BKM120 blocked PI3K-AKT signaling and exhibited cell line-dependent anti-glioma effects
We first investigated the antiproliferative effect of BKM120 using cell viability and colony formation assays across eight GBM cell lines. BKM120 exhibited general growth inhibitory effects in a dose-dependent manner, but limited responsiveness was observed for several cell lines, such as U251, compared with sensitive cell lines like U87 or T98G (Figure 1A and 1B). Next, we selected BKM120 sensitive and insensitive cell lines for further investigation of signaling pathway perturbation. Exposure of U251, U87 and T98G cells to BKM120 resulted in suppression of AKT and S6 phosphorylation in a dose-dependent manner, suggesting that the PI3K-AKT signaling was sufficiently blocked even in the BKM120 insensitive cell line (Figure 1C).
PTEN deletion or mutation is a critical event frequently occurred during GBM development. More than half of the GBM cell lines harbored mutations in the PTEN gene (Additional file 1: Figure S1A), and this could affect the activities of PI3K inhibitor as an anti-cancer agent [19]. We took advantage of the PTEN WT and KO isogenic LN18 cell lines that our lab established previously to investigate the relationship between the PTEN status and the anti-glioma effect of BKM120. As shown in Figure S1B, PTEN loss did not confer sensitivity to BKM120 in LN18 cells. Furthermore, we analyzed the Wooster cell line dataset which comprise over 300 cell lines treated with 19 drugs. Again, whether these cell lines were sensitive or resistant to BEZ235 or GSK1059615, another pan PI3K inhibitor, were independent of the levels of PTEN expression (Figure S1C). We also calculated Spearman’s correlation between PTEN protein expression (data obtain from the cell line reverse-phase protein array in the Cancer Cell Line Encyclopedia project) and IC50 (obtained from the Genomics of Drug Sensitivity in Cancer project). We found that low PTEN expression even confers marked increased resistance instead of sensitivity to BKM120 treatment, for a negative (p < 0.05) correlation between the expression of PTEN and the IC50 of another PI3K inhibitor AS605240 was observed (Figure S1D).
Drug combination screening identified compounds synergizing anti-glioma effect with BKM120
Evaluation of the drug combination effect can be carried out via various methods and models. However, most of them are based on a dose response matrix, e.g., the Highest Single Agent (HSA) and Bliss independence models [20-22], but such approaches are laborious when screening hundreds of drugs. To simplify this process, we established a method for high-throughput drug combination screening in which a Sensitivity Index (SI) score was introduced to quantify the influence of the addition of another drug. SI was defined as the difference between observed combined effect and expected combined effect which is calculated as the product of the relative proliferation ratio of cells treated with BKM120 and a certain library drug (Figure 2A). The expected combined effect is estimated as a minimal additive effect of two drug because it is based on an assumption that cells only received BKM120 and a library drug sequentially and no residual effect of BKM120 could affect the following treatment. This formula was also used for calculation of the influence of siRNA-induced knockdown of gene expression on drug sensitivity [23]. The SI scores ranged from −1 to +1, with positive values indicating BKM120-sensitizing effects.
From the above results, we determined that 1 μM BKM120 was an appropriate concentration for the combination drug screen, and the U251 cell line was chosen for its insensitivity to BKM120. We took advantage of a drug library containing 606 highly selective small molecule inhibitors targeting multiple cancer-related signaling pathways (Additional file 2: Figure S2). To calculate the SI for individual drugs in the drug library, U251 cells received the following treatments: vehicle (DMSO), 1 μM BKM120, 10 μM library drug and a combination of 1 μM BKM120 and 10 μM library drug (Figure 2B).
The first step resulted in the identification of 110/606 (1.8%) inhibitors with SI ≥ 0.12, a threshold at which potential BKM120-sensitizing effects were expected. In the second step, the antiproliferative activity of these inhibitors was then re-assessed at 10 μM on U251 ± 1 μM BKM120 to confirm the candidates identified in the first step. Finally, the 6 top ranked compounds displaying relatively strong synergistic effects with BKM120 in the second step were chosen for further assessment of their antiproliferative effects at 10 μM and 2 μM in the presence of 1 μM BKM120 across 8 GBM cell lines (Figure 2B).
