Efficacy of Carboplatin Alone and in Combination with ABT888 in Intracranial Murine Models of BRCA-Mutated and BRCA–Wild-Type Triple-Negative Breast Cancer Free

Brain metastases are a particularly devastating complication for patients with triple-negative breast cancer (TNBC) with a limited survival and few therapeutic options (1). TNBC lacks expression of the estrogen receptor (ER), the progesterone receptor (PR), and amplification of the HER-2 gene (2). When compared with other breast cancer subtypes, TNBC has an increased potential to spread to the central nervous system (CNS; refs. 3–5); approximately 50% of patients with advanced TNBC will recur in the brain (4). Despite local cranial therapies (i.e., radiation and/or neurosurgery), overall survival (OS) from detection of CNS metastases for patients with TNBC is less than 6 months (1). Currently, there are no FDA-approved systemic or targeted therapies for patients with TNBC brain metastases. The treatment of CNS metastases arising from TNBC is limited by both the lack of defined biologic targets and the inability of the majority of anticancer agents to penetrate the blood–brain barrier (BBB). Thus, there is an urgent need to develop safe and effective treatments for this aggressive disease.

Although TNBC is a distinct clinical subtype, it is a molecularly heterogeneous disease. Given the lack of defined biologic targets in TNBC, it is important to study possible molecular drivers to develop successful therapies. Gene-expression studies have shown that greater than 75% of TNBC are basal-like or claudin-low subtype (6, 7). With approximately 20% of TNBC patients harboring a mutation in either the BRCA1 or BRCA2 genes (8), exploitation of this molecular pathway with brain-penetrant cytotoxic agents could be used as a therapeutic strategy to treat TNBC brain metastases.

The BRCA family is responsible for repairing DNA double-stranded breaks (DSB) via homologous recombination (HR; ref. 9). Thus, tumors harboring BRCA mutations are sensitized to DNA-damaging cytotoxic agents such as platinum derivatives (i.e., cisplatin or carboplatin) that bind to DNA and form DNA cross-links, leading to DNA DSBs (10). Following platinum treatment of BRCA-mutated breast cancer cells, at least two possibilities may occur: (i) the cells are unable to repair widespread DSBs and undergo apoptosis, or (ii) the cells rely on the base excision machinery to rescue DSBs via the enzyme PARP (11, 12). One promising strategy to enhance the cytotoxic effect of platinum therapies in BRCA-mutated TNBC is concomitant PARP inhibition. This strategy may be particularly advantageous for TNBC BRCA–mutated brain metastases as several clinically available PARP inhibitors cross the BBB (13).

Historically, platinums in combination with PARP inhibitors have shown activity in in vitro and in vivo models of BRCA-associated TNBC (14–16), and are brain penetrant. The combination of platinum/PARP inhibitor therapy has never been examined in breast cancer brain metastases. The novelty of this preclinical study is the exploration and efficacy of brain-permeable, clinically available compounds in animal models of TNBC brain metastases, with the goal to translate these findings into the clinical setting. We present a novel study using preclinical murine models of intracranial TNBC, comparing both BRCA-mutant and wild-type (wt) models, as well as basal-like and claudin-low subtypes in response to systemic platinum ± PARP inhibitor combination therapy. This study uniquely demonstrates brain penetration of both systemically administered carboplatin and ABT888, improved survival in BRCA-mutant intracranial TNBC models, and molecular mechanisms of DNA damage and increased apoptosis in response to therapy. Our results suggest that carboplatin ± ABT888 may be a viable therapeutic option for patients with TNBC brain metastases.

SUM149 (BRCA1-mutant and basal-like), MDA-MB-436 (BRCA1-mutant and claudin-low), MDA-MB-468 (BRCA-wt, basal-like), and MDA-MB-231BR (BRCA-wt, claudin-low) were selected for the study and obtained from the ATCC unless otherwise specified. The identity of the cell lines was confirmed by gene expression (September 2010). All media and additives were purchased from Invitrogen. SUM149 (17) was cultured in HuMEC with supplements and 5% FBS. MDA-MB-468 was cultured in RPMI-1640 plus 10% FBS in a flask with plug seal cap (Corning). MDA-MB-231BR and MDA-MB-436 were maintained in DMEM (high glucose) plus 10% FBS. All cell lines except for MDA-MB-436 were transduced with pTK1261 vector carrying firefly Luciferase as previously described (18). Cell lines were maintained at 37°C and 5% carbon dioxide in the presence of antibiotic–antimycotic solution (Invitrogen). Cells were harvested immediately before intracranial implantation.

