Identification and Characterization of NVP-BKM120, an Orally Available Pan-Class I PI3-Kinase Inhibitor

Because of their crucial role in signal transduction, the dysregulated metabolism of phosphoinositides represents a key step in many disease settings. Multiple enzymes participate in the phosphorylation and dephosphorylation of the inositol head group of phosphoinositide lipids (1). Among them, the phosphoinositide 3-kinases (PI3K) have been the focus of extensive research and drug discovery activities for the last 20 years. Genetic alterations at multiple nodes in the PI3K pathway have been implicated in oncogenesis and cancer (2, 3). PI3K activation can occur in response to (i) constitutively active mutants and/or amplification of growth factor receptor tyrosine kinases (RTK) as in breast cancer with HER2 amplification (4), (ii) amplification of PI3K, (iii) gain of activating somatic mutations in the PIK3CA gene encoding the p110α catalytic subunit (5), (iv) overexpression or activating mutations of the downstream effector kinase Akt (6), (v) loss or inactivating mutations of PTEN (the phosphatase that breaks down phosphatidylinositol-(3,4,5)-trisphosphate (PIP3), or (vi) constitutive recruitment and activation by mutant forms of the Ras oncogene. Given this pervasive involvement in many cancers, the development of molecules that inhibit this pathway have recently been initiated in first in man studies (7).

Following the discovery of NVP-BEZ235, our first dual pan-PI3K/mTOR clinical compound (8), we sought to identify additional PI3K inhibitors from different chemical classes with different profiles against targets within the PI3K pathway. Because the PI3K pathway is a complex signaling pathway composed of multiple and interconnected components, elucidation of effectiveness and tolerability of novel therapeutic modalities is likely to depend upon clinical experience with a number of clinical candidates with varying pathway profiles. An example of this pathway complexity is the regulation of the oncoprotein Akt. To be activated, Akt is first recruited to the plasma membrane through a direct interaction between its N-terminal pleckstrin homology domain and PIP3 molecules. For full activity, Akt is then phosphorylated on 2 distinct sites: on Thr308 by PDK1 and on Ser473 by the mTORC2 complex (9). Once activated, Akt regulates key downstream effectors, including the downregulation of mTORC1, the master regulator of protein translation, growth, and autophagy. However, genetic and biochemical studies have shown that mTORC1 provides a negative feedback loop at the level of RTKs causing reactivation of the PI3K and the mitogen-activated protein (MAP) pathway upon mTORC1 inhibition. The same negative feedback loop was showed in tumor biopsies from patients treated with the mTORC1 inhibitor RAD001 (10). Thus, in some but not all clinical contexts, the combined inhibition of both PI3K and mTOR might be necessary for effective therapy. Given that different wiring applies depending on the genetic alterations responsible for PI3K pathway constitutive activation (11, 12), compounds with varied and well understood profiles will be necessary to design and develop tailored chemotherapy.

Here, we report on the biologic characterization of the clinical candidate NVP-BKM120. This molecule is a 2,6-dimorpholino pyrimidine derivative that is a potent pan-PI3K inhibitor. Unlike NVP-BEZ235, it does not significantly inhibit mTOR and is highly selective against other protein or lipid kinases. NVP-BKM120 exhibits potent antiproliferative and proapoptotic activities in tumor cell lines by specifically blocking the biologic function of PI3K signaling components. NVP-BKM120 shows good oral bioavailability in mice and its on-target activity observed in cellular assays translates well in in vivo models of human cancer in which NVP-BKM120 shows significant tumor growth inhibition or regression at tolerated doses.

NVP-BKM120 was synthesized in the Global Discovery Chemistry group (NIBR Oncology; Novartis). NVP-BKM120, RAD001 (Novartis), STI571 (Novartis), Wortmannin (Alexis; #ALX-350–020-M001), LY294002 (Alexis; #ALX-270–038-M025), ZSTK474 (LC Laboratories; #Z-1006), AZD6244 (Selleck Chemicals; #S1008), and AZD8055 (Active Biochem; #A-1008) were prepared as 10 mmol/L stock (except LY294002; 20 mmol/L) solutions in 100% dimethyl sulfoxide (DMSO). Working solutions were freshly prepared before addition to the cell media such that final DMSO concentrations were kept constant at 0.1% in both control and compound-treated cells. Platelet-derived growth factor (PDGF) stock solution was prepared as recommended by the manufacturer (Sigma; #P4306).

