SPOP mutation leads to genomic instability in prostate cancer

Genomic instability is a fundamental feature of human cancer often resulting from impaired genome maintenance. In prostate cancer, structural genomic rearrangements are a common mechanism driving tumorigenesis. However, somatic alterations predisposing to chromosomal rearrangements in prostate cancer remain largely undefined. Here, we show that SPOP, the most commonly mutated gene in primary prostate cancer modulates DNA double strand break (DSB) repair, and that SPOP mutation is associated with genomic instability. In vivo, SPOP mutation results in a transcriptional response consistent with BRCA1 inactivation resulting in impaired homology-directed repair (HDR) of DSB. Furthermore, we found that SPOP mutation sensitizes to DNA damaging therapeutic agents such as PARP inhibitors. These results implicate SPOP as a novel participant in DSB repair, suggest that SPOP mutation drives prostate tumorigenesis in part through genomic instability, and indicate that mutant SPOP may increase response to DNA-damaging therapeutics. DOI: http://dx.doi.org/10.7554/eLife.09207.001

Whole genome sequencing (WGS) has allowed an unprecedented insight into the alterations underlying cancer. Recently, WGS of treatment naive, clinically localized prostate cancer revealed a striking abundance of genomic rearrangements, in some samples comparable to the number of rearrangements in metastatic prostate cancer (Baca et al., 2013). Furthermore, the type (intrachromosomal vs interchromosomal) and complexity of rearrangements in these tumors shows remarkable heterogeneity, potentially suggesting distinct mechanisms of instability in different molecular classes of prostate cancer. However, somatic alterations underlying these phenomena remain unexplained.
Mutations in SPOP (Speckle-type POZ protein) occur in around 10% of prostate cancers and represent the most common non-synonymous mutations in primary prostate cancer (Barbieri et al., 2012). SPOP mutations define a distinct molecular class of prostate cancer; they are mutually exclusive with ETS rearrangements but display distinct patterns of somatic copy number alterations (SCNAs) (Barbieri et al., 2012).
Here, we investigated somatic alterations associated with genomic rearrangements in prostate cancer. We show that SPOP mutation is an early event specifically associated with increased intrachromosomal genomic rearrangements. Mechanistically, in vitro and in vivo data suggest that SPOP participates in repair of DNA double strand breaks (DSB), and SPOP mutation impairs homology-directed repair (HDR), instead promoting error-prone non-homologous end joining (NHEJ).

