|Home | About | Journals | Submit | Contact Us | Français|
The prevalence of BRCA½ mutations in germline DNA from unselected ovarian cancer patients is 11% to 15.3%. It is important to determine the frequency of somatic BRCA½ changes, given the sensitivity of BRCA-mutated cancers to poly (ADP ribose) polymerase-1 (PARP1) inhibitors and platinum analogs.
In 235 unselected ovarian cancers, BRCA½ was sequenced in 235, assessed by copy number analysis in 95, and tiling arrays in 65. 113 tumors were sequenced for TP53. BRCA½ transcript levels were assessed by quantitative polymerase chain reaction in 220. When available for tumors with BRCA½ mutations, germline DNA was sequenced.
Forty-four mutations (19%) in BRCA1 (n = 31)/BRCA2 (n = 13) were detected, including one homozygous BRCA1 intragenic deletion. BRCA½ mutations were particularly common (23%) in high-grade serous cancers. In 28 patients with available germline DNA, nine (42.9%) of 21 and two (28.6%) of seven BRCA1 and BRCA2 mutations were demonstrated to be somatic, respectively. Five mutations not previously identified in germline DNA were more commonly somatic than germline (four of 11 v one of 17; P = .062). There was a positive association between BRCA1 and TP53 mutations (P = .012). BRCA½ mutations were associated with improved progression-free survival (PFS) after platinum-based chemotherapy in univariate (P = .032; hazard ratio [HR] = 0.65; 95% CI, 0.43 to 0.98) and multivariate (P = .019) analyses. BRCA½ deficiency, defined as BRCA½ mutations or expression loss (in 24 [13.3%] BRCA½–wild-type cancers), was present in 67 ovarian cancers (30%) and was also significantly associated with PFS in univariate (P = .026; HR = 0.67; 95% CI, 0.47 to 0.96) and multivariate (P = .008) analyses.
BRCA½ somatic and germline mutations and expression loss are sufficiently common in ovarian cancer to warrant assessment for prediction of benefit in clinical trials of PARP1 inhibitors.
BRCA1 and BRCA2 play a critical role in DNA repair by homologous recombination.1 BRCA½ germline mutations occur in 11% to 15.3% of women with unselected ovarian cancers.2–4 Poly (ADP-ribose) polymerase-1 (PARP1) inhibitors are synthetic lethal with BRCA½ dysfunction in homologous recombination–deficient cancers and are currently in clinical trials in BRCA½ germline mutation carriers with ovarian and breast cancer.5 The preliminary results of these studies are encouraging.6 Because PARP1 inhibitors may also be effective in cancers in which BRCA½ and thus homologous recombination function is compromised by somatic aberrations, the number of women with ovarian cancer who might benefit from PARP1 inhibitors may be greater than predicted by the frequency of germline BRCA½ mutations. However, BRCA½ status has not been comprehensively studied in a large cohort of unselected human ovarian cancers to assess whether loss of BRCA function can also occur due to somatic events. We thus evaluated BRCA½ in 235 unselected ovarian cancers by sequencing, identifying intragenic deletions, determining gene copy number, and quantifying expression of BRCA½ using the assays described below. Germline mutation status was determined in patients whose tumors demonstrated BRCA½ aberrations when normal DNA could be obtained.
Human ovarian cancer tissues (n = 235) were obtained from the Gynecology Cancer Banks at The University of Texas M. D. Anderson Cancer Center and University of California San Francisco under institutional review board–approved protocols (Table 1). The cases were randomly selected from all ovarian cancers with available snap-frozen tumor tissue collected between June 1990 and December 2006. As varying numbers of samples were used in the assays below, Table 2 provides the rationale for why only a subset of cancers were assessed in specific assays.
Sections of 10-μm thickness from frozen cancers in Tissue-Tek OCT (Qiagen, Valencia, CA) were homogenized using a TissueRuptor (Qiagen) after adding QIAzol lysis reagent, followed by RNA isolation using a QIAgen miRNAeasy MiniKit per manufacturers protocol. A QIAamp DNA MiniKit (Qiagen) was used to isolate DNA per manufacturer's protocol with overnight incubation (56°C) and RNaseA treatment.
