BKM-120 (Buparlisib): A Phosphatidyl-Inositol-3 Kinase Inhibitor with Anti-Invasive Properties in Glioblastoma

Glioblastoma is an aggressive, invasive tumor of the central nervous system (CNS). There is a widely acknowledged need for anti-invasive therapeutics to limit glioblastoma invasion. BKM-120 is a CNS-penetrant pan-class I phosphatidyl-inositol-3 kinase (PI3K) inhibitor in clinical trials for solid tumors, including glioblastoma. We observed that BKM-120 has potent anti-invasive effects in glioblastoma cell lines and patient-derived glioma cells in vitro. These anti-migratory effects were clearly distinguishable from cytostatic and cytotoxic effects at higher drug concentrations and longer durations of drug exposure. The effects were reversible and accompanied by changes in cell morphology and pronounced reduction in both cell/cell and cell/substrate adhesion. In vivo studies showed that a short period of treatment with BKM-120 slowed tumor spread in an intracranial xenografts. GDC-0941, a similar potent and selective PI3K inhibitor, only caused a moderate reduction in glioblastoma cell migration. The effects of BKM-120 and GDC-0941 were indistinguishable by in vitro kinase selectivity screening and phospho-protein arrays. BKM-120 reduced the numbers of focal adhesions and the velocity of microtubule treadmilling compared with GDC-0941, suggesting that mechanisms in addition to PI3K inhibition contribute to the anti-invasive effects of BKM-120. Our data suggest the CNS-penetrant PI3K inhibitor BKM-120 may have anti-invasive properties in glioblastoma.


Cell invasion and migration assays. For 3D spheroid cultures, 5,000 cells/well were cultured in Corning
Ultra-Low Attachment Surface 96 well plates (Corning Inc., Corning, NY) in 100 μ l medium. After 24 hours, the medium was replaced with 50 μ l of collagen I (Advanced BioMatrix, Inc San Diego, CA, USA) and neutralized to pH 7.5 using 1N NaOH and supplemented with FCS, penicillin-streptomycin, and 5 x DMEM. After polymerization, the collagen was overlaid with 50 μ l of medium containing drug or vehicle. Wound-healing assays were performed as described 16 . Transwell assays were performed using 3 μ m pore size fluoroblok inserts (Corning Inc., Corning, NY). Poly-ɛ -caprolactone nanofibers (Nanofiber Solutions, Columbus, OH) were prepared and quantitated as described 18 .
Cell viability, cell-substrate adhesion and cell-cell adhesion assays. Cell viability and proliferation were measured using PrestoBlue reagent according to the manufacturer's instructions (Life Technologies, Grand Island, NY). Cell-substrate adhesion assays were performed with adherent G9-copGFP cells in paired 96 well plates. Cells were treated with vehicle (DMSO) as control or BKM-120 using four replicates at each time point for each condition. Drug and vehicle were washed out every 30 min and replaced with fresh medium. One plate was used for microscopic observation; cells were fixed with 4% PFA, treated with 5% normal goat serum and 0.5% Triton-x-100 in PBS for 1 hour at room temperature, and stained with DAPI. The second plate was used to evaluate cell-substrate adhesion with the CellTiter-Glo ® Luminescent Cell Viability Assay (Promega, Madison, Wisconsin, USA) according to the manufacturer's instructions. Cell-cell adhesion was evaluated using a G9-copGFP single cell suspension treated with DMSO or BKM-120 and placed in a circular shaker in the cell incubator. At prefixed time-points 50 μ l of each suspension was transferred to a 96-well plate and fixed with 4% PFA. Each experiment was performed in 4 replicates.
In vivo studies, drug and tissue preparation. 2 × 10 5 G9-copGFP or U1242 cells were stereotactically injected into nude mouse brains (2 mm right lateral, 1 mm frontal to the bregma and 3 mm deep) as described 15 . A short timeline was chosen in order to show a clearer anti-invasive effect, therefore treatments started on day 4 and on the first day of injection three mice were sacrificed to show the tumors at time 0. Starting on day 7 for the G9-copGFP and day 4 for the U1242 post-implantation five mice per group were treated for four days by daily gavage with vehicle (NMP/PEG300, Sigma-Aldrich), 2 mg/kg or 20 mg/kg of BKM-120. Brains were harvested on day 10 (G9-copGFP) and on day 7 (U1242), placed in 4% PFA for 24 hours, followed by 30% sucrose for 48 hours, and frozen. Sections (30 μ m) were blocked with 5% donkey serum/0.5% TX100/ PBS for 1 hour at room temperature, followed by Vimentin SP20 (1:200 O/N, ab16700 Abcam, Cambridge, MA, USA) and Hoechst 33342 staining (1:3000, H3570 Life Technology, Grand Island, NY, USA). BKM-120 was formulated in NMP/PEG300 (10/90). Solutions were freshly prepared on each day of dosing by dissolving the drug first in NMP and then by adding the remaining volume of PEG300. The human equivalent dose (HED) of BKM-120 was calculated according to the US Food and Drug Administration method previously described by Shannon et al. 19,20 . Microscopy. Nikon Eclipse Ti, Nikon TE2000 and Zeiss LSM710 confocal microscope systems with on-stage incubators for live cell imaging were used. Data were edited/quantified using ImageJ (http://rsb.info.nih.gov/ij/), Microsoft Excel and Prism 6 (GraphPad Software Inc., San Diego, CA).
