases and families with at least one case of male breast cancer.
There is some debate about the pathogenic effect of BRCA2 exon 3 alterations due to the particular features of exon 3 in BRCA2. Exon 3 is in phase in the gene and is absent in the physiological alternative transcript (delta3-transcript) that has been detected in normal tissues, and in particular at moderate to low levels in mammary gland and prostate tissues .
We extensively analysed the balance between delta3- and full-length transcript levels, using different methods that include allele specific expression and competitive quantitative PCR by pyrosequencing which avoids misinterpretations associated with fragment analysis. We observed three classes of differential expression of the delta3-transcript.
We detected low levels of the delta3-transcript in wild-type RNA, which represent less than 10% of the total transcript levels, which could be considered to be constitutive expression of the delta3-transcript. Alternative splicing is less common for BRCA2 than for BRCA1. A splice variant lacking exon 12 of BRCA2 has been detected at higher levels in 33% of breast tumours compared to matched normal tissues, suggesting that dysregulated expression of the isoform may contribute to breast cancer progression . However, recent data indicate that exon 12 is redundant and its loss may not have an impact on the disease . We detected low levels (nearly 10%) of the alternative transcript that lacks exon 3 BRCA2, using 185 sporadic breast tumours samples and some head and neck tumours and normal tissues samples. This constitutive low level of expression suggests that the delta3-transcript is not involved in tumorigenesis in sporadic tumours. The physiological role of this low level of the delta3-transcript remains to be established.
In the case of the Δ3-LR, genomic deletion of exon 3 results in an increase of the proportion of the delta3-transcript relative to the full-length transcript, to 40-50% (by fragment analysis and confirmed by competitive QPCR). This corresponds to a total loss of exon 3 on one allele, as expected due to the large rearrangement. The same result was observed for the sample with the c.316+3delA mutation that also leads to the loss of exon 3 in the RNA. Moreover, the exclusive expression of delta3-transcript by the mutated allele, which we demonstrated by allelic discrimination analysis of the c.316+3delA mutation that is heterozygous for the c.-26 G>A polymorphism, confirms our results. Our results are similar to those reported for a mutation in intron 3, c.316+5 G>C (which was analyzed in a mini-gene system ) and the Portuguese Alu insertion , both of which result in total splicing out of exon 3 on the mutated allele. Consequently, these data, taken together, support the hypothesis of a pathogenic effect when there is exclusive synthesis of the delta3-transcript from one allele. Such functional mRNA splice variants have been demonstrated for two splice site mutations in BRCA1 gene, c.212+3A>G and c.135-6T>G. These isoforms, that lead to an in frame deletion of exon 5, were found to be expressed at increased levels in tumour cells .
The hypothesis of a causal effect of exon 3 deletion is supported by the co-segregation analysis of both families, with Δ3-LR and c.316+3delA, respectively. Co-segregation with disease of loss of exon 3, or of total exclusion of the full-length transcript, is strong, and results in high penetrance of breast cancer in mutation carriers of both families. This is also the case for the afore mentioned mutations, c.316+5G>C  found in all affected members of two families with breast and breast/ovarian cancer syndrome (S. Krieger personal communication) and the founder Portuguese mutation c.156_157insAlu [14, 16, 17]. Taken together, these data highlight the clinically relevant effects of the Δ 3 mutant allele and indicate that it is associated with a significant breast cancer risk.
Our analysis of some intron 2 variants also supports our contention that loss of exon 3 is relevant for breast cancer risk. These variants (c.68-7T>A, c.68-7delT, c.68-7_8delinsAA) have increased levels of the delta3-transcripts, which account for 23 to 30% of the total BRCA2 transcripts. There is an allelic imbalance of 30/70 between full-length transcript and delta3-transcript, that was shown with the heterozygous c.-26A>G samples. This proportion suggest that both alleles synthesise the two transcripts (full-length and delta3), with a slight imbalance for the mutated allele. However, previous reports as well as in silico predictions do not favour a causal effect of the intron 2 mutations. The c.68-7T>A variant, initially described in one patient with breast and ovarian cancer in Italy , has been widely detected in France (82 times, UMD-BRCA2 database ) sometimes with co-occurrence of two different deleterious BRCA2 mutations. The c.68-7delT has been previously described in four endometrial carcinomas with microsatellite instability, one of which had another pathogenic mutation in BRCA2 , and in three colorectal cancers with microsatellite instability . It was detected once in our cohort, 54 times in France (UMD-BRCA2 database)  with co-occurrence of deleterious BRCA2 mutations, and 4 times in the BIC database . Consequently, the slight increase in the delta3-transcript could be considered to be neutral in terms of high risk of breast/ovarian cancer. The relevance of intermediate levels of delta3-transcription for cancer risk remains to be established. In fact, in these cases, the loss of the wild-type allele in tumour tissue would lead to an allele which produces two forms of protein with and without exon 3 domain. Thus, we suggest these three distinct intron 2 mutations should remain unclassified variants until co-segregation analysis, evaluation of frequency in control population and/or functional studies are performed, which would clarify whether there is any risk associated with these mutations.
