Cisplatin USP (Teva) was supplied as a 1 g/L solution in saline, and pemetrexed (Sigma) was dissolved in water immediately prior to use. BGB324 (Selleck Chemicals) was dissolved in DMSO, and diluted in water immediately prior to use. Hydrogen peroxide was diluted in water immediately prior to use from a 30 wt. % stock solution (Sigma). Solubility of all reagents were visually verified in experimental systems prior to use.

Data generated by the TCGA Research Network was used to compare RNA expression and ranking among datasets. Values and graphs for comparisons were obtained from cBioPortal (Cerami et al., 2012; Gao et al., 2013).

Protein lysates from MPM cells lines were subjected to a reverse phase protein array. This array was performed at the MD Anderson Functional Proteomics RPPA Core Facility. MPM cells lines used were SPC111, VMC14, VMC31, M38K, VMC40, Meso53, SPC212, and P31 (Marklund et al., 1982; Pelin-Enlund et al., 1990; Schmitter et al., 1992; Janson et al., 2008; Garay et al., 2013; Schelch et al., 2014; Hoda et al., 2016). All cell lines were maintained in Dulbecco Modified Eagle Medium (GenClone) with 10% fetal bovine serum (Sigma). Lysates were added to nitrocellulose-coated slides in five serial dilutions, and subjected to 301 antibodies. Samples probed with antibodies were analyzed by tyramide-based signal amplification approach, and relative protein levels were determined after protein loading normalization. These levels were fitted and analyzed by the SuperCurve GUI method (Hu et al., 2007). The heat map was generated in Cluster 3.0. Nonmalignant mesothelium Met-5a cells (ATCC) were used for comparing basal AXL expression to VMC40 and P31 cells, and in later cell death experiments.

In six-well plates, 3.5 × 10⁵ cells/well incubated at 37°C for overnight attachment prior to any treatment. Proteins were extracted from cells in 2x Laemmli buffer (Bio-Rad) containing protease and phosphatase inhibitors. Except as noted, equal sample volumes were applied to SDS-PAGE gels and transferred to PVDF membranes. Membranes were incubated in Odyssey Blocking Buffer (Li-Cor) overnight at 4°C prior to probing with primary antibodies overnight at 4°C. IRDye rabbit, mouse, and goat secondary antibodies (Li-Cor) were applied to probed membranes for 45 m at room temperature, and visualized using a Li-Cor Odyssey scanning system. Primary antibodies used for standard immunoblot analysis included AXL (R&D Systems), pAktS473 and pan-Akt (Cell Signaling Technology), and beta-actin (Sigma). Based on minimal detection by standard methods, pAXLY702 (Cell Signaling Technology), GAS6 (GeneTex), and beta-actin (as system control) antibodies were used to detect corresponding protein levels by Wes capillary electrophoresis (ProteinSimple) based on the manufacturer's protocol. Chemiluminescent signal is depicted as traditional immunoblot protein bands, which are generated by Compass software (ProteinSimple), as done previously for RTKs (Furugaki et al., 2016).

In six-well plates, 3.5 × 10⁵ cells/well incubated at 37°C for overnight attachment. Medium was removed from each well, and 5 μM chloromethyl-dihydrogen-2′,7′-dichlorofluorescein diacetate (CM-H₂DCFDA) was added to each well for 20 m. After incubation, cells were lifted with trypsin (Sigma) and transferred to centrifuge tubes. Tubes were centrifuged, supernatant aspirated, and fresh medium premixed with indicated reagent(s) was added. Cells were incubated for 30 m at 37°C with gentle shaking. Prior to analysis, propidium iodide was added at a final concentration of 1.3 μg/mL to reduce dead cells during analysis through gating. Fluorescence in cells was detected with a LSR II flow cytometer (BD Biosciences) using FITC filters.

Caspase 3 activity was measured using the DEVD-AFC (Asp-Glu-Val-Asp, 7-amino-4-trifluoromethylcoumarin) fluorescent assay as we have previously described (Bastola et al., 2016). Briefly, cells were prepared as done for immunoblots and lysed using CHAPS lysis buffer (20 mM PIPES, 100 mM NaCl, 1 mM EDTA, 0.1% (w/v) CHAPS, 10% sucrose, 10 mM DTT pH 7.2). Protein concentration was determined using the Pierce BCA Assay (Thermo Fisher). Twenty milligrams of protein was combined with 2 μL of 2 mM DEVD-AFC (Millipore) in 96-well flat-bottom plates. Two hundred microliters/well lysis buffer was added. The plate was covered and incubated at 37°C. After 2 h, fluorescence measurements were taken at Ex 400 nm and Em 510 nm using a Synergy4 plate reader (BioTek).

