SF3B2-Mediated RNA Splicing Drives Human Prostate Cancer Progression

Published OnlineFirst August 20, 2019; DOI: 10.1158/0008-5472.CAN-18-3965

Cancer Molecular Cell Biology Research

SF3B2-Mediated RNA Splicing Drives Human Prostate Cancer Progression Norihiko Kawamura1,2, Keisuke Nimura1, Kotaro Saga1, Airi Ishibashi1, Koji Kitamura1,3, Hiromichi Nagano1, Yusuke Yoshikawa4, Kyoso Ishida1,5, Norio Nonomura2, Mitsuhiro Arisawa4, Jun Luo6, and Yasufumi Kaneda1

Abstract

Androgen receptor splice variant-7 (AR-V7) is a General RNA splicing SF3B2 complex-mediated alternative RNA splicing constitutively active AR variant implicated in U2 castration-resistant prostate cancers. Here, we show U2 snRNA that the RNA splicing factor SF3B2, identified by 3’ 3’ in silico and CRISPR/Cas9 analyses, is a critical 5’ 3’ splice site 5’ SF3B7 AR-V7 5’ A U2AF2 AGA Exon ? determinant of expression and is correlated SF3B6(p14) SF3B4 SF3B1 SF3B4 SF3B1 with aggressive cancer phenotypes. Transcriptome SF3B5 SF3B2 SF3B3 SF3B2 SF3B3 and PAR-CLIP analyses revealed that SF3B2 con- SF3A3 SF3B2 complex SF3A3 SF3A1 SF3A1 SF3b complex trols the splicing of target genes, including AR, to AR pre-mRNA drive aggressive phenotypes. SF3B2-mediated CE3 aggressive phenotypes in vivo were reversed by AR-V7 mRNA AR mRNA AR-V7 knockout. Pladienolide B, an inhibitor of CE3 a splicing modulator of the SF3b complex, sup- Drive malignancy pressed the growth of tumors addicted to high While the SF3b complex is critical for general RNA splicing, SF3B2 promotes inclusion of the target exon through recognizing a specific RNA motif. SF3B2 expression. These findings support the idea © 2019 American Association for Cancer Research that alteration of the splicing pattern by high SF3B2 expression is one mechanism underlying prostate cancer progression and therapeutic resistance. This study also provides evidence supporting SF3B2 as a candidate therapeutic target for treating patients with cancer.

Signiﬁcance: RNA splicing factor SF3B2 is essential for the generation of an androgen receptor (AR) variant that renders prostate cancer cells resistant to AR-targeting therapy. Graphical Abstract: http://cancerres.aacrjournals.org/content/canres/79/20/5204/F1.large.jpg.

Introduction Splicing of precursor mRNA (pre-mRNA) is a critical process involving the alternative use of exons/introns leading to diverse mature mRNAs from a limited numbers of genes (1). Recent 1Division of Gene Therapy Science, Osaka University Graduate School of Med- icine, Suita, Osaka, Japan. 2Department of Urology, Osaka University Graduate genome-wide analyses of cancer transcriptomes have revealed School of Medicine, Suita, Osaka, Japan. 3Department of Otorhinolaryngology- globally aberrant splicing profiles including exon skipping and Head and Neck Surgery, Osaka University Graduate School of Medicine, Suita, intron retention in mature mRNAs (2, 3). Although the molecular Osaka, Japan. 4Graduate School of Pharmaceutical Sciences, Osaka University, mechanisms underlying aberrant splicing in cancer are largely 5 Suita, Osaka, Japan. Department of Gynecology, Osaka University Graduate unknown, one of the causes of aberrant splicing in cancers may be School of Medicine, Suita, Osaka, Japan. 6James Buchanan Brady Urological mutations in splicing factors including SF3B1, U2AF1, SRSF2 and Institute and Department of Urology, Johns Hopkins University School of Medicine, Baltimore, Maryland. ZRSR2, as well as altered expression of splicing factors, such as RBFOX2, MBNL1/2, and QKI (2, 4). Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). Prostate cancer is one of most frequently detected cancers in men. Although the majority of localized prostate cancers are fi N. Kawamura and K. Nimura are the co- rst authors of this article. curable, progressive and metastatic prostate cancer contributes Corresponding Authors: Keisuke Nimura, Osaka University Graduate School of to approximately 307,000 cancer-related deaths each year world- Medicine, Suita, Osaka 565-0871, Japan. Phone: 81-6-6879-3901; Fax: 81-6-6879- wide (5). Androgen deprivation therapy (ADT) is the first-line 3909; E-mail: [email protected]; and Yasufumi Kaneda, therapy for men with metastatic prostate cancer, but almost all [email protected] men develop castration-resistant prostate cancer (CRPC) after – Cancer Res 2019;79:5204 17 first-line ADT. Patients with CRPC may be treated with additional doi: 10.1158/0008-5472.CAN-18-3965 hormonal therapies, including abiraterone acetate and enzaluta- 2019 American Association for Cancer Research. mide, which are newly developed inhibitors of androgen receptor

