The BRCA1-Δ11q Alternative Splice Isoform Bypasses Germline Mutations and Promotes Therapeutic Resistance to PARP Inhibition and Cisplatin

Germline mutations in the BRCA1 gene are associated with an increased risk of developing breast and ovarian cancer (1, 2). Mutations often result in reading frameshifts and nonsense-mediated mRNA decay (NMD; ref. 3). The BRCA1 protein is essential for efficient homologous recombination (HR)-mediated repair of double-stranded DNA breaks (4, 5). Inhibitors of PARP, as well as platinum agents, induce double-stranded DNA breaks that are repaired by the HR DNA repair pathway (6, 7). Consequently, cells that have defective HR DNA repair, such as those with dysfunctional BRCA1 or BRCA2 proteins, are highly sensitive to PARP inhibitor (PARPi) or platinum treatments (8–11).

Although PARP inhibitors have been shown to provide survival improvements, many patients that harbor germline BRCA1 or BRCA2 mutations do not gain benefit from PARPi therapy (12–14). In addition, many patients that first benefit from either PARPi or platinum therapy develop disease progression and resistance (15). PARPi or platinum resistance has been demonstrated to arise by a variety of mechanisms, including reversion mutations (16, 17), loss of 53BP1 pathway activity (18, 19), expression of hypomorphic BRCA1 proteins (20, 21), and drug efflux (22).

BRCA1 mRNA isoforms generated by alternative splicing lack specific exons and have been shown to be expressed in cells and tissues (23–25). In particular, the relative levels of BRCA1 exon 11 splice isoforms differ between normal and cancer tissues and in discrete phases of the cell cycle (26–29). These isoforms include BRCA1 full-length (inclusion of all coding exons), Δ11 (skipping of exon 11), and Δ11q (partial skipping of exon 11). The BRCA1-Δ11q isoform derives from use of an alternative exon 11 splice donor site, resulting in the exclusion of most exon 11 nucleotides (c.788-4096; Supplementary Fig. S1; refs. 27, 28). In human cells and tissues, the BRCA1- Δ11q isoform expression is more readily detectable than the BRCA1-Δ11 isoform (26, 29, 30).

The BRCA1-Δ11 isoform has previously been implicated in both cell death and proliferation in mouse studies. Both Brca1-null and Brca1^Δ11/Δ11 mice that exclusively express Brca1-Δ11 undergo embryonic lethality, but Brca1^Δ11/Δ11 embryos die at a later stage, suggesting that Brca1-Δ11 isoform can partially compensate for the lack of other Brca1 isoforms during embryogenesis (31–33). BRCA1 mutations located in exon 11 represent approximately 30% of the overall number of mutation carriers that develop breast and ovarian cancer in the United States (34–37). Here, we examined the impact of exon 11 mutations on BRCA1 isoform expression and therapy response.

Cells were purchased from Asterand or ATCC. Cycloheximide (CHX), actinomycin D (ACT), puromycin, blasticidin, and DMSO were purchased from Sigma-Aldrich, cisplatin was from APP/Fresenius Kabai USA LLC, and paclitaxel from Sagent Pharmaceuticals. Pladienolide B (Pl-B) was purchased from Calbiochem. Clovis Oncology provided rucaparib (CO-338) and olaparib (AZD2281) was purchased from Selleckchem. BRCA1-mutated cell lines were validated through DNA sequencing and the identification of specific BRCA1 mutations that are uniquely present in individual cell lines as well as DNA fingerprinting. Cell lines tested negative for mycoplasma.

Depending on colony-forming potential, cells were seeded at a density ranging from 500 to 4,000 cells per well in 6-well plates in the presence of increasing concentrations of either rucaparib or olaparib. For cisplatin and taxol treatments, exponentially growing cells were cultured in 24-well plates, treated with increasing concentrations of cisplatin and taxol for 24 hours, and replated in 6-well plates for colony formation. For shRNA or cDNA add back colony formation experiments, cells were treated as above, but with the addition of either puromycin or blastcitidin in the media. For siRNA treatments, exponentially growing cells were reverse transfected in 24-well plates, and 2 days after transfection, cells were treated with rucaparib for 72 hours and then replated in 6-well plates for colony formation. For Pl-B colony assays, cells were treated with Pl-B (1.25 nmol/L) and rucaparib (100 nmol/L) for 72 hours and then replated into 6-well plates for colony formation. Colony formation was assessed 2 weeks postplating with crystal violet staining. Mean colony formation from three experiments was expressed as percentage of colonies ± SE relative to vehicle-treated cells.

