Analysis of matched primary and recurrent BRCA1/2 mutation-associated tumors identifies recurrence-specific drivers

Abstract

Recurrence is a major cause of death among BRCA1/2 mutation carriers with breast (BrCa) and ovarian cancers (OvCa). Herein we perform multi-omic sequencing on 67 paired primary and recurrent BrCa and OvCa from 27 BRCA1/2 mutation carriers to identify potential recurrence-specific drivers. PARP1 amplifications are identified in recurrences (False Discovery Rate q = 0.05), and PARP1 is significantly overexpressed across primary BrCa and recurrent BrCa and OvCa, independent of amplification status. RNA sequencing analysis finds two BRCA2 isoforms, BRCA2-201/Long and BRCA2-001/Short, respectively predicted to be sensitive and insensitive to nonsense-mediated decay. BRCA2-001/Short is expressed more frequently in recurrences and associated with reduced overall survival in breast cancer (87 vs. 121 months; Hazard Ratio = 2.5 [1.18-5.5]). Loss of heterozygosity (LOH) status is discordant in 25% of patient's primary and recurrent tumors, with switching between both LOH and lack of LOH found. Our study reveals multiple potential drivers of recurrent disease in BRCA1/2 mutation-associated cancer, improving our understanding of tumor evolution and suggesting potential biomarkers.

Fig. 1. Integrated analysis of somatic mutations by whole-exome and targeted sequencing.

a Phenotype, BRCA1/2 allele-specific loss of heterozygosity (LOH), RB1 and PTEN status, and somatic mutations found in 67 paired primary and recurrent tumors (from 27 patients) sequenced by whole-exome sequencing (WES, n = 67) and high-depth targeted sequencing (n = 44). Display is limited to genes with ≥1 mutation with alternative allele fraction (AAF) ≥ 0.05 from targeted sequencing and variants with AAF ≥ 0.05 from WES. Tumors are displayed in chronological order by patient, with the primary tumor at the top and latest recurrence at the bottom. LoF loss of function, GoF gain of function. b Mutational signatures derived from WES mutations. c Hallmark Gene Sets enriched in LoF mutations for primary and recurrent tumor groups (all FDR q < 0.25). d MutSigCV results for both primary and recurrent tumors (computed independently, *FDR q < 0.05 per MutSigCV results). e Variants contributing to MutSigCV results in d, by tumor type. SNV single-nucleotide variant.

Fig. 2. Genome-wide and gene-specific copy number variation.

a Allele-specific loss of heterozygosity (LOH) of BRCA1/2 in 67 paired primary and recurrent tumors, with treatment exposure. Tumors are displayed in chronological order by patient, with the primary tumor at the top and latest recurrence at the bottom. PARPi PARP inhibitor. b Comparison of Homologous Recombination Deficiency (HRD) score for tumors with a change in LOH status from primary to recurrence (n = 14 biologically independent samples, comprising 2 samples/patient from 7 patients). For one patient (Patient 6), who had multiple recurrences with LOH, the first recurrence is displayed. Boxplot elements are as follows: median, center line; box limits, first and third quartiles (spanning the IQR, interquartile range); whiskers, 1.5× IQR in each direction; outliers plotted individually. c GISTIC qplot for 90% confidence interval amplifications in all primary tumors (breast and ovarian). d GISTIC qplot for 90% confidence interval amplifications in all recurrences (breast and ovarian). For c and d, all highlighted genes have residual q ≤ 0.05. e PARP1 copy number by tumor in primary/recurrent cohort. Groupwise differences in average copy number were determined by Kruskal–Wallis test, followed by Dunn’s test with Bonferroni correction (α = 0.05, **p < 0.01). NS not significant. f Segments of copy number gains and amplifications encompassing PARP1, by patient. For e and f, total copy number (Sequenza) was binned as follows: Deletion, CN = 0; Loss, CN = 1; Neutral, CN = 2–3; Gain, CN = 4–5; Amplification, CN ≥ 6. CNV copy number variation. g Hallmark Gene Sets enriched in primary and recurrent tumor gains and amplifications (CN ≥ 4, top) vs. losses and deletions (CN ≤ 1, bottom) (all FDR q < 0.25 per gene set enrichment analysis).

