Introduction
A significant proportion of inherited breast cancer is caused by mutations in the
BRCA1 and
BRCA2 tumour suppressor genes which disrupt their role in cellular DNA repair, cell cycle control, apoptosis, and tumour suppression (reviewed in [
1]). Although most mutations that are known to be pathogenic are nonsense or stop mutations and thus are predicted to cause mRNA decay or protein truncation, there are a significant number of missense variants in the
BRCA1 and
BRCA2 genes, the clinical consequences of which are unclear [
2‐
5]. It is important that the pathogenicity of these variants be understood, for the benefit of breast cancer patients and their relatives carrying such unclassified variants (UVs) and the clinicians involved in their treatment.
A wide range of approaches for the classification of
BRCA1 and
BRCA2 sequence variants have been developed, which include analysis of segregation data, sequence conservation, and protein structure [
3‐
13] and functional analysis based on a range of
in vitro assays [
6‐
11,
14]. Recently, multifactorial likelihood prediction methods have been developed to use data from a variety of sources, including histopathological features of tumours, for the clinical evaluation of UVs [
4]. Predictions using this methodology currently rely heavily on data from co-segregation in families and from co-inheritance of variants with known pathogenic mutations in the same gene. Consequently, very rare variants found in single or small families, and also probable pathogenic variants that do not reach the appropriately stringent odds of causality of 1,000:1 suggested for classification as pathogenic, remain formally unclassifiable [
4,
5]. These findings provide a strong rationale for using functional approaches to contribute additional data to support multifactorial predictions, with the caveat that such approaches are most useful for assessment of variants located in known functional domains for which
in vitro functional assays have been developed.
Building on our previous studies, we selected four UVs for additional analyses, including tumour immunohistochemistry using markers known to be associated with
BRCA1 mutation status [
12], and
in vitro assays to examine the effects on BRCA1 function. The study included one variant we had previously classified as pathogenic by multifactorial likelihood analysis (G1738R) and three variants that remained unclassified after multifactorial analysis (R1699Q, A1708V, and A1708E) [
3‐
5]. The A1708E variant acted as a positive control for functional assays since we and others [
3,
9,
13] have previously shown this variant to be functionally compromised. All four variants map to the transcriptional activation domain (TAD) and the putative interaction site for RNA polymerase II, RNA helicase A, and multiple transcription factors [
1]. We present our comparison of multifactorial likelihood predictions of pathogenicity and functional analysis of these BRCA1 variants.
Discussion
Missense amino acid substitutions in BRCA1 and BRCA2 are difficult to interpret. Multifactorial likelihood models incorporating a range of information therefore have been developed to enable predictions to be made regarding their causality. The underlying assumptions of these models are based on information derived from classical high-risk mutation carriers and predominantly truncating mutations, which currently classify variants as either high-risk or of low clinical significance/neutral. These models are therefore of clinical utility for predicting the pathogenicity of the small subset of variants that are associated with a high risk of disease. The ability of this approach to distinguish between variants that are truly neutral and those associated with a low risk of cancer is limited. Moreover, the final predictions of pathogenicity are necessarily driven by the availability of information on segregation and tumour phenotype (numbers of families and tumours available) and thus it is difficult to assess whether apparent UVs may actually be associated with a moderate risk of cancer. In addition, the model assumes that missense mutations exhibit the same characteristics (cancer risk and tumour phenotypes) as observed for truncating mutations.
The results of this study show for the first time that incorporating data on the immunohistochemical characteristics of tumours arising in carriers of these variants can sometimes improve the prediction of pathogenicity. In the case of BRCA1 A1708E, which was previously classified as a UV using multifactorial likelihood analysis based on limited data, revised multifactorial analysis yielded a posterior probability of pathogenicity of 99%. Tumour data contributed considerably to this classification, with the presence of basal cytokeratin markers and the absence of ER staining consistent with the tumour phenotype of a defective BRCA1 gene. Together with the functional data for this variant, the results provide evidence that this missense variant exhibits the features of a classical pathogenic BRCA1 mutation.