Ten-week-old female Foxn1nu/nu mice (UNC Animal Studies Core) weighing at least 20 g before intracranial tumor implantation were used for all studies. Intracranial tumor implantation was performed as previously described (18). All animals were handled and monitored for health conditions according to the Institutional Animal Care and Use Committee (IACUC) approved protocols and entered the study only if recovered following surgery. Bioluminescence imaging was performed weekly to monitor intracranial tumor growth as described previously (18).

For intracranial tumor models expressing Luciferase (SUM149, MDA-MB-468, and MDA-MB-231BR), bioluminescence signal was used to evenly distribute tumor size among the treatment groups within each model. The average bioluminescence signal between treatment groups within each model was not statistically different before treatment (data not shown). MDA-MB-436–injected animals were assigned to treatment groups randomly.

At 14 days following tumor implantation for all models except for MDA-MB-231BR (7 days due to aggressive tumor phenotype), treatment started. Health and weight of the animals were monitored at least three times weekly until sacrifice due to clinical symptoms representing poor health condition (i.e., decreased response to stimuli, neurologic dysfunction, weight loss of 20%, and/or a body composition score of 2 or less) or 12 weeks after intracranial implantation as dictated by prespecified IACUC guidelines for humane reasons. Treatment doses were selected on the basis of previous literature (16, 19, 20) and dose-escalation studies (data not shown), and were: 100 μL PBS control (IP, weekly), 50 mg/kg/wk carboplatin (IP; 10 mg/mL sterile aqueous solution, University of North Carolina Hospital Pharmacy, Chapel Hill, NC), 25 mg/kg/d ABT888 (oral gavage, OG; Chemistry Center for Integrative Chemical Biology and Drug Discovery, UNC) dissolved in PBS (Life Technologies) immediately before OG, or combination 50 mg/kg/wk carboplatin + 25 mg/kg/d ABT888 (IP and OG, respectively). For the SUM149 model, survival data from three independent experiments were combined on the basis of similar median survival for all control groups (data not shown, P = 0.35). No toxicities were found throughout the survival studies in any model or treatment group.

Four mice from each intracranial model were used. Gadolinium enhanced brain MRI on a BioSpec 9.4 T small animal imaging system (Bruker Corporation) under isoflurane anesthesia began 1 week following MDA-MB-231-BR injection and 2 weeks following SUM149, MDA-MB-436, and MDA-MB-468 injection. Precontrast coronal T1 and T2 images were obtained for localization. Precontrast variable flip angle gradient echo T1 sequences (TE = 1.5 msec, TR 15 msec, flip angles 2, 5, 10, and 15 degrees) were performed followed by a dynamic scan using 15 degree flip angle was repeated for 130 scans (8.6 seconds/repeat) for dynamic contrast enhancement. After the fifth scan, gadopentate dimeglumine (Magnevist; Bayer Healthcare) was administered at 0.1 mmol/kg followed by saline flush via tail vein catheter. Coronal post-contrast T1 images were obtained. Standard physiologic monitoring occurred during imaging procedure. Animals were recovered and returned to standard housing. Mice were reimaged each week until time of sacrifice due to 20% weight loss or signs of tumor burden.

Using the T1 MR 20th and third scan of the dynamic sequence, a difference map was generated and relative enhancement (RE) was calculated as follows:

where Δ[Gd]_T and Δ[Gd]_P represent contrast concentration changes in tissue and plasma during time zero and t, respectively, S_T(t) is signal intensity in selected tissue (tumor, contralateral normal appearing brain tissue, or left lateral ventricular choroid plexus) at time point t, S_P(t) is signal intensity in plasma at time point t, and S_T0 and S_P0 are precontrast signal intensity in selected tissue or plasma, respectively.

Multiple regions of interest (ROI) sized 10 to 15 pixel (0.1 mm²) were manually drawn by one blinded reader on T2 images, guided by the lesion shape, avoiding macroscopic vessels, necrosis, and areas of hemorrhage. Region with greatest signal change was selected as the tumor ROI. ROIs were propogated to the subtracted T1 DCE MR images, acquired in anatomic registration. The same size ROIs were drawn on contralateral amygdala, choroid plexus of the contralateral lateral ventricle, and contralateral posterior facial vein lumen.

An in vivo PAR assay was performed using intracranial tissue from the TNBC SUM149 intracranial murine model. Animals were divided into treatment groups, and drug was administered using doses specified in efficacy studies. At harvest, tumors were flash frozen in liquid nitrogen, and stored at −80°C until further analysis. Net PAR levels were determined in SUM149 intracranial tumors collected immediately following 14 days of treatment with either control or ABT888. To determine net PAR levels in the protein extracts from intracranial tumors, a high-throughput chemiluminescent ELISA PARP in vivo Pharmacodynamic Assay was performed (Trevigen). Frozen SUM149 intracranial tumors from Control (n = 11) or ABT888 (n = 8) were homogenized in Lysis buffer, sonicated three times for 10 seconds each, and assayed for PAR according to the manufacturer's protocol. Luminescence signal of the experimental samples and PAR standards were recorded in photons/seconds using an IVIS Lumina camera (Caliper Life Sciences) and corrected for background. PAR concentrations in the experimental samples were calculated using a standard curve, and PAR levels were determined as pg/100 μg of total protein. Experimental samples and freshly made PAR standards were assayed in triplicate. The assay was repeated once for samples with sufficient amount of protein extract. Concentration of the total protein in the extract was measured using the BCA Assay (Thermo Fisher Scientific).

