A cell-based high-throughput screening assay for radiation susceptibility using automated cell counting

PC-3 and DU145 prostate cancer cell lines were maintained in RPMI1640 medium (Lonza, Verviers, Belgium). A549 lung adenocarcinoma and U2OS osteosarcoma cell lines were maintained in DMEM medium (PAA, Cölbe, Germany). All cultures were supplemented with 10% fetal calf serum (Greiner Bio-One, Alphen a/d Rijn,The Netherlands) and 50U/ml penicillin and 50U/ml streptomycin (PAA, Pasching, Austria) and maintained at 37°C and 5% CO2 in a humidified atmosphere.

Stable knockdown cells were made using lentiviral vectors expressing shRNA from the TRC library (Thermo Scientific Open Biosystems). Lentiviral vectors were made by transfection of HEK-293T cells with psPAX2 and pMD2.G packaging constructs (Addgene, Cambridge, MA) together with lentiviral vector clone TRCN0000006256 carrying the PRKDC-silencing shRNA sequence 5′-CCGG-CCGGTAAAGATCCTAATTCTA-CTCGAG-TAGAATTAGGATCTTTACCGG-TTTTT-3′; or negative control pLKO.1 Empty Vector (Cat. No. RHS4080), using FuGENE®6 (Promega Benelux, Leiden, The Netherlands) transfection reagent. Culture medium containing virus particles was harvested 2 and 3 days after transfection. Cleared supernatant was used to transduce PC-3 cells in the presence of 8 μg/ml polybrene (Sigma Aldrich Chemie BV, Zwijndrecht, The Netherlands). Transduced cells were selected using incubation in 5 μg/ml puromycin (Gibco® by Life Technologies^TM, Bleiswijk, The Netherlands).

Cells were irradiated with the indicated dose IR at a dose rate of 6 Gy/minute using a Clinac 2300CD or TrueBeam linear accelerator (Varian Medical Systems, Palo Alto, CA). Microtiter plates (Greiner Bio-One) were placed in a for this purpose specifically designed Perspex (PMMA, i.e. near-water equivalent material in radiotherapy) container. The overall size of the container is 58.6 x 57 x 8.3 cm with a side wall thickness of 9.6 cm and a top and bottom wall of approximately 2.2 cm. The container allows to irradiate at maximum 30 ANSI/SLAS-standard culture plates simultaneously, divided into 3 groups of 10 plates each. Per group, the plates were clustered and stacked in 2 layers. Between groups, 2 cm thick easily removable Perspex rods were situated. Unused plate positions were filled with Perspex blocks. Together with the thick container walls and the removable Perspex rods between the groups, these blocks assure a full-phantom photon scatter condition for all cells. This is a requirement for a correct and reproducible dose delivery to all cells irrespective of their position in the container. Irradiation planning for homogeneous dose distribution was performed as described in the results section.

Cells were seeded in 6-well plates at a density depending on the irradiation dose, i.e., 250 cells/well for 0–3 Gy and 500 cells/well for 4–8 Gy; and irradiated within 16–20 hours after seeding. Seven to 10 days after irradiation, cells were washed with PBS, fixed with 4% formaldehyde and stained using Giemsa (Sigma Aldrich Chemie BV, Zwijndrecht, The Netherlands) at 1:20 dilution in PBS. Colonies containing 50 or more cells were counted. Data shown are means from three independent experiments done in duplicate. Survival fractions were calculated using the formula: SF = (nr colonies/nr of cells plated) irradiated/(nr colonies/nr of cells plated) untreated. Data were fitted with the linear quadratic model: S = exp(−αD-βD²), where S is the surviving fraction and D is the IR dose.

Cells were seeded 500 cells/well in 100 μl medium and irradiated in 96-well plates as described above. Five days after irradiation, 20 μl CellTiter-Blue (CTB) reagent (Promega Benelux, Leiden, The Netherlands) was added to each well and the cells were cultured for 2 hours. Cell viability was determined by measuring fluorescence at 540 nm excitation and 590 nm emission wavelengths using a Tecan Infinite F200 microplate reader. Data shown are means from three independent experiments done in triplicate. Data were fitted using the linear quadratic model.

