Tumor BRCA1 Reversion Mutation Arising during Neoadjuvant Platinum-Based Chemotherapy in Triple-Negative Breast Cancer Is Associated with Therapy Resistance

Individuals with a heterozygous germline mutation in the BRCA1 or BRCA2 (BRCA1/2) genes are predisposed to a range of familial cancers, including breast and ovarian carcinomas (1). The BRCA1/2 genes are tumor suppressors that encode proteins critical for the accurate repair of DNA double-strand breaks by homologous recombination. The interplay between the BRCA1/2 proteins and RAD51, an essential DNA repair protein, preserves genomic stability by avoiding more error-prone DNA repair pathways (2–4). Tumors from patients with deleterious BRCA1/2 mutations usually have loss of the wild-type allele and are thus BRCA1/2-deficient, whereas BRCA1/2 heterozygosity is retained by normal cells (5, 6). Somatic inactivation of the wild-type BRCA1/2 allele and methylation of the BRCA1 promoter domain with subsequent epigenetic silencing are other reported mechanisms (7).

Approximately 10% to 20% of invasive triple-negative breast cancers arise in the setting of BRCA1/2 deficiency (8–10). BRCA1/2-deficient breast cancers have decreased capacity to repair DNA, resulting in enhanced sensitivity to chemotherapy with cross-linking agents, such as cisplatin and carboplatin, and susceptibility to PARP inhibitors due to a synthetic lethal interaction (11–13). In a small proof-of-concept trial, the PARP inhibitor olaparib led to objective tumor response in 41% of patients with advanced breast cancer with a deleterious germline BRCA1/2 mutation (14). Clinical trials are now underway to assess the efficacy of PARP inhibitors as single agents and in combination with chemotherapy in the advanced as well as curative settings for triple-negative and BRCA1/2-associated breast cancers. Likewise, platinum agents have shown efficacy for BRCA1/2-associated breast cancers (15, 16), with a recent phase III study demonstrating substantial improvement in the objective response rate and progression-free survival with carboplatin compared with docetaxel in patients with a deleterious BRCA1/2 mutation and advanced disease (17).

With the increased recognition of DNA-damaging therapies as the forefront agents for BRCA1/2-associated tumors, an unexpected mechanism of drug resistance has garnered clinical attention: the restoration of BRCA1/2 function due to secondary mutations of BRCA1/2 in tumors. Given that defects in homologous recombination secondary to an underlying BRCA1/2 mutation lead to cellular sensitivity to DNA-damaging agents, the restoration of DNA repair proficiency results in acquired resistance to these drugs. To date, this drug resistance mechanism has been reported primarily in patients with advanced ovarian cancer who had disease progression after platinum or PARP inhibitor therapy (7, 18–24). In this study, we evaluated this novel mechanism of drug resistance in newly diagnosed patients with early-stage BRCA1/2-mutant breast cancer who had a poor response to platinum-based neoadjuvant chemotherapy.

PrECOG 0105 was a single-arm phase II neoadjuvant study in patients with newly diagnosed clinical stage I–IIIA triple-negative [estrogen receptor (ER) ≤ 5%, progesterone receptor (PR) ≤ 5%, and human epidermal growth factor receptor 2 (HER2)-negative (0 or 1+ by immunohistochemistry or FISH unamplified)] or BRCA1/2 mutation–associated breast cancer. Eighty patients were treated with carboplatin, gemcitabine, and iniparib, intravenously, for 6 cycles before definitive surgery. A total of 80 patients were enrolled at multiple cancer centers through PrECOG. Details of the clinical trial methods and results of the study endpoints have been published previously (16).

The primary endpoint of the clinical study was complete pathologic response rate, defined as no invasive carcinoma in the breast or axillary lymph nodes. The extent of residual disease was assessed using the residual cancer burden (RCB) index (25). The RCB index is a validated prognostic marker of distant relapse-free survival in patients who have been treated with neoadjuvant chemotherapy [RCB 0: complete pathologic response (pCR); RCB I: minimal residual disease; RCB II: moderate residual disease; and, RCB III: extensive residual disease); ref. 25].