To examine the ability of ABT888 to inhibit PAR independently of BRCA status, in vitro PAR levels of BRCA-mutant SUM149 and BRCA-wt MDA-MB-231BR were examined at the IC₅₀ value of the SUM149 cell line (60 μmol/L). SUM149 and MDA-MB-231BR cells were plated at 1.5 and 1.0 million cells per 6-cm dish, respectively, in triplicate. After 24 hours, media were changed to either ABT888 (60 μmol/L) or 0.1% DMSO vehicle control. After 2 hours of inhibition, cells were harvested, and PAR was measured according to ELISA PARP in vivo Pharmacodynamic Assay protocol (Trevigen).

For IHC, we collected tissue from BRCA-mutant SUM149 and BRCA-wt MDA-MB-468 TNBC intracranial murine models. Animals were divided into treatment groups within each model and treated for 14 days after tumor injection with 3 doses of carboplatin, 15 doses of ABT888, or carboplatin + ABT888 in combination. Brain tissues from 3 animals per group were collected in formalin the day following the last dose of carboplatin and/or immediately after the last dose of ABT888. Whole mouse brains fixed in formalin were cut parasagittaly approximately 2 mm of midline, incubated in 70% Ethanol, embedded in paraffin, and sectioned immediately before staining.

The following antibodies were used to stain 4-μm sections placed on coated glass slides: Rabbit monoclonal anti–phospho-ser139-H2AX [γ-H2AX] (1:2,000, 2 hours; Cell Signaling Technology), rabbit polyclonal cleaved caspase-3 (cC3; 1:50, 1 hour; Biocare Medical). Staining was performed as previously described (21, 22). IHC was carried out on the Bond Autostainer, and all solutions were from Leica Microsystems. Briefly, slides were deparaffinized in Bond dewax solution (AR9222) and hydrated in Bond wash solution (AR9590). Antigen retrieval of γ-H2AX was performed for 30 minutes at 100°C in Bond-epitope retrieval solution 1 pH-6.0 (AR9961), and in solution 2 pH-9.0 (AR9640) for cC3. After incubation with the appropriate antibody, detection was performed with the Bond Polymer Refine Detection System (DS9800). Stained slides were dehydrated and coverslipped. Positive and negative (no primary antibody) controls were included for each antibody. Sections from two to three biologic replicas and two to three technical replicas (total 6 sections) per treatment group were stained.

Hematoxylin and eosin (H&E)– and IHC-stained sections were digitally imaged (×20 objective) using Aperio ScanScope XT (Aperio Technologies) and analyzed within the Aperio Spectrum Database. Folded tissues, nontumor areas, and artifacts were excluded from the analysis using a negative pen. Tumor area identification was guided by H&E staining of an adjacent section. The expression of γ-H2AX and cC3 was measured using the Aperio Nuclear V9 (cell quantification) algorithm. Positive and negative (no primary antibody) controls were included for each antibody. The intensity score and the percentage of γ-H2AX–positive nuclei in the tumor area obtained with the modified nuclear v9 algorithm at each intensity level were used to calculate the H-Score using the formula: H-Score = (% at 1+) × 1 + (% at 2+) × 2 + (% at 3+) × 3. The H-Score was normalized to a γ-H2AX–positive SK-MEL 181 cell line control included in each staining. Expression of cC3 protein was calculated using the percentage of cC3-positive nuclei.

RNA from BRCA-mutant intracranial SUM149 and MDA-MB-436 tumors was purified using the RNeasy Mini Kit (Qiagen) following the manufacturer's protocol, labeled as previously described (23), amplified incorporating cyanine-5 (Cy5) dye for the tumor samples and cyanine-3 (Cy3) dye for the reference (24) using the Agilent Low RNA Input Fluorescent Linear Amplification Kit (Agilent Technologies), and hybridized to 4 × 44 K Customized Human Oligomicroarrays (Agilent Technologies). Arrays were scanned with either the Agilent Scanner (Agilent Technologies) or the GenePix Scanner (Molecular Devices) as previously described (25). All microarray data have been deposited in the Gene Expression Omnibus under the accession number GSE55399.

