A synthetic lethal siRNA screen identifying genes mediating sensitivity to a PARP inhibitor Nicholas C Turner1, Christopher J Lord1, Elizabeth Iorns1, Rachel Brough1, Sally Swift1, Richard Elliott1, Sydonia Rayter1, Andrew N Tutt1,2 and Alan Ashworth1,* 1The Breakthrough Breast Cancer Research Centre, The Institute of Cancer Research, London, UK and 2Breakthrough Breast Cancer Research Unit, King���s College London School of Medicine, Guy���s Hospital, London, UK Inhibitors of poly (ADP-ribose)-polymerase-1 (PARP) are highly lethal to cells with deficiencies in BRCA1, BRCA2 or other components of the homologous recombination path- way. This has led to PARP inhibitors entering clinical trials as a potential therapy for cancer in carriers of BRCA1 and BRCA2 mutations. To discover new determinants of sensi- tivity to these drugs, we performed a PARP-inhibitor synthetic lethal short interfering RNA (siRNA) screen. We identified a number of kinases whose silencing strongly sensitised to PARP inhibitor, including cyclin- dependent kinase 5 (CDK5), MAPK12, PLK3, PNKP, STK22c and STK36. How CDK5 silencing mediates sensi- tivity was investigated. Previously, CDK5 has been suggested to be active only in a neuronal context, but here we show that CDK5 is required in non-neuronal cells for the DNA-damage response and, in particular, intra-S and G2/M cell-cycle checkpoints. These results highlight the potential of synthetic lethal siRNA screens with chemical inhibitors to define new determinants of sensiti- vity and potential therapeutic targets. The EMBO Journal (2008) 27, 1368���1377. doi:10.1038/ emboj.2008.61 Published online 3 April 2008 Subject Categories: genome stability & dynamics Keywords: CDK5 cell cycle DNA repair poly(ADP)ribose polymerase RNAi screen Introduction Poly (ADP-ribose)-polymerase-1 (PARP) is a highly abundant nuclear enzyme involved in the repair of single-strand breaks (SSBs) (Hoeijmakers, 2001). Inhibition of PARP induces accumulation of large numbers of unrepaired SSBs, leading to the collapse of replication forks during S-phase and the consequent generation of double-strand breaks (DSBs). Cells deficient in DNA DSB repair, in particular homologous recombination (HR) by gene conversion, are highly sensitive to chemical inhibitors of PARP (Bryant et al, 2005 Farmer et al, 2005 McCabe et al, 2006). In contrast, cells with intact DNA DSB-response pathways repair damage with high fide- lity and accordingly show very little sensitivity to PARP inhibitors. The breast and ovarian cancer predisposition genes, BRCA1 and BRCA2, encode proteins that are required for efficient HR (Moynahan et al, 1999 Tutt et al, 2001). Tumours arising in the carriers of heterozygous germline BRCA mutations have generally lost the wild-type BRCA allele, resulting in defective HR, which may be targeted in a synthetic lethal approach (Farmer et al, 2005). PARP inhibi- tors have now entered clinical trials and initial results are promising, with frequent sustained responses in BRCA muta- tion carriers (Yap et al, 2007). Despite the clinical promise of PARP inhibitors in the treatment of BRCA-related cancer, extending the utility of these agents to other cancers is challenging. Little is known about the determinants of PARP-inhibitor sensitivity, other than the profound sensitivity of cells with defects in HR (McCabe et al, 2006). The identification of novel mediators of cellular response to PARP inhibitors may highlight additional patient populations that might benefit form this therapeutic approach. Furthermore, mechanisms of drug resistance and potential combination therapies may also be uncovered. RNA interference (RNAi) screens have the poten- tial to identify novel determinants of drug response and hence enhance the application of novel and existing drugs (Iorns et al, 2007), and have already proven highly effective in the unbiased identification of novel genes involved in biological processes (Aza-Blanc et al, 2003 Mukherji et al, 2006). These screens exploit the naturally occurring mechan- ism of RNAi that controls gene expression at the post- transcriptional level by mediating degradation of mRNA transcripts in a sequence-specific fashion (Meister and Tuschl, 2004). With the development of RNAi libraries com- posed of reagents that allow targeting a wide range of transcripts, it is now possible to conduct high-throughput screens (HTS) that simultaneously interrogate phenotypes associated with the loss of function of many genes (Iorns et al, 2007). Here, we have used a high-throughput RNAi screen to identify new determinants of sensitivity to a PARP inhibitor. Results siRNA screen for kinases sensitising to a PARP inhibitor RNAi screens that have previously examined sensitivity to DNA-damaging chemotherapy drugs have been limited by