Genomic DNA was isolated from cells using the DNeasy tissue kit (Qiagen). To determine whether gene rearrangements had taken place that would have excluded the exon 11q region from genomic DNA of cell lines and patient-derived xenograft (PDX) tumors, we carried out PCR using OneTaq Hot Start 2X Master Mix (NEB) and gDNA as templates. Primers were located in exon 10 (forward, F) and 12 (reverse, R): F:aatcacccctcaaggaacca; R:ctcacacccagatgctgcttc. Total RNA was isolated from cell lines using RNeasy Kit (Qiagen). For quantitative RT-PCR, RNA was tested for quality on a Bioanalyzer (Agilent). RNA concentrations were determined with a spectrophotometer (NanoDrop; Thermo Fisher Scientific). RNA was reverse transcribed using Moloney murine leukemia virus reverse transcriptase (Ambion) and a mixture of anchored oligo-dT and random decamers. Two reverse transcription reactions were performed for each sample using either 100 or 25 ng of input RNA. Assays were used in combination with TaqMan Universal Master mix or Power SybrGreen Master Mix and run on a 7900 HT sequence detection system (Applied Biosystems). Cycling conditions were 95°C, 15 minutes, followed by 40 (two-step) cycles (95°C, 15 seconds; 60°C, 60 seconds). C_t (cycle threshold) values were converted to quantities (in arbitrary units) using a standard curve (four points, 4-fold dilutions) established with a calibrator sample. For each sample, the values were averaged and SD of data derived from two independent PCRs. For the BRCA1 splice variants, amplicons quantified by on-chip electrophoresis on an Agilent 2100 Bioanalyzer were used for absolute quantification assays. Primers that specifically recognize the following BRCA1 mRNA isoforms were used: exon +11 containing: F:tagcaaggagccaacataacagat; R:cttattccattcttttctctcacacag; Δ11q: F:gattctgcaaaaaaggctgct; R:cagatgctgcttcaccctga.

For RBFOX2 F:aagcccagtagttggagctgt, R:ttgcctagggacacatctgctt; and POLR2F: F:tgccatgaaggaactcaagg R: tcatagctcccatctggcag. BRCA1 expression values were routinely normalized to a POL2RF housekeeping gene (HKG) control and expressed as a percentage of the values calculated for MDA-MB-231 cells or to a relevant control sample.

Genomic DNA derived from blood lymphocytes of healthy individuals was tested for the absence of mutations by BROCA sequencing. The entire genomic region of BRCA1 from exon 8 to exon 12 was amplified with the following 5 PCR reactions and primer sets:

• 1F: ccgctcgagAACCTTGGAACTGTGAGAACTCTG

• 1R: ccggatatCAATTTGAGAGCCCAGTTTGAAT

• 2F: ccgctcgagAACCTGGGTGACAGAGCAAGA

• 2R: ccggatatcAGGGAAAAGACAGAGTCCTAATAAGA

• 3F: ccgctcgagAGAGCTAAAATGTTTGATCTTGGTC

• 3R: ccggatatcTCTTGATAAAATCCTCAGGATGAAG

• 4F: ccgctcgagATTTGGGAAAACCTATCGGAAG

• 4R: ccggatatcTAATACTGGAGCCCACTTCATTAGT

• 5F: ccgctcgagCCAGCTCAAGCAATATTAATGAAGT

• 5R: ccggatatcGTTAAAATGTCACTCTGAGAGGATAGC

HA-tag was cloned into pENTRA vector using SalI and XhoI sites. Each BRCA1 fragment was first cloned into pENTRA-HA vector using XhoI (shown in lowercase forward primers above) and EcoRV (shown in lowercase reverse primers above) sites. The 5 cloned fragments were assembled into a minigene using the following restriction sites that corresponded with cut sites in each fragment: EcoRI, SpeI, StuI, and ScaI sites. An eGFP fragment was cloned into the 3′end of the exon 12 fragment using EcoRV site. The minigene was then shuttled into pcDNA6.2 Destination vector using the LR Clonase system (Invitrogen). We confirmed by Sanger sequencing that mutations were not introduced into exons or introns within 500 bp flanking each exon. The indicated mutations (see Supplementary Fig. S3A) were introduced using site-directed mutagenesis (Agilent) using the following primers:

• 2288delT F: ccagtgaacttaaagaatttgtcaacctagccttccaaga

• 2288delT R: tcttggaaggctaggttgacaaattctttaagttcactgg

• 2529C>T F: caaatgctgcacactaactcacacatttatttggttctgtttttg

• 2529C>T R: caaaaacagaaccaaataaatgtgtgagttagtgtgcagcatttg

• 3960C>T F: ctaaggtgatgttcctaagatgcctttgccaatattacc

• 3960C>T R: ggtaatattggcaaaggcatcttaggaacatcaccttag

• 1stFOXMut F: tcagggtagttctgtttcaaacttacacgtggagccatgtg

• 1stFOXMut R: cacatggctccacgtgtaagtttgaaacagaactaccctga

• 2ndFOXMut F:gagtaataaactgctgttctcgtgttgtaatgagctggcatgagta

• 2ndFOXMut R: tactcatgccagctcattacaacacgagaacagcagtttattactc

• Ex11qsplice F: tgcaagtttgaaacagaactcccctgatacttttctggatg

• Ex11qsplice R: catccagaaaagtatcaggggagttctgtttcaaacttgca

To measure BRCA1-Δ11q-reporter expression, total RNA was isolated from transfected cells using RNeasy Kit (Qiagen). cDNA was synthesized using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). PCR was performed using OneTaq Hot Start 2X Master Mix (NEB). Primers that specifically recognize BRCA1-Δ11q-reporter mRNA were: F: gattctgcaaaaaaggctgct; R: agtcgtgctgcttcatgtggt; and BSD: F: gcctttgtctcaagaagaatcca; R: tagccctcccacacataacca. In addition, Western blots using HA antibody detected BRCA1-Δ11q-reporter protein expression.

