Combining a PI3K Inhibitor with a PARP Inhibitor Provides an Effective Therapy for BRCA1-Related Breast Cancer

Unresectable triple-negative breast cancer (TNBC) remains an incurable illness that invariably relapses after treatments considered standard-of-care, leading to death, often within months of diagnosis. Current chemotherapeutic regimens induce not only incomplete remissions that are short but also result in toxicity that severely impacts a patient's quality of life. These shortcomings have led to an extensive search for more effective treatments.

Female BRCA1 mutation carriers have an approximately 85% lifetime risk of developing breast cancer. These cancers generally are negative for estrogen receptor, progesterone receptor, and HER2 (e.g., triple negative), making them nonresponsive to therapies that target these pathways. Sporadic TNBCs that emerge in patients without germline BRCA1 or BRCA2 mutations frequently show evidence for epigenetic silencing of BRCA1. Truncating mutations disrupting the C-terminal end of the BRCA1 protein predispose to breast cancer, whereas mutations in the N-terminal two thirds of the protein result in elevated susceptibility to both breast and ovarian cancer (1).

Loss of BRCA1 in breast epithelial cells disables DNA damage repair via homologous recombination. This defect leads to genomic instability but also sensitizes cells to the deleterious effects of other DNA-damaging agents, such as cisplatin or inhibitors of poly-ADP-ribosylation (PAR). Poly (ADP-ribose) polymerase (PARP) is a nuclear enzyme that senses DNA single-strand breaks and is essential for base excision repair. Once base excision repair is disabled, cells rely on homologous recombination for DNA damage repair. Dysfunction of homologous recombination (such as in BRCA1-deficient cells) presents a context in which inhibition of base excision repair (e.g., by treating with PARP inhibitors) is synthetically lethal. Clinically, PARP inhibitors have emerged as promising agents, inducing objective responses in 41% of patients with BRCA1-related breast cancer (2, 3) and 33% of patients with BRCA1-related ovarian cancer (4, 5). However, the remissions achieved with PARP inhibitors have not been durable, and benefit in the subset of TNBCs that are not BRCA1-related is currently uncertain.

Multiple lines of evidence suggest that growth factor signaling may be a sensible target for the treatment of TNBC: EGF receptor (EGFR) overexpression seems to correlate with the basaloid phenotype and is found in 60% to 70% of TNBCs, including BRCA1-related cancers (6). We have previously shown that upregulation of EGFR and the EGF pathway is an early event in BRCA1-related tumorigenesis (7). IGF-1R levels are increased in BRCA1-related breast cancers (8) and genetic variants in the insulin-like growth factor (IGF) pathway are associated with BRCA1-related tumorigenesis (9, 10). However, VEGF receptor (VEGFR) and EGFR inhibitors, alone or in combination with traditional chemotherapy, have not improved survival for patients with TNBC. One explanation for this lack of efficacy is that resistant tumor cells signal through alternate receptor tyrosine kinases, turning the search for new therapeutic angles to nodal points of intracellular signal transduction, such as mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K), in which inhibition may be harder for tumor cells to evade. Here, we examine the mechanism and the efficacy of a PI3K inhibitor, NVP-BKM120, for the treatment of BRCA1-related breast cancer in a mouse model and report on a surprising in vivo synergy with PARP inhibition.

We and others have previously shown that the MMTV-CreBrca1^f/fTrp53^+/− mouse model faithfully recapitulates many aspects of human BRCA1-related breast cancer, including emergence on a background of multiple synchronous hyperproliferative lesions, high proliferative activity, absence of estrogen receptor expression, and the presence of EGFR overexpression (11–14), although exon 11 deletion in this model results in the residual expression of a hypomorphic BRCA1 protein, rather than complete absence of the BRCA1 protein shown in other models (15).

BRCA1 has been shown to suppress AKT (16) and extracellular signal-regulated kinase (ERK) activation in response to estrogen or EGF stimulation (17, 18) in cell-based studies, suggesting that tumors with defects in BRCA1 might have an increase in AKT and/or ERK phosphorylation. Consistently, we found that phosphorylation of AKT at Serine 473 was strongly positive in both the cytoplasm and the nucleus in these tumor cells (Fig. 1, top right, and Supplementary Fig. S1), whereas in the normal adjacent tissue, cytoplasmic AKT phosphorylation was only seen in the basal layer of cells, not in luminal cells (Fig. 1, top left). Similarly, ERK phosphorylation was absent in normal mammary epithelial cells, whereas cytoplasmic ERK phosphorylation was seen in a majority of but not in all tumor cells (Fig. 1, second panel).

Loss of function of PTEN, either through epigenetic silencing or through gross genomic loss, correlates with loss of function of BRCA1 in TNBC (19). Recently, Gewinner and colleagues (20) as well as Fedele and colleagues (21) showed that, similar to PTEN, the tumor suppressor phosphatase INPP4B is lost in approximately 60% of TNBC, including BRCA1-related breast cancers. Consistent with these data in human disease, INPP4B and PTEN expression were strong in normal glands of MMTV-CreBrca1^f/fTrp53^+/− females but lost in tumor tissues (Fig. 1, third panel and bottom).