Reverse transcription was performed using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA) per manufacturer instructions. For preamplification, a 0.2× probe mix was made by combining 1 μL of 91 20X gene expression assays from Applied Biosystems and 9 μL of low-EDTA Tris-EDTA (TE). Preamplification was performed using 2.5 μL of 2× TaqMan PreAmp Master Mix (Applied Biosystems), 1.25 μL 0.2× probe mix, and 1.25 μL of cDNA. Applied Biosystems TaqMan assays (BRCA1: Hs00173233_m1/Hs00173237_m1/Hs01556190_m1/Hs01556191_m1; BRCA2: Hs00609060_m1; housekeepers: Hs99999908_m1 [GUSB]/Hs00188166_m1 [SDHA]/Hs00237047_m1 [YWHAZ]/Hs00824723_m1 [UBC]/Hs00609297_m1 [HMBS]) were used for preamplification and quantitative polymerase chain reaction (PCR) on a Fluidigm (South San Francisco, CA) BioMark instrument. Cycle conditions were 95°C × 10 minutes, 17 cycles of 95°C × 15 seconds, and 60°C × 4 minutes. The PCR products were diluted 1:5 with low-EDTA TE. Samples were assessed on gene expression M48 dynamic arrays (Fluidigm) per manufacturer's protocol. The comparative cycle threshold (Ct) method was used to calculate relative gene expression using the Ct for the BRCA2 assay, the average Cts from the BRCA1 assays, and the average Cts from housekeeper genes. Quantitative PCR was performed in 220 cancers for which high-quality RNA was obtained.
PCR was performed on 2 ng of DNA in a 3-μL reaction using the primers flanking the exons of BRCA½ that are used in the BRCAnalysis (Myriad Genetics, Salt Lake City, UT) clinical test with the following cycling conditions: 95°C × 10 minutes, 35 cycles of 95°C × 30 seconds, 62°C × 30 seconds, and 72°C × 1 minute, finishing with 72°C × 1 minute. Each PCR product was treated with 0.1 U of Shrimp Alkaline Phosphatase (Sigma-Aldrich, St Louis, MO.) The PCR product was diluted 1:9, and 0.8 μL was used for cycle sequencing with Big Dye Sequencing Chemistry and Taq FS (Applied Biosystems). Cycle conditions were 95°C × 3 minutes, 32 cycles of 95°C × 30 seconds, 50°C × 30 seconds, 60°C × 3 minutes, and 72°C × 10 minutes. Sequence products were run on a Megabace 4500 automated sequencer (GE Medical Systems, Milwaukee, WI) per manufacturer's protocol.
BRCA½ mutations were only included in the analyses below if classified as deleterious or suspected deleterious based on established criteria.7 A suspected deleterious mutation typically is treated clinically as deleterious.
TP53 was amplified in 113 cancers with sufficient DNA using nested PCR. Primary PCR was performed using Taq-Platinum and 1 μL of 2 ng/μL of DNA in a 3-μL reaction with primers without M13 tails. The PCR product was diluted nine-fold and used for a secondary reaction with primers that have M13 tails. Sequence products were run on a Megabace 4500 automated sequencer per the manufacturer's protocol.
The array was designed using eArray (Agilent Technologies, Santa Clara, CA) and synthesized on a 8 × 15,000 probe format. The array design included probes spaced at 20–base-pair intervals across the complete genomic region of BRCA½ from 10 kb upstream of the 5′UTR to 5 kb downstream of the 3′UTR avoiding repeats. Additional probes (1,000) were evenly distributed across the genome to form a backbone against which specific genomic gain/loss was estimated.
Sample preparation/array processing was performed using the Oligonucleotide Array-Based CGH for Genomic DNA Analysis kit and protocol (Agilent Technologies). These arrays were run on 65 ovarian cancers. Data were analyzed using DNA Analytics 4.0 (version 4.0.76) software (Agilent Technologies).
