Article Text
Abstract
Background For individuals with ovarian cancer (OC), therapy options mainly depend on BRCA1/2 germline status. What is the prevalence of deleterious somatic variants, that is, does genetic tumour testing identify subgroups of individuals who also might benefit from targeted therapy?
Methods Paired analysis of tumour-derived versus blood-derived DNA to determine the prevalence of deleterious somatic variants in OC predisposition genes (ATM, BRCA1/2, BRIP1, MSH2/6, PALB2, RAD51C/D and TP53) and the PIK3CA and PTEN genes in individuals with OC (AGO-TR1 study, NCT02222883). Results were complemented by BRCA1, PALB2 and RAD51C promoter methylation analyses and stratified by histological subtype; 473 individuals were included.
Results The combined analyses revealed that deleterious germline variants in established OC predisposition genes (all: 125/473, 26.4%; BRCA1/2: 97/473, 20.5%), deleterious somatic variants in established OC predisposition genes excluding TP53 (all: 39/473, 8.2%; BRCA1/2: 30/473, 6.3%) and promoter methylation (all: 67/473, 14.2%; BRCA1: 57/473, 12.1%; RAD51C: 10/473, 2.1%; PALB2: 0/473) were mutually exclusive, with a few exceptions. The same holds true for deleterious somatic PIK3CA and/or PTEN variants (33/473, 7.0%) found to be enriched in endometrioid and clear cell OC (16/35, 45.7%); 84.3 % of the deleterious single-nucleotide/indel germline variants in established OC predisposition genes showed significantly higher variant fractions (VFs) in the tumour-derived versus blood-derived DNA, indicating a loss of the wild-type alleles.
Conclusion Tumour sequencing of the BRCA1, BRCA2, PIK3CA and PTEN genes along with BRCA1 and RAD51C promoter methylation analyses identified large subgroups of germline mutation-negative individuals who may be addressed in interventional studies using PARP or PI3K/AKT/mTOR inhibitors.
Trial registration number NCT02222883
- ovarian cancer
- targeted therapy
- methylation
- somatic variant
- germline variant
Statistics from Altmetric.com
Introduction
According to the Global Burden of Disease Cancer Collaboration,1 worldwide, approximately 254 000 women are diagnosed with ovarian cancer (OC), and 165 000 women die from the disease every year. Despite improvements in therapy, the prognosis for OC is still poor, with an average 5-year survival rate of 47.4% (National Cancer Institute, Cancer Stat Facts: Ovarian Cancer, https://web.archive.org/web/20190201115803/https://seer.cancer.gov/statfacts/html/ovary.html). The genetic characterisation of the tumour is important for tailored therapies. The interest in genetic tumour testing for this entity has strongly increased, particularly since the approval of PARP inhibitors for the treatment of recurrent OC in carriers of deleterious somatic or germline BRCA1/2 variants by the European Medicines Agency in Europe and the Food and Drug Administration in the USA some years ago.2 3
The BRCA1/2 genes are critical in the process of homologous recombination (HR) repair of double-strand DNA breaks. Several studies using gene panel analyses for the detection of pathogenic germline variants in individuals with OC have been published in recent years, revealing the germline status of BRCA1/2 as a valuable prognostic and/or predictive factor currently widely used for clinical decision making.4–7 Additional OC predisposition genes (ATM, BRIP1, MSH2, MSH6, PALB2, RAD51C, RAD51D and TP53) were validated by, for example, Norquist et al and Lilyquist et al in cohorts sufficiently large to make reliable statements.8 9 Disease associations for these genes have been clinically validated using the ClinGen clinical validity framework.10 While associations for ATM and PALB2 are not firmly established and penetrances for genes like RAD51C/D and BRIP1 still have to be investigated in different populations, we defined these eight genes (in addition to BRCA1/2) as ‘established OC predisposition genes’ for this investigation. Data on the prevalence of deleterious somatic alterations in additional OC predisposition genes are still sparse, and the clinical relevance of these findings is a matter of debate.4 9 11–15
It is suggested that the heterozygous germline inactivation of OC predisposition genes may be accompanied by a somatic inactivation of the second allele by another deleterious variant, loss of heterozygosity (LOH) or promoter methylation, resulting in an HR deficiency and limited DNA repair capacity of the tumour cells. However, it has recently been proposed that BRCA1 promoter hypermethylation rarely occurs in individuals with an underlying deleterious BRCA1/2 germline variant, although this phenomenon has not been studied extensively, and the evidence from individual studies is limited.16 Therefore, this investigation was initiated to determine whether additional tumour testing identifies a sufficiently large subgroup of individuals with deleterious somatic variants who may be addressed in upcoming interventional studies. We conducted a paired next-generation sequencing (NGS) analysis of blood-derived and tumour-derived DNA from formalin-fixed paraffin-embedded (FFPE) tumour samples obtained from 473 OC individuals enrolled in the observational AGO-TR1 study.17 18 This investigation was complemented by promoter methylation analyses of the BRCA1, PALB2 and RAD51C genes, which were the genes with the most frequent deleterious germline variants in the AGO-TR1 study sample. The analysis of BRCA2 promoter methylation was omitted based on previous study results.19–21 In addition, we investigated the occurrence of deleterious somatic PIK3CA and PTEN variants and stratified our findings by histological subtypes.