Results
To nominate somatic events associated with structural genomic rearrangements in clinically localized prostate cancer, we examined WGS data from 55 treatment naive prostate cancers (Baca et al., 2013) ( Figure 1A). This analysis revealed a bimodal distribution, with a more common, 'low-rearrangement' population, and a less frequent 'high-rearrangement' population primarily driven by intrachromosomal rearrangements (deletions, inversions, and tandem duplications), rather than balanced interchromosomal rearrangements ( Figure 1B). We then analyzed the association between recurrent eLife digest Prostate cancer is the most common type of cancer in men in the UK and USA.
Cancers develop when cells in the body acquire genetic mutations that allow the cells to grow rapidly and form a mass known as a tumor. Prostate cancer cells from different individuals can carry different genetic mutations, which affects whether the disease progresses and how the tumors respond to medical treatments. This genetic variety arises in cancer cells partly from a phenomenon known as genomic instability, in which DNA mutations accumulate due to defects in DNA repair.
Genetic studies of biopsies taken from human prostate cancers have shown that genomic instability causes chromosomes-the structures in which the cell's DNA is organized-to break and then be stuck back together haphazardly. As a result, fragments of chromosomes can end up in the wrong position, be duplicated, or be lost altogether. All of these mutations could spur on the growth of the tumor. However, it is currently not clear why some prostate cancers are more genomically unstable than others, or what exactly causes this instability.
Boysen, Barbieri et al. studied prostate cancer cells taken from patients before they started medical treatment. The experiments show that the cancer cells with high levels of genomic instability also often had mutations in a gene that encodes a protein called SPOP. These mutations occur in about 10 percent of men with prostate cancer and appear early in the development of the tumors. Next, they studied the SPOP protein in zebrafish (which is nearly identical to human SPOP), as well as in mouse and human cells. The experiments show that SPOP normally helps the cell to accurately repair DNA that has been damaged. Mutations in SPOP change the DNA repair process, which lead to genomic instability by increasing the likelihood that broken chromosomes will be stuck back together incorrectly.
Further experiments tested drugs known as PARP inhibitors on mouse and human prostate cancer cells. The drugs, which have been recently tested successfully in patients with prostate cancer, block a different method of DNA repair that operates separately to the one that involves SPOP. When both of these pathways were inactivated-one by the SPOP mutation, the other by the drug-the cancer cells died more quickly. Therefore, men that are diagnosed with types of prostate cancer in which the gene that encodes SPOP is mutated might benefit from treatment with PARP inhibitors or other therapies that affect DNA repair. . Several recurrent deletions, primarily on chromosomes 5q and 6q, were significantly associated with intrachromosomal genomic rearrangements (Figure 1-figure supplement 1), and these were completely distinct compared to alterations associated with interchromosomal rearrangements (Figure 1-figure supplement 2). Among recurrent point mutations, only a single lesion-mutation in SPOP-was significantly associated with increased rearrangements ( Figure 1C). Consistent with increased intrachromosomal rearrangements, SCNA analysis revealed that SPOP mutant prostate cancers showed significantly higher total copy number alteration burden ( Figure 1D).
SPOP mutation frequently co-occurs with specific SCNAs, designating a molecular class of prostate cancer (Barbieri et al., 2012;Blattner et al., 2014) (Figure 1-figure supplement 3). Independent analysis of SCNAs from three publicly available data sets comprising 430 tumors (Baca et al., 2013;Barbieri et al., 2012;Consortium TCGA, 2015), including 47 SPOP mutant prostate cancers, confirmed that the rearrangement-associated deletions (Figure 1-figure supplement 1) were those enriched in SPOP mutant prostate cancer ( Figure 1E). When individually comparing SPOP mutations and associated deletions (CHD1 and MAP3K7), we did not observe significant differences in SCNA burden for any one lesion ( Figure 1-figure supplement 4). Analysis of clonality (Baca et al., 2013;Prandi et al., 2014) of specific lesions showed that SPOP mutations were highly clonal compared to loci in the associated deletion peaks, supporting that SPOP mutations precede deletions ( Figure 1F). In addition, analysis of dependencies of the lesions supports SPOP mutations preceding CHD1 deletions; no lesions were predicted to precede SPOP mutation (Figure 1-figure supplement 5). Together, these data nominate a distinct prostate cancer class characterized by early SPOP mutations and genomic instability. We posited that the SPOP mutation impacts genome maintenance and prioritized SPOP for functional studies.
We explored the functional role of SPOP in vivo, using zebrafish as a rapidly assessable vertebrate model system. SPOP is highly conserved (97.3% identical at the amino acid level between human and zebrafish,  rearrangements. X-axis shows (−log10) p-value. (D) SPOP mutant prostate cancers harbor increased total somatic copy number aberration (SCNA) burden. The fraction of altered genome, partitioned into bins covering a range from <0.01 to ≥0.5, is shown as a histogram for SPOP WT and SPOP mutant tumors. Inset: the percentage of altered genome is significantly increased in SPOP mutant prostate cancers (p = 0.0016, two-sample Wilcoxon-Mann-Whitney test). (E) Frequency of somatic copy number alterations in 430 clinically localized prostate cancers. SPOP-mutant cancers (orange) and SPOP-wild-type tumors (gray). Length of bars indicates the frequency of copy number alterations. (F) Clonality of selected alterations associated with genomic rearrangements, in 430 clinically localized prostate cancers. DOI: 10.7554/eLife.09207.003 The following figure supplements are available for figure 1:  Source data 1. List of genes contained in blocks A, B, and C in the heatmap in Figure 2D. SPOP-F133V, the most commonly mutated residue in prostate cancer ( Figure 2D). Consistent with recently reported proteomic data (Theurillat et al., 2014) and heterodimerization between mutant and wild-type SPOP in our models ( . Notably, the transcriptional response to F133V correlated highly with BRCA1 inactivation ( Figure 2E). While SPOP has been previously proposed as involved in DNA damage and repair (DDR) signaling based