Fig. 3. Global transcriptomic programs detected by RNA sequencing.

a 66 samples used for RNA sequencing. Tumors are displayed in chronological order by patient, with the primary tumor at the top and latest recurrence at the bottom. Patients 28–31 had recurrent tumors only. Patients 32–43 were BRCA1/2 mutation carriers with no prior history of cancer or cancer treatment; their samples are normal tissue from breast and fallopian tube. LOH BRCA1/2 allele-specific loss of heterozygosity; PARPi PARP inhibitor. b Differential gene expression in breast tumor recurrences vs. normal breast tissue. c Differential gene expression in ovarian tumor recurrences vs. normal fallopian tube tissue. For b and c, a positive log₂(fold-change) indicates genes with increased expression in recurrent tumors. Adjusted p-values were computed based on linear modeling of mean-variance trends (limma). d Hallmark Gene Sets enriched in genes with increased expression in primary and recurrent breast and ovarian tumors compared to normal breast and fallopian tube tissue (all adj. p < 0.05 per gene set enrichment analysis). NES normalized enrichment score.

Fig. 4. Gene fusions and isoform switching detected by RNA sequencing.

a Clinical characteristics of patients in which IGH-@, IGL-@, and MALAT1 fusions were identified. Tumors are displayed in chronological order by patient, with the primary tumor at the top and latest recurrence at the bottom. “1 fusion” refers to a translocation with one other gene. LOH BRCA1/2 allele-specific loss of heterozygosity; PARPi PARP inhibitor. b Example of MALAT1-IGH-@ gene fusion IGV tracks from patient 9. Junction reads (red, middle track) represent split RNA-seq reads used to map the fusion breakpoint. Spanning reads (black, bottom track) represent paired-end reads of fragments that span, but do not directly overlap, the fusion breakpoint. c Expression of total BRCA2 (all isoforms) in recurrences vs. normal samples (n = 66 biologically independent samples from 39 patients). Data are expressed as mean values +/– SD. NS not significant. Adjusted p-values were computed based on linear modeling of mean-variance trends (limma). d BRCA2 isoform usage (BRCA2 isoform expression normalized to total BRCA2 expression) in recurrences vs. normal samples; q-values computed using DEXSeq within isoformSwitchAnalyzer. e BRCA2 isoforms involved in isoform switching event. NMD nonsense-mediated decay, UTR untranslated region. f BRCA2 isoform expression by sample and group for entire RNA sequencing cohort. g Overall survival (OS) curve for patients that expressed BRCA2-001/Short in any (primary or recurrent) breast tumor compared to those that did not. Survival proportions and p-value were calculated using a Cox proportional hazards model tested for significant associations with ER status, age at diagnosis, tumor stage at diagnosis, and patient recurrent status (α = 0.05, see Methods).

Fig. 5. Validation of prior results by PARP1 and RAD51 protein expression.

a PARP1 nuclear positivity (average tumor H-score) by tumor type, for 23 tumors in tissue microarrays. NS not significant. b PARP1 nuclear positivity (average tumor H-score) by PARP1 copy number status across all tumors. For a and b, groupwise differences were assessed by Kruskal–Wallis test, followed by Dunn’s test with Bonferroni correction (α = 0.05). c Percent of RAD51+ cells in primary (n = 3) and recurrent (n = 4) breast tumors based on BRCA2 transcript usage. d Percent of RAD51+ cells in primary (n = 6) and recurrent (n = 7) ovarian tumors based on BRCA2 transcript usage. For c and d, boxplot elements are as follows: median, center line; box limits, first and third quartiles (spanning the IQR, interquartile range); whiskers, 1.5x IQR in each direction; outliers plotted individually. Groupwise differences in c and d were assessed by two-sided Wilcoxon rank sum test (α = 0.05). Tumors without RNA-seq data (n = 2) or expressing both BRCA2 isoforms (n = 1) were excluded from this analysis.

Fig. 6. Evolution of BRCA1/2 mutation-associated breast and ovarian tumors in three patients.

a Genetic and clinical features in Patient 20, a BRCA1 mutation carrier with ovarian cancer. LOH allele-specific loss of heterozygosity. b Genetic and clinical features in Patient 13, a BRCA2 mutation carrier with breast cancer. AAF alternative allele fraction. c Genetic and clinical features in Patient 6, a BRCA1 mutation carrier with breast cancer. For a–c, display is limited to tumors sequenced for this study. Copy number losses refer to genes with total copy number of 0 or 1; copy number gains refer to genes with total copy number ≥4. Loss of function (LoF) mutations reported here are from high-depth targeted sequencing.