In 203 cancers for which sufficient DNA was present, 250 ng of genomic DNA was processed using GeneChip Mapping NspI or StyI Assay Kit (Affymetrix, Santa Clara, CA) per manufacturer's protocol and hybridized to Affymetrix Mapping 500K NspI or StyI microarrays. After hybridization, array wash, stain, and scan procedures were performed per manufacturer's protocol. Copy number/loss of heterozygosity (LOH) analysis were performed using a software package described elsewhere (Abkevich V et al, manuscript in preparation). Ninety-five chips with high-quality data were used for the final analysis.
All analyses were carried out using R, v2.9.0 (www.R-project.org). In all analyses, observations were removed if the response or a covariate was missing. Fisher's exact test was used to make comparisons involving pairs of categorical variables. t test was used to compare differences in means of continuous variables. Cox proportional hazards regression was used to perform univariate and multivariate analysis on the progression-free survival (PFS) and overall survival (OS) times from the date of debulking surgery. Comparisons of survival probabilities for categorical variables were visualized with Kaplan-Meier plots. The partial likelihood ratio test was used to compute P values. Wald statistic-based CIs were calculated for hazard ratio (HR) point estimates.
In DNA extracted from 235 human ovarian cancers (Table 1), 44 mutations in BRCA1 (n = 31) and BRCA2 (n = 13) were detected in 43 tumors, including one small homozygous intragenic BRCA1 deletion that was detected using tiling arrays. In one cancer, both a BRCA1 and BRCA2 mutation was detected. All but one of the mutations included in the analyses are known or predicted deleterious mutations, with only one mutation being a suspected deleterious mutation (BRCA1, G1738R).7 Five novel mutations were not present in the Myriad Genetics or BIC8 germline BRCA½ mutation databases (Appendix Table A1, online only). This BRCA½ mutation frequency of 19% observed in tumor tissue is higher than the expected frequency of germline mutations in an unselected population of patients with ovarian cancer (11% to 15.3%).2–4 In 212 tumors with known grade, no BRCA½ mutations were observed in grade 1 cancers (0 of 13 v 40 of 199; P = .135; see Table 1 for a detailed list of missing data for each clinical variable). No mutations were found in patients with tumors without a serous component as compared with those with tumors of pure serous histology (0 of 13 v 39 of 186; P = .076). When stage was known, BRCA½ mutation status was not significantly associated with stage (stage 1 or 2 disease [three of 25] v stage 3 or 4 disease [37 of 189]; P = .584). The frequency of BRCA½ mutations in high-grade (grade 3) serous cancers was 22.8% (36 of 158).
Of the 43 patients with ovarian cancers harboring a BRCA1 or BRCA2 mutation (note that one patient had an ovarian cancer with two mutations, that is, in BRCA1 and BRCA2), germline DNA was available from 28 patients. In these 28 patients, 11 (39.3%; 95% CI, 22.1 to 59.3) ovarian tumor BRCA1 (nine of 21; 42.9%) and BRCA2 (two of seven; 28.6%) mutations could be demonstrated to be somatic due to an inability to detect the aberration in germline DNA, whereas 17 mutations (60.7%) were found in both tumor and germline DNA. No significant differences were found between germline and somatic BRCA½-mutant cancers in terms of any clinical variables, although the low tumor numbers limited the power of these analyses. Interestingly, somatic mutations were more frequently novel mutations (four of 11), as defined by absence from the Myriad Genetics or BIC8 germline BRCA½ mutation databases, than germline mutations (one of 17; P = .062). No somatic mutations were detected in tumors from patients with germline mutations.
One homozygous intragenic deletion in BRCA1 and none in BRCA2 was detected by high-density tiling arrays in 65 ovarian cancers. Homozygous deletion of both copies of BRCA1 or BRCA2 was not detected by 500K single-nucleotide polymorphism array in 95 patients with high-quality single-nucleotide polymorphism data, confirming the low frequency of deletions in tumors.
LOH in BRCA1 was detected in 82 (87.2%) of 94 ovarian cancers. In contrast, LOH in BRCA2 was detected in significantly fewer (45 [52.9%] of 85; P < .0001) ovarian cancers. The one retained gene copy was replicated (two or three copies) in 28 of 45 cases of LOH of BRCA2 (62.2%) and in 51 of 82 cases of LOH at BRCA1 (62.2%). Interestingly, LOH of BRCA2 was only detected in one sample without LOH of BRCA1 (P = .001).