Western blotting was carried out as described previously and proteins detected using the following antibodies: BRCA1: N-terminal: (MS110, EMD), C-terminal: (D9, Santa Cruz Biotechnology), BRCA1 (9010, Cell Signaling Technology), tubulin (2148, Cell Signaling Technology), HA (23675, Cell Signaling Technology), RAD51 (H-92, Santa Cruz Biotechnology), Pol II [C-18, (Santa Cruz Biotechnology)], CtIP (A300-438-A, Bethyl Laboratories), PALB2 (A301-247A1, Bethyl Laboratories), BARD1 (A300-263A, Bethyl Laboratories), and BRCA2 (OP-95, EMD). HA (23675, Cell Signaling Technology) antibody was used for immunoprecipitation of HA-BRCA1 complexes from 2 mg of nuclear extract using Pierce Classic IP Kit (Thermo Scientific) according to the manufacturer's instructions. Nuclear extracts were derived using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific) according to the manufacturer's instructions. Densitometric analyses were carried out using ImageJ software.

For immunofluorescence HA (HA.11, Covance), BRCA1 (MS110, EMD), γ-H2AX (pS139; N1-4131, EMD) and RAD51 (H-92, Santa Cruz Biotechnology), geminin (10802-1-AP, ProteinTech Group) and geminin (Abnova, E7071-1A8) antibodies were followed by secondary antibodies conjugated to FITC or Texas Red (Jackson ImmunoResearch Laboratories). We acquired immunofluorescence images using Nikon NIU Upright Fluorescent Microscope and generated images using Nikon NIS Elements software. For IR experiments, we routinely fixed cells 8 hours after treatment with 10 Gy. For analyses, we counted a minimum of 200 cells per condition or cell line. Foci-positive cells were expressed as a percentage of geminin-positive cells to account for any differences in S–G₂–M populations between cell lines. Each experiment was carried out three times with biologic replicates. γ-H2AX IRIF was present equally in all cell lines and routinely measured as a positive control for IRIF.

For BRCA1 focus formation analysis, tissue sections were deparaffinized with xylene and hydrated with decreasing concentrations of ethanol. For target antigen retrieval, sections were microwaved in Dako Antigen Retrieval Buffer pH 9.0 for 4 minutes at 110°C [a T/T MEGA multifunctional Microwave Histoprocessor (Milestone)]. Sections were cooled down in distilled water for 5 minutes, and then permeabilized with Dako Wash Buffer containing Tween 20 for 5 minutes, followed by incubation in blocking buffer (Dako Wash Buffer with 1% BSA) for 5 minutes. Primary antibodies were diluted in Dako Antibody Diluent and incubated at room temperature for 1 hour [anti-BRCA1 Abcam MS110 diluted 1:200; anti-Geminin (ProteinTech Group, 10802-1-AP, diluted 1:400)]. Sections were washed for 5 minutes in Dako Wash Buffer followed by 5 minutes in blocking buffer. Secondary antibodies (Alexa Fluor 488 or 568) were diluted 1:500 in blocking buffer and incubated for 30 minutes at room temperature. The two-step washing was repeated followed by 5 minutes incubation in distilled water. Dehydration was performed with increasing concentrations of ethanol. Sections were mounted with DAPI ProLong Gold antifade reagent and stored at −20°C. For analyses, tumors from 2 mice per treatment group were analyzed and counted. For each tumor, a minimum of 100 geminin-positive cells were identified and counted for those that had at least 5 BRCA1 foci. BRCA1 foci–positive cells were calculated as a percentage of geminin-positive cells.

For assessment of Ki-67 and γ-H2AX by IHC, slides were deparaffinized and hydrated. Antigen retrieval was performed using pH 9 EDTA buffer (DAKO, S2368). Endogenous peroxidases were quenched by immersion of slides in 3% hydrogen peroxide solution (30% H₂O₂, Fisher BP2633-500, diluted in methanol). Primary antibody Ki-67 (Clone EP5), Epitomics, (1:1,500), or γ-H2AX (pS139; N1-4131, EMD; 1,20,000) were diluted with DaVinci Green Diluant (Biocare Medical, PD900) and incubated on slides overnight at 4 degrees in a humidified slide chamber. Slides were then washed 3 times in TBST and incubated with Envision+ System HRP Labelled Polymer Anti-Rabbit (Dako, K4003) for 1 hour at room temperature. Specimens were washed 3 times in TBST and then developed with DAB solution (Dako, K3468) and counterstained in Meyer Hematoxylin (Dako, S3309). For analyses of Ki67 and γ-H2AX expression, a minimum of 2 tumors derived from 2 separate mice were used. Mice were treated with rucaparib 150 mg/kg two times per day for 4 continuous days. For cisplatin, mice were treated with a single dose of 6 mg/kg. Tumors were resected and formalin fixed 4 days from the first dose. A minimum of 2 tumors and 3 images per tumor were used to calculate staining intensities. Image staining intensities were measured using ImageJ analysis software.

