Article Text

Download PDFPDF
Short report
The BRCA2 polymorphic stop codon: stuff or nonsense?
  1. J E Higgs1,
  2. E F Harkness2,
  3. N L Bowers1,
  4. E Howard1,
  5. A J Wallace1,
  6. F Lalloo1,
  7. W G Newman1,
  8. D G Evans1,3
  1. 1Manchester Centre for Genomic Medicine, Manchester Academic Health Sciences Centre, Institute of Human Development, University of Manchester and Central Manchester Foundation Trust, Manchester, UK
  2. 2Centre for Imaging Sciences, Institute for Population Health, University of Manchester, Manchester, UK
  3. 3Genesis Breast Cancer Prevention Centre and Nightingale Breast Screening Centre, University Hospital of South Manchester, Manchester, UK
  1. Correspondence to Professor D G Evans, Manchester Centre for Genomic Medicine, MAHSC, St. Mary's Hospital, Oxford Road, Manchester M13 9WL, UK; gareth.evans{at}cmft.nhs.uk

Abstract

Background Despite classification of the BRCA2c.9976A>T, p.(Lys3326Ter) variant as a polymorphism, it has been associated with increased risks of pancreatic, lung, oesophageal and breast cancer.

Methods We have noticed multiple co-occurrences of the BRCA2 c.9976A>T variant with the pathogenic BRCA2c.6275_6276delTT frameshift mutation p.(Leu2092ProfsTer7) and using a cohort study have assessed if this might account for these tumour risk associations.

Results We identified 52 families with BRCA2c.6275_6276delTT, all of which occur in cis with the BRCA2c.9976A>T variant allele as demonstrated by co-segregation in all family members tested. Of 3245 breast/ovarian cancer samples sequenced for BRCA2, only 43/3245 (1.3%) carried BRCA2 c.9976A>T alone, after excluding individuals with BRCA2c.6275_6276delTT (n=22) or other BRCA1 (n=3) or BRCA2 (n=2) pathogenic mutations. The resultant frequency (1.3%) after removal of co-occurring mutations is lower than the 1.7% and 1.67% frequencies from two control populations for BRCA2 c.9976A>T, but similar to the 1.39% seen in the Exome Aggregation Consortium database. We did not identify increased frequencies of oesophageal, pancreatic or lung cancer in families with just BRCA2 c.9976A>T using person-years at risk analysis.

Conclusions It is likely that the previous associations of increased cancer risks due to BRCA2c.9976A>T represent reporting bias and are contributed to because the variant is in LD with BRCA2c.6275_6276delTT.

  • Cancer: breast
  • Cancer: lung
  • Cancer: oesophageal
  • BRCA2 polymorphic stop codon
  • Cancer: pancreatic

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

In 1996, 12 months after the cloning of the BRCA2 gene,1 Mazoyer and colleagues published evidence for the existence of a nonsense mutation c.9976A>T,p.(Lys3326Ter) in the last exon (27) leading to loss of the terminal 93 amino acids, including the potentially important Granin domain.2 However, this initial report noted that a frameshift mutation c.6275_6276delTT, p.(Leu2092ProfsTer7) was present on the same BRCA2 allele in one family. They eventually identified seven families with co-occurrence of the c.6275_6276delTT and c.9976A>T, but the c.9976A>T variant was not associated with breast or ovarian cancer in case–control studies being present in 7/233 controls and 7/242 breast cancer and 7/361 ovarian cancer cases of UK origin. However, since that time, potential associations with an increased risk of familial pancreatic cancer (OR 4.4),3 oesophageal cancer (OR 6.0)4 and lung cancer (OR 1.83)5 have been reported. Furthermore, the recent iCOGs analysis in 2013 showed a highly significant 1.39-fold relative risk (RR) for the c.9976A>T variant and breast cancer risk.6 As we have documented the potential increased risk of other cancers associated with c.9976A>T for many years in our diagnostic mutation reports for BRCA2 and given the recent publication of increased risks of breast and lung cancer, we felt it was important to fully assess the potential impact of carrying the variant with or without other pathogenic BRCA1/2 mutations.

Methods

Mutation analysis

DNA extracted from lymphocytes from EDTA blood samples were either Sanger sequenced (1999–2012)6 ,7 or next-generation sequenced (2013–) for all coding exons of BRCA2 and intron–exon boundaries. Samples also underwent Multiple Ligation-dependent Probe Amplification to test for large rearrangements involving BRCA1 and BRCA2. Samples were submitted from affected individuals with either breast cancer, ovarian cancer or both cancers from families in North West England (Manchester region and Merseyside) and from similarly ascertained individuals from centres in Finland, Germany and the Irish Republic.