Validation of the synergism between BKM120 and the candidate agents
For the six top candidates selected above, SI were re-assessed to ensure their abilities to sensitize cells to the antiproliferative effect of BKM120. All candidate compounds had a BKM120-sensitizing effect (SI>0), and five of them exhibited a strong sensitizing effect (SI>1.0) (Figure 3A). Next, we assessed the antiproliferative effect of the combination treatment of eight GBM cell lines. Although differentiated responses were observed across different cell lines and across different treated concentrations, the mean SI of all compounds were all above 0, suggesting an overall BKM120-sensitizing effect of these candidates. A brief summary of the top six candidates were given (Figure 3B). Finally, the top four candidate compounds, TH588 (an MTH1 inhibitor), CGP57380 (an MNK1 inhibitor), GW9662 (a PPAR inhibitor) and 4-hydroxytamoxifen (an estrogen receptor inhibitor), were chosen for further evaluation. As shown in colony formation assays, these compounds almost eliminated all colonies of U251 in the presence of BKM120, while a single drug or BKM120 failed to show such a strong effect (Figure 3C). Among them, ABT263, CGP57380 and 4-hydroxytamoxifen have already been reported for their synergistic effects in combination with BKM120 [24-26]. Altogether, this evidence supports that our proof-of-principle screening method is useful to identify potential drug combinations that have synergistic interactions with BKM120.
Characterization of the anti-glioma effect of the combination treatment of BKM120 and TH588
Because TH588 was the top candidate and has not yet been reported elsewhere for its combined effect with BKM120, it was chosen for further investigation. U251 and SNB19 cells were treated with a 5-by-5 dose titration matrix of TH588 and BKM120, and the viability of the cells was determined. As shown in Figure 4A and 4D, the expected versus observed dose-response surface plot of the 5-by-5 matrix viability data demonstrated synergistic effect of two agents against U251 and SNB19 GBM cells. In addition, the combination treatment led to marked reduced viability of GBM cells. Moreover, most dots (each dot represents a combination of a specific concentration of two agents) had a CI value below 1.0 in the Fa-CI plots, suggesting the interaction between TH588 and BKM120 were synergistic (Figure 4B and 4E). We also applied classical synergy models including HSA and Bliss [20-22] to determine the HSA and Bliss values, which are a readout for synergistic inhibition and depict the difference between expected inhibition and observed inhibition. Most of the HSA and Bliss values were above 0, indicating synergistic interactions between the two drugs (Figure 4C and 4F). Furthermore, the combination of TH588 and BKM120 also severely suppressed colony formation in eight GBM cell lines (Figure 4G) and 3D spheroid formation in SNB19 and LN229 cells (Figure 4H-K), confirming that the effect of the combination of the two drugs exceeds these of each single agent in suppressing the growth and proliferation of GBM cells in different models.
The combination of BKM120 and TH588 synergistically induces oxidative DNA damage and apoptosis
TH588 has been reported to cause DNA damage by increasing 8-oxodG or 2-OH-dA incorporation into DNA, leading to DNA base mispairing, mutation and cell death [27]. Interestingly, previous works suggest that inhibition of PI3K interferes with DNA synthesis or repair through nucleoside depletion [28] so there might be a link between BKM120 and TH588. To reveal the synergistic mechanism, we used a comet assay, a sensitive method for detecting 8-oxodG or 2-OH-dA at the single-cell level to evaluate the DNA damage caused by the combination treatment [29]. The significant elongated tailing ratio of combination treatment relative to control or single agent treatment suggests that more abundant DNA-damaged fragments and accumulation of 8-oxodG in DNA were presented after treatment of both MTH1 and PI3K inhibitor (Figure 5A and 5B).
The nucleosomal histone protein H2AX is rapidly phosphorylated at serine 139 (γ-H2AX) at the double strand break (DSB) site [30], so γ-H2AX foci formation is usually used to quantify DSBs. Therefore, we reasoned that the PI3K inhibitor also enhanced the DSB damage of DNA caused by TH588. To confirm that, GBM cells treated with BKM120 and/or TH588 was stained with γ-H2AX and analyzed by flow cytometry. As shown in Figure 5C~5D and Additional file 3: Figure S3, the γ-H2AX-positive fraction of U251 cells increased significantly upon treatment with BKM120 and TH588 compared with that in cells that received single-agent treatment, suggesting that the combination of PI3K and MTH1 inhibitors dramatically elevates DSB levels in GBM or lung cancer cells. Accumulating evidence suggests that PI3K pathway is also involved in DNA replication and genome stability [28]. Therefore, PI3K might be an attractive target for combining with targeted agent related to DNA repairing pathway [31].