SUM149 arrays scanned with the GenePix scanner were normalized to be comparable with the Agilent scanner, were adjusted according to Supplementary Materials and Methods (Supplementary Fig. S1). Principle component analysis demonstrates effective removal of scanner bias (Supplementary Fig. S2). Treatment-specific gene sets were identified using R v3.0.2 and the two class unpaired significance analysis of microarray (v3.11; ref. 26) and as further described in Supplementary Materials and Methods (Supplementary Fig. S3). The following comparisons were conducted within each model: Control versus carboplatin, Control versus ABT888, Control versus carboplatin + ABT888, and carboplatin versus carboplatin + ABT888 (Supplementary Table S1B–S1I). All genes with an FDR = 0 in control versus carboplatin intracranial SUM149 or MDA-MB-436 tumors were analyzed with DAVID (27, 28). Pathways are reported with a Bonferroni corrected P value of <0.05 (Supplementary Table S1J–S1K). For comparative visualization, SUM149 arrays were scaled to the same dynamic range as the MDA-MB-436 arrays (Supplementary Fig. S4). Unsupervised and supervised hierarchical clustering were clustered using centroid linkage in Gene Cluster 3.0 v1.52 (29) and viewed with Java Treeview v1.1.6r4 (30).

Results for BBB permeability are reported as mean RE ± SEMs. PAR assay results were normalized to the percentage of the respective control group's median and are presented as mean of the normalized values ± SEM. IHC to evaluate DNA damage and apoptosis is reported as mean values ± SEM. One-way ANOVA tests were used to compare mean values between treatment groups using GraphPad Prism 6.04. Unpaired t tests were performed for all pair-wise comparisons, with Bonferroni correction for multiple comparisons when applicable.

The Kaplan–Meier method and log-rank tests were used to compare OS among treatment groups. Mice were censored at prespecified study endpoints if they had not been previously sacrificed. Overall and pair-wise tests were completed, and unadjusted P values are reported. For bioluminescence imaging, fold changes were calculated relative to the start date of treatment, and if present, negative imaging values (due to correction for background) were recorded and set to zero. The common logarithm of these values was used for this analysis. For every time point where at least two animals were alive in the treatment group, the median level and interquartile range (IQR; 25th–75th percentile) for bioluminescence imaging were calculated. For gene-expression analysis, significance analysis of microarray was used through R v. 2.15.0 (26).

Our preclinical study evaluates efficacy of carboplatin therapy as a single agent and in combination with the PARP inhibitor, ABT888, in intracranial orthotopic murine models of BRCA-mutant and BRCA-wt TNBC.

To assess the inherent differences in therapy accessibility due to the vasculature across all intracranial murine models, we performed MRI with Gadolinium contrast. Because contrast can only enter brain tissue in which the BBB is disrupted, RE estimates the BBB permeability. Tumor mean RE for each intracranial model was not significantly different across models (SUM149, 1.081 ± 0.1084; MDA-MB-436, 0.8711 ± 0.1788; MDA-MB-468, 0.7683 ± 0.2210; MDA-MB-231BR, 0.6835 ± 0.2052; P = 0.48). In addition, RE was significantly higher in all tumors compared with matched normal brain tissue across all models (P < 0.001; Fig. 1A), indicating a similarly compromised, leaky BBB within all intracranial tumor models. There was no significant difference across models within the normal (P = 0.14), tumor (P = 0.48), and choroid plexus (P = 0.82) tissues (Fig. 1A). The small amount of RE in normal tissues is in agreement with previous studies, representing normal cerebral blood volume (31). Representative MRI images are shown for each model and demonstrate that the morphologic appearance of the intracranial tumors formed by each model is a similar, focal, concentrated mass with defined borders in the right hemisphere in vivo (Fig. 1B).

PARP catalyzes the formation of poly(ADP-ribose) (PAR) chains on a variety of proteins, including PARP itself. To assess penetration of the PARP inhibitor ABT888 across the BBB and monitor inhibitory activity, we measured net PAR levels (±SEM) in SUM149 intracranial tumors following 14 days of ABT888 treatment. PAR levels in ABT888-treated SUM149 intracranial tumors (18.76 ± 7.566) were significantly decreased as compared with control tumors (119.8 ± 30.37; P = 0.0091), indicating 84.3% inhibition and intracranial tumor exposure with ABT888 treatment (Fig. 2A).