Our results also emphasise the fact that multifactorial predictions must be viewed in light of the available data and the underlying assumptions of the model, particularly the assumption that
BRCA1 missense mutation carriers are expected to display classical '
BRCA1-like' features. Our previous analysis of tumour immunohistochemical features of UV carriers suggested disparate results for two tumours from G1738R carriers, a result we confirm here with more detailed immunohistochemical analysis of basal markers of the BRCA1 mutation phenotype. While it is possible that some tumours occurring in BRCA1 carriers may be sporadic tumours not driven by abnormalities in BRCA1, the ages at onset of the two G1738R carriers (44 and 43 years) do not obviously point to the possibility of a sporadic tumour on a BRCA1 mutation background as an explanation for the observed data. It remains to be tested whether missense mutation carriers will display features similar to truncating mutations, but this will presumably require a large collaborative effort given the paucity of known pathogenic missense mutations. However, we are cautiously optimistic that at least a subset of BRCA1 missense carriers will be identifiable from histopathological characteristics. We have observed that two of the three tumours from carriers of functionally abrogated BRCA1 alleles display BRCA1-like features. Moreover, access to unpublished core data from the kConFab sample set has identified that breast tumours from carriers of RING-finger domain mutations with available pathology data (nine tumours) are all high-grade and those with known receptor status (four tumours) are all ER-negative/progesterone receptor-negative. It is important to note that our previous pathogenic classification for G1738R was not driven by immunohistopathology results but rather by the observation of loss of the wild-type allele in both BRCA1 G1738R tumours studied, data we excluded in this study due to the observation that our loss of heterozygosity data on classified variants do not support the underlying assumptions of the model. Nevertheless, the revised multifactorial data support the observation that this variant exhibits the same functional deficiencies as the A1708E variant, with a 96% posterior probability of pathogenicity. This also supports other studies that report G1738R as a founder Greek mutation [
26], with reported odds of 11,470:1 for causality from analysis of seven Greek families [
27].
Importantly, our study has highlighted the fact that the multifactorial approach will require development to assess whether a variant is associated with a low or moderate risk of cancer. Revised multifactorial analysis incorporating tumour features did not strongly support a high risk for either R1699Q or A1708V, with posterior probabilities perhaps suggestive of a moderately increased risk. Although this may appear to agree with the fact that functional assays suggest that both variants are at least partially functionally comprised, it should be noted that the posterior probabilities were driven by the sequence alignment component of the analysis. It is likely that the individual components of the multifactorial approach may have different predictive power to assess variants that are associated with a lower risk of cancer, and development of the current multifactorial model and/or alternative statistical approaches will need to be considered to test the hypothesis that variants of intermediate function may be associated with an intermediate risk of cancer.
In this study, we did not include analysis of tumour loss of heterozygosity as a component of the multifactorial analysis, as done previously [
5], since interrogation of published and unpublished data we have generated for a range of UVs has revealed increased loss of the variant compared with that expected for the underlying hypothesis used to calculate likelihood estimates, irrespective of their final classification. This suggests that the underlying assumptions for this component may not be appropriate for the classification of missense variants.
We also carried out a range of assays for the variants under study to provide novel and supporting data toward a more comprehensive description of the functional defects, if any, associated with these variants. Results from functional analyses of variants are largely supported by sequence alignment and protein modelling predictions. Sequence alignment and BRCA1 R1699Q maps to the N-terminal BRCT motif of the TAD suggest that it is deleterious, whereas modelling and
in vitro proteolytic assays indicate that the variant may affect the structure or conformation of the BRCT domain [
23]. Functional analysis of this variant suggests that it may partially compromise BRCA1 function. TAD assays in 293T and T47D mammalian cells indicated an intermediate phenotype, 56% of wild-type in 293T cells and 23% of wild-type in T47D cells (Figure
4). This intermediate activity is consistent with similar analyses performed by Vallon-Christersson and colleagues [
9], who showed approximately 20% activity of R1699Q in 293T cells. The difference in the level of activity may reflect differences in the cDNA construct and reporter system used. In other assays, R1699Q was either defective (nuclear foci formation) or indistinguishable from wild-type (centrosome amplification). Taken together, these findings suggest that this variant causes a significant yet incomplete loss of BRCA1 function.
The A1708V variant is also located in the N-terminal BRCT motif of BRCA1. Replacement of an alanine residue with a valine at this position is predicted to affect the conformation of BRCA1 by causing an incompatibility with a bend structure in the helix [
5] and it is possible that this predicted change may have functional effects. However, to date, no functional data have been presented for this variant. Here, we showed that A1708V possesses reduced transcriptional transactivation activity in two independent cell lines, similar to the R1699Q variant, and that it induces centrosome amplification. In contrast, this variant caused no significant change in nuclear foci formation.
The G1738R is located in the interval between the N- and C-terminal BRCT motifs. This induces a conformational change in BRCA1 which renders it susceptible to tryptic digestion [
23], and a recent report of transcriptional activity from assays in yeast and mammalian cells indicates pathogenicity for this variant [
28]. This is supported by results from our mammalian transcription transactivation assays, in which we observed a reduced transactivation capacity of G1738R similar to that of the deleterious variant A1708E. In addition, we have shown that nuclear foci formation and centrosome amplification are compromised to levels similar to that observed for the A1708E missense variant, commonly considered to be pathogenic [
3]. Collectively, current data from functional analysis of the G1738R variant reported here and elsewhere [
28] suggest that it exhibits functional characteristics of a true pathogenic mutation.
