Buparlisib is a brain penetrable pan-PI3K inhibitor

Characterization of the genomic landscapes of intracranial tumours has revealed a clear role for the PI3K-AKT-mTOR pathway in tumorigenesis and tumour maintenance of these malignancies, making phosphatidylinositol 3-kinase (PI3K) inhibition a promising therapeutic strategy for these tumours. Buparlisib is a novel pan-PI3K inhibitor that is currently in clinical development for various cancers, including primary and secondary brain tumours. Importantly however, earlier studies have revealed that sufficient brain penetration is a prerequisite for antitumor efficacy against intracranial tumours. We therefore investigated the brain penetration of buparlisib using a comprehensive set of in vitro and in vivo mouse models. We demonstrate that buparlisib has an excellent brain penetration that is unaffected by efflux transporters at the blood-brain barrier, complete oral bioavailability and efficient intracranial target inhibition at clinically achievable plasma concentrations. Together, these characteristics make buparlisib the ideal candidate for intracranially-targeted therapeutic strategies that involve PI3K inhibition.

Despite some reports mentioning that buparlisib is BBB penetrable 17,18,21,23 , there are no pharmacokinetic data in the public domain supporting this claim. We here report a detailed analysis of the BBB penetration and oral bioavailability of buparlisib and demonstrate that it is a blood-brain barrier penetrable PI3K inhibitor with excellent oral bioavailability and intracranial target engagement.

Results
Buparlisib is only very weakly transported by murine BCRP in vitro. We first sought to investigate whether buparlisib (Fig. 1A) is transported by P-gp or BCRP in vitro using concentration equilibrium transport assays (CETAs). CETAs are sensitive assays to determine translocation over a cellular monolayer and are frequently used to determine substrate affinity for ABC transporters by comparing cell lines that overexpress ABC transporters with their parental counterparts [24][25][26] . Since P-gp and BCRP are apically located transporters, a prerequisite for detecting substrate affinity for these transporters in a CETA is the capacity of a molecule to penetrate cell membranes. We therefore first confirmed that buparlisib could efficiently diffuse over a cellular monolayer using a conventional transwell set-up. In this set-up, buparlisib plateaued to almost complete equilibrium between the apical and basal compartments within 4 hours, independent of the direction of diffusion (Fig. 1B). This diffusion was studied in the parental LLC-PK1 cell line and in presence of the P-gp inhibitor zosuquidar to avoid any confounding effects of porcine P-gp on buparlisib diffusion.
Next, we studied buparlisib translocation in CETAs using cell lines that overexpress murine P-gp (Abcb1a/ Mdr1a), murine BCRP (Abcg2/Bcrp1) or their human orthologues (ABCB1/MDR1, ABCG2). Only a minimal buparlisib translocation was observed in the MDCK-Bcrp1 CETA, whereas no translocation was found in any of the other CETAs, indicating that buparlisib is not a substrate of human P-gp, murine P-gp and human BCRP and a very weak substrate of murine BCRP (Fig. 1C). Note that functionality of the cell lines was confirmed as these assays were also performed in parallel using other compounds that did show basolateral to apical translocation.

Buparlisib brain penetration is not restricted by P-gp and BCRP in vivo.
To study whether P-gp and BCRP attenuate buparlisib brain penetration in vivo, we measured the brain and plasma concentration in wildtype (WT), Abcg2 −/− , Abcb1a/b −/− and Abcb1a/b; Abcg2 −/− mice 1 hour after i.v. administration of 2 mg/kg buparlisib. No differences could be observed among the mouse strains in both the plasma and brain concentrations of buparlisib ( Fig. 2A,B). These results demonstrate that P-gp and BCRP do not restrict the brain penetration of buparlisib in vivo. Interestingly, the brain-plasma ratio was between 1.5 and 2 in all strains, indicating that buparlisib exhibits excellent brain penetration (Fig. 2C). Remarkably, the brain concentration of buparlisib was even slightly higher than the concentrations in other well-perfused organs that contain fenestrated endothelium such as the liver, kidney and spleen ( Fig. 2D-F).