To assess differences in the ability to form PAR chains in BRCA-mutant as compared with BRCA-wt models, in vitro PAR assays of BRCA-mutant SUM149 and BRCA-wt MDA-MB-231BR models were performed following 2 hours of treatment with 60 μmol/L of ABT888. PAR levels in SUM149 cells were significantly decreased with ABT888 treatment (5.820 ± 2.044) as compared with control (72.89 ± 14.25; P = 0.0002). In parallel, PAR levels in MDA-MB-231BR were significantly decreased in ABT888 treated cells (5.518 ± 2.763) as compared with control cells (114.1 ± 10.01; P < 0.0001). Similar percentage inhibition was observed between the BRCA-mutant and BRCA-wt cell lines (SUM149, 92.0% inhibition; MDA-MB-231BR, 95.2%; Fig. 2B). Taken together, these data demonstrate effective ABT888 penetration into intracranial tumor and similar relative inhibition in vitro between a BRCA-mutant and BRCA-wt model.

Recognizing the BBB permeability of carboplatin and ABT888, the efficacy of platinum agents and PARP inhibitors in advanced extracranial BRCA-mutant TNBC, and the potential rational therapeutic combination to treat breast cancer brain metastases, we evaluated the efficacy of carboplatin and ABT888 as single agents and in combination in both BRCA-mutant and BRCA-wt TNBC intracranial murine models. Median survivals are reported for each model by the treatment group (Table 1).

In the basal-like BRCA1-mut SUM149 intracranial model, the median survival of animals treated with carboplatin ± ABT888 was both significantly longer then control [carboplatin, 58 days (95% confidence interval (CI), 47–67 days]; carboplatin + ABT888: 64 days (95% CI, 59–75 days); Control: 36 days (95% CI, 34–40 days); both treatment groups P < 0.0001 relative to control; Table 1 and Fig. 3A). The difference between median survival of control and single-agent ABT888 treated animals (39 days; 95% CI, 30–46 days) was not statistically significant (P = 0.2). In addition, ABT888 alone was inferior compared with therapies with carboplatin as a single agent (P = 0.007) or carboplatin + ABT888 combination (P = 0.0012). Despite theoretical promise, the carboplatin + ABT888 combination therapy exhibited modest improvement of survival in comparison with single-agent carboplatin, and was not statistically significant (P = 0.4; Table 1 and Fig. 3A). Of note, 2 of 16 and 2 of 17 animals from the carboplatin and carboplatin + ABT888 groups, respectively, remained alive at the prespecified study completion date (12 weeks after intracranial implantation of tumor cells).

Recognizing the heterogeneity of TNBC, we investigated response to carboplatin and carboplatin + ABT888 in a second BRCA-mutant intracranial TNBC model, MDA-MB-436. Median survival following treatment with carboplatin alone (86 days; 95% CI, 61 to undefined) and with combination carboplatin + ABT888 (not reached, ≥65 days) was both significantly longer (both P = 0.001) as compared with control (44 days (95% CI, 35–47 days); Table 1, Fig. 3B). Single-agent ABT888 resulted in a modest survival advantage compared with control (57 days; 95% CI, 44–74 days; P = 0.03). The addition of ABT888 to carboplatin did not yield a significant improvement in survival in the MDA-MB-436 model compared with carboplatin alone (P = 0.2); however, 4 of 5 animals in the combination group versus 2 of 5 animals in the carboplatin alone group remained alive 12 weeks after cell injection, suggesting benefit of the addition of ABT888 to carboplatin.

To examine the efficacy of carboplatin and carboplatin + ABT888 in BRCA-wt intracranial TNBC models, we applied the same experimental design to BRCA-wt intracranial TNBC models, basal-like MDA-MB-468 and claudin-low MDA-MB-231BR. In contrast with the BRCA-mutant TNBC intracranial models, neither single-agent carboplatin, ABT888, nor combination carboplatin + ABT888 resulted in an improved median survival compared with control in either BRCA-wt TNBC intracranial model (Supplementary Fig. S5A and S5B; MDA-MB-468, P = 0.8; MDA-MB-231BR, P = 0.1).

In addition to the survival analysis, we examined the effect of treatment with carboplatin alone or in combination with ABT888 on intracranial tumor growth via bioluminescence in the BRCA-mutant SUM149 model. Consistent with OS results, dynamic changes in the intracranial tumor growth as measured by median log-fold change of the bioluminescence signal intensity (photons/second) from the start of the treatment were lowest following treatment with carboplatin ± ABT888 (Fig. 3C). The median log-fold change in intracranial tumors treated with either carboplatin or carboplatin + ABT888 was similar at 8 weeks of treatment (median 3.3; IQR, 2.85–4.14 vs. median 3.32; IQR, 2.56–3.97, respectively]. Although based on small numbers of animals remaining, at 10 weeks after treatment, the median log-fold change for the carboplatin + ABT888 group remained low until the end of the study, whereas median fold change for the carboplatin single-agent group continued to increase over time (Fig. 3C).

To investigate whether processes of DNA damage and apoptosis are activated in intracranial tumors in response to treatment, we evaluated expression of γ-H2AX protein, a known biomarker of DNA damage, and cC3, the central protein in the terminal phase of apoptosis (15, 32–34).