For PDXs, patient consent for tumor use in animals was completed under a protocol approved by the Vall d'Hebron Hospital Clinical Investigation Ethical Committee and Animal Use Committee. Mice were maintained and treated in accordance with institutional guidelines of Vall d'Hebron University Hospital Care and Use Committee. Tumors were subcutaneously implanted in 6-week-old female HsdCpb:NMRI-Foxn1^nu mice (Harlan Laboratories). Animals were supplemented with 1 μmol/L estradiol (Sigma) in the drinking water. Upon xenograft growth, tumor tissue was reimplanted into recipient mice, which were randomized upon implant growth. Mice were allocated in control/treated groups to deliver similar mean and SE when tumor volume reached between 150 and 300 mm³ without any blinding. Vehicle or olaparib was administered at 50 mg/kg per oral gavage 6 days per week. Tumor xenografts were measured with calipers and tumor volumes were determined using the formula: (length × width²). At the end of the experiment, animals were killed by CO₂ inhalation. Tumor volumes are plotted as mean ± SEM. Relative tumor volume RTV for vehicle-treated mice and individual RTVs are shown for olaparib-treated mice. For xenograft studies, MDA-MB-436 cells were subcutaneously implanted in 6-week-old female NSG mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ). When tumors reached approximately 300 mm³, tumors were harvested and cut up into smaller pieces followed by subcutaneous reimplantation. Treatment was initiated when tumors reached between 150 and 180 mm³. Rucaparib was administered at 150 mg/kg twice daily for 10 continuous days with a 2-day break after the first 5 days. Cisplatin was administered at a single dose of 6 mg/kg. Tumors were measured with calipers and tumor volumes calculated as described above. Measurements were carried out every 2 days and mice were euthanized when tumors reached 1,500 mm³ in accordance with institutional guidelines of Fox Chase Cancer Center (Philadelphia, PA).

Mean and SE values were compared using unpaired t tests (GraphPad Software). P < 0.05 was considered statistically significant. Asterisks indicate statistically significant P values. There were similar variances between statistical groups compared. All statistical tests were unpaired t tests unless otherwise stated next to the P value.

To examine the impact of BRCA1 exon 11 mutations on the expression of BRCA1 proteins, we employed a panel of human cancer cell lines that were either BRCA1 wild-type or harbored known deleterious BRCA1 mutations (Supplementary Fig. S2A). Full-length BRCA1 protein was detectable in MDA-MB-231, MCF7, and MDA-MB-468 wild-type BRCA1 cell lines, but was absent in all BRCA1 mutation–containing cell lines. A band corresponding with the predicted molecular weight (∼89 kDa) of the BRCA1-Δ11q isoform was present in BRCA1 wild-type cell lines, as well as L56Br-C1, SUM149PT, and UWB1.289 cell lines that harbored BRCA1 exon 11 frameshift mutations (Fig. 1A).

To establish the identity of the isoform detected below the 100-kDa mark, we transfected cells with a siRNA designed to specifically target BRCA1-Δ11q and measured protein expression. BRCA1-Δ11q siRNA did not impact the levels of the full-length, but dramatically reduced the expression of the predicted BRCA1-Δ11q isoform (Fig. 1B). Furthermore, we confirmed that cells did not harbor secondary BRCA1 reversion mutations or genomic rearrangements (Supplementary Fig. S2).

Next, we investigated the levels of BRCA1 mRNA containing exon 11 (+11) as well as the BRCA1-Δ11q (Δ11q) isoform expression. L56Br-C1, SUM149PT, and UWB1.289 exon 11–mutant cell lines exhibited 3- (P < 0.001), 3- (P < 0.001), and 2.9-fold (P < 0.001) lower expression of +11 transcripts, respectively, relative to MDA-MB-231 cells (Fig. 1C). In contrast, relative Δ11q expression was 1.46- (P = 0.0148), 1.55- (P < 0.001), and 1.58-fold (P = 0.0136) higher in L56Br-C1, SUM149PT, and UWB1.289, respectively, relative to MDA-MB-231 cells expression levels (Fig. 1C).

To investigate the impact of germline mutations on exon 11 splicing, we introduced a series of exon 11 frameshift mutations into a BRCA1-minigene system (Supplementary Fig. S3). Mutation of the exon 11q splice junction disrupted Δ11q-reporter mRNA and protein expression. However, three different frameshift mutations at various locations throughout exon 11 had no effect on Δ11q-reporter expression (Fig. 1D). Moreover, RNA-seq analyses of mRNA splice junctions did not indicate that higher levels of BRCA1-Δ11q expression in cell lines with exon 11 mutations were due to increased global alterations in splicing (Supplementary Table S1).

To further investigate the effects of exon 11 mutations on BRCA1-Δ11q levels, we used CRISPR/Cas9 technology to generate SUM149PT cells that either had additional out of frame mutations, or reversion mutations that restored the reading frame (Supplementary Fig. S4). In line with minigene data, the introduction of additional out-of-frame or frame-restoring mutations in exon 11 did not affect BRCA1-Δ11q expression (Fig. 1E). However, clones 3 and 4 that had restored BRCA1 reading frames, expressing the BRCA1-long-form protein product, had 2.5- (P < 0.0001) and 2.47-fold (P < 0.0001) increased levels of +11 expression, respectively, relative to sg_GFP control cells (Fig. 1E), likely resulting from premature translation termination codon (PTC) removal, enabling +11 containing transcripts to avoid NMD (3).

We confirmed that NMD contributed to the low levels of +11 mRNA detected in exon 11 mutation containing cell lines by treating cells with CHX, an inhibitor of NMD (Fig. 1F; ref. 38). CHX treatment for 5 hours resulted in a 2.8- (P = 0.0007), 3.2- (P = 0.023), and 2.9-fold (P = 0.0024) increase in +11 levels in L56Br-C1, SUM149PT, and UWB1.289 exon 11–mutant cell lines, respectively, but did not affect +11 levels in wild-type MDA-MB-231 cells, or Δ11q mRNA levels in any of the cell lines (Fig. 1F).

