Introduction

Constitutional pathogenic genetic variants in the BRCA1 gene confer a lifetime risk of breast and ovarian cancer of approximately 70% and 45%, respectively1. In tumour cells, BRCA1 inactivation leads to homologous recombination deficiency that can be targeted therapeutically with PARP inhibitors (PARPi)2. Identifying BRCA1 pathogenic variants is therefore of major importance to guide genetic counselling and therapeutic strategies. However, many genetic variants remain of uncertain significance, with insufficient evidence to be classified as benign (with no clinical consequences) or pathogenic (leading to cancer predisposition and PARPi sensitivity)3,4. Copy gains of one or several exons are challenging, as pathogenicity depends on the frame, size, and localization of the additional exons5,6,7,8,9. As most duplications are intragenic, either deep-intronic DNA sequencing or RNA sequencing (RNAseq) usually allows the precise succession of exons to be determined6,8. In most cases, duplicated exons are in tandem6. However, the BRCA1 sequence can also be disrupted by more complex structural variants (SVs) that cannot be properly described by standard short-read sequencing10. Long-read sequencing with Nanopore adaptive sampling has been shown to be a useful approach for accurate description of cancer-predisposing SVs8,9.

Description of genetic variants always relies on comparison between patient sequence and a reference genome assembly. The Genome Reference Consortium (GRC) released 14 versions of the GRCh37/hg19 reference genome assembly between 2009 and 2013, and 15 versions of the GRCh38/hg38 assembly between 2013 and 2022. These assemblies are widely used in medical genetic practice. In 2022, the Telomere-to-Telomere (T2T) Consortium released the first complete human genome assembly (T2T-CHM13/hs1) using long-read sequencing technologies11. This assembly includes an unprecedented description of complex regions such as centromeres (6% of the human genome) and segmental duplications (7% of the human genome)12,13. These latter are of particular interest as they promote genomic rearrangements that may have clinical consequences13. However, in actual clinical practice, no cases of T2T “Homo Sapiens 1” reference genome input have been reported.

We describe a founder pathogenic SV in six distinct French families, resulting in duplication of exons 8 to 20 of the BRCA1 gene and cancer predisposition in carriers. This event involved a highly complex region, located downstream in the genome (approximately 110 kb away from BRCA1) and incorrectly annotated in genome assemblies prior to T2T-CHM13/hs1. Characterization and pathogenicity assessment of this SV was therefore impossible when using GRCh37/hg19 or GRCh38/hg38 assemblies as reference genomes. This SV provoked various BRCA1 transcriptomic abnormalities.

Results

Patients

Eight patients from six unrelated families (numbered F1 to F6) carried a gain of copy of BRCA1 exons 8 to 20 identified in routine diagnostic setting. Extended analysis was performed in Family 1 – Patient 2 (F1-P2). She was an unaffected young woman (<30 years-old) carrying the exonic copy gain of BRCA1 previously identified in her grandmother (F1-P1) diagnosed with early-onset ovarian cancer.

Genomic characterization

Long-read sequencing in F1-P2, revealed fusion reads between BRCA1 intron 20 and NBR1 gene (breakpoint 1 [BP1]), and between BRCA1 intron 7 and CCDC200 gene (breakpoint 2 [BP2]) (Fig. 1). However, the orientations of BP1 and BP2 were inconsistent relative to each other when reads were aligned on GRCh37/hg19 or GRCh38/hg38 reference genomes (Fig. 1A). It was therefore impossible to localize the supplementary exons and to assess the pathogenicity of this SV. This was resolved when aligning on the T2T-CHM13v2.0/hs1 reference genome: both BP1 (involving BRCA1 intron 20) and BP2 (involving BRCA1 intron 7) orientations were consistent with a large intragenic insertion between BRCA1 duplicated exons (Fig. 1B). At both breakpoints, this insertion matched the reversed sequence of a large genomic segment containing NBR1 and CCDC200 genes. This conclusion was confirmed by OGM which showed that the genomic segment containing NBR1 and CCDC200 genes was duplicated, inverted, and inserted between BRCA1 duplicated exons (Fig. 1C). LiftOver annotations between T2T-CHM13v2.0/hs1 and previous reference genome assemblies showed that the complex region containing BP2 is mostly absent from GRCh37/hg19 and is inverted in GRCh38/hg38 (Fig. 1D). The BP1 and BP2 sequences were then confirmed using Sanger sequencing (Fig. S1-A).