Study sample and methods
Study sample
A total of 523 consecutive individuals with invasive epithelial OC were recruited and counselled in 20 centers of the Arbeitsgemeinschaft Gynaekologische Onkologie (AGO), Germany. All individuals were older than 18 years and provided written informed consent prior to enrolment in the trial. Demographic data, disease characteristics, family history and medical history were documented as described previously.17 FFPE tumour samples were available from 496 of 523 individuals enrolled in the AGO-TR1 study.
Next-generation sequencing
All DNA samples were centrally analysed (Center for Hereditary Breast and Ovarian Cancer, University Hospital Cologne, Germany) by targeted NGS covering the entire coding regions and exon-flanking sequences (±15 nt) of 27 genes (ATM, NM_000051.3; BARD1, NM_000465.3; BRCA1, NM_007294.3; BRCA2, NM_000059.3; BRIP1, NM_032043.2; BUB1B, NM_001211.5; CDH1, NM_004360.4; CHEK1, NM_001330427.1; CHEK2, NM_007194.3; FAM175A, NM_139076.2; FANCM, NM_020937.3; MAP3K1, NM_005921.1; MLH1, NM_000249.3; MRE11A, NM_005591.3; MSH2, NM_000251.2; MSH6, NM_000179.2; NBN, NM_002485.4; PALB2, NM_024675.3; PIK3CA, NM_006218.3; PMS2, NM_000535.6; PTEN, NM_000314.6; RAD50, NM_005732.3; RAD51C, NM_058216.2; RAD51D, NM_002878.3; STK11, NM_000455.4; TP53, NM_000546.5; and XRCC2, NM_005431.1). The methodologies and the results of the germline analysis of 25 genes (excluding PIK3CA and MAP3K1) were described previously.17 The hybridisation capture-based NGS method (Agilent, Santa Clara, California, USA) was suitable for the analysis of DNA derived from either blood or FFPE tumour samples (Agilent SureSelect XT protocol optimised for 200 ng of genomic DNA). For the isolation of DNA from FFPE tumour samples, H&E-stained 3 µm tissue sections were centrally investigated (Institute of Pathology, University Hospital Bonn, Germany); that is, tumour areas containing >80% tumour nuclei were chosen for DNA isolation. DNA isolation from FFPE tumour samples was conducted using standard procedures (see supplementary Materials and Methods). Sequencing was performed on MiSeq or HiSeq4000 devices (Illumina, San Diego, California, USA). Of the 496 tumour samples, 488 samples were successfully analysed by targeted NGS with a mean read coverage of at least 100×. For these samples, the mean read coverage was 570× (range 110× –1802×). Bioinformatic analyses were carried out using the SOPHiA DDM platform (Sophia Genetics, Saint-Sulpice, Switzerland) with a minimum cut-off value of 5% for VFs.