The increase in +11 mRNA resulting from inhibition of NMD led us to predict that exon 11–mutant cell lines have robust BRCA1 gene transcription. To test this possibility, we treated cells with the transcription inhibitor ACT and measured BRCA1 mRNA levels. MDA-MB-231 cells treated with ACT had a 1.4-fold (P = 0.0002) reduction in +11 mRNA levels compared with vehicle-treated cells. However, ACT treatment reduced +11 levels 3-fold (P = 0.0133) in SUM149PT cells compared with vehicle-treated cells, and completely abrogated the CHX-mediated increase in +11 mRNA (Fig. 1G). ACT also reduced Δ11q levels 1.6-fold (P = 0.002) and 2-fold (P = 0.021) in MDA-MB-231 and SUM149PT cells, respectively, compared with vehicle treatments (Fig. 1H).

We next measured the ability of exon 11–mutant cells to form BRCA1 and RAD51 foci. BRCA1 wild-type as well as exon 11–mutant cell lines formed robust BRCA1 and RAD51 γ-irradiation–induced foci (IRIF). In contrast, BRCA1 and RAD51 IRIF were low or undetectable in cells that harbored BRCA1 mutations outside of exon 11 (Fig. 2A).

SUM1315MO2, HCC1395, and MDA-MB-436 cell lines were highly sensitive to treatment with the PARPis rucaparib and olaparib. In contrast, exon 11 mutation containing L56Br-C1, SUM149PT, and UWB1.289 cells displayed intermediate PARPi sensitivity. SUM1315MO2, HCC1395, and MDA-MB-436 cell lines were also more sensitive to cisplatin treatment compared with L56Br-C1, SUM149PT, and UWB1.289 cells. The presence of BRCA1 exon 11 mutations did not impact taxol sensitivity (Fig. 2B; Supplementary Table S2).

In cell growth experiments, BRCA1 wild-type cells continuously cultured in the presence of PARPi or cisplatin proliferated at the same rate as vehicle-treated cells (Fig. 2C). Exon 11–mutant cell lines proliferated at a reduced rate in the presence of PARPi or cisplatin (Fig. 2C). MDA-MB-436 and HCC1395 cell lines harbor BRCT domain mutations and lost viability in the presence of PARPi or cisplatin. PARPi and cisplatin initially had a greater impact on the proliferation rate of UWB1.289 cells, but over time proliferation gradually increased, and corresponded with increased expression of BRCA1-Δ11q (Fig. 2D). Importantly, depletion of the BRCA1-Δ11q using 2 BRCA1 shRNAs reduced RAD51 IRIF 6.8- (P = 0.0021) and 5.7-fold (P = 0.0027) in SUM149PT, as well as 6- (P = 0.0014) and 4.7-fold (P = 0.0016) in UWB1.289 cells, compared with nontarget (NT) shRNA cells (Fig. 2E). Furthermore, BRCA1 shRNA sensitized SUM149PT and UWB1.289 cells to both PARPi and cisplatin treatments (Fig. 2F; Supplementary Table S2).

To more finely assess the ability of BRCA1-Δ11q to provide resistance to therapy, we used CRISPR/Cas9–mediated gene editing to modify BRCA1 in SUM149PT cells (Supplementary Fig. S4). SUM149PT cells were subject to an sg_GFP control, or sg_exon11 targeting the mutation-containing region of exon 11. Mutations introduced by sg_exon11 were either out of frame (out-frame) or restored the BRCA1 reading frame (in-frame). Moreover, cell lines that expressed sg_exon22 demonstrated frameshift mutations in exons 22, resulting in loss of BRCA1-Δ11q expression (Fig. 3A and Supplementary Fig. S4). BRCA1-Δ11q expression in sg_exon11–treated clones with unrestored reading frames was identical to sg_GFP control cells and there were no differences in PARPi and cisplatin sensitivity in colony formation assays. However, sg_exon11 in-frame clone 3 was 226- (P = 0.0174) and 3.7-fold (P = 0.0009) more resistant to PARPi and cisplatin, respectively, and clone 4 was 207- (P < 0.0001) and 3.6-fold (P = 0.0058) more resistant to PARPi and cisplatin, respectively, compared with sg_GFP cells. In contrast, sg_exon22 cells were 31.4- (P = 0.0183) and 2.3-fold (P = 0.0059) more sensitive to PARPi and cisplatin, respectively, compared with sg_GFP cells (Fig. 3B).

We also compared the ability of ectopic BRCA1 full-length and BRCA1-Δ11q to rescue PARPi and cisplatin sensitivity (Fig. 3C). MDA-MB-436 cells harbor a BRCA1^5396+1G>A mutation that results in protein misfolding, undetectable BRCA1 protein and RAD51 IRIF, as well as exquisite PARPi and cisplatin sensitivity (20, 39, 40). MDA-MB-436 cells expressing BRCA1 full-length demonstrated robust rescue and were 179- (P = 0.0063) and 4.6-fold (P = 0.0002) more resistant to PARPi and cisplatin, respectively, compared with mCherry control cells. BRCA1-Δ11q was less effective at rescue, and cells were 19- (P = 0.0493) and 2.8-fold (P = 0.0004) more resistant to PARPi and cisplatin, respectively, compared with mCherry control cells (Fig. 3D). Moreover, when we introduced an L304P mutation (equivalent to L1407P in full-length BRCA1) that blocks the BRCA1-PALB2 interaction (41), BRCA1-Δ11q–mediated PARPi and cisplatin rescue was abolished (Fig. 3D).