Statistical analysis. Two sample t tests (adjusted with Bonferroni's method), one-way and two-way ANOVA (Greenhouse-Geisser correction) were used. IC 50 value was calculated as the drug concentration required to reach an inhibitory effect of 50% compared to the vehicle control. Results BKM-120 inhibits GBM spheroid invasion in a dose-dependent manner. PI3K has been previously linked to cell migration in several studies 5,13 , and is a major oncogenic signaling molecule in GBM 1,5 . Initially, we examined the effects of the CNS-penetrant selective class I PI3K inhibitor BKM-120 in 3D spheroid invasion assays on a panel of GBM cell lines, including low passage GSCs, neural stem cells and astrocytes (Fig. 1A, Supplementary Fig. 1A). BKM-120 inhibited GBM spheroid invasion in a dose-dependent manner, with a significant blockade observed at 1-2 μ M (Fig. 1A). Invasion was inhibited by at least 50% in all cell lines at 2 μ M (Fig. 1B), and in highly sensitive cell lines (G9, U251, G157, G146 and U1242) invasion was strongly blocked at this concentration (>90% inhibition). Figure 1C shows a clear blockade of invasion in G9-copGFP cells treated with 2 μ M BKM-120 over 48 hours. Cell proliferation assays showed that across the cell line panel the IC 50 values for cell viability were higher than those for invasion, confirming the ability of BKM-120 to specifically inhibit GBM invasion (Fig. 1D, Supplementary Fig. 1A).

BKM-120 inhibits GBM cell migration in multiple in vitro assays and its effects are reversible.
We then investigated the effects of BKM-120 on GBM cell migration in a range of additional in vitro migration assays. Wound-healing "scratch" assays confirmed that BKM-120 significantly inhibited GBM cell migration in a dose-dependent manner in U251 and G9 cells ( Fig. 2A, Supplementary Fig. 2A) at similar concentrations to those observed in spheroid assays. We then examined the effect of BKM-120 on GBM cell migration on aligned nanofiber scaffolds, which we previously developed as a topographic substrate for cell migration 18 . Cell migration was inhibited in a dose-dependent manner in multiple cell lines similar to above ( Supplementary Fig. 2B), and was strongly inhibited at 2 μ M BKM-120 (Fig. 2B). Importantly, using the nanofiber assay we confirmed that the effects of BKM-120 were not due to irreversible cytotoxicity; a drug wash-out experiment performed after 48h treatment with 1 μ M BKM-120 ( Supplementary Fig. 2C,D) and 2 μ M BKM-120 showed that the anti-migratory effects of BKM-120 were reversible (Fig. 2B). Transwell assays performed on U251-copGFP (Fig. 2C) and G9-copGFP cells at the initial time-point (6 hours) and at later time-points (24 and 48 hours, data not shown) showed a similar block of migration. In summary, these data demonstrate that BKM-120 potently and reversibly inhibits migration in a dose-dependent manner. Moreover, cell morphology changes induced by BKM-120  observed during the nanofiber experiment were confirmed in U251-copGFP cells treated with 1 μ M BKM-120, in which cells became completely rounded up after 24 hours (Fig. 2D).

BKM-120 blocks GBM cell-adhesion.
We then investigated the impact of BKM-120 on GBM cell-substrate and cell-cell adhesion in vitro. Cell-substrate adhesion in G9-copGFP cells was significantly inhibited by 2 μ M BKM-120 and confirmed by microscopic imaging after 200 min (p = 0.009) (Fig. 3A). Cell-cell adhesion was measured using an aggregation assay in suspension culture, by setting three size ranges and counting the number of cell clusters in each specific range dimension. This assay therefore allows discrimination between aggregates of different size. At time 0 all the cell clusters, both in control or BKM-120 treated samples, have a size < 3000 μ m, indicating small cell-cell aggregates and single cells. After 160 min, 10% of control cells had formed aggregates with size in the 3000-6000 μ m range, whereas all of the BKM-120 treated cells remained in the smaller range (p < 0.001). This pattern continued over the full time course of the experiment (Fig. 3B). Taken together these data demonstrate that BKM-120 blocks cell-substrate and cell-cell adhesion suggesting it affects the ability of cells to form robust contacts.