Moreover, comparing with the independent treatment of BKM120 and TH588, the combination treatment significantly increased the Annexin V-positive apoptotic fraction of cells (Figure 5E-G and Additional file 5: Figure S5), and also caused marked elevation of cleaved caspase3.
On top of the MTH1 inhibition, alternative mechanisms of TH588 have been reported as its anti-cancer effects including tubulin depolymerization and AKT signaling downregulation [32, 33]. To confirm it, we investigated the spindle morphology of mitotic cells upon treatment of 5 and 10 µM of TH588. Indeed, TH588 caused disruption of spindle formation and interrupted the separations of duplicated centrosomes (Additional file 3: Figure S3). In addition, treatment of TH588 also caused reduction of phosphor Akt (S473) and its downstream component 4EBP1 (S65) in U251 and U87 cells (Additional file 3: Figure S3). Together these evidence provide additional information for TH588 mediated GBM cell death besides MTH1 inhibition.
The PI3K/AKT signaling is a determinant of the responsiveness to MTH1 inhibition
To assure that the antiproliferative activity of TH588 is due to its selective targeting of MTH1 but not an off-target effect, stable MTH1 knockdown U251 together with control cells transduced with only empty vector control were established. The expression of MTH1 in two MTH1 knockdown cell lines was validated by Western blotting (Figure 6A). As shown in Figure 6B, MTH1-silenced U251 cells were generally sensitive to BKM120, and the IC50 of two MTH1-silenced cells (2.59 and 1.63) were lower than that of control cells (2.99).
To further investigate whether the TH588 synergizing effect with PI3K inhibitor only limit to BKM120, we also assessed the combined anti-glioma effect of TH588 with other PI3K and AKT inhibitors. For GDC-0941, another pan-PI3K inhibitor, it effectively suppressed the proliferation of GBM cells with TH588, and the Combination Index (CI) values were lower than 1 at most concentrations, suggesting the interaction between two agents were synergistic (Figure 6C). For GDC-0068 and MK-2206, two highly selective AKT inhibitors, most CI values were also below 1 when they were combined with TH588 (Figure 6D and 6E). The above evidence suggested that both PI3K and AKT inhibitors enhanced the antiproliferation effect of MTH1 blockade or knockdown, indicating that the PI3K-AKT pathway activation may be an important determinant for the anti-glioma efficacy of MTH1 inhibition.
To find more evidence supporting this finding, we also checked the levels of PI3K-AKT signaling activation in eight GBM cell lines and categorized them into low, medium or high groups according to the expression levels of p-AKT (Figure 6F). Next, we performed a proliferation assay by treating these GBM cell lines with TH588 and analyzing the corresponding IC50. Although the number of cell lines used was limited, GBM cells with highly activated PI3K-AKT signaling tended to be more resistant to TH588 (IC50 values for the p-AKT low, medium and high groups were >14.56, 9.19 and 8.98 µM, respectively, Figure 6G). Based on this observation, we established LN229 isogenic cell lines with PTEN wild type, PTEN knockout, and PIK3CA carrying constitutively activated mutations (p.N345K and p.E542K) [34] and investigated their responses to MTH1 inhibition. It is known that loss-of-function mutations of PTEN and gain-of-function mutations of PIK3CA can both activate the PI3K-AKT pathway. Indeed, the levels of p-AKT and its downstream component p-S6 were elevated in all three mutant cell lines compared with LN229 parental cells (Figure 6H). In line with the above results, both PTEN knockout and PIK3CA mutant cells were less responsive to TH588 than their parental cell lines (Figure 6I). Finally, we transfected a plasmid carrying the intact AKT gene or an empty vector into HeLa cells, and it led to elevation of p-AKT, p-S6 and p-4EBP1 expression in AKT-overexpressing cells relative to empty vector-transfected cells (Figure 6J). Similarly, AKT-overexpressing cells showed significantly elevated TH588 tolerance compared with that in empty vector-transfected control cells (Figure 6K). Altogether, we found that the anti-glioma effect of TH588 was determined by the activation of the PI3K-AKT signaling pathway, which may provide an explanation for the synergistic interaction between PI3K and MTH1 inhibitors.