In addition to comprehensive germline BRCA1/2 testing using serum samples, all patients underwent research core biopsies of the primary breast tumor before therapy initiation; five to ten 5-μm tissue sections of fixed or frozen tumor tissue were collected and processed at Myriad Genetics, Inc. DNA extraction from formalin-fixed, paraffin-embedded or frozen tumors was conducted, and next-generation DNA sequencing using a liquid hybridization assay (Agilent SureSelect) followed by Illumina sequencing was performed. Residual disease tissue was collected in patients who went to surgery and did not achieve pCR and processed using the same assay methodology (26).

To determine BRCA1/2 mutation status, variant and large rearrangement detection was performed on sequence from BRCA1 and BRCA2. Complete descriptions of the mutation detection methods have been previously described (27). Only mutations classified as deleterious or suspected deleterious on the basis of previously described criteria were included in the analysis (28).

Baseline tumor gene expression profiling, copy number alterations and measures of genomic scars (homologous recombination deficiency assay) were assessed for any correlation with treatment response to neoadjuvant therapy. For BRCA1/2 mutation carriers with unfavorable response to neoadjuvant therapy, e.g., RCB II, tumor BRCA1/2 status was re-sequenced in the residual surgical tissue.

DNA extractions from two independent samples of the pretreatment tumor specimen were also performed for patient 15 to assess for a reversion mutation. In both samples, there were between 800 and 1,800 unique clonal reads through the region where the reversion mutation was located. On the basis of simulations, we would have 95% confidence to detect the reversion mutation in this number of clonal reads, if the mutation occurred in at least 1 in 250 cells.

Furthermore, SNP data from the homologous recombination deficiency assay was used to calculate the percent tumor content in each sample and to reconstruct the genome over the locus. This information allowed for calculation of predicted read frequency of both germline and somatic alleles.

All germline and tumor BRCA1/2 mutation testing was performed by Myriad Genetics, Inc.

Eighty patients received six cycles of gemcitabine, carboplatin, and iniparib in PrECOG 0105. Nineteen of these 80 patients (24%) carried a deleterious germline mutation in BRCA1, BRCA2, or both genes (Table 1). Four of these 19 patients had ER/PR-positive breast cancer (two BRCA1 carriers and two BRCA2 carriers), and the remaining 15 patients (78.9%) had triple-negative breast cancer. Deletion mutations leading to nonfunctional BRCA1/2 proteins were the primary mechanism (13 patients, 65%), including one patient with bilateral breast cancer with deletion of the entire BRCA1 gene, as well as a patient with deletions in both BRCA1 and BRCA2 genes (BRCA1: 1048delA; BRCA2: 1366delA); both patients had excellent responses to neoadjuvant therapy. Insertions and missense mutations were also observed (three and four patients, respectively). The affected gene and specific location of the germline mutations are shown in Table 2.

Among all study patients, 29 [36.3%; 90% confidence interval (CI), 27%–46%] achieved a pCR. The pCR rate was 33.0% among wild-type BRCA1/2 patients (CI, 23%–44%), 47% among BRCA1/2 mutation carriers (CI, 27%–68%), and 56% among BRCA1/2 mutation carriers with triple-negative breast cancer (CI, 33%–77%). Among patients with germline BRCA1/2 mutations, RCB class was as follows: RCB 0 = 10 (53%); RCB I = 6 (32%); RCB II = 4 (21%); and RCB III = 0 (0%). One BRCA1 mutation carrier was lost to follow-up and never had surgery.

BRCA1 sequencing of residual tissue was available in three of four patients with RCB II response (no consent in one case, patient 1). These three patients had BRCA1 3374insGA (patient 2), 1479delAG (patient 15), and W1721X (patient 18) mutations with LOH at these loci in the pretreatment specimens (Table 3). Patients 2 and 15 had triple-negative breast cancer, whereas patient 18 had ER/PR-positive, HER2-negative breast cancer. No reversion mutations were detected in the residual tumor specimen at the time of surgery in patients 2 and 18. However, a new BRCA1 mutation was detected in the residual surgical tumor specimen in patient 15. This reversion mutation comprised a 42-base pair deletion that overlapped with the original 2-base pair deletion. It resulted in a 14–amino acid deletion and restoration of the BRCA1 reading frame. A relapse biopsy of an ipsilateral axillary lymph node 4 months later revealed an identical reversion mutation (Table 3, Figs. 1 and 2).