To examine whether activating PIK3CA mutations are responsible for the strong and uniform activation of AKT, we sequenced the Pik3ca gene of 11 murine Brca1-deleted breast tumors. Consistent with the rarity of mutations in human TNBC, we found no activating hotspot mutations in exons 9 or 20 of PI3K. In human TNBC, activating mutations in Pik3ca are relatively rare and seen in only 8% of TNBC, confirming that the activation of the PI3K pathway in TNBC is mostly driven by regulatory mechanisms, such as loss of PTEN and INPP4B, rather than by activating mutations in PIK3CA.

Collectively, these observations suggest that the MMTV-CreBrca1^f/frp53^+/− mouse model accurately recapitulates the activation of growth factor signaling seen in human BRCA1-related breast cancer, including activation of the PI3K and MAPK pathways and the absence of activating PI3K mutations. On the basis of these data, we decided to study whether inhibition of PI3K would be an effective treatment of BRCA1-related breast cancer.

TNBCs, including the BRCA1-related subtype, exhibit high rates of glucose uptake, as judged by positron emission tomography (PET) using the radioactive glucose analog, 2[¹⁸F]fluoro-2-deoxy-d-glucose (FDG; refs. 22, 23). Consistent with these observations in humans, we found that Brca1-deleted tumors in our mouse model were highly avid for FDG. Tumors of subcentimeter size were easily visualized using this technique (Fig. 2; Supplementary Figs. S2 and S3). In a previous study (24), mouse lung tumors that resulted from transgenic expression of the H1047R mutant of PIK3CA were found to have high rates of glucose uptake as judged by FDG-PET, and the PI3K/mTOR inhibitor BEZ235 caused a reduction in the FDG-PET signal within 2 days, consistent with the known role of PI3K in regulating glucose uptake and glycolysis (25–27). We found that within 48 hours of instituting treatment with NVP-BKM120, tumors in all treated animals showed a median decrease in FDG uptake by 46.7% (range, 38.1–92.3), which was sustained after 2 weeks of continued treatment with NVP-BKM120 (median decrease by 54%, range 45.5%–70.5%) and corresponded to inhibition of AKT phosphorylation (Fig. 2A–D, Supplementary Figs. S2 and S3). These results indicate that the activation of the PI3K pathway contributes to the upregulation of glucose metabolism in BRCA1-related breast cancers and that oral delivery of NVP-BKM120 results in inhibition of this response. Further evidence that NVP-BKM120 inhibits PI3K signaling in the Brca1-defective tumors was provided by the observation that phosphorylation of the downstream protein kinase AKT at Ser-473 was strongly decreased in tumors treated with NVP-BKM120 (Fig. 2B, Supplementary Fig. S1). It was remarkable that all Brca1-related tumors examined showed a decrease in FDG uptake and a decrease in AKT phosphorylation in response to NVP-BKM120 (Fig. 2, Supplementary Figs. S2 and S3), suggesting that a high level of PI3K signaling and the consequent enhanced glucose metabolism is a common event in tumors that result from loss of BRCA 1 function. In addition, our data suggest that inhibition of FDG uptake may be an early and predictive pharmacodynamic marker for response to treatments with PI3K inhibitors.

Tumor growth requires neovascularization of the expanding neoplastic tissue. It was previously shown that BEZ235, a PI3K inhibitor with activity against PI3Kα and mTOR, inhibits the sprouting of new blood vessels in tumors and disrupts the integrity of existing blood vessels (27, 28). Spontaneous tumors in MMTV-CreBrca1^f/fTrp53^+/− mice grow rapidly and are highly vascular (Fig. 3A, left; Fig. 3B). However, after treatment with NVP-BKM120, the gross pathology of tumors was notable for central pallor and, eventually, central necrosis (Fig. 3A, middle). In contrast, blood vessels in the tumor capsule remained initially intact, or became ectatic (Fig. 3A). Consistently, the tumor microvasculature, as visualized with an anti-CD31 stain, was diminished in response to NVP-BKM120 (Fig. 3C) while it was maintained in the tumor capsule (Fig. 3D). The necrotic center of treated tumors was frequently hemorrhagic (data not shown), indicating disorganized collapse of the tumor vasculature. We used the Chalkley count of CD31-positive microvessels (29) to compare the vascularization before and after treatment with NVP-BKM120 and found that both the size and number of blood vessels were starkly reduced in treated tumors (Fig. 3E). Thus, consistent with prior observations with BEZ235 (28) and recent data with NVP-BKM120 (30), our data confirm that NVP-BKM120′s antitumor activity is, in part, due to its antiangiogenic activity, and thus this drug may have preferential activity in rapidly growing, endocrine-resistant tumors with a high degree of tumor angiogenesis.

The upregulation of compensatory pathways in response to tumor cell treatments with inhibitors of mitogenic signaling is now a well-known phenomenon (31). Consistent with these prior observations, we found that NVP-BKM120 induced a compensatory activation of the EGFR/MAPK pathways in the human BRCA1-mutant breast cancer cell lines, HCC1937 (BRCA1 5382C mutation and homozygous deletion of PTEN and TP53; ref. 32), and SUM149 (BRCA1 2288delT, PTEN wild type, TP53 mutant; refs. 33, 34; Fig. 4A, second lane for each cell line). As expected, treatments with the PARP inhibitor olaparib alone did not have a discernible effect on the activation status of EGFR, AKT, or MAPK (Fig. 4A, third lane for each cell line). However, with the combination treatment (Fig. 4A, last lane), we found that compensatory activation of EGFR and MAPK could be blocked by the addition of olaparib. These data suggest that PARP inhibition in tumor cells either restricts mitogenic signaling to PI3K-mediated signaling, or disables mechanisms that would reroute mitogenic signaling via EGFR/ERK when PI3K is inhibited.