Both BRCA1 full-length and BRCA1-Δ11q–expressing cells formed BRCA1 and RAD51 IRIF. However, BRCA1-Δ11q expressing cells exhibited 1.9- (P = 0.0053) and 1.9-fold (P = 0.0141) lower levels of BRCA1 and RAD51 IRIF, respectively, than full-length BRCA1-expressing cells, while BRCA1-Δ11q+ L304P–expressing cells had undetectable BRCA1 and RAD51 IRIF (Fig. 3E). Immunoprecipitation of ectopic BRCA1 from MDA-MB-436 cells suggested that BRCA1-Δ11q interacted, although less efficiently, with functional protein partners for full-length BRCA1 (Fig. 3F). We also demonstrated ectopic BRCA1-Δ11q provided therapy resistance in HCC1395 cells and in a mouse embryonic stem cell system (Supplementary Fig. S5; ref. 42).

To examine the significance of BRCA1-Δ11q expression in tumors, we first measured PARPi responsiveness and BRCA1 isoform expression in two individual BRCA1 exon 11–mutant PDX models. Despite both PDX models harboring the same BRCA1-exon 11 2080delA mutation (Supplementary Fig. S2), PDX124 tumors were sensitive and PDX196 tumors were resistant to olaparib treatment (Fig. 4A). We confirmed that tumors did not have reversion mutations or gene rearrangements (Supplementary Fig. S2). Similar to exon 11–mutant cell lines, both PDX124 and PDX196 expressed higher levels of Δ11q compared with +11 mRNA, relative to MDA-MB-231 control cells. However, BRCA1 mRNA levels were significantly greater in PDX196 tumors than in PDX124 tumors, with 26- (P = 0.0001) and 23-fold (P < 0.0001) higher levels of both +11 and Δ11q, respectively (Fig. 4B). BRCA1-Δ11q protein expression was detectable in olaparib-resistant PDX196 tumors, as well as PDX124 tumors that eventually grew through olaparib treatment (Fig. 4C). BRCA1 foci formation was also readily detectable in PDX196 tumors (Fig. 4D).

To further assess the impact of BRCA1-Δ11q on therapy resistance in vivo, we utilized the MDA-MB-436 isogenic cell line panel for xenograft experiments. At 30 days post tumor implantation, rucaparib or cisplatin treatment delayed mean tumor growth 6.8- (P < 0.0001) and 12.9-fold (P < 0.0001), respectively, compared with vehicle-treated mice with mCherry-overexpressing MDA-MB-436 tumors. In contrast, rucaparib and cisplatin treatment delayed mean tumor growth 1.1-fold (P = 0.4455) and 1.28-fold (P = 0.2537) in BRCA1 wild-type and 1.4-fold (P = 0.0872) and 2.3-fold (P = 0.019) in BRCA1-Δ11q–overexpressing tumors, respectively, compared with vehicle-treated mice (Fig. 4E and F).

Kaplan–Meier analyses indicated that the median overall survival (OS) of mice harboring mCherry-expressing tumors that were treated with rucaparib or cisplatin increased 1.7-fold (P = 0.0016, log-rank test) and 2-fold (P = 0.0016, log-rank test), respectively, compared with vehicle treated mice. In contrast, in mice harboring BRCA1 wild-type tumors that were treated with rucaparib or cisplatin, median OS increased 1.3-fold (P = 0.0071, log-rank test) and 1.4-fold (P = 0.0048, log-rank test), respectively, and in mice harboring BRCA1-Δ11q tumors treated with rucaparib or cisplatin median OS increased 1.2-fold (P = 0.1536, log-rank test) and 1.5-fold (P = 0.0333, log-rank test), respectively, compared with vehicle-treated mice (Fig. 4G).

Short-term assessment of pharmacodynamics markers showed that rucaparib increased γ-H2AX positivity similarly in all tumors, but wild-type and BRCA1-Δ11q tumors had 4.5- (P < 0.0001) and 2.2-fold (P = 0.005) lower γ-H2AX positivity, respectively, after cisplatin treatment compared with mCherry-expressing tumors. In addition, wild-type and BRCA1-Δ11q tumors treated with rucaparib had 2.2- (P = 0.0007) and 2.5-fold (P = 0.0059) higher Ki67 positivity, respectively; and tumors treated with cisplatin had 2.6- (P < 0.0001) and 3.7-fold (P = 0.0003) higher Ki67 positivity, respectively, compared with mCherry-expressing tumors (Fig. 4G). In contrast to in vitro data, the degree of PARPi or cisplatin rescue afforded by BRCA1-Δ11q overexpression was more similar to BRCA1 full-length in vivo.

Because exon 11–mutant tumors were capable of expressing BRCA1-Δ11q, we assessed survival outcomes of patients with frameshift mutations located inside (IE11) versus outside (OE11) of exon 11. We analyzed 5-year OS data from the time of initial diagnosis in patients with serous ovarian carcinoma from a previously reported study (43). Here, participants with BRCA1 frameshift mutations OE11 (n = 231, 43%; 95% CI, 36%–49%) presented better OS than noncarriers (n = 1,333, 33%; 95% CI, 30%–36%; P = 0.002, log-rank test) as well as participants with frameshift mutations IE11 (n = 148, 31%; 95% CI, 23%–39%; P = 0.02, log-rank test) at 5 years of follow up. Moreover, participants with frameshift mutations IE11 presented with 5-year OS similar to noncarriers (P = 0.84, log-rank test; Fig. 5A, Supplementary Tables S3 and S4).