Fig. 1: Genomic characterization in Family 1 – Patient 2 (F1-P2).
figure 1

A Long-read DNA sequencing aligned on GRCh37/hg19 and GRCh38/hg38 reference genome assemblies, and diagrams representing various structural hypothesis. Blue boxes represent non-duplicated BRCA1 exons (ex.), red boxes represent BRCA1 duplicated exons (ex.8 to ex.20). None of the hypothesis were totally consistent with long-read DNA sequencing data. BP: Breakpoint. B Long-read DNA sequencing aligned on T2T-CHM13/hs1 reference genome assembly, and diagram representing consistent structural hypothesis. C Optical Genomic Mapping (OGM) confirming general structure of the structural variant. D Rearranged region (T2T-CHM13/hs1) compared to GRCh37/hg19 and GRCh38/hg38 assemblies. Breakpoint 1 (BP1) occurred in BRCA1 intron 20 and first intron of NBR1 antisense transcript. Breakpoint 2 (BP2) occurred in BRCA1 intron 7 and in a region containing many segmental duplications annotated by Vollger and colleagues13. LiftOver with prior reference genome assemblies showed this region is totally absent from GRCh37/hg19, and incomplete and inverted in GRCh38/hg38. Figures 1-A and 1-B were prepared from BAM file visualization in Integrative Genome Viewer (IGV) software26. Annotation data in 2D Figure 1-D were extracted from UCSC Genome Browser (http://genome.ucsc.edu)27.

Full size image

We then compared these results to other patients carrying the copy gains of exons 8 to 20 in the BRCA1 gene. All eight patients, from six families (Fig. S2), unrelated to each other, shared the same breakpoints and a common BRCA1 haplotype (Fig. S1-B and S1-C). Clinical and histopathological family histories provided a combined likelihood ratio (LR) for pathogenicity of 5.96, cosegregation analysis provided a LR for pathogenicity of 326.33. The multifactorial combined LR was therefore 1943.83. Thus, with a prior probability of pathogenicity set to 50%, estimated posterior probability of pathogenicity was over 99.9% for the SV-carrying haplotype. With a prior probability of pathogenicity set to 10%, posterior odds reached 215.98, yielding an estimated posterior probability of pathogenicity over 99.5%.

Transcriptomic characterization

We additionally characterized the transcriptional consequences of this SV in F1-P2 (Figs. 2, 3). To this end, we analysed allelic ratios of 9 single nucleotide polymorphisms (SNPs) she carried on BRCA1 exonic sequences (Fig. 2A). All those SNPs were benign, with minor allele frequencies ranging from 0.33 to 0.38 in the gnomAD database. Two of these SNPs were located in exon 24, after the duplicated exons. The allelic ratio for these two SNPs was approximately 50% in genomic DNA and 100% in RNA-seq, indicating that only one allele of the end of the BRCA1 gene was transcribed (Fig. 2A). Seven other SNPs, located in the duplicated exons, had an allelic ratio of approximately 33% in genomic DNA, suggesting they were present only on the non-duplicated allele. The RNA-seq results showed an allelic ratio of 50%, indicating that only one set of duplicated exons was transcribed (Fig. 2A).