Buparlisib exhibits excellent oral bioavailability and achieves intracranial target inhibition.
Buparlisib is given orally in both preclinical 19 and clinical studies 21 , since it is claimed to have excellent oral bioavailability 17,18,21 . However, no data supporting this claim are available in the public domain. We therefore set out to determine the systemic exposure and brain concentrations after oral administration of buparlisib in mice.
First, we could confirm that buparlisib also achieves excellent brain penetration after oral administration, since the observed brain-plasma ratio in WT FVB mice was in line with those obtained following i.v. administration (Figs 2C, 3A). Interestingly, a dose-proportional increase in plasma and brain concentrations was observed, yielding similar brain-plasma ratios at three different dose levels. Although we only measured one time point, these data suggest that buparlisib displays linear pharmacokinetics between the 1 mg/kg and 5 mg/kg dose levels in mice.
Next, we studied the intracranial pharmacodynamics of buparlisib in WT FVB mice as measured by target inhibition. Immunoblotting of brain lysates revealed that buparlisib dose-dependently decreased the phosphorylation of AKT, the main target of PI3K, without affecting signalling through ERK (Fig. 3B). Importantly, efficient intracranial target inhibition could be reached in mice at clinically relevant plasma concentrations, since 5 mg/kg oral buparlisib resulted in plasma levels of approximately 1200 ng/mL (Fig. 3A) 21 .
Finally, we studied the pharmacokinetics of buparlisib in male and female mice and assessed the oral bioavailability in both genders by comparing the area-under-the-plasma concentration-time-curves (AUCs) following intravenous and oral administration. Both after oral and i.v. administration there was no difference in pharmacokinetics between male and female mice (Fig. 3C, Table 1). Strikingly, buparlisib displayed complete oral bioavailability since no difference in dose-corrected AUC could be observed between i.v. and oral administration in both genders. Interestingly, buparlisib exhibits a relatively small volume of distribution being less than 1 L/kg (Table 1), which is unusual for a small-molecule kinase inhibitor 27 . We therefore also assessed the buparlisib concentrations in a range of other tissues and indeed found that the levels were equal (liver) or even lower (kidney, spleen) than concurrent plasma levels ( Fig. 2D-F). Thus, the brain contains the highest concentration of all tissues included in this study, suggesting that buparlisib may be a very suitable PI3K inhibitor candidate for treatment of intracranial tumours.