All the biochemical in vitro PI3K and protein kinase assays shown in Table 1 were previously described (13).

Fifty nanoliters of compound dilutions were dispensed onto black 384-well low-volume nonbinding polystyrene plates (Corning; NBS#3676). Then 5 μL of ATP and GFP-4EBP1 with 5 μL mTOR proteins (final assay volume: 10 μL) were added, and the reaction was incubated at room temperature in 50 mmol/L HEPES pH 7.5, 10 mmol/L MnCl₂, 50 mmol/L NaCl, 1 mmol/L EGTA, and 1 mmol/L DTT. Reactions were stopped with 10 μL of a proprietary mixture (IVG), containing the Tb³⁺-α-p4EBP1-(pT46) detection antibody, EDTA, in TR-FRET dilution buffer. Plates were then read 15 minutes later in a Synergy 2 reader using an integration time of 0.2 second and a delay of 0.1 second. The control for 100% inhibition of the kinase reaction was created by replacing the mTOR kinase with an equal volume of reaction buffer. The control for 0% inhibition was created by substituting solvent vehicle (90% DMSO in H₂O) without added test compounds. The TR-FRET mTOR kinase assay components were purchased from Invitrogen Corporation. Details are available upon request.

To determine the potency of NVP-BKM120 on the phosphatidylinositol 3-kinase–related kinases (PIKK) family member DNA-PK, an in vitro assay kit from Promega (SignaTECT; # V7870) was used in combination with the purified DNA-PK enzyme from Promega [DNA-dependent protein kinase (DNA-PK), V5811]. The in vitro kinase assay reactions were conducted according to the manufacturer's protocol but modified as follows: 27 U of purified DNA-PK protein/reaction, 1 μmol/L ATP/reaction, 1% DMSO, or indicated compound/reaction at 37°C for 30 minutes.

PI3Kα was incubated for 60 minutes in coated Maxisorp plates (14) in 50 μL medium containing [γ³³P]-ATP (∼6 kBq/well), 0.5 to 400 μmol/L unlabelled ATP, 5 mmol/L MgCl₂, 150 mmol/L NaCl, 25 mmol/L Tris-HCl pH 7.4, and 1% DMSO. The reaction was started by adding PI3Kα (0.4 μg/mL) and stopped by adding 50 μL of 50 mmol/L EDTA. Plates were washed twice with Tris-HCl buffer and dried, after which 100 μL/well MicroScint PS was added and bound radioactivity was determined using a TopCount (Packard) counter. Curves were fitted to a mixed inhibition model (15) by nonlinear 3 dimensional regression. In this model, the enzyme activity is related to S (ATP) and I (NVP-BKM120) by the following equation:

in which the apparent dissociation constant of ATP is represented by K₁ and the apparent dissociation constants of the inhibitor are represented by K₂ (competitive component) and K₃ (noncompetitive component).

To evaluate the isoform-specific potency of NVP-BKM120 in a cell-based system, an N-terminally myristoylated form of each PI3K class IA isoform was expressed in Rat1 fibroblasts. The retroviral expression plasmid pBabePuro (16) containing human p110α, p110β, and p110δ with an N-terminal myristoylation (myr) signal followed by an HA-tag were generated as described (17). Successfully infected Rat1 cells were selected in medium containing 4 μg/mL of puromycin, expanded and characterized for expression of the p110 isoforms. Transgenic expression of the myristoylated protein was confirmed by increased levels of phosphorylated Akt. The TSC1⁻/⁻-null MEFs mechanistic model for mTORC1 constitutive activation are from Dr. D. Kwiatkowski (Brigham & Women's Hospital, Boston, MA; ref. 18).

The culture conditions used to maintain U87MG, SF268, HT-29, PC3M, HCT116, and BT474 are described elsewhere (8, 19). The U87MG (University of Basel, Basel, Switzerland), HCT116 (American Type Culture Collection; #CCL-247), and BT-474 (Friedrich Miescher, Institute, Basel, Switzerland) cell lines were authenticated by single-nucleotide polymorphism analysis. The SF268 cell line (National Cancer Institute, Bethesda, MD) was not authenticated.