Neither the expression of BRCA1 (P = .684, n = 220) nor BRCA2 (P = .966, n = 220) was significantly different in BRCA1- or BRCA2-mutated cancers as compared with wild-type cancers, respectively (Figs 1A and and1B).1B). In other words, BRCA expression level was not influenced significantly by BRCA mutation status. Similarly, BRCA1 and BRCA2 LOH was not associated with significantly reduced expression of BRCA1 or BRCA2, respectively, possibly because of duplication of the retained copy of the gene in a significant proportion of cases of LOH (as described above).
We also hypothesized that loss of expression of BRCA1 or BRCA2 in ovarian cancer would, as with BRCA1 or BRCA2 mutations, impair the function of BRCA1 or BRCA2. Cancers without BRCA½ mutations were considered to have loss of BRCA expression if the average of BRCA1 and BRCA2 δCT (threshold for expression level of BRCA½ corresponding to the 95th percentile of BRCA½ expression in BRCA½ mutant tumors) was higher than the 95th percentile of a normal distribution fit to the mutants' average δCT. As defined in this way, loss of BRCA expression was present in 24 (13.3%) BRCA½–wild-type cancers, implicating other potential mechanisms (eg, methylation) in loss of BRCA1 and BRCA2 gene expression.
Germline mutations in BRCA1 and BRCA2 have been reported to be associated with improved outcomes for patients with ovarian cancer after surgery and platinum-based chemotherapy.9–11 Likewise, herein, BRCA1 and BRCA2 mutations in ovarian cancer tissue were associated with a significantly improved PFS as compared with BRCA1/BRCA2–wild-type cancers in univariate analysis (Fig 2; P = .032; HR = 0.65; 95% CI, 0.44 to 0.98). This significant association was maintained in a multivariable Cox model that included clinical variables (Table 3). PFS for individuals with germline BRCA mutations was not significantly different from that of individuals with somatic BRCA mutations (P = .690), albeit with low numbers. In contrast, BRCA1 and BRCA2 mutations together in ovarian cancer tissue were not significantly associated with OS. An integrated dichotomous BRCA½-deficiency variable was defined as the presence of mutations or loss of expression of BRCA1 or BRCA2. This BRCA½ deficiency variable was present in 67 (30.0%) of 223 ovarian cancers. In univariate analyses, BRCA½ deficiency was significantly associated with PFS (Fig 3; P = .026; HR = 0.67; 95% CI, 0.47 to 0.96). This significant association of BRCA½ deficiency with PFS was maintained in a multivariable Cox model (Table 4).
Neither LOH of BRCA1 nor of BRCA2 was significantly associated with PFS or OS times compared with other ovarian cancers.
TP53 mutations were present in 81 (71.7%) of 113 cases and were significantly associated with BRCA1 and marginally associated with all BRCA½ mutations (Appendix Table A2, online only). Given the association of mutations in BRCA1 and of both BRCA½ mutations with TP53 mutations, we hypothesized that if LOH represented loss of BRCA function, it should also be associated with TP53 mutations. Indeed, we found that TP53 mutations were significantly associated with LOH in BRCA1 (54 of 63 v 0 of 10; P < .0001) and marginally associated with LOH in BRCA2 (28 of 34 v 18 of 30; P = .057).