Discussion
This study demonstrates that buparlisib has pharmacokinetic properties that make it an attractive candidate for treatment of intracranial malignancies in patients. The accumulation of buparlisib in the brain is higher than in most other organs and buparlisib inhibits PI3K in the brain at clinically achievable plasma concentrations. Characterization of the genomic landscapes of intracranial tumours has revealed a clear role for the PI3K-AKT-mTOR pathway in tumorigenesis and tumour maintenance of these malignancies [3][4][5][6] . Further work also supports that PI3K pathway hyperactivated tumours are vulnerable to PI3K inhition 2,18-20 . Thus, targeting this pathway is expected to be a promising therapeutic strategy for these tumours. Importantly however, sufficient brain penetration of any therapeutic agent is a prerequisite for antitumor efficacy against intracranial tumours [10][11][12][13][14]28 . The ability of a compound to access the brain is largely determined by two factors, namely the ability to diffuse over cellular membranes and the affinity for the drug efflux transporters P-gp and BCRP that are expressed at the BBB. Of all PI3K inhibitors that have thus far been reported, buparlisib shows by far the best brain penetration. The brain-plasma ratio ranges between 1.5 and 2 (Figs 2C, 3A), whereas the brain-plasma ratio of several other PI3K inhibitors that are considered brain penetrable (NVP-BEZ235, ZSTK474 and GNE-317) do not exceed 1 12,13 . Importantly, the buparlisib brain concentration is also higher than in most other tissues, which is in line with the relatively low distribution volume of this drug (<1 L/kg; Table 1) 27 . The reason why buparlisib distributes more to the brain than to other tissues is unclear, but could be related to more preferential protein binding in the brain or substrate affinity for uptake transporters that are present at the BBB. Regardless of the cause, this characteristic makes buparlisib a clear frontrunner among all PI3K inhibitors for development against intracranial tumours provided that the mouse mimics the human in this respect.
The plasma concentration-time profiles of buparlisib in mice and humans are very different. The peak plasma levels in humans receiving the maximum tolerated dose of buparlisib of 100 mg per day is about 1700 ng/ml (4.15 μM) and remains above 800 ng/ml (1.95 μM) for the remaining period of 24 h until the next dose 21 . By contrast, the peak plasma level in mice receiving 5 mg/kg is in the same range as in humans, but steeply declines to below 1 ng/ml (2.4 nM) after 24 h. Notably, a plasma concentration between 500 and 1000 nM was needed for inhibition of PI3K in the brain of WT mice (Fig. 3A). Single agent antitumor efficacy in preclinical models has already been shown, but only at 30 to 50 mg/kg dose levels 17,18 . The lack of efficacy at lower doses is likely due to the short duration of pharmacologically active plasma levels. However, high dose levels will cause high and non-clinically relevant plasma levels during the first hours and create the risk that some of the profound single agent antitumor activities observed in mice are in fact due to off-target effects 29 .
Thus far, only one PI3K inhibitor has received FDA approval. Idelalisib is a class I PI3Kδ inhibitor and has been approved as monotherapy for treatment of follicular lymphoma and small lymphocytic lymphoma 30 . Buparlisib inhibits all class I PI3K isoforms. Next to the canonical function of class I PI3Kα in regulating tumour cell proliferation and survival, roles for other class I PI3K isoforms in the immune system, including tumour-induced immune suppression, are emerging 31,32 . Thus, pan-PI3K inhibition may be advantageous for antitumor efficacy. Buparlisib is at an advanced stage of development, with several phase III trials underway in extracranial malignancies (e.g., NCT01633060, NCT02756247) and phase II trials ongoing for primary and secondary intracranial tumours (e.g., NCT02000882, NCT01339052). There is a growing body of clinical and preclinical evidence that PI3K inhibitors used in combination with other drugs may even be more promising for broader antitumor efficacy 33 . For instance in orthotopic mouse models of glioblastoma, the efficacy of PI3K inhibition as monotherapy has previously been shown to be modest at best, even when sufficient brain penetration and intracranial target inhibition could be reached 12,13 . Several preclinical studies have already pinpointed interesting strategies for combination treatment involving buparlisib, for Buparlisib has excellent intracranial target engagement and oral bioavailability. (A) Buparlisib was orally administered to wildtype mice at a dose of 1 mg/kg, 2 mg/kg or 5 mg/kg. One hour after injection, blood and tissues were collected for LC-MS/MS analysis. Buparlisib plasma and brain levels increase dose-dependently, yielding similar and excellent brain-plasma ratios at all dose levels that were tested. (B) Immunoblotting of brain tissue lysates from (A). Buparlisib efficiently inhibited phosphorylation of Akt S473 in the brains of wildtype mice following oral administration at a dose of 5 mg/kg, without affecting signalling through ERK. (C) Plasma-time curves of male and female wildtype mice following intravenous (2 mg/kg) or oral (5 mg/kg) administration. No differences could be observed in buparlisib plasma pharmacokinetics among both genders. The oral bioavailability of buparlisib is excellent, since oral and intravenous administration yielded similar dose-adjusted AUCs (see Table 1). All data are represented as mean ± SD (n = 4).  Table 1. Pharmacokinetic parameters of buparlisib after oral and i.v. administration to male and female FVB mice. AUC, area under the curve; C max , maximum concentration in plasma; t max , time to reach maximum plasma concentration; t 1/2 , elimination half-life; V z , apparent volume of distribution; CL, apparent clearance, F, oral bioavailability; V z /F, apparent volume of distribution after oral administration; CL/F, apparent clearance after oral administration. Data are represented as mean ± SD (n = 4).
instance combination with a MEK inhibitor 34 , Bcl-2 inhibitor 35 and CSF-1R inhibitor 36 in GBM or a Smo antagonist in medulloblastoma 37 . The finding that pharmacologically relevant plasma levels of buparlisib are achieved in patients warrants further clinical investigation of such combination therapies. In summary, buparlisib is a pan-PI3K inhibitor with excellent brain penetration, complete oral bioavailability and efficient intracranial target inhibition at clinically achievable plasma concentrations, making it the attractive candidate for intracranially-targeted therapeutic strategies involving PI3K inhibition.
Cell culture. All cell lines used in this study were previously generated in our institute and generously provided by Dr. A.H. Schinkel [38][39][40] . All cells were cultured in MEM supplemented with 10% FBS, 1% L-glutamine, 1% sodium pyruvate, 1% MEM vitamins, 1% non-essential amino acids and 1% penicillin/streptomycin (all from Life Technologies, Carlsbad, CA) under 37 °C and 5% CO 2 conditions. Concentration equilibrium transport assays. Conventional bidirectional transport assays (CTAs) and concentration equilibrium transport assays (CETAs) were performed as described previously 25 . Buparlisib was used at a concentration of 100 nM and, when appropriate, specific transport was inhibited using the P-gp inhibitor zosuquidar (5 μM) or the dual P-gp/BCRP inhibitor elacridar (5 μM). Transwell leakiness was determined as Carboxyl-[ 14 C]-inulin translocation exceeding 1.5% per hour and these wells were excluded from the analysis.
To prepare buparlisib transport assay samples for LC-MS/MS analysis, 10 μL medium samples were mixed with 30 μL of acetonitrile:formic acid (100:1 v/v). After centrifugation, the supernatant was 5-fold diluted in water and the buparlisib concentration was measured using an LC-MS/MS system as described below.
Animals. All animal housing and studies were approved by the Animal Experimental Committee of the Netherlands Cancer Institute and conducted according to national law and institutional guidelines. Mice were housed at 20.9 °C on a 12 hour light/dark cycle with food and water ad libitum.
Pharmacokinetic studies. The pharmacokinetics of buparlisib were analysed in WT, Abcg2 −/− , Abcb1a/b −/− and Abcb1a/b; Abcg2 −/− FVB mice. All knockout mice strains have been developed within our institute [41][42][43] . For i.v. administration, buparlisib was dissolved in DMSO and injected at a dose of 2 mg/kg. For oral administration, buparlisib was formulated in DMSO:Cremophor EL:water (1:1:8 v/v/v) and administered at 1, 2 or 5 mg/kg as indicated. Tail vein bleeding was used to collect blood at intermediate time points, whereas at the last time point blood was drawn by cardiac puncture and various tissue were collected. Plasma was obtained from whole blood by centrifugation (5 min, 5,000 rpm, 4 °C). After weighing, tissues were homogenized using a FastPrep ® -24 (MP-Biomedicals, NY) in 1% (w/v) bovine serum albumin in water. Buparlisib was extracted from plasma and tissue homogenate by liquid-liquid extraction with ethyl acetate using AZD8055 as an internal standard.