Supplemental material
Detection of CNVs
The germline analyses covered the detection of CNVs. All 523 blood-derived DNA samples were screened for CNVs in the BRCA1/2 genes by Multiplex Ligation-dependent Probe Amplification (MLPA) using the SALSA MLPA probemixes P002 (BRCA1) and P045 (BRCA2) (MRC-Holland, Amsterdam, The Netherlands). In addition, NGS data were used to identify CNVs in non-BRCA1/2 genes using an in silico CNV detection tool incorporated in the SOPHiA DDM platform. Neither CNV calling using the in silico CNV detection tool of the SOPHiA DDM platform nor MLPA-analysis were suitable for FFPE-derived DNA samples. Thus, CNV detection in tumour samples was not included in this investigation.
Variant classification
Variants were considered somatic if they were not identified in a paired germline analysis of the corresponding blood sample. Variant classification was performed as previously described.18 21 22 All genetic variants were classified using a five-tier variant classification system as proposed by the International Agency for Research on Cancer Unclassified Genetic Variants Working Group, namely, deleterious=class 5, likely deleterious=class 4, variant of uncertain significance=class 3, likely benign=class 2 and benign=class 1. Variants reported to occur in large outbred control reference groups at an allele frequency of >1% were generally considered benign. Class 4/5 variants were subsequently defined as ‘deleterious variants’.
Quantitative methylation analysis
Bisulfite conversion of 2 µg of DNA was performed using the innuCONVERT Bisulfite Basic Kit (Analytik Jena, Jena, Germany). Quantitative methylation-specific PCR assays were designed to allow specific amplification of the bisulfite-converted methylated gene promoter sequences of BRCA1, PALB2 and RAD51C. The assays were located in regions previously described to be methylated in OC23–25 (online Supplementary Figure 1). The assays and the oligonucleotide sequences are described in the online supplementary table 1. PCR was performed in 20 µL reaction volumes with PCR buffer26; per cent methylation was calculated using the ΔΔCT method.27 Triplicate measurements were carried out for each sample, and median methylation levels were computed with values ≥5% considered positive. Of the 496 tumour samples, methylation analyses for all three target genes were successfully performed for 473 samples.
Statistical analyses
Statistical analyses were conducted using SPSS Statistics V.25. Fisher’s exact test and Student’s t-test were used to calculate levels of significance, with p values <0.05 considered significant.
Results
Prevalence of deleterious germline variants in 473 consecutive individuals with OC
We focused on 473 individuals enrolled in the AGO-TR1 study for which NGS data for the tumour samples along with promoter methylation analyses of the BRCA1, PALB2 and RAD51C genes were available (CONSORT-like diagram, online Supplementary Figure 2). In this subgroup, comprising 90.4% of the overall AGO-TR1 study sample (473/523), the prevalence of deleterious germline variants was similar to that observed in the overall group, that is, deleterious germline variants in established OC genes (ATM, BRCA1, BRCA2, BRIP1, MSH2, MSH6, PALB2, RAD51C, RAD51D and TP53)8 9 were identified in 125 of 473 individuals (26.4%, online Supplementary Table 2) versus 140 of 523 individuals (26.8%) in the overall sample.17 18 In the subgroup, deleterious germline variants were most prevalent in BRCA1 and BRCA2 (97 of 473 individuals; 20.5%), followed by RAD51C (13 of 473 individuals, 2.7%) and PALB2 (4 of 473 individuals, 0.8%). Regarding all 27 genes analysed, 142 deleterious germline variants (129 single-nucleotide/indel variants and 13 germline CNVs; online Supplementary Table 2) were identified in 133 of 473 individuals (28.1%; nine individuals carried two deleterious germline variants; online Supplementary Table 2).