A total of 2 × 10⁶ cells were seeded per 10-cm dish. Eighteen hours later, the medium was discarded and replaced with 10 mL of fresh medium containing the test items. Unless specified in the figure legend, the incubation period was 30 minutes. For agonist activation, cells were first starved for 18 hours in 9 mL of medium containing only 0.05% FBS. Then cells were pretreated for 30 minutes by addition of 1 mL of medium containing the test items and subsequently stimulated for 10 minutes with the agonist (final concentrations: 50 ng/mL for PDGF, 100 ng/mL for epidermal growth factor (EGF), 60 ng/mL interleukin (IL)-4, and 1 μmol/L anisomycin). Cells were washed twice, lysed, and processed by Western blot analysis as described (8).

All the procedures have been described elsewhere (8, 20).

NVP-BKM120 and AZD6244 were formulated in NMP/PEG300 (10/90, v/v). Solutions were freshly prepared for each day of dosing by dissolving the powder, first in N-Methyl-2-pyrrolidone (NMP) with sonication and then by adding the remaining volume of PEG300. The application volume was 10 mL/kg. Herceptin/trastuzumab (150 mg; Hammer Apotheke) was prepared as a 0.22 μmol/L filtered sterile stock solution at 15 mg/mL in PBS that was subsequently diluted to 1 mg/mL aliquots and stored at −80°C. For each treatment (3 times per week), a new vial was thawed for immediate intraperitoneal treatment in animals with an application volume of 10 mL/kg (final dose of 10 mg/kg).

A maximum of 10 female Harlan nude (HsdNpa:Athymic Nude-nu) mice (8–10 weeks of age) were kept under sterile conditions (type III cage, in an Optimal Hygienic Conditions zone) with free access to food and water. Subcutaneous tumors were established by injection of 100 μL of PBS containing 2 × 10⁶ tumor cells (Rat1-myr-p110α, U87MG, HCT116, NCI-N87, and PC3M), in left flank of each animal. Orthotopic BT-474 tumors were established by injection of 5 × 10⁶ cells prepared in 30% Hank's Balanced Salt Solution (HBSS) containing 10% Matrigel, in the third mammary gland. Treatments were initiated when the mean tumor volume in each randomized group (n = 6–8) reached 100 to 150 mm³. For BKM120, treatments were carried out orally, once per day, using an application volume of 10 mL/kg. Treatments were stopped and animals sacrificed when the tumor size in the vehicle control group reached 1,000 to 1,200 mm³. Tumor volumes were determined by using calipers for measurement of longest (considered as length) and shortest (considered as diameter) dimensions of each tumor and according to the formula length × diameter² × π/6. Antitumor activity is expressed as T/C% (mean increase of tumor volumes of treated animals divided by the mean increase of tumor volumes of control animals multiplied by 100). Data are presented as means ± one SEM. Comparisons between groups and vehicle control group were done using either one-way ANOVA or ANOVA on ranks followed by the Dunnett tests when data were, respectively, either normally distributed or not. For all tests, the level of significance was set at P < 0.05. Calculations were carried out with SigmaStat v 2.03 (Jandel Scientific). All experimental procedures (approved by the Kantonales Veterinäramt Basel-Stadt under license #1769) strictly adhered to the Eidgenössisches Tierschutzgesetz and the Eidgenössische Tierschutzverordnung. For combination in vivo studies, the combination was determined using the Clarke method (21).

The methodologies used were as described previously (22).

NVP-BKM120 (Fig. 1A) was identified upon optimization of drug-like and in vivo protein kinase properties from a 2-morpholino-6-aminopyridyl-pyrimidine scaffold. The molecule binds in the ATP-binding site of the lipid kinase domain, as determined from its cocrystal structure in the PI3K p110γ isoform (PDB accession code: 3SD5) and by homology modeling in the PI3K p110α isoform (Fig. 1B). The alpha model suggests that there are 3 key hydrogen bond interactions (indicated by the dotted lines), formed by the oxygen of the 2-morpholino group and by the exocyclic nitrogen in the 6-position pyridyl substituent. The morpholine oxygen functions as a hydrogen bond acceptor with the backbone amino group of Val⁸⁵¹ in the hinge domain of the PI3K p110α isoform. A similar interaction is observed in all the published structures with ATP and with known inhibitors (23, 24). The 6-pyridyl exocyclic nitrogen (as a donor) binds to Asp⁸¹⁰ and Asp⁹³³ in the catalytic region. A fourth interaction could be modeled, as high resolution structures of p110γ are frequently observed to contain a water molecule forming a hydrogen bond bridge between Tyr⁸³⁶ and Asp⁸¹⁰ (25). Its presence could provide an additional hydrogen bond formed by its interaction with the 6-pyridyl endocyclic nitrogen acting as an acceptor.