on in vitro experiments (Zhang et al., 2014a), these results for the first time implicate a functional role for SPOP in DDR signaling in vivo and suggest this function is selectively impaired by prostate cancer-derived SPOP mutations. To test this hypothesis in human prostate cancers, we performed unsupervised hierarchical clustering of transcriptional data from 11 SPOP mutant and 53 SPOP wild-type tumors, based on the transcriptional signature of BRCA1 inactivation (MSigDB: M2748). We observed significant segregation of SPOP mutant tumors (p-value = 9.5e − 07 , Figure 2F) using this signature, with less robust segregation of tumors harboring CHD1 or MAP3K7 deletions co-occurring with SPOP mutation (Figure 2-figure supplement 3). These data suggest that human prostate cancers with SPOP mutations show transcriptional effects similar to BRCA1 inactivation, consistent with a role for SPOP in DSB repair. Supporting this hypothesis, we observed a trend of mutual exclusivity between SPOP mutations and BRCA1 alterations in genomic data (p-value = 0.0496, Figure 2-figure supplement 4).
Consistent with the hypothesis that SPOP is involved in DSB repair in prostate cells, SPOP forms nuclear foci in prostate cells after γ-irradiation ( Figure 3A) and stable expression of SPOP-wt conferred resistance to DSB-inducing agents cisplatin and camptothecin (CPT) (Figure 3-figure supplement 1A,B). To further define the impact of SPOP mutation in response to DNA damage, we utilized primary prostate cells isolated from transgenic mice with Cre-dependent conditional expression of SPOP-F133V. After transduction with Tamoxifeninducible Cre (Cre-ERT2), cells were treated with 4-OH tamoxifen or vehicle and exposed to ionizing radiation (IR) (  Figure 3B). No differences were seen in apoptosis after IR, as measured by PARP cleavage (Figure 3-figure supplement 2), subG1 events, or caspase-3 cleavage (data not shown). To functionally characterize how SPOP mutations affect response to DSB, we expressed SPOP-wt and SPOP-F133V in benign prostate epithelial cells (RWPE) and prostate cancer cells (22Rv1), and examined the induction, recognition, and resolution of CPT-induced DSBs. We found that DSB formation (measured by γH2AX foci and protein levels) was not affected by modulating SPOP function with siRNA, or expression of wildtype or mutant SPOP ( Figure 3C,  We next examined specific markers (53BP1, RAD51) of the two major DSB repair pathways, HDR and NHEJ. RAD51 is a component of the HDR pathway and a marker for engagement of the HDR machinery (Baumann et al., 1996). 53BP1 is a positive regulator of NHEJ that blocks 5′-DNA-end resection and therefore functions at the intersection of HDR and NHEJ; if 53BP1 is not cleared from sites of DSB by HDR components, it promotes error prone NHEJ (Panier and Boulton, 2014). Strikingly, SPOP-F133Vexpressing prostate cells showed delayed clearance of 53BP1 from sites of DSB ( Figure 3D  dramatic decrease in RAD51 foci formation ( Figure 3E,F, Figure 3-figure supplement 3B), consistent with impairment of HDR. Induction of SPOP-F133V in primary MPCs similarly decreased Rad51 foci after IR ( Figure 3G,H). Knockdown of SPOP also resulted in a decrease of RAD51 foci, suggesting a selective loss of function of SPOP-F133V in HDR ( Figure 3I). We also observed decreased clearance of 53BP1 foci and decreased RAD51 foci formation in SPOP-F133V-expressing cells after gamma irradiation, indicating that this effect is not specific to one mechanism of DSB induction (Figure 3-figure supplement 3C-H). The observed changes in DSB repair were not accompanied by changes in the cell cycle distribution of cells expressing SPOP-wt or F133V under these conditions (Figure 3-figure supplement 3K).
We next investigated the role of SPOP in DSB repair using the well established DR-GFP and Pem1-Ad2-EGFP reporter assays as functional readouts for HDR and NHEJ, respectively ( Figure 4A,D) (Pierce et al., 1999;Seluanov et al., 2004). In the DR-GFP assay, knockdown of SPOP by siRNA decreased HDR competence in human epithelial cells to a similar level of BRCA1 knockdown ( Figure 4B). Conversely, ectopically expressed SPOP-wt increased the HDR competence in these cells, with partial loss of this function by mutant SPOP ( Figure 4C). In contrast, the Pem1-Ad2-EGFP NHEJ reporter assay indicated an increase of NHEJ activity in SPOP-F133V expressing epithelial cells, while both SPOP siRNA and BRCA1 siRNA increased NHEJ ( Figure 4E,F). Taken together, these results suggest that SPOP promotes HDR, while somatic mutation in SPOP, as observed in prostate cancer with increased genomic rearrangements, impairs HDR and promotes error-prone NHEJ.
Human cancers with underlying defects in HDR (such as BRCA1 inactivated breast and ovarian cancers) (Bryant et al., 2005;Farmer et al., 2005;Audeh et al., 2010;Tutt et al., 2010) show sensitivity to poly (ADP-ribose) polymerase 1 (PARP) inhibition. To test if SPOP inactivation conferred sensitivity to PARP inhibition after DSB induction, we utilized siRNA targeting SPOP in human prostate cancer cell lines (PC-3, LnCap, 22Rv1), followed by irradiation (5GY) and incubation with the PARP inhibitors olaparib or veliparib. Reduction of SPOP expression increased the sensitivity of these prostate cancer cells to both PARP inhibitors ( Figure 5A-C, Figure 5-figure supplement 1B,C). To test if SPOP mutation also sensitizes to PARP inhibition, we treated MPCs (control vs tamoxifeninduced SPOP-F133V) with olaparib followed by IR-expression of SPOP-F133V increased sensitivity to olaparib similar to loss of SPOP in human prostate cancer cell lines ( Figure 5D). This was further confirmed in 22Rv1 cells ectopically expressing SPOP-F133V, which also showed an increased sensitivity to olaparib in viability assays, while cells ectopically expressing SPOP-wt were relatively resistant ( Figure 5-figure supplement 1A). We further confirmed these results in HEK293 cells stably overexpressing the two common SPOP mutants (Y87N, F133V) in clonogenic survival assays ( Figure 5E,F).
To determine if the altered sensitivity to PARP inhibitors was associated with increased DNA damage, consistent with impaired double-strand break repair, we performed single-cell gel electrophoresis (comet assay) in prostate cells treated with olaparib. The comet assay revealed an increase of genomic instability in SPOP-F133V RWPE cells after PARP inhibition, similar to SPOP (and BRCA1) knockdown with siRNA ( Figure 5G). Similarly, primary MPCs expressing SPOP-F133V showed increased damage after olaparib treatment ( Figure 5H). Taken together, these data argue that SPOP mutation confers sensitivity to PARP inhibition due to impaired error-free HDR DSB repair and an increase in error-prone NHEJ.