Detection of deleterious germline variants in the corresponding tumour samples
In the paired NGS analyses of tumour-derived versus blood-derived DNA, all 129 single-nucleotide/indel variants were also detected in the corresponding tumour samples (online Supplementary Table 2). Of the 129 single-nucleotide/indel variants, 115 variants were observed in the established OC predisposition genes (ATM, BRCA1, BRCA2, BRIP1, MSH2, MSH6, PALB2, RAD51C, RAD51D and TP53).8 9 The vast majority of the deleterious single-nucleotide/indel variants affecting the established OC predisposition genes (97/115; 84.3%) showed significantly higher VFs in the tumour-derived versus blood-derived DNA (p<0.05, Fisher’s exact Test; online Supplementary Table 2), indicating a loss of the wild-type (WT) alleles in the neoplastic tissue. The remaining 14 deleterious single-nucleotide/indel variants were observed in nine of the additional genes, and significantly increased VFs in the tumour tissue were observed for only five variants (35.7%; NBN [2/2], BUB1B [1/1], FAM175A [1/1], XRCC2 [1/1], FANCM [0/3], CHEK2 [0/2], MRE11A [0/2], CHEK1 [0/1] and RAD50 [0/1]).
Prevalence of deleterious somatic variants in 473 consecutive individuals with OC
Deleterious somatic TP53 variants (386 variants) were identified in 383 of 473 individuals (81.0%, table 1, figure 1). The majority of the deleterious somatic TP53 variants were missense variants (238/386; 61.7%; figure 1). Of note, deleterious somatic TP53 variants were most prevalent in individuals with high-grade serous OC (88.5%) and serous/papillary OC (94.7%). Excluding TP53, deleterious somatic variants predominantly affected the BRCA1 (18 variants; 18 of 473 individuals, 3.8%; table 1) and BRCA2 genes (12 variants; 12 of 473 individuals, 2.5%; table 1). Deleterious somatic variants in all other established OC predisposition genes were rare (14 variants; 12 of 473 individuals, 2.5%; 4 × ATM, 1 × BRIP1, 1 × MSH2, 4 × MSH6, 3 × PALB2, 1 × RAD51D; table 1). Among the additional genes analysed, deleterious somatic variants were most prevalent in the PIK3CA gene (26 variants; 24 of 473 individuals, 5.1%), mostly missense variants described to be activating (online Supplementary Table S3, figure 1). Deleterious somatic PTEN variants (24 variants) were identified in 16 of 473 individuals (3.4%). Deleterious somatic variants in PIK3CA and/or PTEN were most prevalent in individuals with high-grade endometrioid, low-grade endometrioid or clear cell OC (16 of 35 individuals, 45.7%; table 1). Notably, we observed the co-occurrence of deleterious somatic PIK3CA and PTEN variants in 5 of 22 individuals (22.7%) with high-grade endometrioid OC (online Supplementary Table 3). Deleterious somatic variants in all additionally tested genes were rare. Deleterious somatic MAP3K1 variants (five variants) were identified in 5 of 473 individuals (1.1%; figure 1). All other genes affect less than 1% of all individuals for each gene (figure 1; 1 × BARD1, 2 × CHEK1, 4 × FANCM, 1 × PMS2).
Promoter methylation analyses of the BRCA1, PALB2 and RAD51C genes
Promoter methylation analyses were carried out for the BRCA1, PALB2 and RAD51C genes in the 473 tumour samples. PALB2 gene promoter methylation was not detected in any of our tumour samples (online Supplementary Table S3). PALB2 promoter methylation was previously described in OC with a clear cell subtype23 (representing only 1.3% of our study sample; table 1) but not in high-grade serous carcinoma.28 Methylation of the BRCA1 gene promoter was identified in 57 of 473 individuals (12.1%; table 1). Methylation of the RAD51C gene promoter was identified in 10 of 473 individuals (2.1%; table 1).
Combined analysis of the study results
In summary, 125 of 473 individuals (26.4%) carried deleterious germline variants in the established OC predisposition genes, with LOH representing the second hit necessary for tumour development in a large proportion (84.3%) of these samples (figure 2). Deleterious somatic variants in established OC predisposition genes (excluding TP53) were identified in 39 of 473 individuals (8.2%, figure 2). Promoter methylation of the BRCA1 and RAD51C genes was observed in 57 of 473 individuals (12.1%) and 10 of 473 individuals (2.1%), respectively (figure 2). With a few exceptions, these somatic single-nucleotide/indel variants and methylation of BRCA1 and RAD51C and deleterious germline variants were mutually exclusive. The same holds true for the subgroup of 33 individuals (7.0%) carrying deleterious somatic PIK3CA and/or PTEN gene variants (figure 2).