The results we obtained for the three nonsense mutations in exon 3 that we studied, and that we confirmed by the three transcript analysis methods, have not been described previously. No significant increase in the delta3/wild type transcript ratio was detected. This contrasts with the Portuguese Alu insertion mutation, which creates a stop codon 10 nucleotides after the insertion, but exclusively expresses the delta3-transcript. In fact, our results suggest that the mutated mRNAs, that contain the premature stop codon, are not eliminated or destabilized by NMD [27, 28]. Indeed, if NMD were involved, the expression levels of full-length transcript from the mutant allele would have been significantly decreased, whereas the expression levels of the delta3-trancript issued from both alleles would probably not be modified, thus leading to a global increase of the proportion of the delta3-transcript.
An additional argument towards a pathogenic effect of Δ3-LR would reside in the impact of the loss of exon 3, at the functional level, especially in tumours that exclusively express the transcript due to the genomic deletion on one allele and loss of the wild type allele. The delta3-transcript codes for a putative protein that lacks an 83 amino acids highly conserved bipartite domain, including a primary activating region (PAR, amino acids 23-60) and an auxiliary activating region (AAR, 60-105) [29, 30] whose transcriptional activity could be regulated by phosphorylation through two potential phosphorylation sites [31, 32].
Of most relevance, PAR has been shown to interact with two proteins, EMSY and PALB2 (Partner and Localizer of BRCA2), which are involved in breast carcinogenesis. EMSY has endogenous transcriptional repressor activity and contributes to DNA damage focus formation . Similar to BRCA2, EMSY relocates to double-strand break repair sites following DNA damage  and forms a complex with RAD51. EMSY can silence the transcription activation potential of the BRCA2 protein region encoded by exon 3 and thereby negatively regulate BRCA2 . Our study shows that delta3-transcript expression is not significantly increased in sporadic breast tumours, suggesting that other mechanisms are involved in non-hereditary breast tumorigenesis. To date, the EMSY gene has been found to be amplified in 7-13% of sporadic breast cancers, 17% of ovarian cancers  and 13% of pancreatic cancers . Excess of EMSY protein, that results from gene amplification, could contribute to the tumorigenesis of a substantial proportion of non-inherited sporadic breast and ovarian cancers, by silencing BRCA2 .
PALB2 is an integral component of the BRCA complex. It co-localises with BRCA2 in nuclear foci and promotes its stability in nuclear structures , which is essential for key BRCA2 nuclear caretaker functions. PALB2 also mediates the physical interaction of BRCA2 with a carboxy-terminal fragment of BRCA1 and is required for the cooperation between BRCA1 and BRCA2 in the DNA damage responses . Loss of the exon 3 region could affect the intranuclear stability of a subpopulation of BRCA2 proteins that is required for DNA repair. This hypothesis is reinforced by the demonstration that "clinically unclassified" missense variants, located in exon 3 of BRCA2 (p.Trp31Arg, p.Trp31Cys, p.Gly25Arg), display complete loss or reduction of PALB2 binding activity and inefficient DNA repair . In contrast, the p.Tyr42Cys (Y42C) variant did not differ from wild-type . The p.Tyr42Cys variant has now been excluded as a disease associated BRCA2 variant, from in vitro functional studies , and genetic and epidemiologic data, i.e. absence of co- segregation  and high frequency [42 times in France (UMD-BRCA2) and 141 times in BIC]. The low expression of delta3-transcripts we observed for this variant is in accordance with these conclusions. Taken together, our data lead to the hypothesis that disruption of the balance between BRCA2 transcripts could predispose to disease. BRCA2 protein, without the exon 3 region, may act as a dominant negative inhibitor of its transcriptional function, and could also be modified in its DNA repair activity.