Because of the presence of nontumor DNA in the sample, it was not possible to determine whether there were residual tumor cells after neoadjuvant therapy that retained the original 1479delAG germline mutation and lacked the reversion mutation. However, upon analysis of two samples of extracted DNA in the pretreatment tumor specimen, no reversion mutation was detected. On the basis of our statistical constraints, we conclude that whether the reversion mutation was present in the pretreatment tumor specimen, it was in less than 1 in 250 tumor cells. DNA sequencing of the contralateral breast cancer specimen did not reveal evidence of a somatic reversion mutation. Specifically, 324 unique DNA sequence reads were present, spanning both the germline deletion mutation (1479delAG) and the somatic reversion, with 245 of the DNA sequence reads showing the germline mutation and zero reads showing the somatic reversion mutation.

Predicted and observed allele frequencies were consistent with the original mutation being germline and the subsequent reversion mutation being somatic using SNP data. The presence of the original mutation in the post-reversion mutation samples at levels consistent with residual germline DNA was also confirmation that the original mutation was germline (Appendix 1).

At the age of 25 years, this patient was diagnosed with stage IIIC right breast invasive ductal carcinoma that was ER/PR-negative and HER2-unamplified (i.e., triple-negative). A deleterious 1479delAG BRCA1 mutation was detected at breast cancer diagnosis. She received neoadjuvant chemotherapy with adriamycin and cyclophosphamide, followed by a right breast lumpectomy and axillary lymph node dissection. She was found to have achieved a pCR at the time of surgery with no disease in the lymph nodes. This was followed by adjuvant paclitaxel and whole breast radiation (50.4 Gy).

The patient had an in-breast recurrence 1 year later; this was also high-grade, triple-negative carcinoma for which she proceeded with a right mastectomy, followed by additional cycles of anthracycline- and taxane-based chemotherapy.

Approximately 10 years later, she was diagnosed with a contralateral high-grade, triple-negative breast cancer. She enrolled in PrECOG 0105 and received six cycles of gemcitabine, carboplatin, and iniparib. She proceeded with a left mastectomy upon completion of experimental therapy, and this revealed a 2-cm residual invasive carcinoma of mixed lobular and ductal histology (RCB II) with a negative sentinel lymph node dissection.

Within months of completing neoadjuvant therapy, a recurrent left axillary mass was discovered, prompting additional locoregional and systemic treatment. One year later, she was found to have distant disease spread. Her disease progressed rapidly, and she died less than 2 years after her second primary breast cancer diagnosis.

We report the case of a young woman with a history of bilateral triple-negative breast cancer and a germline BRCA1 1479delAG mutation who, over the course of 18 weeks of neoadjuvant platinum-based therapy, developed a 42-base pair somatic deletion in the BRCA1 gene, leading to a reversion of the underlying 1479delAG deleterious germline BRCA1 mutation. The same somatic BRCA1 reversion mutation was identified in a metastatic focus in an axillary lymph node at the time of locoregional relapse, 4 months after the patient's definitive surgery. To our knowledge, there has been no previous report of a BRCA1 somatic reversion mutation developing over the short span of neoadjuvant chemotherapy in a patient with newly diagnosed, early-stage breast cancer.

Two types of somatic reversion mutations have been described. In the case of our patient, a genetic reversion event leading to deletion of a frameshift mutation resulted in correction of the original frameshift, thereby producing an allele that contains a new mutation, but which now encodes a DNA repair–proficient BRCA1 protein. A second class of mutations involves direct reversion to wild-type of the original mutation. Studies have shown that these types of somatic reversions account for the majority of BRCA1/2 mutation–associated ovarian carcinomas with secondary restoration of the BRCA1 or BRCA2 reading frame (7). These ovarian tumors develop platinum resistance, and may also be resistant to PARP inhibitor therapy (15, 17, 21). Reversion mutations in this setting have been associated with a prior history of breast cancer or with exposure to more than one line of chemotherapy for the treatment of ovarian cancer. This has led to the hypothesis that such somatic mutations may have been present at the time of the ovarian cancer diagnosis in a small number of neoplastic cells that were subsequently selected by additional DNA-damaging therapy (17). Furthermore, somatic reversion mutations have also been reported with cisplatin treatment in breast and pancreatic cancer cell lines that had mutations in BRCA2 (19) and with the PARP inhibitor olaparib for a male BRCA2 mutation carrier with metastatic breast cancer (14).