Expression of γ-H2AX was elevated in all treatment groups compared with control following 14 days of treatment in BRCA-mutant SUM149 intracranial tumors (Fig. 4A, P = 0.0106). DNA damage was highest in the carboplatin group (91.0 ± 3.7, P = 0.001) followed by combination carboplatin + ABT888 (76.6 ± 8.8, P = 0.12), which were both higher than control (58.5 ± 6.0). ABT888 treatment resulted in a γ-H2AX H-score (75.4 ± 4.3) that was higher than control (P = 0.045).

With similar BBB permeability observed between the BRCA-wt and BRCA-mutant models, significant differences in DNA damage were also observed in the BRCA-wt MDA-MB-468 intracranial model for both carboplatin and carboplatin + ABT888 treated groups as compared with control (P = 0.0015). Expression of γ-H2AX, however, was notably weaker in the MDA-MB-468 intracranial tumors as compared with the SUM149 intracranial model (P < 0.0001 MDA-MB-468 Control vs. SUM149 Control, Fig. 4A, C, and D). Attenuated DNA damage after DNA adduct formation rather than differences in exposure to carboplatin may provide an explanation to why carboplatin treatment ± ABT888 was effective in the BRCA-mutant and not in the BRCA-wt intracranial models.

In addition to γH2AX expression, the same tumor samples from BRCA-mutant SUM149 and BRCA-wt MDA-MB-468 intracranial models were analyzed to determine expression levels of cC3 protein. In BRCA-mutant SUM149 intracranial tumors, mean percentages of cC3-positive cells for all treatments were significantly higher in comparison with Control (7.7 ± 0.4, P = 0.0297, Fig. 4B, E, and F); however, mean percentages of cC3-positive cells (±SEM) by the treatment group were not significantly different from each other (ABT888 15.5 ± 2.6, carboplatin 12.8 ± 1.5, and carboplatin + ABT888 12.2 ± 1.5; P = 0.46).

In the BRCA-wt MDA-MB-468 intracranial model, mean values of the percentage of cC3-positive cells for ABT888 (14.5 ± 2.7), carboplatin (7.0 ± 2.0), and carboplatin/ABT888 (13.0 ± 2.0) exhibited overall borderline significance when compared with control (8.2 ± 1.3; P = 0.051; Fig. 4B, Supplementary Fig. S6).

Taken together, these data support previous findings that BRCA status has a significant impact on sensitivity to carboplatin treatment through increased DNA damage and apoptosis levels, and provides a possible explanation to why BRCA-wt intracranial tumors are less responsive to carboplatin treatment ± ABT888.

To examine the dynamic transcriptional changes that occur in intracranial tumors in response to treatment, we performed gene-expression analysis of BRCA-mutant intracranial tumors from mice treated with control, ABT888, carboplatin, and carboplatin + ABT888 in the BRCA-mutant basal-like SUM149 and the BRCA-mutant claudin-low MDA-MB-436 intracranial models.

Unsupervised hierarchical clustering of all samples identified cell line–specific delineations as well as distinct clusters defined by carboplatin treatment ± ABT888 (Supplementary Fig. S7). Independent analyses within each model further demonstrated little to no contribution with the addition of ABT888 to control or carboplatin treatments (Fig. 5A and B).

To quantitate gene-expression changes that occurred with treatment, each treatment group was compared with control within SUM149 and separately within MDA-MB-436 tumors (Supplementary Table S1B–S1I). Treatment with ABT888 as compared with control (Supplementary Table S1B And S1C) resulted in 19 differentially expressed genes for the SUM149 model, and 153 differentially expressed genes for the MDA-MB436 model. In contrast, comparison of carboplatin treatment with control (Supplementary Table S1D and S1E) demonstrated a substantially higher number of genes differentially expressed within each model (carboplatin vs. control: SUM149 = 176; MDA-MB-436 = 5125). Supervised clustering of these genes reveals consistent gene-expression patterns throughout each cluster of samples (Supplementary Fig. S8A and S8B). Similar results were seen in comparing carboplatin + ABT888 with Control, with 53 genes differentially regulated in SUM149 tumors, and 4183 genes altered in MDA-MB-436 tumors (Supplementary Table S1F and S1G). When carboplatin was compared with carboplatin + ABT888 treatment, less than 10 differentially expressed transcripts were identified, and none were shared between the two models (SUM149: n = 9; MDA-MB-436 n = 2; Supplementary Table S1H and S1I). Thus, although not inhibiting the ability of carboplatin to induce dynamic transcriptional changes, ABT888 minimally changes gene expression when combined with carboplatin treatment.