Loss of BRCA1 function results in genome instability due to defects in DNA repair by homologous recombination. As a consequence, BRCA1^−/− cells have high rates of DNA damage and are sensitized to the inhibition of alternative DNA repair mechanisms involving PARP-dependent PAR (34). We examined the possibility that the high sensitivity of BRCA1-mutant tumors to PI3K pathway inhibitors is the consequence of a role for the PI3K pathway in maintaining cell survival during DNA repair or in facilitating DNA repair mechanisms. These experiments were carried out in vivo (Fig. 4B) and with the human BRCA1-mutant cell lines, HCC1937 and SUM149. We first examined the effect of NVP-BKM120 on DNA repair responses in cells grown on plastic. Surprisingly, we found that in both cell lines, histone 2AX (H2AX) phosphorylation on Serine 139 (γ-H2AX, a marker for DNA double-strand damage) increased with increasing concentrations of NVP-BKM120 and that this correlated with diminishing phosphorylation of AKT (Fig. 4C). Similarly, tumors treated with NVP-BKM120 in vivo showed a substantial increase in the percentage of cells that express γ-H2AX (Fig. 4B).

Tumors with loss of BRCA1 rely on PARP-dependent PAR of key proteins involved in DNA damage repair (34). Given the surprising increase in H2AX phosphorylation, we examined if treatment with NVP-BKM120 would also affect PARP activity. Treatment with NVP-BKM120 caused a dose-dependent increase in overall PAR that paralleled the increase in H2AX phosphorylation and the decrease in AKT phosphorylation (Fig. 4C). Importantly, this increase in PAR was initially not accompanied by apoptotic cell death, as cells remained negative for cleaved caspase-3 (CC3; Fig. 4C). The basal and NVP-BKM120–enhanced PAR could be completely blocked by treatment with the PARP inhibitor, olaparib (Fig. 4D), whereas γ-H2AX accumulation was enhanced with the combination of NVP-BKM120 and olaparib (Fig. 4D).

Thus, we observed that PI3K inhibition caused a significant increase in activities indicative of both types of DNA damage: PARP activity, which is required for base excision repair and single-strand break repair, as well as H2AX phosphorylation, indicative of the presence of DNA double-strand breaks (DSB). As H2AX is a substrate for the PI3K-related kinases Ataxia teleangiectasia mutated (ATM) and DNA-dependent protein kinase (DNA-PK), we asked if NVP-BKM120 had an effect on these kinases that would explain our findings. We examined PAR and γ-H2AX accumulation in HCC1937 cells in the absence and presence of the ATM-inhibitor KU-55933 (35) and monitored the response to ionizing radiation. As expected, KU-55933 led to a decrease in autophosphorylation of ATM (Fig. 5A, third and fourth lane of each panel), and prevented the increase in H2AX phosphorylation seen in response to ionizing radiation. However, KU-55933 did not prevent the NVP-BKM120–induced induction of γ-H2AX, which was robust both at baseline and in response to ionizing radiation (Fig. 5A, last lane of each panel), suggesting that an alternative kinase, such as DNA-PK, phosphorylates H2AX in response to PI3K inhibition. As shown in Fig. 5A, we found a strong increase in autophosphorylation of DNA-PK in response to the addition of NVP-BKM120 that corresponds to H2AX phosphorylation. Consistent with prior reports (30), these results clearly show that NVP-BKM120 is not acting through an off-target inhibition of ATM or DNA-PK and suggest that inhibition of PI3K by NVP-BKM120 leads to the activation of DNA-PK through a yet unknown mechanism.

Consistent with the results in Fig. 4C, we found that PAR accumulation in the presence of NVP-BKM120 alone increased (Fig. 5A, left, second lane). In the presence of the combination of NVP-BKM120 and KU-55933, PAR accumulation was attenuated but still greater than in the control, suggesting that the NVP-BKM120–induced increase in PAR was only partially offset by inhibition of ATM, again consistent with an ATM-independent mechanism for PAR-accumulation and its induction by PI3K inhibition.

To determine if PI3K inhibition affected the assembly of DNA damage repair foci, we examined the ability of tumor cells from our mouse model to recruit Rad51 to DNA damage repair foci (Fig. 5B–E), following a protocol established previously (36). We generated cell cultures from tumors of MMTV-CreBrca1^f/frp53^+/− mice and examined their ability to form DNA repair foci 6 hours after exposure to ionizing radiation (10 Gy). We found that there was residual double-strand repair activity as shown by the formation of Rad51 foci in this mouse model with a hypomorphic exon 11 deletion (Fig. 5C). Surprisingly, the formation of Rad51 foci in response to ionizing radiation was completely blocked by pretreatment of these cells with NVP-BKM120 (Fig. 5E). A similar phenomenon was observed in HCC1937 cells: although ionizing radiation induced accumulation of Rad51 and H2AX phosphorylation as reported previously (ref. 37; Fig. 5A, control lanes), pretreatment with the PI3K inhibitor NVP-BKM120 led to a dissociation of this radiation response, as we saw a failure to increase Rad51 but a prominent augmentation of radiation-induced H2AX phosphorylation in the presence of NVP-BKM120 (Fig. 5A, second lane of each panel). The mechanism by which NVP-BKM120 decreases Rad51 recruitment to repair foci is yet unknown. However, this observation of a defective DSB repair response may, at least in part, provide an additional explanation for the in vivo synergy of PARP and PI3K inhibition.