Discussion
It is generally assumed that the heterozygous germline inactivation of cancer predisposition genes may be accompanied by a somatic inactivation of the second allele by deleterious variant, LOH or promoter methylation in the neoplastic tissue. In this large investigation of 473 individuals with OC, we demonstrated that the majority of the deleterious germline single-nucleotide/indel variants affecting established OC predisposition genes (97/115; 84.3%) showed significantly higher VFs in tumour-derived versus blood-derived DNA samples (online Supplementary Table 2), indicating the loss of the WT alleles by either large deletions (ie, LOH with copy number losses) or copy number neutral LOH.29 Regarding individuals with deleterious BRCA1/2 germline variants, only 1 of 97 individuals (1.0%, figure 2) carried a deleterious germline variant plus a different deleterious somatic variant in the BRCA2 gene (individual #92; online Supplementary Table 3). Regarding other established OC predisposition genes, only 1 of 28 individuals (3.6%, figure 2) carried a deleterious germline variant plus a different deleterious somatic variant in the MSH6 gene (individual #119; online Supplementary Table S3). The BRCA1/RAD51C promoter methylation was not found in individuals carrying a deleterious germline or somatic variant in the same gene. Thus, deleterious somatic single-nucleotide/indel variants or promoter methylation rarely represent the second hit in the development of hereditary OC.
Deleterious somatic variants in OC predisposition genes (excluding TP53) were predominantly found in the BRCA1 (18 variants; 18 of 473 individuals, 3.8%) and BRCA2 genes (12 variants; 12 of 473 individuals, 2.5%). The prevalence of deleterious somatic BRCA1/2 variants is comparable with those reported in previous studies.4 11 13 30 Deleterious somatic variants in other established OC predisposition genes (excluding TP53) were generally rare, affecting less than 1% of all individuals for each gene (figure 1). Consequently, our analyses demonstrate that the vast majority of deleterious variants found in the tumour were of germline origin. Regarding BRCA1/2, 128 deleterious variants were found in the tumour, of which 98 (76.6%) were proven germline variants. For all genes significantly associated with OC (excluding TP53), 168 deleterious variants were found in the tumour, of which 126 (75.0%) were proven germline variants. Thus, tumour testing without prior germline testing may require genetic counselling due to its potential implications for the stratification of the risk of the individual for other malignancies (ie, breast cancer) and, most importantly, the stratification of the cancer risks of family members who may benefit from predictive testing and prophylactic measures. Moreover, genetic tumour testing alone may miss deleterious germline CNVs in OC predisposition genes.31
Due to the high prevalence of deleterious germline variants in OC predisposition genes, germline testing is generally recommended in individuals with OC. We suggest that germline mutation analysis of established or proposed OC predisposition genes should be complemented with genetic tumour testing in patients without deleterious germline variants in these genes. Genetic tumour testing of OC predisposition genes, the PIK3CA and PTEN genes and BRCA1 and RAD51C promoter methylation analyses identified a large subgroup of germline mutation-negative individuals (125/348, 35.9%), which may be addressed in future interventional studies using poly ADP ribose polymerase (PARP) or phosphoinosotide 3-kinase (PI3K)/ protein kinase B (AKT)/ mechanistic target of rapamycin (mTOR) inhibitors.32 Matulonis et al 33 already described the activity of olaparib in combination with a PI3K-inhibitor. However, they focused only on high-grade serous OC. Our data suggest that treatment options with PI3K inhibitors could also be further investigated in other histological subtypes. In our series, the highest rates of deleterious somatic PIK3CA and/or PTEN variants were shown in the subgroup of individuals with endometrioid and clear cell tumours (table 1).