To annotate the transcriptional changes occurring with carboplatin treatment, pathway analysis of genes altered with carboplatin treatment within each model was analyzed with DAVID (27, 28). Upregulated genes in both SUM149 and MDA-MB-436 were enriched for “signal peptide,” “glycopeptide,” and “SH3 domain” (Supplementary Table S1J). Interestingly, in MDA-MB-436, neuronal development, neuronal differentiation, and neuronal morphogenesis were significantly upregulated (Supplementary Table S1J). Downregulated pathways in DAVID were only significant in the MDA-MB-436 model, with 16 of the top 30 most significant terms involving processes in the mitochondria, including “electronic chain transport,” “respiratory chain,” and “mitochondrial inner membrane” (Supplementary Table S1K).

To define the common transcriptional program that is altered with carboplatin treatment, the intersection of upregulated and downregulated genes common to both models was used. Thirty-seven genes were commonly upregulated (Supplementary Fig. S9A), and one gene was commonly downregulated (Supplementary Fig. S9B) with carboplatin treatment in SUM149 and MDA-MB-436 tumors (Supplementary Fig. S9C). Interestingly, we found three oncogenes upregulated with carboplatin treatment, including N-myc (NDRG4), Ras (RAB39B), and Kit, and genes previously demonstrated to be potential therapeutic targets in TNBC, including αB-crystallin (CRYAB) and folate receptor (FOLR3; Fig. 5C; refs. 35, 36). In addition, multiple neurologic genes were upregulated, including CNTNAP2, a member of the neurexin family that provides adhesion in the CNS, CES1, a possible member of the BBB, and two CNS-specific G-protein–coupled receptors, GPR98 and NPY1R (37, 38).

This study has demonstrated efficacy of carboplatin ± ABT888 treatment in two murine models of BRCA-mutant intracranial TNBC with little to no efficacy in models of intracranial BRCA-wt TNBC despite similar BBB permeability and relative PAR inhibition across both BRCA-mutant and BRCA-wt models. With confirmed brain penetrance, the addition of PARP inhibition to platinum therapy via ABT888 exhibited modest impact on efficacy as measured by tumor burden, survival, and molecular changes within the intracranial tumors themselves.

In BRCA-mutant intracranial TNBC models, carboplatin treatment not only resulted in reduction of intracranial tumor volume, but in several cases, in growth arrest of intracranial tumors as evidenced by bioluminescence imaging. The mechanism of increased cytotoxicity within BRCA-mutant intracranial TNBC is hypothesized to be a failure to effectively repair DNA damage induced by carboplatin due to impaired HR repair pathways with BRCA loss of function (9). This hypothesis is confirmed by augmentation of γ-H2AX activation and apoptosis mediated by increased levels of cC3 following treatment with carboplatin. Interestingly, the addition of a PARP inhibitor, which impairs base excision repair, failed to yield a significant survival benefit as a single agent in these model systems. Moreover, the PARP inhibitor did not induce widespread or common gene-expression alterations in the BRCA-mutant models either alone or when combined with carboplatin. In contrast, carboplatin treatment led to substantial gene-expression changes observed in BRCA-mutant SUM149 and BRCA-mutant MDA-MB-436 intracranial models with 37 genes commonly upregulated, suggesting possible compensatory pathways that are activated upon treatment with carboplatin. These included oncogenes, previously described brain metastasis therapeutic targets, and neural genes. Interestingly, several recent studies report similar upregulation of neural genes in breast cancer brain metastases, including genes related to GABA (39), cellular adhesion (40), and several neurobiologic functions (41, 42). Although our results could be due in part to normal mouse brain contamination, these complementary and recent studies have excluded this caveat by assessing expression in purified tumor cells (39), passaging metastases cells in vitro (41), or by using species-specific microarrays (43). Exploration of potentially targetable upregulated genes in combination with carboplatin treatment that are low or absent in normal brain tissue will be critical to future rational combination therapies to avoid neurologic side-effects.

Our results demonstrate efficacy of well-tolerated therapeutics in model systems for a patient population with few treatment options. Currently in the field of TNBC brain metastases, a lack of therapeutic targets leaves clinicians with few beneficial choices aside from neurosurgical resection or crainal radiation. In addition, TNBC brain metastases are commonly (>80% of the time) accompanied by extracranial metastases, warranting effective systemic therapies capable of controlling both CNS and non-CNS disease (4). Other studies have demonstrated that irinotecan and iniparib yield an approximately 30% clinical benefit rate in patients with progressive TNBC brain metastases (44). In addition, a recent phase II study showed an approximately 65% intracranial response rate as measured by volumetric imaging following treatment with carboplatin and Bevacizumab in patients with progressive HER2-negative brain metastases (45). Despite these promising results, there remains no FDA-approved systemic therapy for TNBC brain metastases. Our data continue to support the role of carboplatin for the treatment of breast cancer brain metastases, specifically in TNBC with BRCA mutations, and the rationale of using carboplatin ± ABT888 combination in the clinical setting to determine whether ABT888 provides an additional benefit in patients with brain metastases.