Given the unanticipated and striking effects of the pan-class IA PI3K inhibitor NVP-BKM120 on the DNA damage response, we asked whether these effects were specific to a single class IA PI3K isoform or required inhibition of multiple PI3Ks or could be an off-target effect of NVP-BKM120. In the BRCA1-mutant cell line SUM149, downregulation of PI3Kα, but not PI3Kβ, with siRNA led to a stark increase in phosphorylation of DNA-PK, γ-H2AX, and PAR, and a stark decrease in Rad51 accumulation (Fig. 5F). These data confirm that it is the inhibition of PI3Kα that is decisive for the disruption of the DNA damage response in these cells.

We first examined the effect of NVP-BKM120 and olaparib on the growth on plastic of the 2 BRCA1-mutant cell lines. HCC1937 cells, with a genetic loss of PTEN, showed greater sensitivity to NVP-BKM120 than SUM149 cells, which have wild-type PTEN (Supplementary Fig. S4A). SUM149, on the other hand, showed greater sensitivity to olaparib (Supplementary Fig. S4B). The drug combination did not have much benefit beyond that of the most effective single agent in either cell line (Supplementary Fig. S4) and isogenic reconstitution of PTEN in HCC1937 did not significantly alter drug sensitivities (Supplementary Fig. S4C and S4D), indicating that under the artificial conditions of growth on plastic with high levels of nutrients and oxygen, and in the absence of the native tumor microenvironment, this drug combination does not result in synergy. We next addressed whether NVP-BKM120 and olaparib might have a more dramatic effect in vivo on endogenous BRCA1-deleted tumors. We first showed that, consistent with the observations with the human BRCA1-mutant cell lines, NVP-BKM120 treatment of mice with Brca1-deleted breast tumors (MMTV-CreBrca1^f/frp53^+/−) resulted in an increase in phosphorylated H2AX in the recurrent tumors (Fig. 4B).

We next compared the effects of NVP-BKM120 and olaparib as single agents and the combination of both drugs on tumor growth. Female virgin MMTV-CreBrca1^f/frp53^+/− mice were observed for the development of spontaneous tumors, which typically occurs at age 8 to 12 months. Once tumors reached a diameter of 5 to 7 mm, mice were randomized to either vehicle control treatments, treatments with NVP-BKM120 via oral gavage, olaparib intraperitoneally, or the combination of NVP-BKM120 with olaparib, all once a day continuously. An initial set of mice was treated with NVP-BKM120 at 50 mg/kg/d, alone or in combination with olaparib (50 mg/kg/d) and a second set at NVP-BKM120 30 mg/kg/d alone or in combination with olaparib (50 mg/kg/d). No significant difference was seen with regard to efficacy or phosphorylated AKT (p-AKT) suppression between the 2 dose levels of NVP-BKM120 and data were pooled (Fig. 6A–D). Tumors were measured at least 3 times a week, and relative tumor volume, as a ratio to baseline tumor volume, was calculated for each treatment modality (Fig. 6A–D). Trend lines were determined on the basis of the best fit to the data in vehicle control (red line) and NVP-BKM120 only (green line). Once tumors were established, their doubling time was rapid if treated with vehicle only, on average 5 days (Fig. 6A). Treatment with NVP-BKM120 alone significantly prolonged tumor-doubling time by a factor of 5 (26 days vs. 5 days, Fig. 6B); however, tumors eventually grew (Fig. 6B). In this mouse model, tumor growth was delayed 3-fold with the use of olaparib (tumor doubling time 16 days vs. 5 days; Fig. 6C). When olaparib and NVP-BKM120 were combined, we found a surprising in vivo synergistic activity, with a tumor doubling time of over 70 days, a 14-fold increase over control (red trend line). The dual combination of NVP-BKM120 and olaparib did not result in measurable toxicity, such as weight loss (Fig. 6E), even in mice that were treated for over 3 months. To ensure target inhibition, we obtained pretreatment biopsies and matched tumor specimens within 2 hours of the last dose of NVP-BKM120 and found that NVP-BKM120 potently reduced AKT phosphorylation (Fig. 6F and Supplementary Fig. S1). In tumor tissue lysates from the combination treatment, we observed inhibition of p-AKT with the combination treatment and induction of γ-H2AX (Fig. 6F), consistent with results observed in the in vitro studies with cell lines (Fig. 4). Interestingly, olaparib alone led to an induction of AKT phosphorylation in vivo (Fig. 6F), an observation consistent with an increased FDG uptake in olaparib-treated tumors (Supplementary Fig. S3) as opposed to NVP-BKM120 or the combination, both of which strongly reduced FDG uptake (Fig. 2; Supplementary Figs. S2 and S3).