In summary, germline, somatic or epigenetic alterations that likely impair OC predisposition gene activity were identified in 224 of 473 individuals (47.4%, figure 2), with an even higher prevalence of (epi)genetic alterations in the subgroup of individuals with high-grade serous OC (figure 2B). Recently, the phase 3 SOLO1 trial (NCT01844986) revealed that maintenance therapy with the PARP inhibitor olaparib provided a substantial benefit with regard to progression-free survival among women with newly diagnosed advanced OC and a deleterious BRCA1/2 variant, with a 70% lower risk of disease progression or death with olaparib compared with placebo.34 The question emerges, however, whether the presentation of BRCA1/RAD51C promoter methylation or a deleterious somatic BRCA1/2 variant represents a robust predictive marker for therapy response, especially regarding PARP inhibitor therapy. The response to PARP inhibition was found to depend on HR deficiency and is not restricted to deleterious BRCA1/2 germline variants,16 and in vitro analyses suggested that BRCA1 hypermethylation and BRCA1 deleterious variants conferred the same degree of sensitivity to PARP inhibition.5 Further prospective studies will help clarify the clinical significance of these somatic (epi)mutations as biomarkers to direct targeted therapy.
Limitations of this study
Due to the mean read coverage of 570×, we used a VF cut-off value of at least 5%. Therefore, we cannot exclude the existence of low-abundance variants in the tumour samples investigated in this study.
Acknowledgments
We would like to thank Astra Zeneca Germany for the financial support of the trial.
References
Footnotes
Contributors JH and EH contributed equally to this work; study concept and design: JH, EH, ADB, RKS and PH; acquisition, analysis and interpretation of data: JH, EH, StS, AR, LR, SK, AH, FM, SaS, KP, HG, AdB, JB, NDG, PN, AE-B, HT, FH, JA, WM, DD, RK, BS-M, KK, EB, KB, CJ, T-WP-S, CE, LH, JP and AS; drafting the manuscript: JH and EH; critical revision of the manuscript for important intellectual content: StS, AR, LR, SK, AH, FM, SaS, KP, HG, AB, JB, NDG, PN, AE-B, HT, FH, JA, WM, DD, RK, BS-M, KK, EB, KB, CJ, T-WP-S, CE, LH, JP, AS, AdB, RKS and PH; accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved: all authors; final approval of the submitted manuscript: all authors; obtained funding: PH; study supervision: EH and PH; administrative, technical or material support: all authors.
Funding The study was funded by AstraZeneca, AGO Research GmbH.
Competing interests EH: honoraria: AstraZeneca; consulting or advisory role: AstraZeneca; research funding: Astra Zeneca (Inst). SS: honoraria: Roche; consulting or advisory role: Tesaro. KK: honoraria: Roche, Pfizer and AstraZeneca. JP: honoraria: Roche Pharma AG; consulting or advisory role: Amgen, AstraZeneca, Coherus Bioscience, DeciBio, F. Hoffmann-La Roche, Pharmamar, Shield Therapeutics, Simon Kucher & Partner, Taylor Wessing LLP and Tesaro; research funding: Astra Zeneca, Hoffmann-La Roche and Tesaro; travel, accommodations and expenses: Amgen, Astra Zeneca, F. Hoffmann-La Roche and Tesaro. AS: honoraria: TEVA and Gedeon Richter; consulting or advisory role: Astra Zeneca and Roche. SK: honoraria: AstraZeneca and Roche; consulting or advisory role: Roche. KP: travel and research funding: Medac Oncology; honoraria: AstraZeneca and Roche. DD is a consultant for AJ Innuscreen GmbH (Berlin, Germany), a 100% daughter company of Analytik Jena AG (Jena, Germany) and receives royalties from product sales (innuCONVERT kits). AdB: consulting or advisory role: AstraZeneca, Pfizer, Pharmamar, Roche/Genentech, Advaxis, Tesaro, Genmab, Clovis and Biocad. RKS: honoraria: AstraZeneca; consulting or advisory role: AstraZeneca; research funding: AstraZeneca (Inst); patents, royalties, other intellectual property: University of Cologne. CK: consulting and advisory role: AstraZeneca and Roche/Genentech. PH: consulting or advisory role: AstraZeneca, Roche/Genentech, Tesaro, Clovis, Pharmamar Lilly and Sotio; research funding: AstraZeneca (Inst); travel, accommodations and expenses: Medac.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.