VMC40 and P31 cells with AXL shRNA were generated using human GIPZ lentiviral shRNAmir individual clones (Dharmacon) according to the manufacturer's protocol. Transfected cells were selected in puromycin for 2 weeks prior to use. Knockdown was determined by immunoblots using methods described above using three different gel loading volumes for estimation of shRNA efficacy compared to nontargeting control (NTC). Cell proliferation was determined by the sulforhodamine B (SRB) assay by plating equal cells, 5,000 cells/well in 96-well plates, and incubating for 4 days. Quantification with SRB stain was done as described below, based on our previous methods (Bastola et al., 2016).

Short-term SRB survival assays and synergy calculations were performed as previously described (Bastola et al., 2016). Briefly, 5,000 cells were plated in 96-well plates and attached by overnight incubation. The following day, cells were treated with different concentrations of BGB324 and cisplatin or pemetrexed, and in combinations as indicated. After 72 h, cells were fixed with 10% trichloroacetic acid (Sigma). Fixed cells were washed with tap water and stained using SRB (Sigma) at room temperature for 45 m. Cells were washed with 1% acetic acid solution, and then were dissolved in 10 mM Tris, pH 10, and fluorescence measurements were taken at Ex 488 nm and Em 585 nm using a Synergy4 plate reader. Where specified, the concentration that causes 50% growth inhibition (GI₅₀) was estimated using Prism software (GraphPad). Drug synergy was determined by calculating the combination indices (CIs) obtained from the fluorescence measurements. The CI values were calculated based on dividing the combination expected effect by the observed effect: CI = (D₁ +D₂ ^*(1-D₁))/D_observed, where D represents the cell death from drug 1, drug 2, and the combination (observed) (Bastola et al., 2016). The mean CI values were determined from multiple drug combinations in duplicates that produced 20–80% effect from independent experiments.

Long-term survival assays were performed as previously described (Bastola et al., 2016). Briefly, 500 cells/well were plated in 6-well plates and cells were attached by overnight incubation. Regents were added to cells and incubated for 72 h. Following treatment, medium was gently aspirated and replaced with regular growth medium. Cells were then incubated for 4–7 additional days. Once the colonies were optimal for visualization, cells were fixed with 10% trichloroacetic acid. Fixed cells were washed with tap water and stained using SRB at room temperature for 45 m. Cells were then washed with 1% acetic acid solution. The plates were air-dried, and pictures were taken using the Bio-Rad Imager System. For quantification, stained cells were dissolved in 10 mM Tris, pH 10 and the solution was transferred to 96-well plates for fluorescence measurement.

Densitometry values from immunoblots were determined by Image Studio software (Li-Cor). Dose-response data and non-linear regression curves were generated using Prism software. Bar charts were generated via Excel (Microsoft), and t-test statistical analyses used with a significance threshold of p < 0.05.

To uncover RTKs that may be promoting MPM cell survival, the expression levels of common RTKs were ranked based on the expression among the 30 TCGA datasets accessed through the cBioPortal bioinformatics web tool (Cerami et al., 2012; Gao et al., 2013). AXL expression was the fourth highest in MPM samples among other cancer types, with a mean log₂ value of 11.40 (Figure 1A and Supplementary Figure 1). This expression value for AXL was higher than IGF1R (11.04), EGFR (10.14), and the other “TAM” kinases Tyro3 and Mer. The VEGFR1 kinase expression was higher than AXL in MPM, however, the VEGF ligand (VEGFA gene) was not expressed in MPM as high as most other cancers in the dataset (ranking #20 of 30). The AXL ligand, GAS6, had the third highest expression in MPM, versus the other cancer datasets, with a mean log₂ value of 13.06 (Figure 1A and Supplementary Figure 2). Moreover, AXL was demonstrated in prior studies to regulate MPM proliferation (Ou et al., 2011), and ROS was shown to activate AXL and increase GAS6 binding (Konishi et al., 2004). Based on literature, and AXL and GAS6 expression in MPM, the AXL kinase was selected for further study.

Eight MPM cell lines were used in a reverse phase protein array to detect differences in protein expression including AXL (Figure 1B and Supplementary Figure 3). Of these cell lines, AXL expression was highest in VMC40 cells (normalized log₂ median-centered value of 1.35), and P31 cells had AXL expression near the median (−0.07). The immunoblot analysis confirmed that AXL expression and phosphorylation were higher in VMC40 than the nonmalignant mesothelial cell line Met-5a (Figure 1C). In addition, GAS6 expression was also higher in VMC40 than in Met-5a cells. These results suggest autocrine AXL pathway is active in MPM cells.

Cisplatin and pemetrexed each have been shown previously to induce elevated ROS in cells (Berndtsson et al., 2007; Hwang et al., 2015), but have not been evaluated for cellular ROS increases as a combination. Cisplatin can induce detectable ROS increases in melanoma cells at concentrations of 10 and 20 μM, measured at 3 h (Berndtsson et al., 2007). However, at MPM clinical doses (e.g., 75 mg/m²) plasma concentrations of cisplatin is below 20 μM (Andersson et al., 1996). In contrast, MPM pemetrexed doses of 500 mg/m² can achieve C_max values near 250 μM (Sweeney et al., 2006). Moreover, pemetrexed could induce ROS at a concentration of 2 μM in MPM cells, measured at 24 h (Hwang et al., 2015). In both VMC40 and P31 cells, higher concentrations of cisplatin (7.5 and 15 μM), multiple concentrations of pemetrexed (2, 4, and 8 μM), and multiple concentrations of combined cisplatin and pemetrexed (3.8/2, 7.5/4, and 15/8 μM, respectively) induced significant increases in ROS after 30 m versus control cells when using the CM-H₂DCFDA indicator (Figure 2). The combination of 15 μM cisplatin and 8 μM pemetrexed produces significantly higher ROS than cisplatin alone (p = 0.049 and 0.047 for VMC40 and P31, respectively), but not significantly different from pemetrexed alone.

Prior studies have shown AXL can be phosphorylated by hydrogen peroxide (Konishi et al., 2004), but it has not been evaluated if chemotherapy agents can activate AXL. Based on our observation that chemotherapy agents increase ROS levels (Figure 2), VMC40 and P31 MPM cells were treated with 15 μM cisplatin and 8 μM pemetrexed for short (minutes) and relatively long (6 h) time points. Increases in AXL phosphorylation were detected after 10 m of exposure to cisplatin and pemetrexed (Figures 3A,B, Supplementary Figure 4A). Consistent with prior studies indicating that AXL inhibitor BGB324 attenuates ROS-induced phosphorylation of AXL (Smolock and Korshunov, 2010), AXL phosphorylation was blocked when MPM cells were co-treated with BGB324. Again, consistent with prior studies indicating that pro-survival kinase Akt1 was activated downstream of AXL (Huang et al., 2013), both cisplatin and pemetrexed induce Akt phosphorylation of S473 at variable time points. Akt phosphorylation was blocked by BGB324. The pAkt levels did not change from adding BGB324 inhibitor alone compared to no treatment (Supplementary Figure 4B), suggesting that BGB324 does not inhibit Akt directly. In parallel, cells treated with 150 μM hydrogen peroxide show AXL and Akt phosphorylation, and this phosphorylation was blocked by BGB324 (Figure 3C). Cells co-treated with 4 mM N-acetylcysteine antioxidant did not show an increase in AXL and Akt phosphorylation (Figure 3D). Collectively, these results suggest cisplatin and pemetrexed induce ROS which then activate AXL and downstream Akt. Finally, this signaling cascade can be blocked by BGB324 or N-acetylcysteine.

Based on the phosphorylation of Akt, we next determined changes in downstream caspase-3 activity with and without BGB324 after 24 h of chemotherapy exposure (Figure 3E). The presence of BGB324 significantly increased the caspase-3 activity in cells treated with cisplatin and pemetrexed. Overall, these results indicate the possibility that BGB324 could enhance the cytotoxicity of these chemotherapy agents in VMC40 and P31.

Based on the results that cisplatin and pemetrexed enhance pAXL and pAkt (Figure 3), we hypothesized AXL inhibition may enhance sensitivity to these chemotherapy agents. AXL was knocked down via AXL shRNA in both VMC40 and P31 cells (Figure 4A). AXL knockdown efficiency was estimated by immunoblots using loading volumes of 5, 15, and 30 μL. Detected AXL of VMC40 cells averaged 37.7% for shRNA#1 and 36.1% for shRNA#2 compared to NTC controls, and of P31 cells averaged 33.0% for shRNA#1 and 28.4% for shRNA#2. The AXL shRNA decreased cell proliferation (Figure 4B). Using the SRB assay and comparing to NTC, cells with AXL shRNA were more sensitive to cell death at increasing concentrations of cisplatin and pemetrexed (Figure 4C, Supplementary Figure 5). To analyze the drug response following AXL knockdown, drug response was compared at each concentration respective to the absence of drug. A significant decrease in cell survival was observed in AXL knockdown cells compared to NTC at 1.88 and 15 μM for cisplatin in VMC40 cells, and at 1 and 8 μM for pemetrexed in both VMC40 and P31 cells (Supplementary Figure 5). Consequently, we verified these results using secondary long-term clonogenic assays where the drug is washed out after 72 h (Figure 4D and Supplementary Figure 6). These data suggest AXL may be targeted to enhance sensitivity to cisplatin and pemetrexed.

Building upon the observation that AXL knockdown enhances sensitivity to cisplatin and pemetrexed, we next evaluated if inhibiting AXL phosphorylation would result in enhanced sensitivity to cisplatin and pemetrexed. The selective AXL inhibitor, BGB324, has been studied as a monotherapy and with other chemotherapeutic agents (Holland et al., 2010; Wnuk-Lipinska et al., 2014; Wang et al., 2016). We used the SRB assay to determine the extent of cytotoxicity induced by BGB324 in the presence of incrementing concentrations of cisplatin and pemetrexed (Figure 5A and Supplementary Figure 7). In these SRB assays, combined effects of BGB324 with cisplatin and pemetrexed were mostly additive with CI values of approximately 1 (Figure 5B). CI values <1 indicate synergistic effects, while values >1 indicate antagonistic effects. In contrast, the addition of BGB324 did not have a significant effect on GI₅₀ values with incrementing concentrations of cisplatin and pemetrexed in nonmalignant Met-5a cells (Figure 5C). To further validate results from VMC40 and P31 SRB assays, we performed clonogenic assays in these cell lines (Figures 5D,E). Both cell lines showed increased cell death at relatively low concentrations compared to SRB assays, suggesting a fraction of cells surviving in SRB assays could be in senescence. Moreover, mean CIs from the clonogenic assay were 0.95 for 1.4 μM cisplatin with BGB324 and 0.91 for 0.1 μM pemetrexed with BGB324 in VMC40 cells, which suggests marginal synergistic cytotoxicity. For P31 cells, the mean CIs for concentrations shown were 0.93 for cisplatin with BGB324 and 1.02 for pemetrexed with BGB324.

The mainly additive drug-drug interaction between BGB324 and cisplatin or pemetrexed raises the question if the sequence of drug exposure must be optimized. We performed essentially the same assays in Figure 5, except the BGB324 inhibitor was added 4 h prior to any addition of cisplatin and pemetrexed (Figure 6). In SRB assays, cell death was enhanced when pre-treatment of BGB324 was added to incrementing concentrations of cisplatin and pemetrexed (Figure 6A and Supplementary Figure 8). Calculated CI means for these combinations suggest more cell death synergy for both VMC40 and P31 cells (Figure 6B), compared to when agents were added to cells at the same time (Figure 5B). In Met-5a cells, adding BGB324 as a pretreatment did not significantly alter the GI₅₀ values for cisplatin and pemetrexed. The SRB results for VMC40 and P31 cells were validated using clonogenic assays (Figures 6D,E). The mean CIs from the clonogenic assay were 0.91 for 1.4 μM cisplatin with BGB324 and 0.74 for 0.1 μM pemetrexed with BGB324 in VMC40 cells, suggesting synergistic cytotoxicity. For P31 cells, the mean CIs at concentrations shown were 0.88 for cisplatin with pretreated BGB324 and 0.94 for pemetrexed with pretreated BGB324.

The combination of cisplatin and pemetrexed are the current chemotherapy for MPM (Benedetti et al., 2015; Yap et al., 2017). As a proof-of-concept, we performed clonogenic assays using this combination (at half the concentrations used previously in Figures 5D,E, 6D,E) with BGB324 (Figure 7A). Also, we performed the essentially the same assay with BGB324 being added 4 h prior to chemotherapy agents (Figure 7B). These results were quantified by solubilizing and measuring the stain (Figure 7C). For VMC40 cells, the mean CIs were 0.91 for co-treatment with BGB324 and 0.74 for pretreated with BGB324, indicating synergistic cytotoxic effects. For P31 cells, the mean CIs were 1.04 for co-treatment with BGB324 and 0.92 for pretreatment with BGB324.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.