This is the first comprehensive study of BRCA½ status in ovarian cancer tissue. Although thought previously to be uncommon, we demonstrate that somatic BRCA½ mutations account for at least one third of BRCA½ mutations in ovarian cancer specimens.12 In fact, BRCA½ mutations in total occur in approximately 18% of all ovarian cancers and 23% of high-grade serous cancers, compared with previous reports that BRCA½ germline mutations occur in 11% to 15.3% of unselected women with ovarian cancer.2–4 Using cases for which germline DNA was available, we estimate germline and somatic mutation rates of approximately 11.5% and 7%, respectively. Mutations of BRCA½ in ovarian cancers are associated with improved PFS after surgery and platinum/taxane-based chemotherapy. This is consistent with several previous reports for germline BRCA1 and BRCA2 mutations in women with ovarian cancer and likely represents, at least in part, increased effectiveness of platinum drugs in cancer cells with deficient homologous recombination.1,7,9,10 We also hypothesized that loss of expression of BRCA1 or BRCA2 in ovarian cancer would, as with BRCA1/BRCA2 mutations, impair BRCA½ function and thus significantly improve PFS after surgery and platinum-based chemotherapy. Indeed, BRCA½ deficiency (mutations plus expression loss) was also associated with PFS, suggesting that loss of BRCA1 and BRCA2 expression likely occurs for reasons other than mutations and rare homozygous deletions and may also impair homologous recombination in cancer cells. Of note, the numbers of low BRCA½ expressors in our study were not consistent with reported rates of methylation of these genes in our studies or in the literature (approximately 20%), perhaps because methylation may not in all cases lead to the very low level of BRCA expression defined in our article.13 However, it is possible that published studies have overestimated the methylation rate or that our study may underestimate the rate of low BRCA expression in ovarian cancer.
In contrast, and although some studies suggest an OS benefit in patients with germline BRCA½ mutations,14 mutations of BRCA½ in ovarian cancers were not significantly associated with OS herein. This may reflect, at least in part, a lack of impact of BRCA½ mutations on ovarian cancer responsiveness to nonplatinum chemotherapy drugs used as second-line therapy or on the natural history of ovarian cancer in the absence of therapy. After all, platinum sensitivity, but not sensitivity to other cytotoxic chemotherapy drugs commonly used in ovarian cancer treatment, is a surrogate marker for impaired homologous recombination in cancer cells. Alternatively, we may have seen an OS benefit with BRCA½ mutations with larger patient numbers. Although neither PFS nor OS was significantly different in individuals with germline versus somatic BRCA½ mutations, our current data set is not large enough to enable us to definitively determine whether there is a biologic difference between germline and somatic mutations.
Mutations of BRCA1 are almost universally associated with TP53 mutations. This is consistent with genetically engineered mouse models in which BRCA1 deletion is an early lethal, whereas embryos with combined BRCA1 and TP53 mutations survive significantly longer.15
Several reasons may explain the lack of an association between BRCA transcript expression levels and mutations or LOH. In the case of BRCA½ LOH, in approximately 62% of cases, there were two (and, in the case of BRCA1, sometimes three) copies of the gene. In the case of mutations, a significant proportion of the mutations identified do not result in nonsense mediated transcript decay. We also identified LOH in many tumors without mutations (87% and 53% of tumors had LOH at BRCA1 and BRCA2 but only 13% and 6% had BRCA1 and BRCA2 mutations, respectively). As in all ovarian tumors with BRCA½ LOH, in tumors with BRCA1 or BRCA2 LOH but without mutations, we observed approximately 60% of cases with two or more gene copies and almost 40% with only one copy. Thus if loss of one gene copy does affect gene expression, then this would affect both BRCA½-mutant and nonmutant ovarian tumors. These and likely other factors may explain the lack of a correlation between BRCA½ expression and either BRCA½ mutation status or LOH.
Two studies have proposed that platinum chemoresistance can arise from mutations that restore the BRCA2 open reading frame and thus homologous recombination.16,17 In this study, there was no evidence of somatic mutations in ovarian cancers from patients with germline mutations. Because the cancer tissues used in this study were mostly collected before chemotherapy administration, our study does not impinge on the concept that selective pressure during chemotherapy could select for mutations that restore BRCA function.
Currently, germline BRCA½ mutation screening is performed in women with ovarian cancer judged to be at high risk for carrying an inherited mutation based on clinical models (eg, BRCAPRO).18,19 PARP1 inhibitor trials are underway in BRCA½ germline mutation carriers with ovarian cancer, and the preliminary results of these studies are encouraging.5,6 However, because PARP1 inhibitors are selectively active in BRCA½-deficient cancers, assessment of BRCA½ mutation status in all ovarian cancers could identify a higher number of women who might benefit from these novel drugs. This approach should be investigated in future trials of PARP1 inhibitors. Whether loss of expression or LOH of BRCA1 or BRCA2 will contribute to sensitivity to PARP1 inhibitors also warrants exploration. Further, homologous recombination is a complex process with multiple components. Thus the frequency of homologous recombination aberrations will be greater than that indicated by studies of BRCA½ alone. Although that frequency is difficult to speculate at this time, it is likely that additional patients with ovarian cancer may benefit from PARP inhibitors.