Although patients with exon 11 mutations demonstrated worse 5-year survival, we hypothesized that, similar to our PDX data, BRCA1-Δ11q might only be highly expressed in a fraction of exon 11–mutant tumors, and so not all exon 11–mutant patients would necessarily be resistant to therapy. Analyses of primary tumors (unrelated to the above studies) with mutations IE11 showed variable levels of BRCA1-Δ11q expression. Tumor #317 expressed 2.7- (P = 0.0227) and 4-fold (P = 0.0117) greater levels of Δ11q compared with tumor #325 and tumor #411, respectively, despite all tumors harboring deleterious exon 11 mutations (Fig. 5B).

Because high levels of BRCA1-Δ11q expression were required for residual function, we hypothesized that limiting splicing-dependent excision of exon 11q might provide an opportunity to enhance PARPi sensitivity. We first identified two binding elements for the FOX2 splicing site selection factor at position +18 and +63 downstream from the cryptic splice site location (Supplementary Fig. S6; refs. 44, 45). Depletion of FOX2 using 2 individual siRNAs did not impact +11 expression levels, but reduced Δ11q expression 1.7- (P = 0.0043) and 1.6-fold (P = 0.0179) in MDA-MB-231, 1.5- (P = 0.048) and 1.6-fold (P = 0.0301) in UWB1.289, 2.2- (P = 0.0056) and 3.1-fold (P = 0.0017) in SUM149PT cells, compared with scrambled siRNA–treated cells, respectively (Fig. 6A). At the protein level, densitometric analyses of Western blots indicated FOX2 siRNA did not impact full-length BRCA1 protein expression, but BRCA1-Δ11q levels were reduced 2.9- and 2.2-fold in MDA-MB-231, 2- and 2.8-fold in UWB1.289, 1.9- and 1.8-fold in SUM149PT cells, compared with scrambled siRNA–treated cells (Fig. 6B). FOX2 siRNA had no impact on MDA-MB-231 cells PARPi sensitivity, but reduced the LC₅₀ value of rucaparib 3.8- (P = 0.0063) and 27.4-fold (P = 0.0017) in UWB1.289 cells, as well as 1.9- (P = 0.0237) and 5-fold (P = 0.0002) in SUM149PT cells, compared with scrambled siRNA–treated cells (Fig. 6C). Furthermore, BRCA1-minigene analyses demonstrated that mutation of FOX2-binding motifs also partially decreased Δ11q-reporter mRNA and protein levels (Fig. 6D).

We next investigated the ability of Pl-B, a small-molecule inhibitor of the U2 snRNP spliceosome machinery (46), to disrupt BRCA1-Δ11q production. Pl-B treatment reduced both +11 and Δ11q mRNA expression; however, the relative reduction in Δ11q levels was 7.7- (P = 0.002) and 7-fold (P = 0.016) greater compared with the reduction in +11 levels in MDA-MB-231 and UWB1.289 cells, respectively (Fig. 6E). Furthermore, full-length BRCA1 protein levels were reduced but remained readily detectable after treatment with up to 10 nmol/L Pl-B in MDA-MB-231 cells. However, BRCA1-Δ11q protein levels were barely detectable at 1.25 nmol/L Pl-B in both MDA-MB-231 and UWB1.289 cells (Fig. 6F). Simultaneous incubation with both Pl-B and rucaparib did not significantly impact MDA-MB-231 cells, but resulted in a 2.8- (P = 0.0296) and 2.1-fold (P = 0.0426) reduction in colony formation compared with cells treated with rucaparib only in UWB1.289 and SUM149PT cells, respectively (Fig. 6G). Furthermore, ectopic overexpression of a BRCA1-Δ11q cDNA that did not depend on exon 11q splicing, resulted in a 1.9-fold (P = 0.001) rescue in colony formation in cells treated with the Pl-B and rucaparib combination, compared with GFP-expressing control cells (Fig. 6H), suggesting that Pl-B–induced PARPi sensitization was mediated, in part, through the reduction in BRCA1-Δ11q levels.

In the current study, we provide evidence that BRCA1 splice isoforms lacking exon 11 are capable of producing truncated but hypomorphic proteins. It was previously shown that the exon 11–deficient Brca1-Δ11 isoform partially compensates for the lack of other Brca1 isoforms during embryogenesis (31–33, 47). Here, we show that BRCA1-Δ11q can also partially compensate for full-length BRCA1 in response to homologous recombination targeting therapeutics. BRCA1-Δ11q retains, although less efficiently, many of the protein–protein interactions carried out by full-length BRCA1. We identified the BRCA1–Δ11q–PALB2 interaction as critical for BRCA1-Δ11q–mediated RAD51 IRIF and resistance.