Assessment of cancer risk

Person-years at risk analysis was carried out using data from Public Health England, National Cancer Intelligence Network (NCIN), Knowledge and Intelligence Team, North West. Person-years describes the accumulated amount of age-related time that each individual in the study was followed up and was calculated in Microsoft Excel. The number of person-years was multiplied by the incidence rates of each of three cancers previously linked to c.9976A>T, from the NCIN rates, to give the expected number of cancers in the time period. Assessment started in 1980 when robust incidence rates were available. Cancers occurring prior to 1980 were excluded. Patients were censored at date of appropriate cancer or date of death, whichever was the earlier.

Results

Our laboratory has carried out primary screening of 1850 unrelated DNA samples for full assessment of the BRCA2 gene in high-risk breast/ovarian cancer families from the Manchester region of North West England as previously described.6 ,7 Forty-four samples (2.4%) carried the c.9976A>T variant, but 18 of these (41%) also carried the c.6275_6276delTT mutation and 2 further BRCA2 and 1 BRCA1 pathogenic mutations were also found on complete sequencing. Thus only 23 (1.24%) carried the c.9976A>T variant alone (table 1).

Table 1

Summary of cohorts, numbers of variants and mutations identified

A further 1576 samples from familial breast/ovarian cancer cases in the North West were screened as part of research studies where polymorphisms were not reported. This increased the number of samples with the c.6275_6276delTT mutation from 18 to 43. Subsequent genotyping revealed that the c.9976A>T variant was present in all 43 index cases and 47 relatives testing positive for c.6275_6276delTT, but in none of the 34 relatives testing negative. As such there was perfect segregation of the c.9976A>T variant in cis with the c.6275_6276delTT mutation.

In one of the families with the c.9976A>T variant co-occurrence with a different BRCA2 mutation, it was possible to assess segregation. The family with a BRCA2 c.1929delG mutation was shown to have the mutation in trans. We have exome screened 735 population controls from North West England for the c.9976A>T variant and found 13 (1.7%) instances, none of which had the c.6275_6276delTT mutation.

We have also screened 1395 samples from Liverpool (UK), Irish Republic, Finland and Germany with 26 having the c.9976A>T variant (1.9%); 4 of these also have the c.6275_6276delTT mutation (0.3%) and 2 have a concurrent BRCA1 mutation with only 20 with c.9976A>T alone (1.4%). The Exome Sequencing Project cohort in European Americans has this genotype as present in 72/4300 (1.67%), which is almost the same frequency in our cases after removing the c.6275_6276delTT haplotype or other clearly pathogenic BRCA mutations, as is the frequency seen in the Exome Aggregation Consortium of 1.39%.8 The c.6275_6276delTT accounts for 9% (43/471) of BRCA2 mutations in the Manchester families compared with 21/471 (4.5%) for c.1929delG.

A further five samples tested by our laboratory as positive controls for presymptomatic testing from other centres in Europe with the c.6275_6276delTT mutation also have the c.9976A>T variant, bringing the total to 52 independent families.

To assess the effects of the c.9976A>T variant on cancer risk, pedigrees of the index cases were assessed. In 43 families with the c.9976A>T variant and c.6275_6276delTT mutation, there were 342 mutation carriers or first-degree relatives. Of these, 258 (164 women and 94 men) were within the study period of 1980–2015, and 4 developed lung cancer (1.55%), 1 pancreatic (0.39%) and none oesophageal cancer. Person-years at risk analysis gave an expected 0.48 pancreatic cancers, giving an RR of pancreatic cancer of 2. Person-years at risk analysis estimated an expected 2.82 lung cancers (RR 1.4) and 0.55 oesophageal cancers.

In families with BRCA2 mutations, excluding the c.6275_6276delTT mutation, there were 34/3674 (0.9%) with pancreatic, 17 (0.5%) with oesophageal and 45 (1.2%) with lung cancer.

In the 23 families containing only the c.9976A>T variant, there were 168 variant carriers and untested first-degree relatives of the carriers (96 women and 72 men), in which two lung cancers (1.2%) occurred, but no pancreatic or oesophageal cancers. Person-years at risk analysis calculated an expected 0.47 pancreatic, 3.06 lung and 0.49 oesophageal cancers, which are consistent with the observed figures (table 2).