PI3K-independent effects of BKM-120 contribute to the migratory phenotype. We then investigated the anti-invasive effects of another potent selective PI3K inhibitor, GDC-0941 21 , that is in phase I clinical studies for advanced solid tumors 22 . This showed a modest effect on cell invasion compared with BKM-120 even though PI3K inhibition was stronger after GDC-0941 treatment as shown by Akt phosphorylation (Fig. 3C). GDC-0941 did not fully block migration even at concentrations as high as 10 μ M (Fig. 3C), cell-substrate adhesion in the presence of 2 μ M GDC-0941 was not significantly altered compared with controls (data not shown, p = 0.1) and was significantly less compared with BKM-120 (p = 0.009). In vitro kinase selectivity profiling revealed no significant differences between BKM-120 and GDC-0941 in a panel of 140 protein kinases and 15 lipid kinases (Supplementary Table 1). Similarly, we did not see any differences between the effects of either drug in an ELISA-based array of 39 major phospho-proteins (Supplementary Figure 3A). In fact, our Western blotting data suggests that GDC-0941 has a stronger inhibitory effect on PI3K than BKM-120. Overall, these data suggest that PI3K inhibition alone is not sufficient to block glioblastoma cell migration.
To better understand these observations and the potential PI3K-independent effects of BKM-120, we examined the previously described potential effects of BKM-120 on microtubules 12 . In U251 cells, after 24 hours of BKM-120 , the cells became rounded up with possible damage to the microtubule cytoskeleton (Fig. 3D). To further examine microtubule behavior we used U251 and U1242 cells expressing EB1-GFP. EB1 preferentially localizes to the plus-ends of growing microtubules and is pro-migratory in GBM 23 . A time-lapse study revealed a reduction in microtubule dynamics as indicated by EB1 trafficking after BKM-120 treatment, compared to a modest effect of GDC-0941 ( Fig. 3D and Supplementary 3B). Due to the effects on cell adhesion observed above, we also investigated the effects of BKM-120 and GDC-0941 on focal adhesions using U251 cells expressing a paxillin-GFP fusion. This showed that 2 μ M BKM-120 completely eliminated focal adhesions after 24 hours, whereas GDC-0941 had a much less pronounced effect (Fig. 3D). These data suggest that the effects of BKM-120 on the inhibition of GBM invasion and migration could be due to off-target effects on focal adhesions and microtubule dynamics.

BKM-120 reduces GBM tumor spread in vivo in orthotopic xenografts. The effect of BKM-120
in vivo was examined using two orthotopic xenograft models. We performed a 10 day treatment in G9-copGFP cells and a 7 day study in U1242 cells. The G9-copGFP tumor-bearing mice (5 for each group) were treated with vehicle and BKM-120 20 mg/kg ( Supplementary Fig. 3D). The U1242 tumor-bearing mice (5 for each group) were treated with vehicle and 2 different concentrations of BKM-120, 2 mg/kg and 20 mg/kg, corresponding to a human equivalent dose (HED) of 1.62 mg/kg and 0.16 mg/kg respectively (Fig. 4B). The tumor perimeter was measured after the treatments at day 7, and and the average tumor size was normalized to day 4 showing a strong and dose-dependent effect of BKM-120 in vivo (Fig. 4A). Microscopic analysis of tumor sections revealed a marked difference in tumor distribution compared with vehicle-treated control mice and high magnification analysis of tumors revealed a well-defined tumor/normal brain interface after treatment, and reduced migration ( Fig. 4B and supplementary Fig. 3D).
Thus, BKM-120 has anti-invasive properties that alongside BBB penetration may render this a useful molecule for limiting GBM invasion. These effects are not mediated completely by PI3K inhibition, and future mechanistic studies will investigate the additional targets involved in these effects.

Discussion
Our study shows that the CNS-penetrant PI3K inhibitor BKM-120 is a candidate anti-invasive drug for GBM treatment. BKM-120 inhibited invasion in vitro in a panel of GBM cell lines, in a range of assays, and this was not associated with cytotoxic effects. Moreover, BKM-120 is also able to inhibit invasion in neural stem cells and astrocytes at different concentrations, confirming the presence of common migratory mechanisms between cancer and normal cell lines. In vivo we were able to show reduced invasion in G9-copGFP and U1242 GBM cells after treatment with BKM-120.