5204 Cancer Res; 79(20) October 15, 2019

SF3B2-Mediated RNA Splicing in Prostate Cancer Progression

(AR) signaling with proven survival benefit in patients with volume reached approximately 100–300 mm3 for 22Rv1 or CRPC (6–9). A significant proportion of men with CRPC are approximately 200–400 mm3 for LNCaP95, the pladienolide B resistant to abiraterone acetate and enzalutamide, and nearly all derivative (5 mg/kg) or vehicle was administered intraperito- men develop acquired resistance to these agents over a period of neally to each mouse on days 0, 2, 4, and 6, as described 1–2 years. Because both abiraterone acetate and enzalutamide previously (12) with minor modification. The relative tumor exhibit anticancer effects through inhibiting of the ligand-binding volume was calculated as the ratio between the tumor volume domain at the C-terminus of the AR, they may not suppress AR at time t and the tumor volume at the start of treatment. signaling mediated by truncated AR splice variants (AR-V) lacking the ligand-binding domain (10). Among the many AR-Vs that Flow cytometry analysis and cell sorting have been characterized (11), AR-V7 is most compatible with AR-V7-GFP cells were cultured in medium containing detection due to its high frequency and abundance relative to 0.5 mg/mL puromycin for selection after transfection with guide other AR-Vs. Detection of AR-Vs has been associated with aggres- RNA (gRNA) targeting to genes of interest or control gRNA for sive prostate cancers and CRPC progression (10). These findings 5 days. FACSAria II (BD Biosciences) was used for flow cytometry. suggest a critical role of splicing in modulating the activity of AR, a At least 50,000 events were collected per sample. After negative key prostate cancer drug target associated with CRPC progression. selection for doublet cells, the remaining gated cells were ana- However, key splicing factors that are critical for AR-V7 generation lyzed and sorted on the basis of GFP expression levels. Dead cells have not been definitively identified and characterized. were stained by propidium iodide (Dojindo) and Annexin V In this study, we identified SF3B2 (also known as SF3b145 or (BioLegend). The data were analyzed using FlowJo software SAP145) as a positive splicing regulatory factor for AR-V7 by (Version 10). in silico analyses of RNA-binding factors associated with AR-V7 expression using transcriptome data from prostate cancer patients Quantitative RT-PCR followed by examination of AR-V7 expression after the candidate RNA was isolated using the Isogen RNA extraction kit (Wako splicing factors were knocked out by the CRISPR/Cas9 system. Pure Chemicals). cDNA was synthesized from total RNA using the By genome-wide RNA splicing analysis and photoactivatable High-Capacity cDNA Reverse Transcription Kit (Applied Biosys- ribonucleoside-enhanced cross-linking immunoprecipitation tems). qPCR was performed using the SYBR Green PCR Master (PAR-CLIP), we show that SF3B2 controls the inclusion of Mix (Applied Biosystems) and the CFX384 Real-Time System SF3B2-bound exons and an exclusion of SF3B2-bound introns. (Bio-Rad). The primers used in this study are listed in Supple- Moreover, pladienolide B, an inhibitor of a splicing modulator mentary Table S1. Genomic DNA contamination was examined SF3B complex, repressed SF3B2-addicted tumor growth under by evaluating reverse transcription reaction samples lacking castration conditions in vivo. Collectively, these results indicate reverse transcriptase. that SF3B2 is a critical determinant of RNA splicing and gene RNA-seq expression patterns and controls the expression of key genes Sequencing libraries from at least two biological replicate associated with CRPC progression, such as AR-V7. RNA samples were prepared using NEBNext Ultra RNA Library Prep Kits for Illumina (NEB) following the manufacturer's Materials and Methods instructions. Cell culture and transfection RNA precipitation using Halo-tag The 22Rv1 prostate cancer cell line was purchased from the The Halo-tag was knocked-in immediately before the stop ATCC. LNCaP95 was a gift from J. Luo (Johns Hopkins University codon of the SF3B2 gene in 22Rv1 cells. The total cell extracts School of Medicine, Baltimore, MD). Cell line authentication was of 5 106 SF3B2-Halo knocked-in cells were incubated with fi Mycoplasma not performed. All cells were con rmed to be negative 50 mL of Halo-Link Resin (Promega) overnight at 4 C to precip- before inoculation of cells into mice (TaKaRa). The length of time itate SF3B2–RNA complexes. Contaminated genomic DNA was between thawing and use is less than 3 months. The detailed removed from the extracted RNA using 1 ml of TURBO DNase method is described in extended experimental procedures. (Thermo Fisher Scientific). Enrichment of RNAs was determined by quantitative RT-PCR. Xenograft prostate cancer model All experiments using mice was approved by Osaka Univer- PAR-CLIP sity Animal Experiments committee and was performed fol- PAR-CLIP was performed according to the original proto- lowing the guidelines. The in vivo tumor growth of human col (13) and another protocol with some modifications (14). prostate cancer cells was determined using a subcutaneous The antibodies used in this study are listed in Supplementary transplant xenograft model. Male NOD/SCID mice (Charles Table S2. The detailed method is described in extended experi- River) were castrated surgically at 7 weeks of age. Cancer cells mental procedures. (2 106 cells) in a PBS/Matrigel mixture were injected sub- cutaneously into these castrated NOD/SCID mice at 8 week of Tandem affinity purification and mass spectrometry age under deep anesthesia, and the mice were maintained in a The SF3B2-TAP–expressing prostate cancer cell line was gen- temperature-controlled and pathogen-free room. All animals erated from 22Rv1 cells. The SF3B2–TAP complex was purified were handled according to approved protocols and the guide- from nuclear extracts by TAP technology as described previous- lines of the Animal Committee of Osaka University (Osaka, ly (15). The purified proteins were precipitated with trichloroa- Japan). The tumor size was measured weekly, and the tumor cetic acid and separated on SDS-PAGE gels. Protein bands stained volume was calculated according to the following formula: with silver were excised from the gel, in-gel digested with trypsin, tumor volume (mm3) ¼ length (width)2/2. When the tumor and analyzed by LC/MS-MS.

www.aacrjournals.org Cancer Res; 79(20) October 15, 2019 5205

Kawamura et al.