BRCA1-Δ11q was expressed in BRCA1 wild-type as well as exon 11–mutant cell lines. Although frameshift mutations in exon 11 resulted in the NMD of exon 11 containing transcripts, they did not impact the rate of alternative splicing or directly affect BRCA1-Δ11q levels. In support of this, PDX124 and PDX196 tumors both harbored identical 2080delA exon 11 mutations; however, both +11 and Δ11q transcript levels were significantly higher in PDX196 relative to PDX124 tumors, suggesting that overall BRCA1 gene transcription, rather than the rate of alternative exon splicing, was elevated. Both patients whose tumors were used to derive PDX124 and PDX196 initially demonstrated clinical responsiveness to olaparib therapy; while PDX124 was derived prior to the patient starting olaparib therapy, PDX196 was derived at the time of clinical tumor progression on olaparib. We anticipate that high levels of BRCA1-Δ11q expression may be selected for in response to chemotherapy. Primary patient tumors harboring exon 11 mutations also demonstrated variable BRCA1-Δ11q levels, and this isoform might only be overexpressed in a subset of exon 11 mutation carriers. Additional resistance mechanisms such as secondary reversion mutations have also been shown to occur in tumors with exon 11 mutations (48, 49).

Intriguingly, our analyses indicated that exon 11 mutation carriers with ovarian cancer had a worse 5-year overall survival compared with non-exon 11 mutation carriers (Fig. 5A). Another recent study of mutation-specific cancer risks identified the exon 11 region of BRCA1 as an ovarian cancer cluster region, where mutation carriers had a relative decrease in breast relative to ovarian cancer risk (50). Further work is required to determine whether BRCA1-Δ11q has any impact on long-term survival outcomes or breast versus ovarian tissue–specific phenotypes.

Currently, an understanding of PARPi and platinum resistance is incomplete. Our findings provide evidence for a role of the BRCA1-Δ11q alternative splice isoform in promoting resistance. Moreover, we show that inhibition of the spliceosome reduced BRCA1-Δ11q levels and sensitized exon 11–mutant cell lines to PARPi, potentially offering a strategy to resensitize tumors that express high levels of BRCA1-Δ11q.

C. Gourley reports receiving a commercial research grant, has received speakers bureau honoraria, and is a consultant/advisory board member for AstraZeneca. G.I. Shapiro reports receiving commercial research support and is a consultant/advisory board member for Clovis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Wang, C. Cruz, D. Lambrechts, A.D. D'Andrea, G.I. Shapiro, E.M. Swisher, N. Johnson

Development of methodology: Y. Wang, A.J. Bernhardy, C. Cruz, E. Nicolas, V. Serra, N. Johnson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Wang, A.J. Bernhardy, C. Cruz, J.J. Krais, J. Nacson, E. Nicolas, I. van der Heijden, S.W. O'Brien, M.I. Harrell, S.F. Johnson, F.J. Candido dos Reis, P.D.P. Pharoah, B. Karlan, C. Gourley, D. Lambrechts, G. Chenevix-Trench, H. Olsson, M.H. Greene, M. Gore, R. Nussbaum, S. Sadetzki, S.K. Kjaer, kConFab Investigators, G.I. Shapiro, D.C. Connolly, M.B. Daly, E.M. Swisher, P. Bouwman, J. Jonkers, J. Balmana, V. Serra, N. Johnson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Wang, C. Cruz, E. Nicolas, S. Peri, I. van der Heijden, Y. Zhang, F.J. Candido dos Reis, P.D.P. Pharoah, B. Karlan, J.J. Benitez, S.K. Kjaer, G.I. Shapiro, D.L. Wiest, P. Bouwman, J. Jonkers, J. Balmana, V. Serra, N. Johnson

Writing, review, and/or revision of the manuscript: Y. Wang, C. Cruz, F.J. Candido dos Reis, P.D.P. Pharoah, B. Karlan, C. Gourley, G. Chenevix-Trench, H. Olsson, M.H. Greene, M. Gore, R. Nussbaum, S. Sadetzki, S.A. Gayther, S.K. Kjaer, kConFab Investigators, A.D. D'Andrea, G.I. Shapiro, D.L. Wiest, D.C. Connolly, M.B. Daly, E.M. Swisher, J. Balmana, V. Serra, N. Johnson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.J. Bernhardy, B. Karlan, C. Gourley, D. Lambrechts, R. Nussbaum, P. Bouwman, V. Serra

Study supervision: S.K. Kjaer, P. Bouwman, J. Balmana, V. Serra, N. Johnson

Other (performed experiments): H. van der Gulden

The authors thank the FCCC Biostatistics and bioinformatics, Biorepository and Genomics facilities and Yasir H. Ibrahim, Pilar Antón, Ana Vivancos, Orland Diaz, Michael Slifker, Yusheng Li for helpful experimental support and data analysis.

This work was supported by US NIH grants P50 CA083638 (Fox Chase Cancer Center (FCCC) Specialized Program of Research Excellence (SPORE) in Ovarian Cancer and R21CA191690, 5P30 CA006927 (FCCC Developmental New Investigator funds), Susan G. Komen Career Catalyst Award CCR12226280, and OC130212 Department of Defense Ovarian Academy Award, Basser Center for BRCA pilot project award (N. Johnson), P50 CA83636 (Pacific Ovarian Cancer Research Consortium SPORE in Ovarian Cancer), the Wendy Feuer Ovarian Cancer Research Fund (E.M. Swisher). M.H. Greene was supported by the Intramural Research Program of the US National Cancer Institute. J. Balmana is recipient of an Instituto de Salut Carlos III (ISCIII) grant PI12/02606. C. Cruz is supported by a Sociedad Española de Oncología Médica (SEOM)- BUCKLER 0′0 grant; Conquer Cancer Young Investigator Award; and Asociación Española contra el Cáncer (AECC). This work was also supported by a GHD/FERO (V. Serra).

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.