Discussion
In summary, here we report that SPOP, the substrate recognition component of an E3 ubiquitin ligase, is a regulator of the HDR-based DSB repair machinery. Using functional genetic approaches, we find that prostate-specific SPOP mutants deregulate DSB repair by promoting the error-prone NHEJ pathway ( Figure 6). Importantly, after DSB induction, this loss of function leads to an increased sensitivity of mutant SPOP prostate cancer cells to PARP inhibition. These observations provide the first mechanism for the increased genomic instability of SPOP mutant prostate cancer, but also suggest that similar to other cancers with impaired HDR, this distinct class of cancer might benefit from treatment with clinically established DNA damaging therapeutics. This study therefore provides a rationale for hypothesis-based biomarker-driven clinical trials using PARP inhibitors or other DNA damaging agents in patients with prostate cancer.
Genomic rearrangements represent critical events deregulating prostate cancer genomes and driving tumorigenesis, although among individual cancer samples, there is marked variability in number and character of structural rearrangements. Here, we identified a genomically unstable subclass of primary prostate cancer characterized by dramatically increased intrachromosomal rearrangements. This subset of clinically localized, treatment-naive prostate cancers display a degree of genomic instability previously thought only to occur in late stage, metastatic tumors. Importantly, increased rearrangements likely result in higher total SCNA burden, which is associated with more aggressive clinical behavior Lalonde et al., 2014). Increased rearrangements were associated with mutations in SPOP, encoding the substraterecognition component of an E3 ubiquitin ligase complex. SPOP mutations are the most common point mutations in prostate cancer, but their role in promoting prostate cancer pathogenesis remains unclear. SPOP mutations occur early in the history of prostate cancer, based on clonality analysis reported here and presence in the prostate cancer precursor high grade prostatic intraepithelial neoplasia (HG-PIN) (Barbieri et al., 2012), potentially consistent with a 'gatekeeper' role in genome maintenance.
Consistent with increased intrachromosomal rearrangements in SPOP mutant cancers, in vivo data nominated a functional role for SPOP in DSB repair, similar to BRCA1. Interestingly, SPOP loss of function induced a developmental phenotype of neuronal degeneration and apoptosis; similar effects have been reported with other genes modulating DNA repair, including BRCA1 (Pulvers and Huttner, 2009).
SPOP encodes the substrate recognition component of an E3 ubiquitin ligase, and here we identified a role for SPOP in regulating the HDR-based DSB repair machinery. Notably, many established components of the DNA damage response are components of enzymes regulating ubiquitylation (Ceol et al., 2011). A number of substrates have been reported as deregulated by prostate cancer-derived SPOP mutations, but the relevance of these in vitro findings to human prostate cancers is still unclear (Geng et al., 2013(Geng et al., , 2014An et al., 2014;Theurillat et al., 2014;Zhang et al., 2014b). In contrast, we focused on the phenotypes observed in human prostate cancers harboring SPOP mutations. Using functional genetic approaches, we find that prostate-specific SPOP mutants deregulate DSB repair by promoting the error-prone NHEJ pathway ( Figure 4F).
The clinical importance of our findings is timely as potent PARP inhibitors, inducing synthetic lethality in cancers with alterations in DSB repair, are being evaluated and utilized in human cancers (Fong et al., 2009;Tutt et al., 2010;Schiewer and Knudsen, 2014). Here, we define a novel biomarker for increased genomic instability in clinically localized prostate cancer, provide the first mechanism by which SPOP mutations induce these alterations, and also suggest that similar to other cancers with impaired HDR, this distinct class of cancer may benefit from treatment with clinically established DNA damaging therapeutics, providing a rationale for genotype-based clinical trials.