Bioinformatics and AR-V7–positive prostate cancers (Supplementary Fig. S2E Deep-sequencing data are available under DRA accession and S2F) and AR-V7-GFP negative cells were increased by U2AF2 DRA006189, which includes the following: 3 PAR-CLIP libraries gRNA transfection (Supplementary Fig. S2B), the correlation and 8 RNA-seq libraries. The detailed methods are described in between U2AF2 expression and AR-V7 expression in the RNA-seq extended experimental procedures. data of CRPCs was not consistent compared with SF3B2 (Sup- plementary Fig. S2F). These results support the conclusion that SF3B2 is a strong candidate splicing factor involved in the regu- Results lation of AR-V7 expression. SF3B2 is identified as a splicing factor associated with AR-V7 Increased SF3B2 expression was detected in CRPCs and pros- expression by in silico analysis of transcriptome data and by tate cancers with high AR-V7 expression (Fig. 1F). SF3B2 expres- disruption of candidate genes by CRISPR/Cas9 in prostate sion was also increased in prostate cancers with high Gleason cancer scores (8–10), a pathologic indicator of prostate cancer malig- Because increased AR-V7 expression has been detected in nancy, consistent with a similar pattern of AR-V7 expression CRPCs (16), we first sought to identify splicing factor– (Fig. 1G). In contrast, SF3B2 expression did not correlate with encoding genes with increased expression in CRPCs and AR- full-length AR expression (Supplementary Fig. S2G). Notably, V7–positive prostate cancers (Fig. 1A and B). Using previously mutations in SF3B2 did not affect AR-V7 expression or the published microarray-based gene expression data in patients with Gleason score (Supplementary Fig. S2H). Higher SF3B2 expres- benign, localized prostate cancers and CRPCs (17), we compared sion was associated with poor progression-free survival in pros- the expression level of 309 genes encoding splicing factors and tate cancer (P ¼ 0.0042 by log-rank test; Fig. 1H). Interestingly, RNA-binding proteins, selected by gene ontology (GO) terms in high SF3B2 expression was also associated with poor overall AmiGO (Fig. 1A; ref. 18). This set of genes differentiated the three survival in bladder cancer (P ¼ 0.0237), acute myeloid leukemia sample groups by cluster analysis, with a general trend toward an (AML; P ¼ 0.0316), lung adenocarcinoma (P ¼ 0.0109), head and upregulation in CRPCs compared with localized prostate cancers neck squamous cell carcinoma (P ¼ 0.0491), and breast cancer and benign prostate samples (Fig. 1A). We further examined this (P ¼ 0.0441; Supplementary Fig. S3). Thus, high SF3B2 expres- subset of genes according to AR-V7 expression levels in the The sion was associated with aggressive phenotypes of various human Cancer Genome Atlas (TCGA) prostate cancer dataset (19). Using cancers, including prostate cancer, and SF3B2 may regulate the this in silico approach, we identified 21 genes encoding splicing expression of AR-V7 as well as other genes involved in cancer factors that were upregulated in both CRPC and TCGA prostate progression. tumors with high AR-V7 expression (Fig. 1B). To determine the functional effect of the candidate splicing Depletion of SF3B2 decreases the expression of AR-V7 and factors on AR-V7 expression, we generated CWR22Rv1 (hereafter androgen-responsive genes in human prostate cancer cells called 22Rv1) prostate cancer cells with stable knock-in of GFP To examine whether SF3B2 regulates AR-V7 in cells with intact immediately before the AR-V7 stop codon in AR cryptic exon 3 AR-V7 sequences, AR-V7 protein expression in SF3B2 gRNA– (AR-V7-GFP cells) using CRISPR/Cas9 (Fig. 1C; Supplementary transfected prostate cancer cells was analyzed in 22Rv1 and Fig. S1). The GFP cassette does not include polyA sequence to LNCaP95 cells, another prostate cancer cell line expressing endog- avoid affecting the use of the cryptic exon 3 polyA site (20). 22Rv1 enous AR-V7 (23). Transfection of SF3B2 gRNA decreased AR-V7 cells have a tandem duplication of exon 3 and cryptic exon 3 in AR protein expression in both cells, concomitantly correlating with gene (Fig. 1C; ref. 21). To avoid unexpected effect on further reduced levels of SF3B2 (Fig. 2A and B). analysis, we selected a clone that lost the tandem duplication To further confirm the role of SF3B2 in the regulation of AR-V7, (Fig. 1C; Supplementary Fig. S1A–S1C). Fusion of GFP with AR- we used the AR-V7-GFP signal as a marker for isolating SF3B2- V7 was confirmed by Western blotting (Supplementary Fig. S1A depleted cells in SF3B2 gRNA–transfected AR-V7-GFP cells, since and S1B) and FACS analysis of AR-V7 gRNA–transfected AR-V7- SF3B2 knockout cells could not be established and siRNAs against GFP cells (Supplementary Fig. S1D). Full-length AR (AR-FL) was SF3B2 could not efficiently decrease SF3B2 to examine its func- consistently expressed in AR-V7-GFP cells, compared with wild- tion. SF3B2 protein was significantly depleted in AR-V7-GFP– type 22Rv1 cells, while AR-V7 expression was increased in AR-V7- negative cells in SF3B2 gRNA–transfected AR-V7-GFP cells (top), GFP cells (Supplementary Fig. S1B and S1E). It may be due to while AR-V7-GFP–positive cells (top) had comparable amounts increasing stability of AR-V7 RNA, because AR-FL expression was of SF3B2 relative to control cells (Fig. 2C and D). Thus, we defined consistent (Supplementary Fig. S1B and S1E). We could not detect the isolated AR-V7-GFP–negative cells from SF3B2 gRNA– AR-V7-GFP protein by AR-V7–specific antibody (Supplementary transfected AR-V7-GFP cells as SF3B2-depleted cells. Indeed, Fig. S1B), although we could detect the protein by GFP antibody, SF3B2-depleted cells lost not only SF3B2 but also AR-V7 protein FACS analysis, or qPCR (Supplementary Fig. S1A, S1D, and S1E). expression, while a weak decrease in AR-FL protein was detected An addition of GFP at the end of AR-V7 may prevent the AR-V7– (Fig. 2E). In contrast, AR-V7 depletion did not diminish specific antibody from recognizing the C-terminus region of AR- the amounts of SF3B2 protein, suggesting that AR-V7 was V7. This cell line allowed for the examination of endogenous AR- not involved in the upregulation of SF3B2 expression (Fig. 2F). V7 expression through the evaluation of GFP. Transfection of Collectively, these results support a key role for SF3B2 in the gRNAs against the 21 candidate genes revealed that a distinct AR- regulation of AR-V7 expression. V7-GFP–negative population that was the most striking in We next performed genome-wide mRNA expression analysis to SF3B2 gRNA–transfected cells without cell death (Fig. 1D and determine whether SF3B2 depletion led to altered expression of E; Supplementary Fig. S2A–S2D). U2AF2 (also known as androgen-responsive genes (Fig. 2G and H; Supplementary U2AF65) is known to be correlated with AR-V7 expression (22). Fig. S4A). SF3B2 depletion resulted in the decreased expression Although an increase in U2AF2 expression was detected in CRPCs of 67 out of 100 androgen-responsive genes belonging to the gene