Kinetic fittings of enzymatic results using a Michaelis–Menten model revealed that NVP-BKM120 inhibits PI3Kα via a mixed-type mode (15), affecting both the V_max and the K_m for ATP (Fig. 1C). Because the K_m for ATP increases with compound concentration (and correspondingly, the compound IC₅₀ value increases with ATP competition), NVP-BKM120 has greater affinity for the free, relative to the ATP bound, PI3K enzyme. While further study of the ATP-noncompetitive aspect has not been yet pursued, the ATP-competitive facet of this inhibition is consistent with the binding mode established by the crystallography results described earlier.

NVP-BKM120 is approximately equipotent against the class IA PI3Ks α, β, and δ and modestly less potent against the class IB γ isoform (Table 1). The compound also shows comparable potency against activating p110α somatic mutations that have been described in a wide array of human cancers (26). NVP-BKM120 is significantly less potent in biochemical assays against the PI3K class III family member Vps34, the related class IV PIKK protein kinases mTOR, ATR, and DNA-PK, and the distinct lipid kinase PI4Kβ. NVP-BKM120 was shown to be mostly inactive against all the kinases tested in an in-house selectivity panel with the exception of colony-stimulating factor 1 receptor (CSF1R). This inhibitory activity was confirmed at a concentration of 5 μmol/L in the Invitrogen functional kinase panel Supplementary Table S1 but no effect was observed in a cell-based CSF1R autophosphorylation assay (Supplementary Fig. S1A). NVP-BKM120 was further profiled in the Ambit kinase competition panel (Supplementary Table S2), in which EphA2 and fibroblast growth factor receptor 2 (FGFR2) kinases were found to be inhibited by more than 90% at a concentration of 1 μmol/L. Such as for CSF1R, these hits were not reconfirmed in FGFR2 (Supplementary Fig. S1B) and EphA2 (Supplementary Fig. S1C) cellular autophosphorylation assays. Therefore, NVP-BKM120 can be considered as a selective pan-class I PI3K inhibitor with a profile distinct from the previously described dual PI3K/mTOR inhibitor NVP-BEZ235 (8).

Rat1 cell lines that overexpress activated versions of each class IA PI3K isoform were engineered and pathway activation was confirmed by the concomitant increase in S473P-Akt levels (Supplementary Fig. S2). NVP-BKM120 inhibited S473P-Akt in a concentration-dependent manner (Fig. 2A) with IC₅₀ values (RPPA determination) of 104 ± 18, 234 ± 47, and 463 ± 87 nmol/L (n = 7) in lines that express myr-p110α, β, and δ, respectively. These reductions in S473P-Akt levels are probably not attributable to mTOR inhibition, as the IC₅₀ value for S6RP S235/236 phospho levels was 922 ± 142 nmol/L (n = 5) in the mTORC1 constitutively activated TSC1^−/− cells. In contrast, the dual PI3K/mTOR NVP-BEZ235 as well as the mTOR catalytic inhibitor AZD8055 (27) and NVP-BEZ235 were, as expected, extremely active in this assay with IC₅₀ values of 3.5 ± 0.6 (n = 4) and 2.0 ± 0.8 (n = 4), respectively (Fig. 2B).

In PTEN-null U87MG cells, NVP-BKM120 treatment reduced both S473P-Akt (IC₅₀ value: 130 ± 44 nmol/L, ELISA) and T308P-Akt (IC₅₀ value: 229 ± 40 nmol/L, ELISA) levels with similar potency comparable with what is observed in the mechanistic Rat1-myr-p110 models (Fig. 2C). This effect is correlated in a dose-dependent manner with inhibition of downstream direct (FOXO3A/FKHRL1) and indirect (p70S6K) Akt effectors, showing profound Akt pathway modulation upon NVP-BKM120 exposure. Compound washout studies in U87MG cells showed that these inhibitory effects are fully and rapidly reversible upon compound withdrawal (Supplementary Fig. S3), suggesting rapid binding kinetics and a short residence time of the compound on the target.