In summary, loss of BRCA function due to frequent somatic aberrations in ovarian cancers likely deregulates homologous recombination and thereby increases sensitivity to platinum drugs and possibly also to PARP1 inhibitors. This is consistent with prior studies of germline BRCA½ mutations. The novel observation that somatic BRCA½ aberrations occur frequently could significantly increase the ability to identify patients who will benefit from PARP1 inhibitors in ovarian cancer clinical trials. Somatic and germline mutations and BRCA½ expression loss are sufficiently common in ovarian cancer to warrant assessment in clinical trials for prediction of benefit from PARP1 inhibitors.
|Mutation||AA Change||Significance||Somatic or Germline|
NOTE. These novel mutations were not present in either the Myriad Genetics or BIC8 germline BRCA½mutation databases.
Abbreviation: AA, amino acid.
|BRCA||TP53 Mutant||TP53 Nonmutant||P|
Supported by the Kleberg Center for Molecular Markers at the University of Texas M. D. Anderson Cancer Center, National Cancer Institute Grant No. PO1CA099031 (G.B.M.), The Susan G. Komen Foundation Biomarkers Identification and Validation Award FAS0703849 (B.T.H., G.B.M.), The M. D. Anderson Cancer Center Physician Scientist Program, the McNair Scholars Program supported by the Robert and Janice McNair Foundation, and an American Society of Clinical Oncology Career Development Award (B.T.H.).
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment or Leadership Position: Kirsten M. Timms, Myriad Genetics (C); Alexander Gutin, Myriad Genetics (C); Darl D. Flake II, Myriad Genetics (C); Jennifer Potter, Myriad Genetics (C); Jerry S. Lanchbury, Myriad Genetics (C) Consultant or Advisory Role: None Stock Ownership: Kirsten M. Timms, Myriad Genetics; Alexander Gutin, Myriad Genetics; Darl D. Flake II, Myriad Genetics; Jennifer Potter, Myriad Genetics; Jerry S. Lanchbury, Myriad Genetics Honoraria: None Research Funding: Gordon B. Mills, Myriad Genetics Expert Testimony: None Other Remuneration: Dmitry Pruss, Myriad Genetics
Conception and design: Bryan T.J. Hennessy, Kirsten M. Timms, Mark S. Carey, Jerry S. Lanchbury, Gordon B. Mills
Financial support: Kirsten M. Timms, Jerry S. Lanchbury
Provision of study materials or patients: Bryan T.J. Hennessy, Mark S. Carey, Pat Glenn, Karen Smith McCune, Russell R. Broaddus, Karen H. Lu
Collection and assembly of data: Bryan T.J. Hennessy, Kirsten M. Timms, Alexander Gutin, Victor Abkevich, Jennifer Potter, Dmitry Pruss, Yang Li, Jie Li, Jerry S. Lanchbury, Gordon B. Mills
Data analysis and interpretation: Bryan T.J. Hennessy, Kirsten M. Timms, Alexander Gutin, Larissa A. Meyer, Darl D. Flake II, Victor Abkevich, Jennifer Potter, Dmitry Pruss, Jerry S. Lanchbury, Gordon B. Mills
Manuscript writing: Bryan T.J. Hennessy, Kirsten M. Timms, Alexander Gutin, Larissa A. Meyer, Darl D. Flake II, Victor Abkevich, Ana Maria Gonzalez-Angulo, Jerry S. Lanchbury, Gordon B. Mills
Final approval of manuscript: Bryan T.J. Hennessy, Kirsten M. Timms, Mark S. Carey, Alexander Gutin, Larissa A. Meyer, Darl D. Flake II, Victor Abkevich, Yang Li, Jie Li, Ana Maria Gonzalez-Angulo, Karen Smith McCune, Maurie Markman, Russell R. Broaddus, Jerry S. Lanchbury, Karen H. Lu, Gordon B. Mills