5206 Cancer Res; 79(20) October 15, 2019 Cancer Research

SF3B2-Mediated RNA Splicing in Prostate Cancer Progression

A B Genes encoding splicing factors and RNA-binding proteins (309 genes) 468 2 −4−20 Microarray data set (n = 122) TCGA data set (RNA-seq n = 498) Grasso et al. (17) AR-V7 low PC vs highly positive PC Localized prostate cancer vs mCRPC Genes whose expression are increased in AR-V7 Genes whose expression are increased in CRPC highly positive PCa

41 Genes 21 Genes 12 Genes

C AR-FL AR-V7 1 234567 8 1 23 Cryptic exon 3

UGA AUG UGA AUG AR in 22Rv1 Cells Cryptic exon 3 Cryptic exon 3 Exon 1 2343 5678 5’ 3’ 309 Genes encoding splicing factors and RNA-binding proteins Tandem duplication CRPC Localized PC Stop codon Benign Stop codon 5’ CE3 3’ 5’ CE3 3’

5’ Homology arm 3’ Homology arm Cryptic exon 3 Knock-in Exon 1 2 3 4 5 6 7 8 5’ 3’ GFP T2A Neo R 3’ GFP-2A-Neo R Stop codon

D Control gRNA E F P < 0.0001 15,000 SF3B2 gRNA 40 P = 0.0083 P < 0.0001 100 35 13 P = 0.6047 30 12,000 80 25 12 9,000 60 20 15

Count 40 10 11 6,000 SF3B2 Expression 20 5 3,000 Percentage of AR-V7– negative population 0 10 0 0 1 2 3 10 10 10 10 Benign CRPC

AR-V7-GFP Localized PC Control gRNA AR-V7 NegativeAR-V7 Positive SF3B2 gRNASF3B2 #1 gRNASF3B2 #2 gRNA #3 AR-V7 High Positive

G H Prostate cancer (TCGA) P < 0.0001 3,000 P = 0.9856 P < 0.0001 1.0 SF3B2 High (n = 242) P = 0.0069 P = 0.0718 P = 0.9208 16,000 7.5 n P = 0.0044 0.8 SF3B2 Low ( = 242) 2,000 12,000 5.0 0.6 P = 0.9844 1,000 8,000 2.5 0.4 P = 0.0042 AR-V7 Expression AR-FL Expression SF3B2 Expression 0 4,000 0 0.2 Progression-free survival 678−10 678−10 678−10 0.0 Gleason score Gleason score Gleason score 0 20 40 60 80 100 120 140 160 180 Postoperative months

Figure 1. Identiﬁcation of SF3B2 as a splicing modulating factor associated with AR-V7 expression. A, Heatmap of 309 genes encoding splicing factors or RNA-binding proteins. Published microarray data for prostate cancers were reanalyzed (17). Specimen types were color-coded and included CRPC (n ¼ 35), localized prostate cancer (PC; n ¼ 59), and benign prostate tissue samples (n ¼ 28). B, Schematic of the screening for candidate genes associated with the splicing for AR-V7. C, Schematic illustration of AR-FL and AR-V7 splicing and AR-V7-GFP knock-in in CWR22Rv1 prostate cancer cells. D, FACS analysis of AR-V7-GFP expression in control or SF3B2 gRNA-transfected AR-V7-GFP cells. Light blue, SF3B2 gRNA; red, control gRNA; dark blue, overlap area of SF3B2 gRNA and control gRNA. E, Percentages of AR-V7-GFP–negative cells following transfection with control gRNA and SF3B2 gRNA (n ¼ 3). F, Correlation between SF3B2 expression and CRPC in published microarray data (17) and correlation between SF3B2 and AR-V7 expression in the TCGA data (19). AR-V7 negative [n ¼ 437, reads per kilobase million (RPKM) < 1], AR-V7 positive (n ¼ 49, RPKM >1), and AR-V7 high positive (n ¼ 12, RPKM > 3). G, Correlation between SF3B2/AR-FL (full length)/AR-V7 expression and Gleason scores in TCGA data (19). Gleason score 6 (n ¼ 45), 7 (n ¼ 247), and 8–10 (n ¼ 205). H, Kaplan–Meier curves of progression-free survival in TCGA prostate cancer patients with SF3B2 high or low expression divided by the median. log, log-rank test.

www.aacrjournals.org Cancer Res; 79(20) October 15, 2019 5207

Kawamura et al.

ABC22Rv1 LNCaP95 Control gRNA SF3B2 gRNA 250K

200K

150K

Control-gRNA SF3B2-gRNA FSC Control-gRNA SF3B2-gRNA (bulk) (bulk) (bulk) (bulk) 100K SF3B2 SF3B2 10.2 10.8 50K Control Lower Upper AR-FL AR-FL 11.1 11 0 AR-V7 AR-V7 101 102 103 104 101 102 103 104 1 0.06 10.7 ACTB ACTB AR-V7-GFP 11 11

DEF

Control SF3B2-Depletion Control AR-V7 KO AR-FL SF3B2

Control-gRNA SF3B2-gRNA (upper) SF3B2-gRNA (lower) SF3B2 AR-V7 ACTB SF3B2 ACTB ACTB

GH0.20 0.15 0-100 0.10 0.05 0.00 Control 0-100

0-100

RNA-seq SF3B2 dep 0-100 Log10 FPKM + 1 2 TMPRSS2 TMPRSS2 1 KLK3 0-300

Control 0-300 NKX3-1 KLK4 0-300

RNA-seq SF3B2 dep 0-300

Control KLK3

SF3B2 Depletion

Figure 2. SF3B2 depletion decreases the expression of AR-V7 and androgen-response genes. A and B, Western blots of the indicated proteins in SF3B2 gRNA-transfected CWR22Rv1 cells (A) and LNCaP95 cells (B). The intensity of each band was normalized to that of ACTB. C, FACS analysis of SF3B2 gRNA-transfected AR-V7-GFP cells showing the "upper" and "lower" cell populations. D, Western blots of lower SF3B2 protein expression in AR-V7-GFP cells with "lower" GFP expression compared with those with "upper" GFP expression. E, Western blots of the indicated proteins in SF3B2-depleted AR-V7-GFP cells. F, Western blots of SF3B2 in AR-V7–depleted AR-V7-GFP cells. G, Heatmap of expression data showing decreased expression of the AR-target genes in SF3B2-depleted or nontarget gRNA-transfected AR-V7-GFP cells. H, TMPRSS2 and KLK3 mRNA expression in SF3B2-depleted AR-V7-GFP cells.