To examine if there was a pharmacokinetic interaction between NVP-BKM120 and olaparib, we examined NVP-BKM120 levels in animals treated with NVP-BKM120 at 30 mg/kg/d and the combination of NVP-BKM120 and olaparib (30 mg/kg/d and 50 mg/kg/d, respectively). For these studies, tissue extracts were processed for mass spectrometry 3 hours after the last dose (Figs. 6G, Supplementary Fig. S5). We found that although NVP-BKM120 levels in tumor tissues were variable, they were consistently in the micromolar range and were not affected by concurrent administration of olaparib.

The mouse model used here for BRCA1-related breast cancer MMTV-CreBrca1^f/fTrp53^+/− results in the residual expression of a hypomorphic BRCA1 protein, and we did find residual Rad51 recruitment to repair foci (Fig. 5C). This residual homologous recombination activity may also explain the incomplete responses of the BRCA1-exon 11–deleted–expressing mammary tumors to olaparib monotherapy (Fig. 6C).

To test the applicability of our results to human BRCA1-related breast cancer, we treated xenograft tumors established from patients with BRCA1-related breast cancer (Fig. 6H and I). The first patient-derived tumor (Fig. 6H) was derived from a patient with an N-terminal germline mutation in BRCA1 (185 del AG). At the time of tissue acquisition, this tumor had developed resistance to standard chemotherapy as well as olaparib, which had been administered in the context of a clinical trial. Growth of this tumor was modestly attenuated by either NVP-BKM120 or olaparib alone in HsdCpb:NMRI-Foxn1nu (nude) mice. However, the combination induced stability over a period of 8 weeks (Fig. 6H), confirming the in vivo synergy that we observed in our genetically engineered mouse model of BRCA1-related breast cancer. The second human tumor (Fig. 6I) was derived from a patient with a C-terminal BRCA1 germline mutation (2080 del A). The patient who donated this tumor specimen had not yet been treated, and the tumor showed exquisite sensitivity to the PARP inhibitor, NVP-BKM120, and the combination of both drugs. These human ex vivo data confirm the sensitivity of BRCA1-related breast cancer to NVP-BKM120, olaparib, and their combination, and, taken together, justify the exploration of this combination in an early-phase clinical trial.

Eventually, even in tumors that received dual treatments, resistance was observed and at that point, tumors regrew rapidly (Fig. 6A–D). To determine the nature of resistance to the NVP-BKM120 and olaparib combination, we examined pretreatment biopsies, on-treatment biopsies at the time of response on day 10, and posttreatment tissue at the time of progression (Fig. 7A). Target inhibition, that is, suppression of AKT phosphorylation, was maintained even in resistant tumors (Fig. 7A, panel 1),suggesting that resistance to NVP-BKM120 is not due to PI3K pathway activation but to relief of feedback inhibition of alternative pathways, including MAPK activation as suggested earlier (31). The “pushing margin,” that is, a highly proliferative rim of tumor cells that rarely infiltrate the surrounding tissue, is a hallmark of BRCA1-related tumors (38), yet its biologic basis is not understood. Interestingly, we found an increase in the number of cells with high phospho-ERK levels especially at the “pushing margin” of the tumor, paralleled by an increase in proliferating, that is, Ki67-positive cells (Fig. 7A and B, Supplementary Fig. S6). This phenomenon, the concentration of phosphorylated ERK (p-ERK)–positive cells at the “pushing margin,” was seen in tumors before treatment (Fig. 4B, Fig. 7A), at the time of progression on NVP-BKM120 alone (Fig. 4B), or at the time of progression on the combination of the PARP inhibitor with NVP-BKM120, whereas in responding tumors (day 10 biopsy), p-ERK–positive cells were conspicuously absent (Fig. 7A). As expected with PI3K inhibition and consistent with the p-ERK status of tumor cells, we found that tumors initially showed a stark decrease in proliferative activity (Ki67, Fig. 7A and B), and that resistant tumors were characterized by high mitotic activity (Figs. 4B, 7A and B). Thus, the activation of proproliferative MAPK signaling may be a major driver for the resistance of tumors treated with PI3K inhibitors.

We report here on a surprising in vivo synergy of NVP-BKM120 in combination with olaparib for the treatment of BRCA1-mutant breast tumors, which suggests an important role of PI3Kα in the DNA damage response. Kumar and colleagues (39) showed that PI3K β is required for the recruitment of NBS1 to DNA DSBs and for the assembly of repair foci in response to ionizing radiation. It was shown previously that loss of PTEN, frequently seen in TNBC, leads not only to activation of the PI3K pathway, but also to an accumulation of DNA DSBs (40). In addition, NVP-BKM120 enhances production of PAR and phosphorylation of H2AX, suggesting increased DNA damage when the PI3K pathway is inhibited in the context of a BRCA1 mutation. In vivo H2AX phosphorylation in tumors increased when mice were treated with the combination of NVP-BKM120 and olaparib during the period of response (day 10), and was highest at the time of treatment failure (Fig. 7), suggestive of a progressive accumulation of unrepaired DNA DSBs, which would contribute to the reliance on PARP activity for DNA damage repair and would explain the sensitivity to combined PARP and PI3K inhibtion.