Fig. 2: Transcriptomic characterization.
figure 2

A Variant allele frequencies of Single Nucleotide Polymorphisms (SNPs) carried in BRCA1 by Family 1 – Patient 2 (F1-P2). Seven SNPs were detected in the duplicated region, all with an allelic ratio of approximately 33% in genomic DNA and approximately 50% in RNA-seq, indicating that only one set of duplicated exons was transcribed. Two SNPs were detected downstream of duplicated exons, with allelic ratios of approximately 50% in genomic DNA and approximately 100% in RNA-seq, indicating that only one allele of the end of the gene is transcribed. B Coding DNA (cDNA) Sanger sequencing showing back-splicing between BRCA1 exon 20 and exon 2. C BRCA1 strand-specific short-read RNA sequencing (RNA-seq) in Family 1 – Patient 2 (F1-P2) and 12 merged controls. Depth of coverage scales are indicated on the left. D Fusion reads involving BRCA1 in long-read direct RNA sequencing (RNA-seq). Represented on rearranged breakpoint 1 (BP1). The two reads aligning on the forward strand of the rearranged region are shown at the top: both started in NBR1 antisense transcript first exon (red) and included BRCA1 exonic sequences (blue). Twenty-one reads aligned on the reverse strand of the rearranged region are shown at the bottom. They all started in BRCA1 (blue) and continued with full NBR1 gene (red) with various alternative splicing. Figures 2-C and 2-D were prepared from BAM file visualization in IGV software26.

Full size image
Fig. 3
figure 3

Founder BRCA1 rearrangement and its transcriptomic consequences.

Full size image

Moreover, short-read RNA-seq and RNA Sanger sequencing revealed an abnormal splicing junction between BRCA1 exons 20 and 2 (Fig. 2B) suggestive of a back-splicing event, likely resulting in a circular RNA (circRNA), which was not observed in the 12 control samples.

The BRCA1 gene is physiologically transcribed from the reverse strand. In F1-P2, strand-specific RNA-seq showed that exons 1 to 20 underwent a significant increase of forward-strand transcription (Fig. 2C). Long-read RNA-seq detected two reads linking NBR1 antisense non-coding RNA with forward strand BRCA1 sequences (Fig. 2D), suggesting that forward strand BRCA1 transcription occurred from the NBR1 bi-directional promoter.

Short-read RNA-seq also detected abnormal junctions between BRCA1 exon 20 and NBR1 exon 2 (both with physiological transcription orientation). Long-read RNA-seq confirmed the existence of BRCA1::NBR1 fusion transcripts with various alternative splicing (Fig. 2D). Among 65 reads spanning BRCA1 exon 20 in long-read RNA-seq, 21 reads (32%) supported fusion with NBR1 exon1 (1 read), exon 2 (19 reads), or exon 3 (1 read) (Fig. 2D).

Altogether, RNA-seq revealed various BRCA1 transcriptomic abnormalities caused by this complex SV: back-splicing, forward strand transcription, and fusion transcript with NBR1 (Fig. 3). Mostly, analysing allelic ratios of several exonic SNPs demonstrated the mono-allelic expression of the final part of BRCA1, giving additional argument for the pathogenicity of this complex SV.

Discussion

We report a complex rearrangement of the BRCA1 gene involving a duplication of exons 8–20 coupled with a duplicated and inverted insertion of a large segment, located downstream in the genome and containing NBR1 and CCDC200 genes (Fig. 3). This SV was identified in eight patients from six families and produced abnormal transcripts, leading to BRCA1 loss of function and increased cancer risk. The accurate characterization of this SV was made possible using advanced molecular technologies, including long-read Nanopore sequencing, and the alignment of sequencing data to the latest T2T-CHM13/hs1 version of the human genome.

Genomic profiling technologies are continually evolving to offer more powerful tools for genome exploration. Nanopore sequencing is a particularly valuable innovation. Its unique method — real-time detection of ionic current changes as nucleic acids pass through a nano-scale pore — allows for the reading of long to ultra-long sequences. This capability makes Nanopore sequencing especially useful for resolving complex SVs, as it can accurately reconstruct these regions whereas short-read sequencing struggles with repetitive or complex sequences. Consequently, interest in long-read DNA sequencing technologies, including Nanopore sequencing, has grown in molecular biology, with their applications in routine practice expanding accordingly14.

In this study, the use of long-read DNA sequencing was decisive for precise breakpoints characterization and ensure the absence of additional breakpoints. By confirming the global structure of the SV, OGM also provided decisive insights. Then, Sanger sequencing allowed to easily assess that all patients from all families carried the same SV. The comparison between DNA-seq and RNA-seq supported the pathogenicity of this SV by showing that, on the recombined allele, the final BRCA1 exons were not transcribed.