5208 Cancer Res; 79(20) October 15, 2019 Cancer Research

SF3B2-Mediated RNA Splicing in Prostate Cancer Progression

set of NELSON_RESPONSE_TO_ANDROGEN_UP (24), includ- CLIP signals of SF3B2 were detected across the meta-exon or ing KLK3 [also known as prostate-specific antigen (PSA)] and -intron regions (Fig. 4D and E). The SF3B2 signals were stronger transmembrane protease serine 2 (TMPRSS2), but not ACTB in exons than in introns, possibly reflecting that the removed (Fig. 2G and H; Supplementary Fig. S4C). Thus, SF3B2 depletion introns were degraded. Notably, transcripts with SF3B2-bound in prostate cancer cells led to significant decreases in the expres- exons were significantly decreased by SF3B2 depletion (P ¼ sion of AR-V7 and the majority of androgen-responsive genes. 1 10 11, Fig. 4F), while transcripts with SF3B2-bound introns demonstrated increased intron retention (P ¼ 2 10 15, PAR-CLIP analysis identifies global SF3B2-binding sites Fig. 4G). Collectively, these results support a critical role for To investigate the genome-wide molecular functions of SF3B2, SF3B2 in the inclusion/exclusion of exons and introns. we determined the binding sites of SF3B2 by PAR-CLIP analysis (13). A single band at the expected size was detected in PAR- The SF3B2-recognizing motif in cryptic exon 3 is critical for the CLIP of tandem affinity purification-tagged SF3B2 (SF3B2-TAP) generation of the AR-V7 transcript from whole-cell extracts of 22Rv1 cells with stable SF3B2-TAP To further demonstrate the critical role of SF3B2 binding to expression (Fig. 3A). SF3B2-bound RNAs from three independent cryptic exon 3 (as shown in Fig. 3) in the splicing of AR-V7, we experiments were analyzed by deep sequencing, yielding approx- designed two gRNAs for the SF3B2-recognizing motif to disrupt imately 460 M of reads in total. SF3B2-binding sites were found the SF3B2 binding sequence (Fig. 5A). While control gRNA not only in coding regions, including exons and introns, but also recognizing the sequence between the cryptic exon 3 and the intergenic regions including promoters (Fig. 3B; Supplementary SF3B2-binding site did not increase the GFP-negative population Fig. S4D and S4E). in AR-V7-GFP cells compared with the control gRNA that does not To discover the SF3B2-recognizing RNA motif, the reads with a contain the matched sequence in the human genome (Figs. 1E T-to-C conversion by the PAR-CLIP reaction were analyzed by and 5B and C), both gRNA #1 and #2 targeting the SF3B2- PARalyzer (25) and cERMIT (26). cERMIT identified the CNNGU recognizing motif significantly increased in the AR-V7-GFP– RNA sequence as the SF3B2-recognizing RNA motif (P < 1 10 8, negative population (P ¼ 0.0001 or P < 0.0001; Fig. 5C). The compared with 1,000 random scores), following analysis of 3,154 data indicate that the SF3B2-binding site is necessary for inclusion clusters identified by PARalyzer and the exclusion of clusters of cryptic exon 3 into AR mRNA, that is, the generation of the assigned to repeats and rRNA (Fig. 3C and D). mature AR-V7 transcript. Notably, SF3B2 bound to AR exon 1 and cryptic exon 3, the We next examined splicing between exon 3 and cryptic exon inclusion of which would lead to the AR-V7 transcript (Fig. 3E). 3 or between exon3 and exon4 in AR by quantitative PCR SF3B2 depletion significantly decreased transcription from these (qPCR) in the AR-V7-GFP–negative population isolated from exons (Fig. 3E), although SF3B2 depletion did not significantly the SF3B2 core motif in cryptic exon 3–targeting gRNA- affect the expression of AR-FL protein (Fig. 2E). This difference transfected AR-V7-GFP cells (Fig. 5D–F). Splicing between exon might be caused by the stability of AR-FL protein. To examine 3 and cryptic exon 3 was significantly decreased in the isolated whether endogenous SF3B2 binds to the SF3B2-binding sites in AR-V7-GFP–negative population, while splicing between exon AR, we established 22Rv1 cells carrying a Halo-tag knock-in 3 and exon 4 was not affected by disruption of the SF3B2- immediately before the stop codon of SF3B2 (Fig. 3F). RNA binding site in cryptic exon 3 (Fig. 5E). Conversely, SF3B2 precipitation analysis of SF3B2-Halo revealed an enrichment of overexpression increased splicing between exon 3 and cryptic RNAs in AR exon1 and cryptic exon 3 (Fig. 3F). Thus, PAR-CLIP exon 3 without having a significant effect on splicing between analysis of SF3B2 revealed the genome wide SF3B2-binding sites, exon 3 and exon 4 (Fig. 5F). Thus, SF3B2 binding to AR cryptic and confirmed SF3B2 binding to AR pre-mRNA including cryptic exon 3 of the AR plays a key role in the inclusion of this target exon3, a critical exon for AR-V7. exon into AR-V7 mRNA.