the small relative sensitivity, or therapeutic window, that exists between cells that are sensitive and resistant, limiting screens to identification of genes that cause profound effects when silenced (Bartz et al, 2006). DNA DSB repair-deficient cells are potentially more than a thousandfold more sensitive than resistant cells to PARP inhibitor (Bryant et al, 2005 Farmer et al, 2005 McCabe et al, 2006), probably due to the Received: 10 September 2007 accepted: 4 March 2008 published online: 3 April 2008 *Corresponding author. The Breakthrough Breast Cancer Research Centre, The Institute of Cancer Research, 237 Fulham Road, London SW3 6JB, UK. Tel.: �� 44 0 20 7153 5333 Fax: �� 44 0 20 7153 5340 E-mail: alan.ashworth@icr.ac.uk The EMBO Journal (2008) 27, 1368���1377 | & 2008 European Molecular Biology Organization | All Rights Reserved 0261-4189/08 www.embojournal.org The EMBO Journal VOL 27 | NO 9 | 2008 &2008 European Molecular Biology Organization EMBO THE E M B O JOURNAL THE EMBO JOURNAL 1368

specificity of the DNA damage induced, increasing the ability of a screen to detect significant but less sensitising effects. We performed a PARP inhibitor synthetic lethal screen with a short interfering RNA (siRNA) library targeting 779 human protein kinase and kinase-associated genes. We selected kinases as they represent drugable targets. CAL51 cells were used for the screen, which are a diploid, TP53 wild-type breast cancer cell line. The HTS assay in- volved transfecting CAL51 cells with siRNA in a 96-well plate format and dividing the cells the day after transfection into replica plates, treating half with the PARP inhibitor KU0058948 and half with the vehicle (Figure 1A). The screen was optimised to detect modestly sensitising effects by using a dose of KU0058948 sufficient to inhibit the repair of SSBs (data not shown) and equivalent to the SF80 (80% survival after KU0058948 administration). Furthermore, cells were exposed to drug continuously for 5 days to allow multiple cell cycles to occur, allowing effects of PARP inhibition to accumulate and modelling chronic exposure to these drugs in the clinic. The screen was completed in duplicate. Comparison of the two duplicates revealed the screen to be highly reproducible (Figure 1B and C). The duplicates of the screen were com- bined in the final results of the screen displayed in Figure 1D. A robust significance or ���hit��� threshold of a combined Z-score of 3 or less was selected, with 24 gene-specific siRNA pools (SMARTPoolss Dharmacon) fulfilling this criterion (Table I). Full results of the screen are supplied as Supplementary Table 1. Internal validation of the high sensitivity of the screen was provided by the demonstration that siRNAs targeting the key DNA-damage response genes, ataxia���telan- giectasia related (ATR), ataxia���telangiectasia mutated (ATM) and CHK1, significantly sensitised to the PARP inhibitor (Table I), as we have previously reported (McCabe et al, 2006), emphasising the importance of intact DNA DSB- response pathways in tolerance to PARP inhibitors. Validation of siRNA screen hits In addition to silencing a target gene, siRNAs potentially suppress the expression of a large number of other genes Figure 1 PARP-inhibitor synthetic lethality screen with protein kinase siRNA library. (A) HTS method. CAL51 cells plated in 96-well plates were transfected with siRNA. Each transfection plate contained 80 experimental siRNAs (SMARTPools of four different siRNA targeting the same gene) supplemented with four wells of non-targeting siCON, and two wells of siRNA directed against BRCA1 (positive control). Transfected cells were divided into six replica plates, half treated with DMSO vehicle alone and half with PARP inhibitor KU0058948 at 1 mM, the SF80 of CAL51. Cell viability was assessed after 5 days of KU0058948 exposure using CellTiter-Glo Luminescent Cell Viability Assay (Promega). (B) Reproducibility of HTS method. Correlation of the effect of siRNA on cell growth in vehicle-treated plates from two replicates of the entire screen. Spearman correlation coefficient, r �� 0.83. (C) Correlation of KU0058948 sensitivity Z-scores from two replicates of the entire screen. r �� 0.54. (D) Scatter plot of averaged Z-scores from PARP inhibitor sensitivity screen carried out in duplicate with KU0058948. Red line indicates 3 averaged Z-score significance threshold. Black, siRNA (SMARTPools) targeting 779 protein kinase genes red, siCON and blue, siBRCA1. Reflecting the reproducibility and sensitivity of the screen, siCON and siBRCA1 Z-scores were widely separated, with a screen Z0-factor (Zhang et al, 1999) of 0.34. siRNA screen for PARP-inhibitor sensitivity genes NC Turner et al &2008 European Molecular Biology Organization The EMBO Journal VOL 27 | NO 9 | 2008 1369