LC-MS/MS analysis.
Buparlisib was measured in samples from in vitro transport assays and in vivo pharmacokinetic studies using an LC-MS/MS system comprised of an UltiMate 3000 LC Systems (Dionex, Sunnyvale, CA) and an API 4000 mass spectrometer (Sciex, Framingham, MA). Samples were run through a Securityguard C18 pre-column (Phenomenex, Utrecht, The Netherlands) prior to separation on a ZORBAX Extend-C18 column (Agilent, Santa Clara, CA). Elution was done using a in a 5 minute gradient from 20% to 95% B (mobile phase A was 0.1% HCOOH in water (v/v) and mobile phase B was methanol). 95% B was maintained for 3 min followed by re-equilibration at 20% B. Multiple reaction monitoring was performed at 411.3/367.2 (buparlisib) and 418.2/138.4 (AZD8055). Analyst ® 1.6.2 software (AB Sciex; Foster City, CA) was used for system control and data analysis. Pharmacokinetic and statistical analysis. As described in detail before, CETA results were analysed with the General linear model repeated measures procedure of SPSS (v22; SPSS Inc, Chicago, IL) 25 . PK solver was used to determine pharmacokinetic parameters 44 . The standard error of the oral bioavailability was calculated using the formula below: Buparlisib concentrations in WT and transporter knockout mice in vivo were compared using one-way analysis of variance followed by post hoc Bonferroni tests. Statistical differences between the buparlisib PK parameters in male and female mice were calculated as Bonferroni corrected p-values using multiple Student's t-tests. Differences were considered statistically significant when p < 0.05.