SF3B2 regulates the splicing and expression of target genes SF3B2 overexpression enhances tumor growth in vivo and SF3B2 is a component of the SF3B1 (also known as SF3b155 increases the expression of AR-V7 and its target genes or SAP155) complex, known as SF3b, which functions as a Because high SF3B2 expression is associated with aggressive component of U2-small nuclear ribonucleoprotein (snRNP) in cancer phenotypes in patients (Fig. 1; Supplementary Fig. S3), we pre-mRNA splicing (27–29). We next explored the genome- next examined whether SF3B2 overexpression could alter phe- wide functions of SF3B2 in the regulation of splicing and gene notypes in prostate cancer cells. Consistent with the results expression. SF3B2 depletion affected approximately one-fourth showing that SF3B2 increased AR-V7 splicing (Fig. 5), SF3B2 of the detected splicing events (Supplementary Fig. S5A and overexpression increased AR-V7 protein expression in two pros- S5B) and increased intron retention (IR) and exon skipping tate cancer cell lines, while the full-length AR protein was not (ES) (Supplementary Fig. S5C and S5D). The expression of signiﬁcantly affected (Fig. 6A–C). Because AR-V7 promotes tumor SF3B2-target genes was decreased according to the location and growth in vivo, along with an increase in its target gene expres- number of SF3B2-binding sites (Fig. 4A and B), implicating sion (23, 30), we examined the effect of SF3B2 overexpression on that SF3B2 binding to an intron iscriticaltocontrolling gene tumor growth in vivo under an androgen-depleted conditions expression, whereas the peak height of the SF3B2 binding site (Fig. 6D and E). We demonstrated that SF3B2 overexpression (xRPM) was not associated with a decrease in expression of signiﬁcantly promoted tumor growth in 22Rv1 and LNCaP95 SF3B2 target genes (Fig. 4C). cells under castration conditions (P < 0.05, Fig. 6D and E). To examine whether SF3B2 modulates both an inclusion of Notably, SF3B2-mediated aggressive cancer phenotype was exons and an exclusion of introns in the SF3B2-binding sites, we reversed by AR-V7 knockout (Fig. 6D and E; Supplementary performed meta-exon or -intron analysis of transcripts or PAR- Fig. S6), indicating that regulation of AR-V7 splicing is one of CLIP signals of SF3B2 at SF3B2-bound sites (Fig. 4D–G). PAR- the critical roles of SF3B2 in CRPC.

www.aacrjournals.org Cancer Res; 79(20) October 15, 2019 5209

Kawamura et al.

A TAP B Other 2.1% Promoter 9.3% 5’UTR 0.75% Exon 9.5% GFP-TAP SF3B2-TAP Intergenic 39.5% 150 kDa Intron 31.8%

100 kDa TTS 3’UTR 3.4% 75 kDa 3.6%

50 kDa C 2.0 P < 1 × 10-8 37 kDa 1.5

Bits 1.0

25 kDa 0.5 0.0 SF3B2 1 2345 D Gene name Representative PAR-CLIP clusters xRPM FPKM AR ------AGCAACUGUGUCUGUCUGAGGUUCCUGUGGCCAUCUUU------9.0 100.9 WDR74 ------CCACUCCACGCAUCGACCUGGUAUUGCAGUACUUCC------139.0 25.1 NEMF --UUGCCUAAGGAGGGGUGAACCGGCCCAGGUCGGAAACGGAGCAGGUCAAAACUCCCGUGCUGAUCAGUAGUGGGAUUGCGCCUGUG- 157.0 8.0 TRAF4 ------AAGCUGGCAAUGGCACGGCAUGUGGAGGAGAGUGUGA------12.9 55.7 POLR3GL ------AGAUGUCAGGUCCGAUUGA------13.9 11.7 MALAT1 ------AGACAGGUGGGAGAUUAUGAUCA------22.9 332.5 NOTCH1 ------AGGAGGCCGCCGCCCGGGCGCAGAGGGCAGCCGGUGGGGAGGCA------79.2 18.6 EIF4G3 ------UUCUCGGGUUGGAAUUU------13.7 12.4 RAB3B ------CCUGGUUAGUACUUGGAUGGG------1.8 5.4 ZC3H3 UUGAUUCGGCUGAUCUGGCUGGCUAGGCGGGUGUCCCCUUCCUUCCUCACUGCUUCAUGUGCGUCCCUCCCGAAGCU------420.7 11.7 MATR3 ------UACACUGCUGGGAGAACAGCAGCCAAUAGCUGGUUGGCAUUCUGGCCCUGGUU------12.4 97.9 NAGLU ------AGGUGCUGGCUAGUGACAGCCGCUUCUUGCUGGGCA------55.6 23.3 SETD1B ------UCUAGAGCCUCGUGGGUAACGGCAUAGUGUGGCCAGGUGUGUAGGGAAUG------44.1 22.5 RNF187 ------AACUCAGGUCUUCAGGGAGAG------18.4 128.7 ATP5O ------GUUCAGGUAUACGGUAUUGAAGGUC------6.9 32.5 CCNK ------AGUAGGAACUGGUAUGGAUG------18.3 16.7 RIT1 ------GAGGGGUCACAGGUCAGUGCCGGAGCCUCCGCGAGUGAAGGAA 3.8 22.0 ANKS3 ------ACUCCCCAGCCCCCAGCCCUCCUCAUCCUUGUC------2.4 3.8 DHX34 ------CUGGCCCUGGCUGGUGCCACACACAGGCCU------3.0 14.6 MLL5 ------AAAGACAUCAAAUUAAUCUUGUGAAUGCAG------2.4 17.5

AR E F F1 F1 F1

0-3,000 0-3,000 Exon 1 2 3 4 5 6 7 8 9 Control 0-3,000 0-3,000 5’ 3’

0-3,000 0-3,000

RNA-seq SF3B2 R1 Cryptic exon 3 R1 R1 dep 0-3,000 0-3,000 * 0-100 0-2,000 2.5 SF3B2 PAR CLIP 0-100 0-2,000 2.0 AR 1.5 Exon 1 Exon 3 Cryptic exon 3 1.0 0.5

Relative SF3B2 binding 0 ex1 CE3 ex8 ex1 CE3 ex8 SF3B2-Halo + −

Figure 3. Identiﬁcation of SF3B2-binding RNA regions. A, Top, autoradiograph of 32P-labeled RNA crosslinked with TAP-tagged SF3B2 (SF3B2-TAP) protein separated by SDS-PAGE. Bottom, Western blot analysis of SF3B2. B, Pie chart of approximately 63,000 SF3B2-binding sites with T-to-C mutations above threshold. C, The SF3B2-recognizing RNA motif. A total of 3,154 clusters excluding repeats and rRNA were analyzed. D, Sequence alignment of representative SF3B2-binding sites. Gray characters indicate the SF3B2-recognizing motif. E, mRNA expression and SF3B2 PAR-CLIP signals in AR. F, RNA precipitation analysis of endogenous SF3B2 with Halo-tag knock-in. Top, primer positions. Bottom, enrichment of RNAs by RNA precipitation of SF3B2-Halo. , P < 0.05; n ¼ 3. FPKM, fragments per kilobase of transcript per million mapped reads; xRPM, crosslinked-reads per million.