Of particular interest was our observation that, despite the increase in phosphorylation of H2AX in response to NVP-BKM120, both NVP-BKM120 and depletion of PI3Kα greatly reduced Rad51 incorporation into foci in cells treated with radiation. These results suggest that class IA PI3K catalytic activity is required for recruitment of Rad51 into sites of DNA damage and raise the possibility that the increase in DNA-PK phosphorylation is a feedback response to this failure to form proper DNA damage repair complexes. BRCA1 is known to play a role in recruitment of Rad51 to sites of DNA damage (41), and thus it is possible that in BRCA1-defective cells, a PI3K-dependent pathway becomes more critical for this recruitment. Clearly, additional studies will be needed to understand the interactions between PI3K, Rad51, and DNA-PK in DNA repair processes.

Regulated PARP activity allows for DNA damage repair required for the maintenance of genomic stability. However, massive PARP activation leads to the depletion of its substrate NAD⁺ and consecutively depletion of ATP in an effort to replenish NAD⁺, resulting in energy loss and eventually cell death. Activation of PI3Kα leads to increased energy production via glycolysis. Glycolysis and PAR both consume NAD⁺, and may compete for NAD⁺ available in the cytosol. Such metabolic competition makes sense for decisions on the fate of cells: if energy supply and glycolysis are high, the amount of NAD⁺ diverted into PAR is limited, and cell death as a consequence of massive PARP activation is avoided. Conversely, if glucose supply and glycolytic activity are low, NAD⁺ is consumed by PARP, and the ensuing massive PAR may lead to cell death (42). PARP inhibition spares NAD⁺, which becomes available for glycoloysis and can protect cells from death, such as myocardial or central nervous system ischemia (43, 44), sepsis (45), or pancreatic islet cell damage (46). Consistent with this model, we saw in vivo enhancement of glucose uptake (Supplementary Fig. S3) and phosphorylation of AKT in response to PARP inhibition, which was reversed by addition of the PI3K inhibitor (Fig. 6E). Thus, a possible explanation for the in vivo synergy of PI3K and PARP inhibitors is that PI3K inhibition reverses the prosurvival effect of PARP inhibition and thereby makes these drugs more effective, a combination that one would predict to be particularly effective in cancers with defects in homologous recombination, such as BRCA1/2-related breast and ovarian cancers.

Finally, it is noteworthy that the in vivo approach allowed us to make several observations that could not be made in vitro: much greater efficacy of the NVP-BKM120/olaparib combination was observed in vivo than in vitro, suggesting that tumor microenvironment and metabolism may be important. Sequential tumor biopsies allowed us to monitor target inhibition in combination with tumormetrics, which allowed us to discover a potent synergy of the PI3K inhibitor NVP-BKM120 with the PARP inhibitor olaparib to treat BRCA1-related breast cancer that may warrant exploration in an early-phase clinical trial.

The PI3K inhibitor NVP-BKM120 was obtained through a Material Transfer Agreement with Novartis Pharmaceuticals. Olaparib was purchased from LC Laboratories and KU-55933 was purchased from Selleck. BRCA1-mutant human breast cancer cell line HCC1937 was from American Type Culture Collection, CRL-2336, and maintained in Dulbecco's Modified Eagle's Medium (DMEM)/10% FBS, and SUM149 was a gift from Dr. Christina Gewinner, Division of Signal Transduction, Beth Israel Deaconess Medical Center (Boston, MA), maintained in Ham's F-12 with 5% FBS, 5 μg/mL insulin, 2 μg/mL hydrocortisone, 5 μg/mL gentamicin, and 2.5 μg/mL fungizone. Cell lines were authenticated by immunoblotting for BRCA1 and PTEN and tested for the absence of mycoplasma.

Animal experiments were conducted in accordance with Institutional Animal Care and Use Committee (IACUC)–approved protocols at Beth Israel Deaconess Medical Center, and at the University of Vall d'Hebron (Barcelona, Spain). Female MMTV-CreBrca1^f/frp53^+/− mice were obtained by breeding Brca1 conditional knockout mice (01XC8, strain C57BL/6), originally generated by Drs. Xiaoling Xu and Chu-Xia Deng (12), who made these mice available to us via the National Cancer Institute (NCI) repository with MMTV-Cre [Jackson Laboratory B6129-TgN (MMTV-Cre) 4Mam; ref. 47)] and Trp53 knockout (Taconic Farms, P53N12-M, C57BL/6; ref. 48). At the time of the study, mice had been inbred for 4 years (>7 generations). The floxed or wild-type status of Brca1, the presence of the MMTV-Cre transgene, and the Trp53 heterozygosity were determined by PCR as previously described (12). Mice were examined for the occurrence of tumors twice weekly. When tumormetrics were conducted, the length and width of the tumor was determined using calipers, and the tumor volume was determined (width² × length/2). Tumor volume was used as a measure of growth and was recorded as ratio to tumor volume at diagnosis. Tumor doubling times were calculated using the functions of the best-fit curves for all data points in each treatment modality. NVP-BKM120 was resuspended in 5% methylcellulose solution (Fluka) and administered via oral gavage at 50 mg/kg/d or 30 mg/kg/d. Olaparib was resuspended for intraperitoneal administration as described (15) and dosed at 50 mg/kg/d. For patient-derived tumor grafts, consent for tumor use was obtained from patients under a protocol approved by the Vall d'Hebron Hospital Clinical Investigation Ethical Committee. Tumors were subcutaneously implanted in 6-week-old female HsdCpb:NMRI-Foxn1nu mice (Harlan Laboratories, Italy). Animals were supplemented with 1 μmol/L estradiol (Sigma) in the drinking water. After tumor graft growth, tumor tissue was reimplanted into recipient mice, which were randomized upon implant growth.