RNA-seq also revealed several unusual transcriptomic features. First, we discovered that this SV causes constitutional fusion transcripts between BRCA1 and the neighbouring gene NBR1. Long-read RNA-seq enabled a detailed characterization of these fusion transcripts. More than 90% of detected BRCA1::NBR1 fusion transcripts had a junction between BRCA1 exon 20 and NBR1 exon 2 (19 reads among 21 reads spanning both BRCA1 and NBR1). BRCA1 exon 20 ends with a full Lysine codon (AAG) while NBR1 exon 2 starts with a short untranslated region (UTR) containing 9 nucleotides (CCTCACAGC) before the initiating Methionine codon (ATG). Therefore, it is likely that the majority of detected BRCA1::NBR1 fusion transcripts encode an in-frame fusion protein containing: (1) amino-acids encoded by BRCA1 from its initiating codon to the end of exon 20, then (2) three amino-acids corresponding to the final part of NBR1 5’UTR (Proline-Histidine-Serine), and finally (3) the full in-frame coding sequence of NBR1 gene. This fusion protein could retain some function of the wild type NBR1 protein or, more hypothetically, gain additional function. However, it is very unlikely that it retains significant function of the wild type BRCA1 protein. Indeed, a single amino acid change in the BRCT domains of the terminal part of BRCA1 can be considered as pathogenic by the international ENIGMA consortium (for instance: c.5516 T > C / p.(Leu1839Ser), c.5513 T > A / p.(Val1838Glu), c.5509 T > C / p.(Trp1837Arg), c.5363 G > T / p.(Gly1788Val), etc.).

Second, strand-specific RNA-seq revealed a marked increase in forward-strand transcription at the BRCA1 locus. Long-read RNA-seq linked this atypical transcriptional orientation to the bi-directional promoter of NBR1.

Finally, both short-read RNA-seq and Sanger RNA sequencing identified a back-splicing event between BRCA1 exon 20 and exon 2, likely resulting in a circRNA15,16. As this class of long non-coding RNAs lacks polyadenylation (poly-A), the back-splicing event was not detected by long-read RNA sequencing, which was performed after poly-A capture. To our knowledge, this represents the first documented case of a SV inducing a back-splicing event. Globally, the integration of all techniques mentioned above underscores the importance of a multimodal approach in genomic research.

In routine clinical genetic testing, time and cost constraints, driven by the large number of patients and the need for rapid diagnostic answers, make it challenging to adopt such multimodal approach. Fortunately, the vast majority of exonic duplications can be resolved using more straightforward strategies, such as deep-intronic DNA sequencing or RNA sequencing alone. For more complex SVs, optimal strategy will depend on each case. In any case, databases and literature reviews can be decisive. For example, the SV reported here can now be readily identified in any patient through targeted Sanger sequencing of the breakpoints we have characterized. This underscores the importance of sharing such findings and experiences with the broader community. In our effort to describe this SV, the main challenge we encountered was the limited reliability of the GRCh37/hg19 and GRCh38/hg38 reference genome assemblies.

The Human Genome Project remains one of the greatest scientific achievements in the history of biology. This international project was launched in 1990 to comprehensively determine all of the base pairs composing the human genome. The commonly used GRCh37/hg19 and GRCh38/hg38 reference genomes still have limitations that can impair proper genetic diagnosis for genes located in complex regions17. Our work demonstrates that incorrectly annotated regions can also impair genetic diagnosis for distant genes located in genomic regions with no particular complexity, such as BRCA1. The T2T-CHM13/hs1 assembly provides a more complete description of challenging genomic regions11,12,13. This was crucial for accurate mapping of the structural variant reported here. Although limitations remain, which can impair, for example, the detection of alterations involving CFHR-Factor H cluster genes involved in complement disorders18, our findings emphasize the need for the latest genome builds in clinical diagnostics.

Methods

Patients

Copy gain of BRCA1 exons 8–20 were identified in routine diagnostic setting in each family by Next-Generation Sequencing performed in four distinct medical centres. Exonic copy gains were confirmed by Multiplex Ligation-dependant Probe Amplification (MLPA, MRC Holland probe mix P002-D1) on DNA extracted from blood.

