signature=4481f76f83d9510a14511c67a909703e,A functional genomics screen reveals a strong synergistic...

Cell culture

LNCaP, DU145 and PC3 were cultured in RPMI 1640 medium supplemented with glutamine and 10% fetal bovine serum. LNCaP-AI cells were cultured in charcoal-stripped DMEM medium minus phenol red.

siRNA transfection and screen

Human OnTarget plus siRNAs were ordered from Horizon Discovery (Dharmacon), either individual (screening library) or as pool (AR). Cells were transfected with siRNAs using a reverse transfection protocol with Lipofectamine RNAiMAX (Invitrogen), following the manufacturer’s instructions. The same protocol was used for all cell lines transfected with siRNA.

Viability screen

SiRNA screening was performed in triplicate with a custom library of 48 human genes, and a set of four individual siRNAs per gene (Horizon). Cell stocks were grown to confluency, harvested, counted and seeded in 96-well plates at the following densities to reflect differing growth rates: PC3 at 3000 cells per well (cpw), DU145 at 2500 cpw and LNCaP at 5000 cpw. A reverse transfection method was used in order to transfect cells with siRNA. Subsequently cell suspensions were added to the plates using an XRD automated reagent dispenser (FluidX). Plates were incubated at 37 °C, 5% CO2 for 24 h, after which either EC30 DCT (PC3: 1.36 nM), (DU145: 3.28 nM), (LNCaP: 1.80 nM) or DMSO was added to cells. After incubation for a further 48 h, the medium was removed and cells were fixed with 4% formaldehyde and stained with DAPI/Tx-100 in 1x PBS (0.25 µg/ml/0.001% final concentration/well). Images of nuclei were acquired using the Operetta High Content Imaging system (PerkinElmer) and the number of nuclei in each well were quantified as a measure of cell growth using Columbus High Content Imaging and Analysis Software (PerkinElmer).

Incucyte growth assay

Cells were seeded in 96-well Greiner black glass bottom plates (Greiner 665090), and transfected with siRNA at seeding where needed. Depending on cell type, wells were seeded with 5000 to 12000cpw in 100 ul medium. Plates were incubated overnight and transferred to the Incucyte. Drugs were added with additional 50 ul medium (spike-in) between scans. CytoxGreen reagent was added according to manufacturer’s protocol along with drugs, where required. In each experiment, plates contained three or more replicate wells per condition and two pictures were taken per well per scan. Confluence data were downloaded, plotted and analysed using MATLAB custom routines. In particular, for relative GI, we first normalized confluence growth curves by the confluence at time of drug addition in order to balance out variation in seeding. We then calculated the ratios of confluence measures for drug over DMSO for each condition, in each experiment. We then plotted the mean of experiments with the SEM indicated in error bars. In order to analyse differences between the curves we calculated area under curves and compared these values using a Student’s T-test.

Western blots

Cells were seeded in 6-well plates and transfected at seeding as needed. Cells were seeded at 300,000 cpw in 2000 ul medium. Cells were incubated overnight and drugs (DMSO, DCT) were added in additional medium (200 ul) and were left on for required periods (24 h). Cells were washed once with PBS and lysed in buffer (50 mM TRIS ph7.5, 0.5% SDS) with cOmplete Mini preparation (Roche). Lysates were collected in Qiagen shredder columns (Qiashredder) and centrifuged through the column for 2 min at high speed. Lysates were buffered in NuPage LDS sample buffer with 5% mercaptoethanol, and heated to 100 °C for 5 min. Western blotting was performed using the NuPage system, with NuPage Bis-Tris 4–12% precast gels. Gels were run in MES buffer (NuPage MES SDS running buffer, 135 V, 80 min), and were transferred onto nitrocellulose membranes (GE Healthcare Amersham Protran 0.2NC) in NuPage transfer buffer plus 20% methanol (100 V, 90 min). Membranes were stained with Poinceau solution, washed in TBS, blocked in 5% milk in TBS for 1 h and incubated with primary antibodies (see below) at manufacturer’s recommended concentrations overnight. Blots were washed with TBS and incubated with secondary antibodies (Licor) for 1 h. Blots were washed in TBS followed by distilled water and were screened on Licor System. Licor software was used to perform quantitation.