Although our study adds to the literature surrounding TNBC brain metastases treatment, there are several limitations to our study. Direct intracranial implantation, as opposed to hematogenous spread of tumor cells to the CNS, may not recapitulate human spread of metastases; however, the purpose of this study was to evaluate established brain metastases, not prevention, in response to drug delivery and its impact on the intracranial tumor. In addition, although ABT888 reduced PAR levels in intracranial tumors, suggesting effective BBB penetration, limited efficacy and few gene-expression changes were observed. Future studies may aim to investigate other PARP inhibitors with varying degrees of catalytic activity, and thus may be more effective and synergistic with carboplatin treatment in brain metastases (46).

In conclusion, to our knowledge this study is the first to demonstrate a significant survival advantage in response to systemic carboplatin ± ABT888 treatment in BRCA-mutant TNBC intracranial murine models, as well as a mechanistic explanation of these results. Moreover, we have shown that PARP inhibition as a single agent, although effectively penetrating the BBB, does not yield significant improvement in survival or have a dynamic impact on gene expression. These results provide strong rationale to translate our findings into the design of early-phase clinical trials for patients with BRCA-mutant TNBC brain metastases testing carboplatin ± ABT888, with the ultimate goal of improving the survival of patients with an incurable disease who, at present, have limited systemic therapeutic options.

Y.Z. Lee reports receiving a commercial research grant from Carestream, Inc., and is a consultant/advisory board member for Merrimack Pharmaceuticals. C.K. Anders is a consultant/advisory board member for Novartis, Sanofi/BiPAR, to-BBB, GERON, Angiochem, Merrimack, and Lilly. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: O. Karginova, M.B. Siegel, A.E.D. Van Swearingen, C.K. Anders

Development of methodology: O. Karginova, M.B. Siegel, A.E.D. Van Swearingen, A.M. Deal, S. Bazyar, Y.Z. Lee, C.R. Miller, C.K. Anders

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): O. Karginova, M.B. Siegel, A.E.D. Van Swearingen, B. Adamo, M.J. Sambade, N. Nikolaishvili-Feinberg, R. Bash, K. Sandison, C. Santos, W. Zamboni, Y.Z. Lee, C.K. Anders

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): O. Karginova, M.B. Siegel, A.E.D. Van Swearingen, A.M. Deal, B. Adamo, M.J. Sambade, S. Bazyar, N. Nikolaishvili-Feinberg, R. Bash, S. O'Neal, J.S. Parker, Y.Z. Lee, C.K. Anders

Writing, review, and/or revision of the manuscript: O. Karginova, M.B. Siegel, A.E.D. Van Swearingen, A.M. Deal, B. Adamo, K. Sandison, D. Darr, W. Zamboni, Y.Z. Lee, C.R. Miller, C.K. Anders

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): O. Karginova, M.B. Siegel, A.E.D. Van Swearingen, M.J. Sambade, S. Bazyar, N. Nikolaishvili-Feinberg, R. Bash, D. Darr, Y.Z. Lee, C.K. Anders

Study supervision: O. Karginova, A.E.D. Van Swearingen, C.K. Anders

The authors thank Toshiyuki Yoneda, PhD, as the source of the MDA-MD-231-BR cell line, and Drs. Nancy Thomas and Bill Kaufmann for providing the SK-MEL cell line for IHC staining control. The authors thank Yongjuan Xia in the UNC Translational Pathology Laboratory (TPL) for expert technical assistance. The UNC Translational Pathology Laboratory is supported in part, by grants from the National Cancer Institute (3P30CA016086) and the UNC University Cancer Research Fund (UCRF). The authors thank the Small Animal Imaging Core facility at the UNC Biomedical Imaging Research Center for providing the MRI service, and the imaging core is supported in part by an NCI cancer center Grant #P30-CA016086-35-37.

This research is supported by the 2010 Breast Cancer Research Foundation-AACR grant for Translational Breast Cancer Research, grant number 10-60-26-ANDE (to C.K. Anders) as well as NIH K23157728 (to C.K. Anders). Research reported in this publication was supported by the National Cancer Institute of the NIH under award number K23CA157728. M.B. Siegel is supported by the National Cancer Institute Cancer Cell Biology Training Grant (T32-CA071341-17). C.K. Anders and C.R. Miller are Damon Runyon Clinical Investigators supported (in part) by the Damon Runyon Cancer Research Foundation (CI-64-12; to C.K. Anders, CI-45-09; to C.R. Miller).

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.