All patients provided informed consent and were included in the COVAR (COsegregation of VARiants) study (NCT01689584), authorized by Ethics Committee in 2011 (Comité de protection des personnes Ile de France III, Am5677-1-2940). All procedures involving human participants were conducted in accordance with the Declaration of Helsinki and its amendments.

DNA and RNA extraction

Blood DNA was extracted with the QiaSymphony DSP DNA Midikit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions.

RNA was extracted from B lymphoblastoid cell lines established by in vitro infection with Epstein Barr Virus. Cells were treated by puromycin to inhibit nonsense mediated decay. After storage in 1 mL Trizol (Invitrogen, ref. 15596026), RNA was extracted using the standard chloroform/isopropanol procedure.

Family analysis

Haplotype determination was performed by amplifying five microsatellite regions surrounding BRCA1 gene: D17S1327 downstream BRCA1 in the genome (5’-mCTAAGGAGGTTTCTCTGGAC-3’, 5’-TTCACAACTCAAGGTAAGATAGG-3’), D17S1323 in intron 12 (5’-mTAGGAGATGGATTATTGGTG-3’, 5’-AAGCAACTTTGCAATGAGTG-3’), D17S1322 in intron 19 (5’-mGCAGGAAGCAGGAATGGAAC-3’, 5’-CTAGCCTGGGCAACAGAACGA-3’), D17S855 in intron 20 (5’-mACACAGACTTGTCCTACTGCC-3’, 5’-GGATGGCCTTTTAGAAAGTGG-3’) and D17S1185 upstream (5’-mGGTGACAGAACAAGACTCCATC-3’, GGGCACTGCTATGGTTTAGA-3’). PCR was performed with AmpliTaq GOLD DNA Polymerase according to the manufacturer’s recommendations (Applied Biosystems, ref. 4311818) for 30 cycles (Hybridization: 55 °C). Amplified DNA (2μL) was then mixed with 0.5μL 500 LIZ dye size standard (Applied Biosystem, ref. 4322682) for fragment size determination by capillary electrophoresis.

As previously described4, likelihood ratios (LR) for pathogenicity were computed from clinical and histopathological family histories19,20, as well as on cosegregation analysis using a Bayesian statistical model described by Thompson et al. and updated by Belman et al.21,22. Probability for pathogenicity was computed using the multifactorial model defined by Goldgar et al.23. As prior probabilities of pathogenicity have not been calibrated for complex structural variants3,24, we tested prior probabilities ranging from 0.5 (prior odds = 1, so the posterior odds equal the LR, making the posterior probability depend only on the LR25) to 0.1 (a conservative prior).

Short-read and long-read DNA sequencing (DNA-seq)

Short-read DNA-seq was performed on a NextSeq 500 (Illumina) after enrichment using a custom SureSelect QXT kit (Agilent), as described previously8. Mapping on GRCh37/hg19 was performed with Bowtie2.

Long-read DNA-seq was performed for the first run on a Oxford Nanopore Technologies Minion Flow Cell R9.4.1 (ref. FLO-MIN106D) after library preparation by manufacturer ligation kit (Oxford Nanopore Technologies, ref. SQK-LSK110) on 2 μg DNA, as described previously8. For the second run, a long-read DNA library was prepared using the new chemistry with the long sequencing kit SQK-LSK114 on 2 µg DNA, as per the supplier’s recommendations. The DNA library was then injected into a Minion Flow Cell R10.4.1 (ref. FLO-MIN114). Computational enrichment was performed by adaptive sequencing (GRCh37/hg19). The first run targeted coding sequences of 120 genes including BRCA1 (49 Mb)8. The second run targeted the whole long arm of chromosome 17 (84 Mb) including BRCA1. Bioinformatics analysis was performed with a custom NanoCliD pipeline (https://github.com/InstituteCurieClinicalBioinformatics/NanoCliD) including Minimap2 for alignment on GRCh37/hg19, GRCh38/hg38, or T2T-CHM13/hs1.

Optical Genomic Mapping (OGM)

Ultra-high-molecular weight DNA was isolated and purified using the Bionano Prep SP-G2 Blood and Cell Kit as per the manufacturer’s instructions. Direct DNA labelling on CTTAAG sequence was conducted according to the DLS-G2 protocol with the DLE1 enzyme.