Cells were seeded at the indicated number per well in 96-well plates. Where indicated, caffeine (Sigma-Aldrich) was added at 2 mM final concentration 1 hour before irradiation and cells were irradiated as described above. Five days after irradiation, the culture medium was removed and cells were washed with PBS and fixed with 7% formaldehyde for 30–60 minutes. After a second wash with PBS, cells were optionally stored at RT until analysis. Prior to analysis on an Acumen eX3 (TTP LabTech, Melbourne, UK) microplate cytometer integrated with a Twister II robotic system (Caliper, Teralfene, Belgium) for unattended high-throughput data acquisition, cells were stained with 0.3 μg/well Hoechst 33342 (Sigma Aldrich Chemie BV, Zwijndrecht, The Netherlands) in PBS for at least 30 minutes. The plates were scanned in Cytometry Mode using a scan resolution of 1x4 μm and a laser power of 6 mV, with a 405 nm excitation laser and a photomultiplier tube detector equipped with 500–530 nm detection filters. Identified fluorescent objects were used to calculate cell numbers using the Acumen eX3 software and user-defined parameters. These can be optimized for each individual cell line. Here, we used a single algorithm for all four cell lines included in the study. For three of these cell lines we observed that when cells grew to high density, adjacent nuclei were sometimes recognized as one object. Inspection of these objects revealed that they consisted of on average 3 nuclei. Therefore, single cell objects and cluster objects representing 3 cells were defined separately as follows. Small objects with a width and depth of 5–50 μm were defined as single cells and larger objects with a width and depth of 50–250 μm were defined as clusters of on average 3 cells. The cell number was calculated by the sum of the number of small objects plus 3-times the number of larger objects. Upon irradiation, a small proportion of nuclei were enlarged and had increased fluorescence intensity, i.e., increased DNA content. Due to their enlarged size, these nuclei typically representing polyploid cells were categorized among the cluster objects. Consequently, the cell count at high IR dose was slightly overestimated. This had, however, minimal effect on the dose–response curves and was therefore deemed acceptable. Data were fitted using the linear quadratic model.

Whole human genome siRNA library screens for molecular radiosensitization targets in PC-3 cells were performed using the Thermo Fisher Scientific Dharmacon siARRAY library. PC-3 cells were seeded 1,500 cells per well in 96-well microtiter plates. The next day, they were transfected with 20nM siRNA using 0.02% DharmaFECT1 (Thermo Fisher Scientific Dharmacon, Lafayette, CO) transfection reagent. Two arrayed screens were done, each comprising a total of 544 96-well plates, which were run in two separate sessions of 272 plates each, comprising a set of 136 plates in duplicate. One set of replicates was irradiated with 4Gy two days after siRNA transfection; the other set was not. To allow assay quality assessment, on each plate two wells with a non-targeting siRNA control (siNT#2, Cat. No. D-001206-14-20) and two wells with PRKDC siRNA (Cat. No. M-005030-01-0020) as positive control were included. Thus, four screening sessions were done, each including 272 irradiated positive controls, 272 irradiated negative controls, 272 non-irradiated positive controls and 272 non-irradiated negative controls. The counted cell numbers were corrected for plate, session and screen effects by means of a linear regression model fitted to the log-2 intensities of the entire experiment simultaneously. This normalization ensures estimation of technical effects is robust, since estimates are unlikely to be affected by deviations observed on a few wells, or for only one replicate. In addition, it preserves the irradiation effect, as each technical factor corrected for always involves both irradiated and non-irradiated observations. In order to find siRNAs that yielded a significantly larger difference in cell viability before and after irradiation, compared to the difference in cell viability measured for negative controls, a linear regression model with treatment effect was fitted to the normalized data, and this treatment effect was compared between each siRNA and all negative controls (i.e. 1,088 with IR and 1,088 without IR). The regression model thus includes main effects for treatment and for siRNA type (negative control or the siRNA chosen), as well as an interaction effect between treatment and siRNA that is used to test for siRNAs with different treatment effect, compared with controls. As the model is fitted for each siRNA, a list of t-statistics for the interaction effects, and corresponding p-values corrected for multiple testing by controlling the false discovery rate (FDR), is produced. Those siRNAs that are associated with significantly less cell survival in combination with IR are considered radio-sensitizers.