Human cell lines
Cells were purchased from ATCC and cultured according to the manufacturer's instructions. All cells were maintained with 1% penicillin/streptomycin (15140-122, Gibco, Grand Island, NY, United States). Cells were treated with camptothecin (C9911, Sigma-Aldrich, St. Louis, MO, United States), olaparib (AstraZeneca, United Kingdom), veliparib (Selleckchem, United Kingdom), or cisplatin (Sigma-Aldrich) at the indicated concentrations. For SPOP expression, lentiviral vectors (pLenti6/V5-Dest Gateway vector, V496-10, Invitrogen, Grand Island, NY, United States) coding for mutant or wild-type SPOP were used. Viruses were titrated to equalize infectivity and protein expression. Viral titers were measured by a p24 ELISA (Lenti-X p24 rapid titer kit, PT5002-2, Clontech, Mountain View, CA, United States). Viral transduction was performed by using Polybrene transfection reagent (TR-1003-G, Millipore, Billerica, MA, United States) at a final concentration of 8 ng/μl. Stable cell lines for WT or mutant SPOP were generated by pLenti viral transduction and subsequent selection by cell sorting for RFP using an Aria-II cell sorter (BD Biosciences, San Jose, CA, United States). Stable selection was monitored over time by confirming RFP-and MYC-tagged SPOP expression.