PI3K isoforms have multiple roles in cell migration and, in particular, class IA and class IB subtypes are activated by, interact with, and contribute to activation of integrins and FAK 24,25 . However, all our GBM cell lines express PI3Kα and β isoforms and mostly all the others (Supplementary Fig. 4) and the effects of BKM-120 are likely to be directly on cell migration processes as opposed to invasion/remodelling, as the drug was equally as effective at blocking invasion in the presence and absence of matrix. The PI3K/Akt signaling pathway is activated by loss or mutation of the PTEN tumor suppressor. Previous papers support the PTEN-independence of the effects of BKM-120 on cell proliferation in vitro 10,11 . We also did not observe any correlation between BKM-120 Scientific RepoRts | 6:20189 | DOI: 10.1038/srep20189 The total area of the paxillin-GFP signal and the velocity (microns/sec) of EB1 were measured using ImageJ. Each measurement was made in triplicate. One-way ANOVA was used for statistical analysis (**p < 0.01, ***p < 0.001, ****p < 0.0001). and PTEN mutations (Supplementary Fig. 4). Most importantly in terms of the goal of this study, which was to investigate the potential of BKM-120 as an anti-invasive drug, orthotopic xenografts demonstrated a reduction in tumor spread and were less invasive after BKM-120 treatment. It should be noted that unequivocal monitoring of anti-invasive effects is challenging under conditions where growth is also slightly inhibited, as alterations in tumor metabolism in rapidly growing tumors may promote invasion 26 . In addition, BKM-120 treated tumors clearly show a more defined tumor/brain interface, which strongly supports our in vitro observations. Moreover, in contrast to high BKM-120 dose used for in vivo experiment by previous papers 10,12,27 , our experiments were performed using the mouse equivalent of the current maximum tolerated doses used in patients and a 10-times less concentrated dose. Photographs illustrate tumor growth (low resolution, bar = 1 mm) and the tumor normal brain interface indicating the degree of invasion (high resolution, bar = 20 μ m). One-way ANOVA was used for statistical analysis (**p < 0.01, ***p < 0.001). Clinical trial data on recurrent GBM treated with BKM-120 are not yet published, but invasion was not part of these studies, so is unlikely to have been addressed in patients. Combination studies of BKM-120 may be critical to investigate its anti-invasive potential in human patients. It is of interest that a clinical trial is underway combining BKM-120 with avastin (NCT01349660). In the context of this combination, potential anti-invasive effects of BKM-120 may be addressed.
Interestingly, experiments with the similar PI3K inhibitor GDC-0941 showed that the effects of BKM-120 may be only partially mediated via PI3K inhibition. BKM-120, but not GDC-0941, strongly blocked invasion, and potently induced alterations in cell morphology, microtubule dynamics and eliminated focal adhesions. Further experiments are necessary to better understand direct and off-target effects of BKM-120, and to determine their relative contribution to the anti-migratory effects observed, and whether these are additive or synergistic with PI3K inhibition. Our data seems to indicate a direct effect on microtubule stability as previously observed 12 , given that the general structure of the microtubule network seems to be altered in our cells under conditions that block migration, however further experiments are necessary to elucidate the exact mechanism. EB1 tracking of microtubule plus ends revealed potential effects on microtubule dynamics. This may be related to the enhanced effect of BKM-120 on focal adhesions compared with GDC-0941. Others have established a link between microtubules and focal adhesion regulation, which may be relevant to our observations 24,25,28 . The differential effects of BKM-120 and GDC-0941 on GBM invasion suggest that class I PI3K functions in GBM migration, but also that class I PI3K inhibition alone is not sufficient to fully block cell motility. However, given that both BKM-120 and GDC-0941 are pan-class I inhibitors and we found a partial effect of PI3K on cell migration, further experiments will be necessary to underline the importance of each isoform and test them generating a drugs resistant allele 29 . There is data showing that Src 30 and Akt kinases 6 play a role in glioblastoma invasion. Although Src inhibition blocks migration in our cell line panel (data not shown), we did not see any changes in Src activation in the phospho-ELISA assay ( Supplementary Fig. 3) which we have confirmed by Western blotting (data not shown). The observation that GDC-0941 does not block migration would rule out a role for Akt in our observations on GBM migration.
Overall, these support the important role of the off-target and direct effects of BKM-120 that are not yet fully defined. It is likely that further global profiling studies may uncover additional protein or lipid kinases that could underlie these observations. At present, the nature of the effects which may be involved in mediating some of the effects of BKM-120 is not clear, therefore elucidation of the underlying mechanism is of importance and may lead to the development of improved anti-invasive approaches.
In summary, the CNS-penetrant drug BKM-120 is able to inhibit GBM cell invasion in vitro and in vivo, and its potential anti-invasive properties should be investigated clinically in GBM and potentially other invasive tumor types. Further studies are necessary to completely define additional targets of BKM-120.