RETRACTED ARTICLE: The BRCA2 variant c.68-7 T>A is associated with breast cancer

The variant BRCA2 c.68-7T>A has been demonstrated to cause variant splicing, but not invariably so [1, 2]. It has been discussed that such ‘leaky’ splicing may cause lower risk for cancer than truncating pathogenic BRCA2 variants [1], and it is demonstrated to cause low penetrance in PMS2 [3]. We have previously identified the BRCA2 c.68-7T>A in a breast cancer kindred, and we then expanded the family to show multiple cases of breast cancer cases with the variant, categorized the variant as pathogenic, and subjected the variant carriers to health care according to the accepted standard [4].

The outpatient genetic clinic at The Norwegian Radium Hospital, part of Oslo University Hospital, has invited breast cancer kindreds for genetic examinations and prospective follow-up of high risk patients since 1988. We have complete files of all activities and results. We examined the files for information on the pathogenicity of BRCA2 c.68-7T>A. We extracted the following information from our files: Prevalence of BRCA2 c.68-7T>A in the breast cancer kindreds we have examined, segregation analysis was undertaken, and the annual incidence of cancer in female carriers of BRCA2 c.68-7T>A at prospective follow up was determined.

We have previously described our filing system holding all data obtained from the start onwards [5], with a detailed description on how patients/families were selected, examined, followed-up, as well as the results of follow-up [6]. The study was approved by the Ethical review board (ref. S02030) and by The Norwegian Data Inspectorate (ref. 2001/2988–2).

Seventeen out of 714 (2.4%, 95% confidence interval 1.4% to 3.8%) unrelated breast cancer kindreds not having another pathogenic BRCA1/2 variant were sequenced for BRCA2, and were demonstrated to have the variant BRCA2 c.68-7T>A. This was significantly more than expected when compared to both a Norwegian population prevalence (3/1588) [7], ExaC-provided non-Finnish European prevalence ([8, 9]) or Finnish prevalence (36/6594) [8, 9] (Fishers’ exact p < 0.0001 for all comparisons).

Initially, when seeing the variant for the first time in our clinic, we expanded the first family detected for segregation analysis (Fig. 1), and concluded it was actionable for clinical use. We are now aware that the variant is not concluded as actionable by all, and searched our files for what information we presently had available. Likelihood segregation analysis recently established of the family presented in Fig. 1 [10] gave an inconclusive result (likelihood ratio = 0.36). The other families did not have enough informative meioses to be subjected to segregation analysis. All available relevant information on first degree female relatives in all families are listed in Table 1. Except for one family, all female relatives with cancers known to be associated with pathogenic BRCA2 variants were either carriers of the variant or not tested. Although not being statistically conclusive, the results were not in conflict with an association between the variant and breast cancer.

Twenty-four patients were subjected to follow-up for a total of 134.4 years (with a mean of 5.6 years). Two patients were prospectively demonstrated to have breast cancer (one had synchronous contralateral carcinoma in situ), arriving at an annual incidence rate of 1.5% (95% confidence interval of 0.15% to 5.4%). This point estimate was as expected for a pathogenic BRCA2 variant, but the confidence interval overlapped the incidence rate in a general population [11]. Additionally, one patient had breast cancer at first prospective (prevalence round) examination, and one patient who did not have a prior prospectively arranged examination did demonstrate a borderline ovarian cancer at prophylactic surgery. Details are given in Table 2. Borderline ovarian cancer is commonly not considered an expression of pathogenic BRCA2 variants, and was not included in the discussion on pathogenicity below.

Annual incidence estimates based on prospective follow-up needs larger numbers of patients included, or more follow-up years [12]. We here present our limited observations, anticipating that others having similar observations may combine theirs with ours.

Retrospective segregation analysis may be confounded by additional (interacting) genetic causative mechanism(s) in the families examined, and especially so when the other affected family members are examined neither for the variant in question nor for other causative genetic variants. Also, likelihood segregation analysis may be sensitive to ascertainment biases and assumed penetrance of the variant in question [10].

The verified aberrant splicing produced by BRCA2 c.68-7T>A [1, 2] supports the notion that the variant may be pathogenic. However, the variant also allows some level of normal splicing, and such a ‘leaky’ splicing is in itself not evidence for pathogenicity, at least not with high penetrance for disease.

The advocated classification systems for pathogenicity of variants causing inherited cancer [13, 14] are based on the assumption that variants will either be normal (not associated with cancer), or have high penetrance (pathogenic). The scoring system is considering the probability for a given variant to be either normal or pathogenic: and is thus not referring to penetrance (i.e. how strong the association with disease may be, meaning the lifetime cumulative incidence for a carrier to contract cancer). High-penetrance variants are by definition infrequent, and an upper threshold of 1% allelic population prevalence for a variant to cause cancer with high penetrance is commonly used [14]. Lower-penetrance alleles may have higher population prevalence. The reported population prevalence for BRCA2 c.68-7T>A is lower than 1%, but higher than most other pathogenic variants causing cancer. This is why it is justified to more closely examine not only whether or not the BRCA2 c.68-7T>A variant is pathogenic; but also the degree of penetrance, if pathogenic.

It is well known that pathogenic variants of the same genes may have different penetrance, such as a PMS2 variant reportedly causing the recessively inherited congenital mismatch-repair disease without manifestations in monoallelic carriers [3], while another variant of the same gene causes dominantly inherited Lynch syndrome [15]. Interestingly, the former, having lower penetrance, was demonstrated to have partially aberrant splicing. We have previously reported a case with Fanconi syndrome caused by two different pathogenic BRCA2 variants, where the one variant displayed high penetrance, while the lineage in the family carrying the other variant (c.7964A>G) had no cases of breast or ovarian cancer, being consistent with possibly lower penetrance [16].

The relevant part of BRCA2 with respect to the BRCA2 c.68-7T>A causes a cryptic RNA splice site, encoding a variant with an altered protein domain that is ordinarily associated with PALB2 protein interaction. PALB2 is another gene recognized to cause breast cancer when disrupted [17]. PALB2 was not studied in our series.

Combining all the above arguments, we have demonstrated that BRCA2 c.68-7T>A is associated with familial breast cancer, to the consequence that in such families, the carriers may have increased risk for cancer. On disclosure of results of genetic testing in breast cancer kindreds, carriers of the variant should be informed that they probably have a clinically actionable pathogenic variant and referred to health care accordingly [13, 14]. It is a possibility that the examined families do have other modifying factors that could increase the penetrance of BRCA2 c.68-7T>A, and it is a recognized challenge to identify modifiers of risk for pathogenic BRCA1/2 variants [18].

We demonstrate BRCA2 c.68-7T>A to be associated with breast cancer in breast cancer kindreds based on increased incidence in the families. According to the prevalence of BRCA2 c.68-7T>A there are many carriers in the populations of this variant. Recognition of BRCA2 c.68-7T>A as disease associated will, because of its prevalence, have practical implications for how to interpret and disclose the result of genetic testing results. We have not excluded that the selected kindreds may have additional genetic factors contributing to the results, and the pathogenicity BRCA2 c.68-7T>A remains to be validated outside breast cancer kindreds.