Our functional analysis of the A1708E TAD variant confirms our previous findings [
3] that this variant is deleterious. In addition, we have extended our previous multifactorial likelihood analysis and now provide convincing evidence for pathogenicity using this approach, with a posterior probability of pathogenicity of 99%.
The overall results from multifactorial and functional analyses highlight the known limitation of multifactorial analysis in that it was not designed to distinguish between variants that are truly neutral and those associated with a moderate or low risk of cancer. The need to address or circumvent this limitation is obvious from the increasing number of reports of low to moderate risk genetic variants contributing to breast cancer [
29‐
33] and recent evidence that rare variants of known breast cancer genes (including
BRCA1 and
BRCA2) act additively or multiplicatively to significantly increase cancer risk at the individual level [
34].
Our multifactorial analysis indicates that both R1699Q and A1708V are not
high-risk variants, but functional analyses have shown that each displays either intermediate or defective phenotype in some but not all assays (Table
2), raising the possibility that these variants may be associated with a low to moderate risk of cancer. Interestingly, whereas both variants display intermediate activity with respect to transcriptional transactivation, only R1699Q appears defective in nuclear foci formation and only A1708V induced centrosome amplification. Foci formation is known to be dependent on an intact BRCT structure [
7], and our results using trypsin sensitivity analysis of BRCT structure are consistent with this, with R1699Q displaying BRCT destabilisation and defective foci formation whereas A1708V is normal on both counts. Our results also suggest that transcriptional transactivation activity is likely to involve a region of the C terminus other than that targeted by trypsin digestion and suggest that the domains involved on foci formation and the regulation of centrosomes are likely to be independent. Given that these specific functional assays to some extent examine discrete activities of the BRCA1 protein, it would be preferable in the shorter term to use a battery of tests to assess altered function. However, given that the ultimate aim of an effective functional test is to have the minimum number of robust test results, ultimately an assay that measures general tumour suppressor activity is required. Such an assay would be expected to reflect the contribution of multiple overlapping and independent activities and also to establish whether loss of a single activity may be sufficient to disrupt its overall function as a tumour suppressor. Unfortunately, although the most reliable functional assays may improve estimates of the level or type of compromised function, they are unlikely at this stage to provide a direct translation to measures of cancer risk. An alternative study design such as large collaborative case-control studies [
35], or pooled family studies assessing risk associated with variant of similar functional capacity, may be required to provide better estimates of cancer risk associated with variants of intermediate functional phenotype.
Acknowledgements
We wish to thank Heather Thorne, Eveline Niedermayr, kConFab research nurses and staff, the heads and staff of the Family Cancer Clinics, and the Clinical Follow Up Study (funded by National Health and Medical Research Council [NHMRC] grants 145684 and 288704) for their contributions to this resource, and the many families who contribute to kConFab. We thank Amie Deffenbaugh for providing data on variant co-occurrence with mutations from the database of Myriad Genetic Laboratories, Inc. kConFab is supported by grants from the National Breast Cancer Foundation, the NHMRC, the Queensland Cancer Fund, the Cancer Councils of New South Wales, Victoria, Tasmania, and South Australia, and the Cancer Foundation of Western Australia. This research was supported by a grant from the Susan G. Komen Breast Cancer Foundation and the NHMRC. GC-T is an NHMRC Senior Principal Research Fellow, and ABS is funded by an NHMRC Career Development Award. BRH is an NHMRC Senior Research Fellow. MTSM was supported by a grant from the Cancer Council of New South Wales. FJC is supported by National Cancer Institute Breast Cancer Specialized Program in Research Excellence grant P50 CA116201, American Cancer Society award RSG-04-220-01-CCE, and the Breast Cancer Research Foundation, and DJF is supported by U.S. Army Medical Research and Materiel Command grant W81XWH-06-1-0480. SVT and DEG were supported in part by the INHERIT BRCAs programme from the Canadian Institute for Health Research and by a Subaward Agreement from the Mayo Clinic Rochester.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
PKL carried out experimental work and drafted the manuscript. ABS secured funding for the study, participated in project conception and supervision of experiments, carried out the multifactorial analysis, and contributed substantially to the drafting of the manuscript. MTSM, FJC, DJF, and BRH carried out or supervised aspects of the experimental work. SRL, SH, SA, and DB participated in the tumour characterisation component of the study. DEG and SVT provided data and advice regarding multifactorial analysis. GC-T participated in project conception and project supervision and critically reviewed the manuscript. MAB secured funding for the project, participated in project conception, supervised the functional arm of the study, and contributed substantially to the drafting of the manuscript. All authors approved the final manuscript. PKL and ABS contributed equally to this work.