Because SF3B2 overexpression increased AR-V7 splicing, we (P < 2 10 16, Fig. 6F; Supplementary Fig. S7A) and promoted next examined whether SF3B2 overexpression promoted genome- the splicing efﬁciency of SF3B2-bound introns (P < 2 10 16, wide changes in splicing. Consistent with AR-V7 splicing, SF3B2 Fig. 6G). SF3B2 depletion led to a decrease in SF3B2-bound gene overexpression increased the inclusion of SF3B2-bound exons expression, while SF3B2 overexpression led to both a decrease and

5210 Cancer Res; 79(20) October 15, 2019 Cancer Research

SF3B2-Mediated RNA Splicing in Prostate Cancer Progression

n P ABC-7 1 Binding site ( = 2,987) = 1.00 Intron (n = 4,522) P = 2.4 × 10 2 Binding sites (n = 1,737) P = 0.00552 Exon (n = 1,049) P = 1.00 3 Binding sites (n = 953) P = 0.00079 < 1 xRPM (n = 6,580) P = 2.4 × 10-9 Intergenic region (n = 1,885) P = 1.00 4 Binding sites (n = 524) P = 1.2 × 10-6 1-10 xRPM (n = 771) P = 1.00 Intron and Exon (n = 1,314) P = 0.00048 ≥5 Binding sites (n = 687) P = 9.0E-10 > 10 xRPM (n = 400) P = 0.851 Control (n = 4,511) Control (n = 6,396) Control (n = 6,396) 1.00 1.00 1.00

0.75 0.75 0.75

0.50 0.50 0.50

0.25 0.25 0.25 Cumulative distribution Cumulative distribution Cumulative distribution 0.00 0.00 0.00

0.0 2.5 5.0 0.0 2.5 5.0 -5.0 -2.5 0.0 2.5 5.0 -5.0 -2.5 -5.0 -2.5 Log2 -fold change of gene expression Log2 -fold change of gene expression Log2 -fold change of gene expression (SF3B2 depletion/control) (SF3B2 depletion/control) (SF3B2 depletion/control) DFE 6 150 Control SF3B2 Depletion 200 4 100

P × −11 100 = 1 10 2 50 SF3B2 Binding SF3B2 Binding expression level SF3B2-Bound exon 0 0 0 5’ 3’ Exon G 12.5 Control SF3B2 Depletion 10.0

7.5

5’ 3’ 5’ 3’ 5.0 P = 2 × 10−15 Exon Intron expression level

SF3B2-Bound intron 2.5

5’ 3’ Intron

Figure 4. Depletion of SF3B2 results in alterations of splicing and gene expression. A–C, Changes in gene expression in SF3B2-depleted cells compared with control cells. Cumulative plots of changes in gene expression binned by SF3B2-binding regions (A), numbers of SF3B2-binding sites (B), and numbers of xRPM of SF3B2- binding sites (C). P values were calculated by the pairwise Wilcoxon test with Bonferroni adjustment. D–G, Meta-analysis of SF3B2-bound exons or introns. Meta-exon analysis of SF3B2-binding sites along with the heatmap of SF3B2 binding (D) and transcripts (F) in the SF3B2-binding exons (n ¼ 2,240). Meta-intron analysis of the SF3B2-binding sites along with the heatmap of SF3B2 binding (E) and transcripts (G) in SF3B2-binding introns (n ¼ 14,621). P values were calculated using the Kolmogorov–Smirnov test.

an increase in SF3B2-bound gene expression (Supplementary tance and prostate cancer invasion (31, 32). Depletion of SF3B2 Fig. S7B). Consistent with the increase in AR-V7 expression, AR led to decreased NOTCH1 expression by altering the splicing of target genes, such as KLK3 and KLK4, were significantly upregu- the NOTCH1 mRNA (Supplementary Fig. S7D). Consistent with lated in response to SF3B2 overexpression (Fig. 6H; Supplemen- these altered gene expression patterns, GO terms related to cell tary Fig. S7C). Notably, AKT1 and UBE2C, which are the AR-V7– activation, including mitochondrion organization and ribosome specific target genes and regulate cell proliferation, were increased biogenesis, were significantly enriched in genes upregulated by in response to SF3B2 overexpression (Fig. 6H; Supplementary SF3B2 overexpression (P < 0.001, Fig. 6I). Thus, high SF3B2 Fig. S7C; refs. 23, 30). SF3B2 overexpression also increased the expression leads to aggressive phenotypes in prostate cancer with expression of NOTCH1, which is implicated in castration resis- increased expression of genes associated with cancer progression.

www.aacrjournals.org Cancer Res; 79(20) October 15, 2019 5211

Kawamura et al.

ACAR P < 0.0001 Exon 1 2 3 4 5 6 7 8 9 P 20 = 0.0001 5’ 3’

Cryptic exon 3 16

12 Genomic region for a potential SF3B2-binding RNA sequence 8 Stop codon gRNA #2 4 Percentage of AR-V7– depleted population Control gRNA gRNA #1 0 Coding sequence of cryptic exon 3 PAM gRNA #1 gRNA #2 gRNA #2 GGACACCGGTAGAAATAAACACA 5’ AGCAACTGTGTCTGTCTGAGGTTCCTGTGGCCATCTTTATTTGTGT 3’ Control gRNA |||||||||||||||||||||||||||||||||||||||||||||| 3’ TCGTTGACACAGACAGACTCCAAGGACACCGGTAGAAATAAACACA 5’ gRNA #1 GTGTCTGTCTGAGGTTCCTGTGG PAM B Control gRNA gRNA #1 gRNA #2 1,000 1,000 1,000

800 800 800

600 600 600 FSC 400 2.25% 400 16.4% 400 16.1%

200 200 200

0 0 0 0 1 2 3 0 1 2 3 0 1 2 3 10 10 10 10 10 10 10 10 10 10 10 10 AR-V7-GFP

D AR Exon 3 (ex.3) Exon 4 (ex.4) F1 Cryptic exon 3 (c.ex.3)

5’ 3’

R2 R1

EFN.A. 2.5 GFP-OE P = 0.0095 NT gRNA SF3B2-OE 1.6 CE3 gRNA-lower 1.4 2 1.2 P = 0.0002 P = 0.0528 1.5 1 N.A. 0.8 1 0.6 0.4 Relative splicing

Relative splicing 0.5 0.2 0 0 F1+R1 F1+R2 F1+R1 F1+R2 Exon 7-8 ex.3-ex.4 ex.3-c.ex.3 ex.3-ex.4 ex.3-c.ex.3

Figure 5. Validation of the function of SF3B2 for the regulation of splicing in AR. A, Schema of the cryptic exon 3 of AR with the position of two gRNAs or control gRNA and the SF3B2-binding site. Gray characters show the SF3B2-recognizing motif. B and C, FACS analysis of the indicated gRNA-transfected AR-V7-GFP cells (n ¼ 3). D, The position of primers for detecting splicing of AR-FL and AR-V7. E and F, Relative splicing of AR-FL and AR-V7 in the GFP-negative population in gRNA against the cryptic exon 3-transfected AR-V7-GFP (E) or SF3B2 stably expressed-22Rv1 (F) cells. N.A., not applicable.