Labelled molecules were linearized into Saphyr chip G3.3 nanochannels to allow simultaneous direct imaging on the Saphyr instrument. A de novo assembly was carried out using the Bionano serve 3.7 and Access software version 1.7.

Genomic DNA and complementary DNA (cDNA) Sanger Sequencing

Breakpoints were confirmed by DNA Sanger sequencing after PCR amplification with primers 5’-GCTGTTTGCGTTGAAGAAGT-3’ and 5’-CTGCCATTTCTTTTCACTCTGG-3’ for breakpoint 1 (BP1); and 5’-ACCCCAGCACTCCTAAGAAC-3’ and 5’-GGGACCACTATCAGCTGACT-3’ for breakpoint 2 (BP2).

For RNA Sanger sequencing, RNA was reverse-transcribed using SuperScript II reverse-transcriptase (Invitrogen, ref. 18064014) as per the manufacturer’s instructions, with 1U/μL RNAse inhibitor (Applied Biosystems, ref. N8080119) and 2.5 μM Random Hexamer Primers (Invitrogen, ref. N8080127). cDNA was then amplified using a forward primer specific to BRCA1 exon 20 (5’-AGAAACCACCAAGGTCCAAAG-3’) and a reverse primer specific to BRCA1 exon 9 (5’-GCCTTATTAACGGTATCTTCAG-3’).

PCR reactions were performed with Taq DNA Polymerase (VWR, ref. 733–1301) as per the manufacturer’s instructions over 35 “touchdown” cycles (Hybridization: 58 °C x2; 57 °C x2; 56 °C x2; 55 °C x3; 54 °C x3; 53 °C x; 52 °C x4; 51 °C x5; 50 °C x10). Sequencing reactions were performed using Big Dye Terminator as per the manufacturer’s instructions (ThermoFisher, ref. 4337452).

Strand-specific short read RNA sequencing (RNA-seq)

Strand-specific RNA-seq was performed on a NextSeq 500 (Illumina) after library preparation with custom SureSelect XT HS2 RNA probes (Agilent). We followed the manufacturer’s protocol for strand-specific library preparation. Briefly, after initial preparation and fragmentation of 200 ng RNA, first-strand and second-strand cDNA were synthesized in two distinct steps with two distinct mixes. The second-strand cDNA mix contained dUTPs for specific second-strand marking. Reads were mapped on GRCh37/hg19 using STAR.

The sequencing depths of forward-strand and reverse-strand transcripts were compared to a merged bam file containing data from 12 distinct controls. These controls were patients suspected of carrying a genetic variant causing a splicing defect in a gene involved in paediatric cancer predisposition (n = 2), ataxia-telangiectasia or ataxia-telangiectasia-like disorders (n = 3), or digestive cancer predisposition (n = 7). All controls had provided informed consent for genetic analysis for diagnostic and research purposes.

Long-read direct RNA-seq

Long-read RNA-seq libraries were prepared from 1 µg of total RNA using the Oxford Nanopore Direct RNA Sequencing Kit (ref. SQK-RNA004). After an initial hybridization step, polyadenylated (polyA) messenger RNAs were captured, reverse-transcribed, and sequencing adaptors were ligated. Sequencing was performed using PromethION flowcell RNA (Oxford Nanopore Technologies, ref. FLO-PRO004RA) and reads were mapped to GRCh37/hg19 using Minimap2. For Fig. 2D design, alignments were displayed in Integrative Genomics Viewers (IGV 2.15.4) software, and fusion reads spanning BRCA1 were exported for manual reconstruction on BP1 structure.

Reference transcripts

Represented transcripts correspond to NM_007294.4/ENST00000357654.9 for BRCA1, ENST00000657841.1 for NBR2, NM_005899.5/ENST00000590996.6 for NBR1, NM_145041.4/ENST00000612339.4 for TMEM160A, NM_001363254.2/ENST00000636331.2 for CCDC200, and ENST00000635600.1 for NBR1 antisense transcript AC060780.1 (sharing the same first exon as NR_110868/ LOC101929767).