To further evaluate the specificity of NVP-BKM120, the compound was tested in cellular models in the presence of various agonists that activate a spectrum of well-characterized signaling pathways. In the presence of mitogenic stimuli such as PDGF (Fig. 2D) or EGF (Supplementary Fig. S4A), NVP-BKM120 blocked S473P-Akt induction by the respective growth factors but failed to affect receptor activation (Y751P-PDGFR and Y845P-EGFR, respectively) and activation of the mitogen-activated protein kinase (MAPK) pathway. This is in contrast to the effects observed with PDGF receptor (STI571) or EGF receptor (NVP-AEE788) inhibitors. Similarly, NVP-BKM120 did not inhibit IL-4–induced activation of STAT6 (Supplementary Fig. S4B), anisomycin-induced activation of c-jun-NH₂-kinase (JNK) and p38 (Supplementary Fig. S4C) nor class IV PIKK (ATM and DNA-PK) activation in response to DNA double-strand breaks (Supplementary Fig. S4D). Cumulatively, these results further show the specificity of NVP-BKM120 in a cellular context.

Activation of the PI3K pathway is a common feature in human cancers, arising through various genetic abnormalities (28). To test whether NVP-BKM120 would be an effective treatment for tumors presenting these abnormalities, the antiproliferative activity of the molecule was determined against a panel of 353 cell lines that vary with respect to key genetic determinants such as the status of the PIK3CA, PTEN, and KRAS genes. Cell lines characterized by the presence of an oncogenic mutation in the PIK3CA gene showed greater and statistically significant (P = 0.01) sensitivity to NVP-BKM120 than cells bearing a nonmutated wild-type form of the gene (Fig. 3A). No differences in sensitivity could be observed when comparing cells carrying a deletion or a mutation in the PTEN gene (Fig. 3B) or an activating mutation in the KRAS gene (Fig. 3C) when compared with cells bearing wild-type PTEN and KRAS genes, respectively. However, in the latter case, a trend toward more sensitivity in KRAS wild-type cells is seen. PIK3CA and KRAS mutation often co-occur in human colorectal adenocarcinoma. A multivariate analysis looking at status of both genes revealed that PIK3CA mutant/KRAS wild-type tumors are most sensitive to NVP-BKM120 versus other possible combinations (Supplementary Fig. S5A) and most notably in comparison to PIK3CA mutant/KRAS mutant models (P = 0.00005). These data show that the presence of a KRAS mutation could overcome the sensitivity to a PI3K inhibitor, probably by activation of alternate, KRAS-dependent signaling pathways such as the extracellular signal-regulated kinase (ERK) pathway. Interestingly, cotreatment of NVP-BKM120 with a MAP/ERK kinase (MEK) inhibitor could induce cell death, as revealed by PARP cleavage, in KRAS mutant/PIK3CA mutant HCT116 cells when single-agent treatment was essentially ineffective (Supplementary Fig. S5B). A similar observation was made for ZSTK474, another pan-PI3K inhibitor (29), suggesting that this phenomenon is not unique to NVP-BKM120.

To determine whether NVP-BKM120 could reach exposure leading to pathway modulation in tumor tissue, Rat1-myr-p110α tumor-bearing animals were treated orally, once, at dose levels of either 30 or 60 mg/kg. Phosphorylated Akt levels (Ser 473) as well as compound concentrations were then determined, at different time points postdrug administration (Fig. 4A). Maximum inhibition of p-Akt levels in tumor tissue was achieved at the 1 hour time point when the concentration in both tumor and plasma compartments was highest, for both dose levels. Phosphorylated Akt levels remained strongly inhibited 8 hours postdose (the inhibition was equal or superior to 80%, at 30 and 60 mg/kg, respectively) and at 16 hours postdose for the 60 mg/kg dose level (70% inhibition). A similar relationship is observed in the PTEN-null U87MG model (Supplementary Fig. S6A). Overall, these data show that NVP-BKM120 oral administration can achieve pharmacodynamic effect in the tumor in a concentration-dependent manner, showing pharmacokinetic and pharmacodynamic relationship.