BRCA1 is a large multidomain protein that includes the “really interesting new gene” (RING) finger, nuclear localization signal (NLS), coiled-coil, and BRCA1 C-terminus (BRCT) domains. Prior BRCA1 somatic reversion mutations have been reported in the RING finger domain and in downstream nucleotide sites (Fig. 3). In ovarian cancer, BRCA1 reversion mutations have been reported at several sites, including six patients with a 185delAG mutation and two patients with a 2594delC mutation (21, 24); however, reversion of the germline BRCA1 1479delAG mutation has not previously been reported in either breast or ovarian cancer.

Our study has limitations. Here, we report an association between a somatic BRCA1 reversion mutation and therapeutic failure after neoadjuvant platinum-based therapy for breast cancer. However, we cannot establish a causal relationship, in particular, as our patient's breast cancer history was complex, given she had been treated for a contralateral high-grade breast cancer 10 years prior to her new triple-negative breast cancer diagnosis, and, as such, she was exposed to chemotherapy in the past. Her clinical response to gemcitabine, carboplatin, and iniparib may not be generalizable given her past oncologic history, and, as a result of past therapeutic exposures, she may have had other molecular triggers, besides the BRCA1 reversion, leading to chemoresistant tumor cells. Furthermore, this is a relatively small study, and corroboration in a larger cohort of BRCA1/2 carriers treated with neoadjuvant therapy for breast cancer is warranted.

We report a novel, acquired somatic frameshift mutation in BRCA1 leading to reversion of a 1479delAG germline BRCA1 mutation, which developed during 18 weeks of neoadjuvant platinum-based chemotherapy for triple-negative breast cancer. We conclude that this reversion mutation led to reduced treatment sensitivity, manifested as significant disease burden at definitive surgery and rapid metastatic recurrence shortly after completion of therapy. The timing and etiology of such a reversion mutation are unclear and may have stemmed from genomic instability of the primary tumor given an underlying deficiency in homologous recombination DNA repair. How frequently somatic reversion mutations occur in clinical practice is not well known but may have significant implications for rational treatment selection in BRCA1/2 mutation–associated cancers. Given the growing number of DNA repair–targeted therapies, characterization of the prevalence, time course, and correlates of reversion mutations in BRCA1/2 and other DNA repair genes is essential to guide patient care and drug development.

S. Vinayak reports receiving company-sponsored meals from Tesaro during a clinical trial investigator's meeting. M.L. Telli is a consultant/advisory board member for Myriad Genetics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Afghahi, S. Vinayak, R.W. Carlson, J.M. Ford, M.L. Telli

Development of methodology: A. Afghahi, K.M. Timms, J.M. Ford, M.L. Telli

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.C. Jensen, R.W. Carlson, P.-J. Chang, E.A. Schackmann, A.-R. Hartman, J.M. Ford, M.L. Telli

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Afghahi, K.M. Timms, S. Vinayak, K.C. Jensen, A.-R. Hartman, J.M. Ford, M.L. Telli

Writing, review, and/or revision of the manuscript: A. Afghahi, K.M. Timms, S. Vinayak, K.C. Jensen, A.W. Kurian, R.W. Carlson, E.A. Schackmann, A.-R. Hartman, J.M. Ford, M.L. Telli

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.M. Timms, E.A. Schackmann, M.L. Telli

Study supervision: K.M. Timms, M.L. Telli

The authors gratefully acknowledge research support from the ASCO Conquer Cancer Foundation, Breast Cancer Research Foundation, BRCA Foundation, Stanford Cancer Institute, and Susan G. Komen for the Cure.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.