Primary mouse prostate cells from conditional SPOP-F133V transgenic mice Gene targeting
All mouse studies were approved by the WCMC IACUC under protocol 2012-0065. To generate mice conditionally expressing flag-myc tagged human SPOP F133V pBTGinv (kind gift from Dr Yu Chen) was used to clone the cDNA subsequently to two inverted loxP sites, as previously described by Chen et al. (2013). This vector, pBigT-invloxP-SPOP F133V -IRES-nlsEFP was cloned into Rosa26-Pam1 (Addgene 15036). Vector was targeted into ES cells derived from albino C57B6 mice and injected into C57B6 blastocysts. Cre induction of SPOP-F133V was confirmed in ES cells. R26 F133V/+ mice were generated and confirmed though standard genotyping.

Primary mouse prostate cell isolation
Murine prostates were isolated from 10-week-old R26 F133V/+ mice and primary MPC lines were generated as described by Lukacs et al. (2010). Digested tissue was immediately mixed with growth factor reduced matrigel, plated in 36 drops and covered with organoid media as described by Karthaus et al. (2014), for 3-4 days. After 3-4 day of 3D culture cells were plated onto collagencoated plates and grown as 2D cultures.

Cre-induction and sorting
Stable MPC cells were transduced with a retrovirus expressing Cre-ERT2 (kind gift from Yu Chen) followed by puromycin selection. Following 48-hr treatment of the MPC cells with 1 μM 4-hydroxytamoxifen (Sigma-Aldrich) or vehicle, cells were sorted for GFP positivity using FACSAria II cell sorter (BD Biosciences). Vehicle-treated cells were sorted through a similar process.

In vitro drug sensitivity assay
Human prostate and mouse prostate-epithelial cells (1000-5000) were seeded into 96-well microtiter plates (CLS3610-48EA, Sigma Aldrich). After irradiation (5GY) cells were treated with olaparib or veliparib in a concentration range of 0.0001 μM-100 μM for 72-96 hr. DMSO was used as control. Cell viability was measured by using the CyQuant NF cell proliferation assay (Life Technologies,C35006) according to the manufacturer's instructions. Data normalization and nonlinear regression to calculate IC50 values was performed with Graph Pad Prism.

Clonogenic survival assays
Stable SPOP-expressing HEK293 or 22Rv1 cells were treated with cisplatin (24 hr, concentration as indicated) or olaparib (1 μM, 4 days). 50 cells were seeded per well of a 12-well plate, grown for 10-12 days, fixed and stained with ethanol (10%) and crystal violet (0.09%). Colonies containing 50 cells or more were counted, and the drug cytotoxicity was calculated as the ratio of colonies formed after drug treatment relative to colonies from untreated cells.

Single-cell gel electrophoresis
Single-cell gel electrophoresis (comet assay) was performed according to the manufacturer's instructions (Trevigen, Gaithersburg, MD, United States). 10 5 cells/ml were mixed with molten LM agarose in a ratio of 1:10 and spread on a comet slide. Cells were subsequently lysed and subjected to electrophoresis. Cells were fixed and stained with SYBR Green. Images were collected with a 20× objective lens using a monochrome camera (CV-M4+CL, Jai, Japan) fitted in an Olympus BX51 microscope (Olympus, Center Valley, PA, United States). COMETscore v1.5 was used for image analysis, and 50-100 individual cells were analyzed per condition. The results are shown as comet tail moments, which is a representation of the fluorescence intensity in the tail relative to the head.

DR-GFP (HDR) and Pem1-Ad2-EGFP (NHEJ) reporter assay
Cells were electroporated using the cell line nucleofactor kit V (VCA-1003, Lonza, Allendale, NJ, United States) with positive control plasmids (pNZE) and combinations of HDR-(pDRGFP) or NHEJ-(pPEM1-Ad2-EGFP) reporter constructs and an expression vector for the restriction enzyme I-SceI as described previously (Pierce et al., 1999). All the plasmids used in the GFP-reporter assay were a kind gift of Maria Jasin, Memorial Sloan Kettering Cancer Center, New York. The GFP expression induced by the positive control plasmid was used to normalize the electroporation efficiency. Negative controls contained the reporter construct and empty expression vectors lacking I-SceI. Cells were grown for 48 hr and processed for further flow cytometry analysis.