To show that the observed pharmacodynamic effects could translate into effective antitumor activity, Rat1-myr-p110α tumor-bearing mice were treated orally with NVP-BKM120 once per day at doses of 40, 50, or 60 mg/kg. At these dose levels, the compound was well tolerated (right) and a robust reduction in tumor mass (left) was observed (T/C values of −25%, −48%, and −46%, respectively; Fig. 4B), showing the ability of NVP-BKM120 to treat PI3K-addicted tumors. A similar response was observed in the MCF7 PIK3CA mutant model MCF7 (Supplementary Fig. S6C). Consistent with the in vitro data (Fig. 3), the compound was less efficacious against U87MG PTEN-null tumors in mice compared with PIK3A-mutated cell lines, even at the dose of 60 mg/kg (Supplementary Fig. S6B).

Because clinical benefit in the treatment of solid tumors is often obtained as a combination therapy, NVP-BKM120 was combined with a number of standards of care agents or with targeted inhibitors to model rational combinatorial treatments. Genetically engineered lung KRAS mutant tumors have been shown to be almost insensitive to PI3K inhibitor treatment, but complete efficacy was obtained when combined with a MEK inhibitor (30). Similarly, when grown in vivo, KRAS mutant/PIK3CA mutant HCT116 tumors were found to be relatively insensitive to suboptimal dose levels of either NVP-BKM120 or the MEK inhibitor AZD6244. However, significant tumor regression was observed in a very well-tolerated manner when both drugs were combined (Fig. 4C and Supplementary Fig. S6D). These data show that the combination of both drugs lead to activity superior to that seen with either agent alone. In agreement with this, more apoptosis was observed in vitro, when HCT116 were exposed to both drugs (Supplementary Fig. S5B).

As true for other pathways, the PI3K/Akt/mTORC1 pathway is subjected to regulation by negative feedback loops. Efficacy of PI3K inhibitors might, therefore, be limited by reactivation of critical upstream nodes. Such a phenomenon has recently been described in RTK-addicted tumor lines such as breast HER2-amplified models, in which PI3K inhibition led to ERK and PI3K pathway reactivation through HER2/HER3 overexpression and overactivation (31–34). Cotreatment with both PI3K and RTK inhibitors in RTK-addicted tumors is, therefore, warranted to avoid this adaptive mechanism of resistance. To test this hypothesis, orthotopic breast HER2-amplified BT474 tumor-bearing mice were treated with NVP-BKM120 in presence or absence of the HER2 antagonist trastuzumab (Fig. 4D and Supplementary Fig. S6E). Only partial responses could be obtained as single agents (T/C of 49.7% and 18.7% for Herceptin and BKM120, respectively) but significant tumor regression was obtained when both drugs were combined (T/C of −15%) with no impact on body weight. Similar data were obtained for the HER2-amplified gastric cancer tumor model NCI-N87 (Supplementary Fig. S7A).

The alkylating agent temozolomide or the microtubule stabilizer taxotere are 2 standard of care agents used to treat glioblastoma multiforme (GBM) or prostate cancer, respectively. In both indications, the PTEN gene is frequently inactivated, suggesting some dependencies on the PI3K signaling. Suboptimal dosage regimen of NVP-BKM120 (30 mg/kg) was used to check whether it could give advantages in combination studies to these agents. Whereas only partial activity is observed as single agent (as was the single agent use of temozolamide or taxotere), the combination of temozolamide or taxotere with NVP-BKM120 against GBM U87MG (Supplementary Fig. S7B) and prostate PC3M (Supplementary Fig. S7C) tumors, respectively, caused robust tumor shrinkage in a well-tolerated manner. These results show the potential for at least additive activities with these cytotoxic agents and NVP-BKM120 in selected settings.

Overall, these in vivo efficacy data show that NVP-BKM120 is an orally available PI3K inhibitor that displays a good correlation between its pharmacokinetic and specific pharmacodynamic readouts, resulting in a strong antitumor activity either alone or in combination with targeted or cytotoxic therapies in relevant disease models.