Table 2

Observed and expected oesophageal, pancreatic and lung cancers in mutation carriers and first-degree relatives (FDR) with c.6275_6276delTT and c.9976A>T alone

Discussion

This study does not support any increased tumour risk effect with the BRCA2 variant c.9976A>T alone. The increased risk of breast cancer is likely to be attributable to LD with the pathogenic BRCA2c.6275_6276delTT mutation. Previous haplotype analysis of the c.6275_6276delTT mutation in families from Belgium, the Netherlands, Sweden and the UK showed a probable single haplotype.7 Given the co-occurrence of the common c.9976A>T variant, it is likely that this mutation has occurred only once on the background of the polymorphic stop codon, intimating a founder event thousands of years ago given its spread across Europe. With this mutation comprising 9% of pathogenic BRCA2 mutations, this approximates to a rate in the general population of 0.000135 or 1 in 7400 people in the Manchester region given our previous estimate of 0.0015 for all BRCA2 mutations.9 This would mean that 1 in 126 instances of c.9976A>T will have co-occurrence of the c.6275_6276delTT mutation in our population and that this is likely similarly high in most of Europe. The c.6275_6276delTT mutation is now more frequently found than our original local founder BRCA2 mutation, c.1929delG, which also occurs on a single haplotype.10 It is unsurprising therefore that the one identified in an individual with c.9976A>T was on the opposite BRCA2 allele in a patient with no evidence of Fanconi anaemia. The RR of 1.39 for breast cancer with c.9976A>T found in the iCOGs study6 did not appear to take into account co-occurrence of the c.6275_6276delTT mutation, although the study acknowledged that the variant could be ‘tagged’ to a pathogenic mutation. Taking an approximate frequency of 1.5% for BRCA2 mutations in unselected breast cancer11 in this study, there would only have to be a 5% proportion of BRCA2 mutations being c.6275_6276delTT (below the 9% in Manchester samples) translating into a frequency of 1 in 22 breast cancer cases with c.9976A>T having co-occurrence of the c.6275_6276delTT mutation. This would be sufficient alone to account for a 1.39 OR found in the iCOGs analysis6 assuming at least a 10-fold risk of breast cancer with a pathogenic BRCA2 mutation.11

The association of c.9976A>T with pancreatic cancer remains intriguing. In the original report, 8/144 (5.6%) of familial pancreatic cancer cases had the variant compared with 3/250 controls (1.2%) giving a significant OR of 4.84 (95% CI 1.27 to 18.55, p<0.01). Excluding a concurrent pathogenic BRCA2 mutation still left a significant 4.24-fold RR, although the CIs were very large. However, this was not replicated in sporadic cases with only 7/250 (2.8%) having the variant (OR 2.37, 95% CI 0.61 to 9.27). Our data do not support an increased risk of pancreatic, lung nor oesophageal cancer with c.9976A>T. Neither pancreatic nor oesophageal cancer was observed in our study group, consistent with the expected numbers calculated by person-years at risk analysis. Less lung cancers were observed than expected with person-years at risk analysis. The CIs are nonetheless wide and do include the previous estimates at the upper 95% CIs. We have previously shown that both pancreatic and oesophageal cancers occur with a fourfold incidence in individuals with a pathogenic BRCA2 mutation (RR 4.1, 95% CI 1.9 to 7.8).12

Functional assays have found that the c.9976A>T truncation is not deleterious.13 ,14 Nonetheless, a compound heterozygote for BRCA2 c.9976A>T with BRCA2 3033delAAAC frameshift mutation (c. c.2808_2811del) was found in a patient with Fanconi anaemia.15 This could be explained by the presence of another BRCA2 mutation on the allele with c.9976A>T that was not identified by the mutation detection strategy, such as a deep intronic mutation.16 However, this case was subsequently identified to have a mutation in the FANCB gene17 which explains the Fanconi anaemia phenotype.17 The fact that our own case with c.1929delG in trans with c.9976A>T also throws doubt on the likelihood that c.9976A>T was responsible for Fanconi in that family.

There is often bias in publication, where positive associations are more likely to be reported.18 Indeed, initial reports of associations are usually followed by much smaller effect sizes in subsequent studies.17 It is of note that no studies have validated the increased cancer risk reported for each cancer site to date. Associations of BRCA2 variants with cancers have been studied extensively, so it would not be surprising if some associations were found due to multiple testing generating false positive results. The high frequency of c.9976A>T makes it an attractive variant for such analyses. The question is how many negative studies have not been reported?

In conclusion, the present report does not provide any evidence for an increased risk of cancer related to the c.9976A>T polymorphic stop codon alone, and the recently reported association with breast cancer risk6 is likely to be caused by LD of the variant with c.6275_6276delTT.

Acknowledgments

We thank clinicians in Merseyside, Finland, Germany and the Irish Republic for sending samples for mutation analysis.

References

Footnotes

  • Contributors According to the definition given by the International Committee of Medical Journal Editors (ICMJE), the authors have made substantial contributions to the intellectual content, conception and design, acquisition of data and/or analysis and interpretation of data. They have participated in drafting of the manuscript and/or critical revision of the manuscript for important intellectual content. DGE is an NIHR senior investigator.

  • Competing interests None declared.

  • Ethics approval North Manchester Research (08/H1006/77) and University of Manchester ethics committees (08229).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement This manuscript nor one with substantially similar content under our authorship has been published or is being considered for publication elsewhere. The data upon which the manuscript is based can be provided upon request to the editors or their assignees.