A total of 0.3 to 0.4 mCi of FDG was injected intravenously through the retro-orbital vein of the anesthetized mouse. After a “washout” period of 1 hour, the mouse was imaged on a NanoPET/CT (Bioscan/Mediso) scanner. The NanoPET/CT is a high-resolution small-animal multimodality scanner consisting of 12 lutetium yttrium oxyorthosilicate (LYSO) detector blocks. The blocks comprise a total of 39,780 crystals each with a dimension of 1.2 mm × 1.2 mm × 13 mm³.

Images were acquired in 3 dimensions. The mice remained supine and maintained their position throughout the procedure. First, a computed tomography (CT) scan was conducted, and second, a whole-body FDG-PET scan was acquired covering the same area as the CT scan. Counts per minute (cpm) were obtained, converted to mCi, and values were normalized for region of interest volume and injected dose. To correct for metabolic variability between exams and to determine tumor-specific uptake changes, FDG-uptake rates were corrected for cardiac FDG uptake (μCi/μCi injected/voxels^tumor)/(μCi/μCi injected/voxels^heart). For studies involving repeat scanning, the change in tumor-specific FDG uptake was determined in percentage [1 − (FDG-uptake^post/FDG-uptake^pre) ×100]. Animals were housed in the Longwood Small Animal Imaging Facility between scans.

For immunohistochemistry (IHC), we used anti-CC3 (9661S, Cell Signalling, Rabbit polyclonal Asp175) and anti-Ki67 (9106-S; Thermo Scientific, Rabbit monoclonal SP6). All other antibodies used are described in Immunoblotting. All IHCs were done as described previously (11), including antigen retrieval with a citrate-buffer.

Cells were treated with mock, NVP-BKM120, olaparib, KU-55933, or the combination and lysed in cell lysis buffer (9803, Cell Signaling) per the manufacturer's instructions. Immunoblots were conducted using the Nupage System (Invitrogen). A total of 20 μg of protein was loaded, except for PAR, phospho-ATM, and phospho-DNA-PK/PRKDC Western blot analysis, where 40 μg were loaded. Tumor tissue lysates were prepared similarly with the exception of tissue homogenization by using an electric homogenizer for 30 milliseconds after addition of the lysis buffer. Primary antibodies used for Western blot analysis were total AKT (9272), CC3 (9661), total ERK (4695), phosphorylated AKT Ser473 (4058), phosphorylated ERK Thr202/Tyr204 (9106), phosphorylated histone H2AX Ser139 (2577), and PTEN (9559) from Cell Signaling. Phospho-ATM Ser1981 (2152-1) and phospho-DNA-PK/PRKDC Ser2056 (3892-1) were obtained from Epitomics; CD31 (ab28364), actin (ab6276), and INPP4B (ab81269) from Abcam; pADPr (sc56198) from Santacruz Biotechnology; and Ki67 (RM-9106) was purchased from Thermo Scientific. Rad51 antibody was a gift from Dr. Ralph Scully.

Cells were plated on coverslips in 6-well plates and incubated overnight at 37°C with 5% CO₂ before drug treatment. Cells were exposed to NVP-BKM-120 for 24 hours followed by irradiation (10 Gy). Cells were fixed with 3% paraformaldehyde and 2% sucrose diluted in PBS 6 hours postirradiation and subsequently permeabilized with 0.5% TritonX-100 buffer (20 mmol/L HEPES pH 7.4, 50 mmol/L NaCl, 3 mmol/L MgCl, and 300 mmol/L sucrose) for 3 minutes on ice. Cells were incubated with a primary rabbit antihuman Rad-51 antiserum at 1:500 dilution in hybridization buffer (5% goat serum, 0.5% NaN₃, 1 × PBS) for 30 minutes at 37°C. Secondary antibody used was a donkey antirabbit Alexafluor 488 conjugate (Invitrogen) at a concentration of 1:50. Images were acquired using a Zeiss 710 NLO laser scanning confocal microscope.

The siRNAs were obtained from Dharmacon. SUM149 cells were transfected with either 10 or 30 nmol/L pool of 4 siRNA sequences targeting PIK3CA (catalog number L-003018-00-0005) or PIK3CB (catalog number L-003019-00-0005) siRNA using HiPerFect Transfection Reagent (QIAGEN) according to the manufacturer's protocol. Control cells were treated with HiPerFect alone. Cells were grown and harvested 48 hours after the transfection using cell lysis buffer (9803, Cell Signaling) per the manufacturer's instructions and analyzed by immunoblotting.

For cell viability assays, breast cancer cells were seeded at a density of 250 cells per well in 96-well plates in the absence or presence of drugs, and cell viability was determined using the CellTiter-Glo Luminescent Cell Viability Assay (Promega) according to the manufacturer's instructions, using a Wallac 3 plate reader.