PI3K can be actively recruited and activated by angiogenic factors (35, 36). PI3K inhibitors are, therefore, expected to display some degree of antiangiogenic activity. As described for NVP-BEZ235 (22), NVP-BKM120 treatment readily blocks VEGF induced neovascularization in vivo (Supplementary Fig. S8A). Angiogenic tumors are characterized by tortuous, chaotic, and highly permeable vessels. This phenomenon is in part due to VEGF-induced eNOS activation, through a PI3K-dependent mechanism involving Akt (22). Treatment of highly angiogenic rat mammary BN472 tumor-bearing rats with NVP-BKM120 led to significant reduction of tumor vasculature leakiness from the tumor tissue, as reflected by the strong reduction of Evans blue–mediated fluorescence (Supplementary Fig. S8B). As a consequence, NVP-BKM120 treatment also produced a strong reduction in tumor interstitial fluid pressure (IFP), which depends on the vasculature permeability (Fig. 5A). Reduction in permeability and IFP could be measured in a noninvasive manner by dynamic contrast enhanced (DCE)-MRI, in the presence of the contrasting agent Vistarem. Treatment of BN472 tumor-bearing rats with NVP-BKM120 strongly reduced tumor permeability as reflected by the rapid drop in K_trans factor (Fig. 5B). These data confirm that NVP-BKM120 possesses strong antiangiogenic properties and these can be visualized and quantified by DCE-MRI imaging technology.

A plethora of preclinical evidence predicts that inhibitors targeting various components of the PI3K pathway will have therapeutic benefit in cancer. However, a key question that remains to be answered is: what is the most effective and tolerated inhibitory profile of a small molecule cancer therapeutic targeting this pathway? Because cancer is not one disease, but many, it is unlikely that all tumor types, encompassing different lineages and genetic status, would be similarly “wired” in this signal transduction machinery. Therefore, to maximize chances of clinical success, following the discovery of the dual PI3K/mTOR inhibitor and imidazoquinoline derivative NVP-BEZ235, we pursued the development of other clinical candidates from a different chemical space and with a different biologic profile.

This effort led to the discovery of NVP-BKM120, identified from the pan-PI3K inhibitor lead series 2-morpholino-4-amino-6-pyrimidinyl pyrimidines (MAPP) through a combination of structure-based drug design and optimization of in vivo properties (37). NVP-BKM120 inhibits PI3K in an ATP competitive manner, but this is the result of a mixed effect on K_m and V_max. NVP-BKM120 spares tyrosine or serine/threonine protein kinases, and unlike NVP-BEZ235, is not an effective mTOR inhibitor. This distinction can be explained by the binding modes of the 2 molecules: the proposed binding mode of NVP-BEZ235 in PI3Kα showed a key interaction with the Ser 774 residue of PI3K that was proposed to also be important for mTOR inhibition, due to the H-bond interaction with the mTOR corresponding residue, Ser 2665 (8). NVP-BKM120, on the other hand, lacks this contact with the Ser 774 residue providing a rationale why it is a much less effective mTOR inhibitor.

The effects of NVP-BKM120 on the PI3K pathway downstream biomarkers p-Akt and pRPS6 in mechanistic PI3K (Rat1-transduced cells) and mTORC1 (TSC1^−/− minus MEFs) cell systems, respectively, show the predictability of the biochemical profile in a cellular environment. These effects were also observed in disease models, showing the potency, specificity, as well as the dose dependency of the compound to block the PI3K pathway in relevant cell-based models. NVP-BKM120 inhibitory effects in cells can be rapidly reverted upon drug washout, in agreement with the fast kinetics of the drug on PI3K observed in vitro. Pathway inhibition by NVP-BKM120 led to proliferation inhibition in a variety of cell lines, representing different lineages and genetic abnormalities that promote PI3K pathway activation. Interestingly, PIK3CA mutant lines were found to be statistically more sensitive than PIK3CA wild-type lines. However, equivalent sensitivity was observed between cells bearing a functional PTEN gene versus cells with phosphatase inactivating mutations or deletions. Identical observations were made for the pan-class I PI3K inhibitor GDC-0941, in an analysis restricted to a panel of 54 breast cancer lines (38). However, another study conducted with different PI3K inhibitors carried out in a panel of 39 lines from 9 different lineages did not show enhanced activity in either PIK3CA or PTEN mutant lines (39). These conflicting data might reflect the fact that, in the latter study, only a small number of lines for each lineage and genetic alteration were analyzed, with a strong imbalance in the representation of each characteristic. Moreover, evidences for an important function for p110β in PTEN-null cell and tumor models have been reported (40, 41). BKM120 is slightly less potent on p110β and the concentration used to establish the sensitivity profile (0.8 μmol/L) might have been insufficient to completely inhibit this isoform in our study.