Pladienolide B represses growth of SF3B2-addicted tumors SF3b components in prostate cancer cells. SF3B2-associated pro- in vivo teins were purified from the nuclear extracts of 22Rv1 cells with Because SF3B2 has been shown to be involved in the SF3b stable SF3B2-TAP expression by tandem affinity purification complex (27–29), we examined whether SF3B2 is associated with (Fig. 7A). SF3B2-TAP was localized to both the nucleus and the

5212 Cancer Res; 79(20) October 15, 2019 Cancer Research

SF3B2-Mediated RNA Splicing in Prostate Cancer Progression

ABC 22Rv1 LNCaP95 * P = 0.0028 1.6 1.4 1.2 GFP SF3B2 1.0 GFP #1 GFP #2 SF3B2 #1 SF3B2 #2 Exogenous SF3B2 0.8 Exogenous SF3B2 0.6 AR-FL 0.4 AR-V7/AR-FL AR-FL 11 0.2 101.2.8 0.9

AR-V7 Relative expression of 0 AR-V7 GFP SF3B2 111.1.4 1.4 12.4 ACTB ACTB 22Rv1 1111 11

DE 22Rv1 LNCaP95 )

1,200 ) 3

3 700 GFP GFP 1,000 SF3B2 600 SF3B2 800 SF3B2/V7KO 500 SF3B2/V7KO 600 400 300 P = 0.0371 400

P = 0.0133 200

200 P = 0.0002 100 P = 0.0029 P < 0.0001 N.S. Tumor volume (mm 0 Tumor volume (mm 0 0 7 14 21 28 35 0147 2128354249 Days after innoculation Days after innoculation

F GFP Overexpression H SF3B2 Overexpression KLK3 AKT1 150 0-300 0-1,500 GFP-OE 0-300 0-1,500

100 0-300 0-1,500

-16 P < 2 × 10 RNA-seq SF3B2-OE 0-300 0-1,500 50 expression level SF3B2-Bound exon 0 I 5’ 3’ Exon Mitochondrion organization in upregulated genes GO Terms enriched G GFP Overexpression SF3B2 Overexpression Ribosome biogenesis P < 2 × 10-16 Cellular response to oxygen- 10 containing compound

Translational termination 5 Cholesterol biosynthetic process enriched in downregulated genes GO Terms expression level SF3B2-Bound intron 0 −4 −3 −2 −1 0 −1 −2 −3 −4 5’ 3’ Multi-test adjusted P value (log ) Intron 10

Figure 6. Effects of SF3B2 overexpression on tumor growth in vivo and gene expression. A and B, Western blots of SF3B2 stably overexpressing 22Rv1 (A) or LNCaP95 cells (B). The intensity of each band was normalized to that of ACTB. C, Relative expression of AR-V7 normalized to the full length of AR (n ¼ 3). Tumor growth of SF3B2 stable overexpression or SF3B2 stable overexpression and AR-V7 KO 22Rv1 (n ¼ 11–12; D) or LNCap95 cells (n ¼ 6–8; E) in NOD/SCID mice under castration conditions. Meta-exon (n ¼ 2,240; F) or -intron (n ¼ 14,621; G) analysis of transcripts in SF3B2-binding exons or introns. H, mRNA expression of KLK3 and KLK4 in SF3B2 stable overexpression or GFP stable expression 22Rv1 cells. I, Representative GO terms with P < 0.001 in SF3B2 stable-overexpression 22Rv1 cells. N.S., nonsigniﬁcant.

cytoplasm, consistent with the localization of endogenous SF3B2 or SAP61; Fig. 7B). These results indicate that SF3B2 is a key (Supplementary Fig. S8A). Mass spectrometry identiﬁed an asso- component of the SF3b complex. ciation of SF3B2 with SF3B1, SF3B3 (also known as SF3b130 or Several compounds have been shown to inhibit splicing and SAP130), SF3B4 (also known as SF3b49 or SAP49), SF3A1 (also exhibit potential as antitumor drugs (33). Among them, pladie- known as SF3a120 or SAP114) and SF3A3 (also known as SF3a60 nolide B, a commercially available macrolide, has an antitumor

www.aacrjournals.org Cancer Res; 79(20) October 15, 2019 5213

Kawamura et al.

A B DMSO PLA-B Complex Calmodulin binding peptide 250 kDa

SF3B2 IgG Beads SF3B1 Protein A SF3B2 150 kDa TEV protease cleavage site SF3B3 SF3A1 1st Elution from IgG beads 100 kDa by TEV protease cleavage 75 kDa SF3B2 Calmodulin beads SF3A3 50 kDa SF3B4

DMSO Pladienolide B 37 kDa Figure 7. Wash Therapeutic effect of pladienolide B on prostate cancer. A, Schematic of SF3B2 Calmodulin Calmodulin 25 kDa beads SF3B2 beads ﬁ 20 kDa the puri cation of the SF3B2 2nd Elution complex and pladienolide B from calmodulin beads treatment. B, Top, staining of the 15 kDa puriﬁed SF3B2 complex with or

SF3B2 SF3B2 without 20 mmol/L pladienolide B SF3B2 (PLA-B). Proteins identiﬁed by MS are shown on the left. Bottom, Western blot of SF3B2 as an CD internal control. C, FACS analysis 22Rv1 DMSO of the AR-V7-GFP–negative P = 0.0271 population in pladienolide B– 100 P = 0.0092 PLA-B n ