Kinome-wide siRNA screening identifies molecular targets mediating the sensitivity of pancreatic cancer cells to Aurora kinase inhibitors

Chemicals and Reagents

VX-680 (MK-0457), sorafenib, and imatinib were purchased from ChemieTek, LLC (Indianapolis, IN). ZM447439 was purchased from Tocris Bioscience (Ellisville, MI). Aurora kinase inhibitor-1 (AKI-1) and MP235 were synthesized in our lab (16, 31). PHA-739358 was purchased from Selleck Chemicals (Houston, TX). Etopside was purchased from Sigma-Aldrich (St. Louis, MO). The chemical structures of the Aurora kinase inhibitors used in this study are shown in Supplementary Figure S1.

The Human Validated Kinase siRNA Set (HVKS) V2 was purchased from Qiagen (Valencia, CA). This siRNA library contains two validated siRNA oligonucleotides for each of 588 kinase and kinase related genes (a total of 1,176 siRNA oligonucleotides; see www.qiagen.com). Additional siRNA oligonucleotides targeting individual genes or negative (non-targeting) siRNA oligonucleotides were also purchased from Qiagen. The siRNA oligonucleotides were dissolved in RNase free water at 10μM stock concentration and stored at −80°C until use.

Cell culture

BxPC-3, Mia PaCa-2, AsPC-1, CFPAC-1, PANC-1 and SU 86.86 pancreatic cancer cell lines were purchased from American Type Tissue Culture Collection (ATCC, Manassas, VA) and cultured in RPMI 1640 supplemented with 10% (v/v) fetal bovine serum, 100 units/ml penicillin, and 100μg/ml streptomycin (Invitrogen, Carlsbad, CA).

Cell line identities were verified by STR profiling (32) using the AmpFISTR Identifiler PCR amplification kit (Applied Biosystems, Foster City, CA). This method simultaneously amplifies 15 STR loci and Amelogenin in a single tube, using 5 dyes, 6-FAM™, JOE™, NED™, PET™, and LIZ™ which are then separated on a 3100 Genetic Analyzer (Applied Biosystems). GeneMapper ID v3.2. Software was used for analysis (Applied Biosystems). AmpFISTR control DNA and the AmpFISTR allelic ladder were run concurrently. Results were compared to published STR sequences from the ATCC. The STR profiling is repeated once a cell line has been passaged more than 6 months after previous STR profiling.

Optimization of transfection conditions for HT-siRNA screen

To find the most optimal transfection reagent and conditions for pancreatic cancer cells, we first tested a panel of transfection reagents with two siRNA oligonucleotides, a non-silencing negative control siRNA (Qiagen) and a positive control siRNA (UBB1, Qiagen) (33) in a panel of pancreatic cancer cell lines, including AsPC-1, BxPC-3, CFPAC-1, Mia PaCA-2, PANC-1, and SU86.86. The panel of transfection reagents includes Lipofectamine 2000 (Invitrogen), Lipofectamine RANiMax (Invitrogen), siLentFect (Bio-Rad, Hercules, CA), Oligofectamine (Invitrogen). The siRNA (9 ng in 2μl) was first printed onto solid white 384-well plates using a Biomek FX liquid handling system (Beckman Coulter, Fullerton, CA). The transfection reagents were diluted in OptiMEM (Invitrogen) at five different ratios (1:2, 1:3, 1:5, 1:7, and 1:8) from 200 nl/well. The final volumes of the transfection reagents tested were therefore 100, 66.7, 40, 28.6, and 25 nl/well. Diluted transfection reagents (20μl) were added to the 384-well plates containing siRNA oligonucleotides and were allowed to complex for 20 minutes. Equal volume of cells was added in growth media resulting in 1,000-1,200 cells per well depending on growth characteristics of the cell lines (determined in separate experiments). The cells were then incubated in a CO₂ incubator at 37°C for 96 hours at which point 25μl of CellTiter-Glo® reagent (Promega, Madison, WI) was added to each well to determine cell viability. The luminescence intensities were obtained for each plate using an Analyst GT microplate reader (Molecular Devices, Sunnyvale, CA). Percent viability values were calculated by comparing the intensity units from each treatment condition with that of the untreated controls. The transfection reagent and conditions that give the highest difference in cell viability between the Non-silencing siRNA (negative control purchased from Qiagen) and the lethal siRNA (positive control) were then chosen for the subsequent HT RNAi screening in combination with AKIs.

Selection of cell lines and AKIs for HT-siRNA screening

To select a cell line and an AKI that would maximize our chances of finding siRNA hits that are specific to Aurora kinase inhibition, we first evaluated three different AKIs in a panel of pancreatic cancer cells, including AsPC-1, BxPC-3, CFPAC-1, Mia PaCa-2, PANC-1, and SU86.86, using the same growth and assay conditions as those for the siRNA transfection. The three AKIs were VX-680 (34), MP235(35, 36), and AKI-1 (30). All three AKIs have been shown to inhibit Aurora kinases in cell-free assays with nM IC₅₀s and induce phenotypes in cancer cells that are consistent with the inhibition of Aurora kinases (30, 34, 36). The cells were treated with varying doses of AKIs. Cell viability was determined 96 hours after adding the drug using CellTiter-Glo® Assay. The cell line that provided the most consistent dose response results with a modest sensitivity crossing all the AKIs tested was selected as the screening cell line (BxPC3). The AKI showing relatively smooth dose response curves with modest activity crossing all cell lines tested would be selected as the screening compound (AKI-1).

High-throughput RNAi screening

The siRNA library was printed onto 384-well cell culture plates (Corning, Lowell, MA) at 2μl/well (the final concentration of siRNA in the screening assay was 16nM). A set of control siRNA oligonucleotides including GFP siRNA, All Star Negative Control siRNA, Non-silencing siRNA, and the UBB1 positive control siRNA (all purchased from Qiagen) were also included. Each of library siRNA sequences was printed in duplicates and the control siRNAs were printed in quadruplicates. A set of siRNA buffer only wells were also printed for inclusion of negative controls such as buffer and transfection reagent only controls. For each compound concentration one set of the plates printed with siRNA library were used. The assay procedure is shown in Supplementary Figure S2. Briefly, on Day 1, 20μl of SilentFect diluted in serum free medium was added onto the pre-printed siRNA library plates and incubated for 30 minutes at room temperature. 20μl of BxPC-3 cells (1,200 cells/well) were then added to each well of the plates. Following an overnight (16 hours) incubation in a CO₂ incubator, 5μl of the Aurora kinase inhibitor, AKI-1, at appropriate concentrations was added into each well of the plates. To ensure the quality of the positive hits to be identified, we designed a screening scheme with five different AKI-1 drug concentrations, EC₁₀, EC₂₀, EC₃₀, EC₅₀ (concentrations required to achieve 10%, 20%, 30% and 50% of the maximal growth inhibitory effect, respectively), as well as a vehicle control, all of which were calculated based on the non-regression curve fitting equations of the dose-response curves using the Prism 5 software (GraphPad Inc., La Jolla, CA) in the screening cell line. The zero, EC₁₀ and EC₃₀ concentration sets were performed in duplicate to ensure the screening quality (this makes a total of nine screening sets of the siRNA library). After adding drugs, the cells were further incubated for 96 hours in a CO₂ incubator. On Day 6, the viability of cells in each well was measured by the CellTiter-Glo® Luminescent Cell Viability Assay (Promega) as per manufacturer’s instructions.

Hit selection

Hit selection was based on the detection of changes between two drug-dose response curves (DDRC) generated from individual siRNA and the negative siRNA control. In general, the following steps were involved: data normalization, filtration of toxic siRNA and outliers, IC₅₀ calculation, and ranking. The data normalization was done by fixing the DDRC of plate median and shifting the sample DDRC so that both curves have identical origins at drug dose 0 (siRNA only control). Such normalization allowed us to calculate the size of enclosed area formed by two DDRCs (sample and plate median). Toxic siRNA oligonucleotides (percentage cell survival for siRNA only control is less than 50%) were removed from further analysis. EC₅₀s and EC₃₀s (drug concentration required to achieve 50% and 30% of the maximal cell growth inhibitory effect) for plate median and each siRNA were calculated by fitting the data to a sigmoid dose-response model using nonlinear regression with the Matlab software (2007a, The MathWorks Inc). The EC₃₀ and EC₅₀ shift between sample DDRC and the DDRC of plate median was then used to rank the siRNA. For the validation experiments, since more potential sensitizer hits were tested, we used a negative siRNA control (Non-silencing siRNA from Qiagen) as a reference instead of plate median in data normalization.

Confirmation screening

From primary screening, we identified kinase genes targeted by siRNA that mediate sensitivity of AKI-1 in the BxPC-3 cell line. To exclude the possibility of siRNA with biological off-target effects, we performed a confirmation screen using four siRNA sequences per gene in combination with AKI-1 in the BxPC-3 cell line and defined confirmed hits as those kinases whose inhibition was synthetically lethal with AKIs in pancreatic cancer cells with concordant results from two or more unique siRNAs.

Drug combination treatment

Cells were seeded at 2,000 cells/well in 96-well plates and allowed to grow overnight. On the second day, a serial dilution of the Aurora kinase inhibitors (ZM447439 or PHA-739358) combined with fixed concentrations of the second drug (imatinib or sarofenib) as indicated in the figures was added to cells and incubated for 96 hours. At the end of drug incubation, cell viability was determined using the SRB (Sulforhodamine B) assay.

Sulforhodamine B (SRB) assay

After drug treatment, culture media were removed from the 96-well plate and the cells were fixed by adding 65 μl of 10% trichloroacetic acid (TCA) solutions and incubating for 30 minutes at 4°C. Cells were then rinsed five times with deionized water and stained with 0.04% SRB solution (40 μl/well) for 30 minutes at room temperature. Cells were then washed five times with 1% acetic acid to remove unbound dye, and left to air dry. The bound SRB dye was then solubilized by adding 50 mM Tris-base solution (100 μl/well), and plates were incubated at room temperature for 40 minutes with shaking. Plates were finally read at OD 564 nm using a BioTek plate reader (BioTek, Winooski, VT). Cell viability was calculated by dividing the average of the reading number for the drug treated wells by the average of the reading number for vehicle treated wells. The IC₅₀ values (concentration required to achieve 50% growth inhibition) were determined using the Prism 5 software (GraphPad Software).

Cell cycle analysis using flow cytometry

Cells were seeded in T-25 tissue culture flasks (6×10⁵ cells per flask) (Corning, Lowell, MA) and grown overnight before drug treatment. For cell cycle analysis, AsPC-1 cells were treated with PHA-739358 (3μM), imatinib (15μM), or PHA-739358 (3μM) plus imatinib (15μM) for 24, 48, and 72 hours. The drug-treated cells and untreated control samples were harvested by trypsinization and stained with propidium iodide (Sigma-Aldrich, St. Louis, Mo) in a modified Krishan buffer for 1 hour at 4°C. The propidium iodide-stained samples were then analyzed with a FACSCalibur Flow Cytometer (BD Immunocytometry Systems, San Jose, CA). Histograms were analyzed for cell cycle compartments, and the percentage of cells at each phase of the cell cycle was calculated using CellQuest Pro Software (BD Immunocytometry Systems).

Caspase 3/7 activity based apoptosis assay

Cells were seeded in 6-well plates (0.3 × 10⁶ cells/well) and incubated for 24 hours at 37°C to allow attachment. Then cells were treated with various concentrations of drugs as indicated in the figure legends. Culture media were collected at 72 hours after drug treatment. After washing with Phosphate Buffer Saline (PBS) solution, the cells were detached by trypsinization and combined with the culture media for each sample. The cell suspension was pelleted by centrifugation at 1,000 rpm for 5 min. 200 μl of NP40 lysis buffer (10 mM Tris-Cl (pH 7.4), 10 mM NaCl, 3 mM MgCl_2, 0.5% NP40) was then added into the cell pellet and mixed by pipetting and incubated on ice for at least 30 min. The lysed cell mixture was then spun down at 13,000 g for 10 min to remove cell debris. Protein concentrations were determined using the BCA protein assay kit (Pierce, Rockford, IL). Caspase 3/7 activity was measured using the Caspase-Glo® 3/7 Assay kit (Promega) according to the manufacture instructions. Briefly, an equal volume (100 μl) of Caspase-Glo® 3/7 reagent was added to each cell lysate sample (100 μl) in a 96-well assay plate with a final assay volume of 200 μl. Samples were incubated at room temperature for one hour (protected from light) with shaking, and the luminescence of each sample is measured using a Veritas™ Microplate Luminometer (Turner BioSystems, Sunnyvale, CA). The Caspase 3/7 activity was normalized to the amount of total protein contained in the cell lysate as determined by the BCA protein assay (Thermo Scientific, Rockford, MI).

Western blotting analysis

The cells were treated with PHA-739358 (3μM), imatinib (15μM), or PHA-739358 (3μM) plus imatinib (15μM), for 72 hours and then harvested by trypsinization. The cell lysates were prepared as described for the Caspase 3/7 activity assay. Cell lysates containing equal amount of protein (20 μg) were resolved on 4-12% SDS-PAGE (polyacrylamide gel electrophoresis) gels. The separated proteins were transferred to nitrocellulose membranes. Membranes were then probed with primary antibodies against Phospho-PDGFRA, Bcl-xL, Bcl-2, PI3K, Phospho-PI3K, ERK, Phospho-ERK and β-actin (All from Cell Signaling Technology, Inc., Danvers, MA). β-actin was included to serve as a protein loading control. The bound primary antibodies were detected using peroxidase-conjugated secondary antibodies (Cell Signaling) and chemiluminescence by the Immobilon^TM Western Chemiluminescent HRP Substrate (Millipore, Billerica, MA) according to manufacturer’s instructions. The luminescent signal of the membrane was then detected by photographic film.

Optimization of conditions for HT-siRNA screening

Inhibition of PDGFRA by small molecule inhibitors sensitizes pancreatic cancer cells to AKIs

To further validate PDGFRA as a sensitizing target for AKIs in pancreatic cancer, we examined the anti-proliferation activity of combination treatment of PDGFR inhibitors (imatinib and sorafenib) and different AKIs.

Various concentrations of imatinib combined with a serial dilution of two AKIs (ZM447439 and PHA-739358) were first evaluated in three pancreatic cancer cell lines (AsPC-1, BxPC-3, and SU.86.86). As shown in Figure 2, addition of 9 or 13μM of imatinib to ZM447439 (a selective Aurora B inhibitor) resulted in a left shift of the dose response curves in all 3 cell lines (Fig. 2). Imatinib at 13 μM reduced the IC₅₀ values of ZM447439 by 2 and 3 fold in the AsPC-1 and SU.86.86 cell lines, respectively (Fig. 2A and 2C). Addition of imatinib to PHA-739358 (a pan-Aurora kinase inhibitor) also increased the sensitivity of two of the cell lines. Imatinib (20 μM) reduced the IC₅₀ of PHA-739358 by 2 fold in AsPC-1 and 9 fold in SU.86.86 (Fig. 2D and 2F). In addition to the IC₅₀ decrease in the AsPC-1 cell line, this combination enhanced the cytotoxicity effect at the higher concentration of PHA-739358 (Fig. 2D). Table 2 summarizes the IC₅₀ values of the AKIs in combination with imatinib after normalization with the imatinib only treatment (assuming the Bliss independence criterion (40)) and their ratios to the IC₅₀ values of AKI only treatments in the three pancreatic cancer cell lines. A ratio of less than 1 indicates a synergistic interaction between the AKIs and imatinib at the concentrations tested.

Since imatinib is known to inhibit other kinases besides PDGFR, to further confirm that the synergism observed is specific to PDGFR inhibition we tested another known small molecule inhibitor of PDGFR, sorafenib. Similar to imatinib, sorafenib (2 μM and 4 μM) caused a left shift of PHA-739358 dose response curves in AsPC-1 and SU.86.86 cell lines but not in BxPC-3 (Supplementary Fig. S6).

Effects of imatinib and PHA-739358 combination on cell cycle progression

Since Aurora kinase inhibition has been shown to induce cell cycle arrest we examined the effects of the combination treatment of imatinib and PHA-739358 on cell cycle progression in AsPC-1 cells. As expected, PHA-739358 alone induced significant G2/M arrest and polyploidy. PHA-739358 (3 μM) significantly increased the G2/M population from 19.37% to 30.56% and the population of polyploidy cells from 5.80% to 15.61% within 24 hours (Table 3). Imatinib (15 μM) does not affect the cell cycle distribution of at 24 hours. However, the combination treatment of both drugs resulted in further induction of G2/M arrest (45.72% of total cell population) compared to PHA-739358 alone (30.56%)). Similar synergistic effect was observed at both 48 and 72 hour time points where the combination treatment significantly increased G2/M arrest when compared to either drug alone (Table 3). Interestingly, the addition of imatinib to PHA-739358 reduced the polyploidy (>4N DNA content) population induced by PHA-739358 at all 3 time points (Table 3). For instance, at the 24 hour time point, the cell population with >4N DNA increased from 5.8% in untreated control and 5.6% in imatinib only treatment to 15.6% in PHA-739358 only treatment, and reduced back to 5.4% in the imatinib plus PHA-739358 combination treatment.

Induction of apoptotic cell death by the combination treatment of imatinib and PHA-739358 in pancreatic cancer cells

Consistent with its inhibitory activity against both Aurora A and B, PHA-739358 (3 μM) as a single agent reduced the proliferation of AsPC-1 cells and increased the formation of multinucleated cells (top right panel in Fig. 3A). Imatinib (15μM), as a single agent, did not significantly affect the growth of AsPC-1 cells (Bottom left panel in Fig. 3A). However, combination treatment of PHA-739358 and imatinib induced dramatic cell death (Bottom right panel in Fig. 3A). Caspase 3/7 activity assays indicated that PHA-739358 (3 μM) alone significantly induced apoptosis at 72 hours compared to vehicle control (P < 0.001) whereas imatinib (15 μM) did not (Fig. 3B). When the two drugs were combined, the induction of apoptosis further increased significantly when compared to PHA-739358 only treatment (P = 0.00003, Fig. 3B), indicating that PHA-739358 and imatinib act synergistically in inducing apoptosis. Furthermore, combination of another AKI, ZM447439, and imatinib also showed a significant increase in the induction of caspase activity in comparison to either drug alone in the BxPC-3 cell line (Supplementary Fig. S7). To explore the mechanism of action of the increased apoptotic effect of the combination treatment, the expression of the two anti-apoptotic proteins, Bcl-2 and Bcl-xL, were examined by Western blotting. As shown in Figure 3C, treatment with either PHA-739358 or imatinib alone did not significantly affect the level of either protein whereas the combination treatment reduced the expression of Bcl-2 and Bcl-xL by 73% and 68%, respectively, compared to the untreated control, indicating that the increased anti-apoptotic effect of the combination treatment might be a result of the synergistic down-regulation of Bcl-2 and Bcl-xL expression by the two drugs.

Combination treatment of imatinib and AKIs decreased the phosphorylation of PI3K but not ERK

Two major effector pathways of PDGF/PDGFR signaling are the Ras/Erk pathway and the PI3K/Akt pathway. To investigate the effect of combination treatment of imatinib and AKI on these two pathways, we examined the phosphorylation of PI3K and Erk1/2 upon the drug treatment. As shown in Fig. 4, AsPC-1 cells treated with single agent PHA-739358 (3μM) or imatinib (15μM) did not significantly affect the phosphorylation of either Erk1/2 or PI3K. However, combination treatment of PHA-739358 and imatinib resulted in decreased phosphorylation of PI3K but not the ERK1/2 kinases. Similarly, combination of ZM447439 and imatinib resulted in a significant decrease of PI3K phosphorylation level, but not the phosphorylation of Erk kinase in the BxPC-3 cell line (Supplementary Fig. S8). These results suggest that AKIs and imatinib might act synergistically in inhibiting the PI3K/Akt induced cell survival in pancreatic cancer cells.