Overall, pharmacologic inhibition of PI3K shows that: (i) a PIK3CA mutation is certainly not identical to a PTEN mutation/deletion in terms of PI3K addiction and (ii) PTEN mutant or deleted tumors might have addiction to other pathways for proliferation and survival. Understanding what these pathways would be could provide insightful information on which combination partners could be used to complement the efficacy of PI3K inhibitors. Although PIK3CA mutant tumors seemed to be more sensitive to PI3K inhibitors, one cannot conclude that all PIK3CA mutant tumors will respond to them. For instance, the cooccurrence of KRAS activating mutations dramatically reduces the sensitivity to NVP-BKM120. One explanation to this phenomenon is that pathways often converge into similar and important downstream key effectors, and efficacy is dependent upon silencing of all critical inputs to such downstream node. This was recently elegantly illustrated in KRAS mutant/PIK3CA mutant colorectal cancer models, in which efficacy could be achieved only when cyclin D1 levels could be reduced by cotreatment with MEK and Akt inhibitors (42). Similarly, apoptotic events in the HCT116 lines could be detected only when NVP-BKM120 was combined with the MEK inhibitor AZD6244, and these effects translated in strong enhanced in vivo antitumor activity.

Target modulation and good pharmacokinetic and pharmacodynamic correlation was achieved with NVP-BKM120 in tumor-bearing mice. Significant and dose-dependent antitumor activity was observed at dose levels sufficient to shut down the PI3K pathway, showing the strong relationship between compound concentration, pathway inhibition and efficacy. Suboptimal dose of NVP-BKM120 was enough to strongly enhance the antitumor activity of standard of care (cytotoxic or targeted agents) in various cancer types such as prostate (with Taxotere), HER2-amplified breast cancer, or gastric cancer (with Herceptin/trastuzumab) and GBM (with temozolomide). The fact that enhanced activity was also observed in vivo against KRAS mutant HCT116 tumors when BKM120 was combined with the MEK inhibitor AZD6244 and opens new opportunities in indications with high and unmet medical need, in which KRAS is frequently mutated (such as pancreatic cancers), but confirmatory studies are warranted to test and eventually validate this paradigm in the relevant preclinical models.

As found with other PI3K inhibitors (22, 43), NVP-BKM120 possesses strong antiangiogenic activity. Hence, highly angiogenic tumors could be very sensitive to the compound. NVP-BKM120 could, therefore, be used efficiently as second line treatment upon failure of approved antiangiogenic drugs such as sutent, or sorafenib, in the case of metastatic renal cell carcinoma. Of interest is the fact that NVP-BKM120 possesses excellent brain penetration, hence, there is the possibility to treat advanced nonresectable highly angiogenic GBM, mostly in combination with temozolomide-based therapies.

In conclusion, we have identified NVP-BKM120 as a selective pan-class I PI3K inhibitor. This compound possesses excellent drug-like properties that have allowed its selection for full clinical development. NVP-BKM120 is currently in phase II trials, in parallel with our dual PI3K/mTOR inhibitor NVP-BEZ325. The outcome of these studies is eagerly awaited, as they will be of primary importance in understanding the clinical relevance of PI3K or dual PI3K/mTOR inhibition in different cancer indications.

S.-M. Maira, S. Pecchi, A. Huang, M. Burger, M. Knapp, D. Sterker, C. Schnell, D. Guthy, T. Nagel, M. Wiesmann, S. Brachmann, C. Fritsch, P. Chène, K. Shoemaker, A.D. Pover, D. Menezes, G. Martiny-Baron, D. Fabbro, C.L. Wilson, R. Schlegel, F. Hofmann, W.R. Sellers, and C.F. Voliva are Novartis employees. M. Dorsch and C. Garcia-Echeverria are Sanofi-Aventis employees.

The authors thank Marc Hattenberger, Fabian Stauffer, Sandra Mollé, Laurent Laborde, and Juliane Vaxelaire for excellent technical assistance.

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.