Cell cycle analysis
RWPE cells were transduced with LacZ, WT SPOP, or F133V SPOP lentiviral particles. Transduced cells were treated with Nocodazole (40 ng/ml) for 18 hr to synchronize cell cycle. Cell cycle distribution was determined by FACS 24 hr after culturing synchronized cells in normal growth media and labeling cells with propidium iodide (PI).

Zebrafish
Wild-type zebrafish (strain AB/T) were maintained as described previously (Ceol et al., 2011). For functional studies, embryos were harvested at the one cell stage and yolk injected with SPOPtargeting morpholino (final 2-4 ng) or SPOP mRNA at the indicated concentrations. For phenotypical analysis, embryos were monitored for 5 days. RNA was isolated using Trizol reagent (15596-026, Life Technologies) at 24 hpf for RNA sequencing after dechorionating the embryos with pronase. For cell cycle analysis, embryos were homogenized and individual cells were stained with propidium iodide. All protocols were performed with prior approval of the WCMC IACUC (2011-0026).

Morpholinos
Antisense oligos (morpholinos) targeting Spop were synthesized by Gene Tools (Corvalis, OR, United States). Morpholino 5 (MO5) was targeted to the sequence flanking the border between intron 1 and exon 2. Morpholino 7 (MO7) was targeted to the sequence flanking the border between intron 2 and exon 3. Morpholino sequences are shown in Supplementary file 1. Morpholino antisense oligos were dissolved in water to a final concentration of 1 mM. Primers used to validate morpholino-induced splicing of Spop are shown in Supplementary file 1.

In vitro mRNA transcription
Human SPOP mRNA for rescue experiments in the zebrafish was generated using the mMessage mMachine SP6 kit (AM1340, Life Technologies) according to the manufacturer's instructions. RNA was analyzed quantitatively and qualitatively using the Agilent 2100 bioanalyzer.

RNA sequencing
RNA was isolated from zebrafish embryos at 24 hpf using Trizol. 50 water-or morpholino-injected embryos were used. RNA libraries were created and sequenced using the GA2X or HighSeq genome analyzer (Illumina, United Kingdom) as described previously (Barbieri et al., 2012). The files first underwent quality control using FASTQC (v2_1.2.10). The vast majority of the reads from all the samples were of good quality (phred quality score >30). Therefore, no trimming or filtering of reads was required. The reads were then aligned with STAR (v2.3.0e) to zebrafish reference genome (Zv9).

Bioinformatic analyses
Genomic analysis: SCNA profiles for 402 prostate cancers were obtained through processing of highdensity oligonucleotide array data upon signal segmentation and then combined for genome-wide analysis for SPOP mutant and SPOP wild-type tumors. Clonality analysis for selected lesions was performed from sequencing data as in Baca et al. (2013).
Differential Gene Expression (DGE) and Gene Set Enrichment Analysis: After sorting and indexing, the aligned files were used with bedtools (v2.16.1) and Ensembl Zebrafish Transcriptome (v70) to generate read counts for genes for all samples. These read counts were used with DESeq (2_1.2.10) for differential expression. Orthology was also added to the differential expression table in this step. Read counts were also counts per million (cpm) normalized using EdgeR (v3.4.2) package and then pre-ranked using Log2 Fold Change. GSEA (v2-2.0.13) was run in pre-ranked mode to identify enriched signatures. Primers used for validation of the expression of selected genes are shown in Supplementary file 1.

Human samples
Clinically localized prostate cancers were selected for transcriptome sequencing as described (Barbieri et al., 2012). All samples were collected with informed consent of the patients and prior approval of the institutional review boards (IRB) of respective institutions. Additionally, the sequencing and data release of all transcriptome-sequenced samples was reviewed and approved by local IRB.

Statistics
The results are represented as means of at least three experiments (standard error of the mean (s.e.m.) are indicated by error bars). Student's t-test was used to calculate the statistical significance of differences in the HR-and NHEJ-reporter assays. Fisher's exact test (two sided) was used to assess the statistical significance of differences in the foci assays. Nonlinear regression was performed to determine the IC50 values in the in vitro drug sensitivity assays.

Accession codes
All RNA-seq data are deposited in the NCBI sequencing read archive (SRA) under the accession number SRP063952.