Genomic DNA was isolated and PCR amplification conducted for regions in the murine Pik3ca gene that are homologous to the regions frequently mutated in human breast cancer, that is, E542K and E545K in the helical domain and H1047R in the kinase domain. Primers used were for exon 9: forward CGCATACCTGCATCTGTTCTA; reverse AAATGATGTGTGTGCTGGGT, exon 20: forward AGCAGCTCACTGACCAGATGT; reverse ACTCACTGCCATGCAGTGGA. PCR products were subjected to direct sequencing at Genewiz.

Determination of the Chalkley score was done as described (29, 49). Briefly, the 3 most vascular areas (hot spots) with the highest number of microvessel profiles in each tumor were photographed under an Olympus light microscope at × 200; a digital mask representing the Chalkley grid area, 0.196 mm², was used to count the CD31-positive spots in a blind fashion and the mean value of the 3 grid counts was obtained. A 2-sided t test was used to determine significance.

Metabolites from 100 mg of mouse tumor samples were extracted using 80% methanol according to Yuan and colleagues (50), and 10 μL were injected and analyzed using a 5500 QTRAP hybrid triple quadrupole mass spectrometer (AB/SCIEX) coupled to a Prominence UFLC HPLC system (Shimadzu) via selected reaction monitoring for the Q1/Q3 transition of 410.8/367.0 for NVP-BKM120. ESI voltage was +4,900 V in positive ion mode using a dwell time of 4 milliseconds and collision energy of 45. Approximately 15 data points were obtained for NVP-BKM120 per liquid chromatography/tandem mass spectrometry (LC/MS-MS) experiment. Samples were delivered to the MS via hydrophilic interaction chromatography (HILIC) using a 4.6 mm i.d. × 10 cm Amide Xbridge column (Waters) at 350 μL/min. Gradients were run starting from 85% buffer B (HPLC grade acetonitrile) to 42% B from 0 to 5 minutes; 42% B to 0% B from 5 to 16 minutes; 0% B was held from 16 to 24 minutes; 0% B to 85% B from 24 to 25 minutes; and 85% B was held for 7 minutes to reequilibrate the column. NVP-BKM120 eluted at approximately 3.50 minutes. Buffer A was composed of 20 mmol/L ammonium hydroxide/20 mmol/L ammonium acetate (pH 9.0) in 95:5 water:acetonitrile. Peak areas from the total ion current for the NVP-BKM120 metabolite selected reaction monitoring transition was integrated using MultiQuant v2.0 software (AB/SCIEX). For the concentration curve data, NVP-BKM120 was prepared at concentrations of 1 nmol/L, 10 nmol/L, 100 nmol/L, 500 nmol/L, 1 μmol/L, and 10 μmol/L in 40% methanol. Five microliters of each sample were injected using the parameters described earlier.

L.C. Cantley and J. Baselga have consulted for Novartis Pharmaceuticals, which is developing NVP-BKM120 for cancer treatment, and both authors are consultant/advisory board members of Novartis. J. Baselga has consulted for AstraZeneca, which is developing olaparib for cancer treatment. J. Balmañà is a consultant/advisory board member of Clovis. No potential conflicts of interest were disclosed by the other authors.

The Editors-in-Chief of Cancer Discovery are the authors of this article. In keeping with the AACR's Editorial Policy, the paper was peer reviewed and a member of the AACR's Publications Committee rendered the decision concerning acceptability.

Conception and design: A. Juvekar, R. Scully, L.C. Cantley, G.M. Wulf

Development of methodology: A. Juvekar, L.N. Burga, H. Hu, Y.H. Ibrahim, A. Papa, G.M. Wulf

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Juvekar, L.N. Burga, E.P. Lunsford, Y.H. Ibrahim, J. Balmañà, A. Rajendran, K. Spencer, C.A. Lyssiotis, C. Nardella, J. Baselga, J.M. Asara, G.M. Wulf

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Juvekar, L.N. Burga, H. Hu, E.P. Lunsford, Y.H. Ibrahim, P.P. Pandolfi, J. Baselga, R. Scully, J.M. Asara, L.C. Cantley, G.M. Wulf

Writing, review, and/or revision of the manuscript: A. Juvekar, L.N. Burga, E.P. Lunsford, C.A. Lyssiotis, C. Nardella, P.P. Pandolfi, J.M. Asara, L.C. Cantley, G.M. Wulf

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.P. Lunsford, A. Rajendran, A. Papa, G.M. Wulf

Study supervision: A. Papa, L.C. Cantley, G.M. Wulf

The authors thank the women who allowed us to study their tumor tissues. We are grateful to Min Yuan for help with mass spectrometry experiments. The authors also thank Novartis Pharmaceuticals for providing NVP-BKM120, Dr. Celina Garcia-Garcia, University Hospital Vall d'Hebron, for her work in the tissue collection study, and Drs. Ursula Matulonis and Eric Winer, Dana-Farber Cancer Institute, for their support and discussion of the data.

L.C. Cantley and G.M. Wulf are supported by a Stand Up To Cancer Dream Team Translational Research Grant, a Program of the Entertainment Industry Foundation (SU2C-AACR-DT0209), and by the Breast Cancer Research Foundation (BCRF). G.M. Wulf was supported by a Susan G. Komen for the Cure grant BCTR0601030. J.M. Asara is supported by NIH NCI grants 5P01CA120964-05 and 5P30CA006516-46. L.C. Cantley is supported by NIH grant GM41890 and NCI grant P01CA089021.