Antibodies:

CDC20: CDC20 (D6C2Q) Rabbit mAB #14866, Cell Signaling Technology

DLGAP5: HPA005546, Sigma Aldrich

AR: AR(441):sc-7305, Santa Cruz Biotechnology

TUB: Anti-Tubulin Antibody YL1/2 ab6160, abcam

CCNB1: Anti-Cyclin B1 Antibody ab32053, abcamFKBP5: FKBP5 (D5G2) Rabbit mAB #12210, Cell Signaling Technology

GAPDH: GAPDH Rabbit PolyAB CatNo 10494-1-AB, Proteintech Europe

ICC/IF

For immunofluorescence, LNCaP cells were plated on Greiner black myClear Cellstar 96-well plates (Greiner 655090) and were transfected with siRNA at seeding. Cells were seeded at 12,000 cpw. After incubation for 36 h, 0.5 nM DCT or DMSO was added to the wells and plates were incubated for an additional 12 h. Cells were then washed once with PBS and fixed in 4% para-formaldehyde for 20 min. Plates were washed three times in PBS, blocked for 1 h in PFS (0.5 L PBS, 3.5 g fish skin gelatin, 1.25 mL 10% saponin stock) and incubated with primary antibodies (DLGAP5, AR, TUB, same as above) in PFS at manufacturer’s recommended doses overnight. Cells were washed in PFS and incubated with secondary antibodies (Alexa 488, 594, 647, DAPI) in PFS for 1 h. After another round of washing, wells were filled with 200 ul PBS and plates were screened using the Opera system at ×20 and ×63 resolution. Images were analysed using the system software (Columbus High Content Imaging and Analysis Software, PerkinElmer). Meta- and post-metaphase cell counts were performed manually using the ×20 resolution images with Tubulin and DAPI channels (36 image fields per well, three replicate wells per condition per experiment). Ratios of post-meta/metaphase cells were calculated for each well, ratios were averaged over replicate wells in each experiment, and mean of experiments was plotted with error bars indicating SEM.

Computational analyses

Gene expression data were downloaded from the Gene Expression Omnibus (GEO) (GSE21034, GSE16560, GSE25136, GSE35988, GSE44905, GSE60721, GSE4636, GSE69330) and from The Cancer Genome Atlas (TCGA) portal (TCGA-PRAD). TCGA, GSE21034 and GSE16560 data sets were classified using our previously described classification schemep-values of enrichment in the data sets, their expression (present) in the cell lines, and their annotations. To evaluate the list of 48 genes as a signature, we used ssGSEA to calculate signature scores for each sample in a set, and then clustered the scores into three groups using kmeans. Survival data was then analysed using Kaplan–Meier analysis and logrank p-values. It was tested in a fourth cohortp-values were calculated using MATLAB standard script. All data were handled using Matlab and Unix shell script.

Statistical analyses

Screening data were analysed for quality control, and Z prime scores were calculated for each plate using the Dotmatics Studies software (all Z prime scores were greater than 0.5). Each plate contained eight samples negative and positive controls (siNTC, siAS). The median of the eight negative controls on each plate was used to calculate sample values for the siRNA probes (value = (median(siNTC) − SAMPLE) / (median(siNTC)) × 100). Mean of three experiments was calculated for each cell line, each gene, each sequence and each condition (DMSO, DCT). Window sizes were calculated across all genes and sequences in a given cell line, and mean plus two standard deviations was used as a cutoff to determine significant sequences. Genes with two or more significant sequences were considered significant in that cell line. Furthermore, combination indices (CI) were calculated using the Bliss independence model, for each gene, sequence and experiment in each cell line. For each cell line, we plotted the distribution of all CI scores and determined the 0.05 quantile in order to obtain a significance value relative to the screening data.

Incucyte data were downloaded as percent confluence in phase and green channels (CytoxGreen). Mean of technical replicates (three or more wells per condition, two images per well per scan) was used for all calculations and data were normalized by the confluency at time of drug addition for all conditions. Relative GI curves were then calculated as DCT/DMSO and plotted using mean of experiments and error of means. Significant differences between groups of curves were determined by calculating AUC using the trapezoidal rule for each curve, followed by Student’s T-test.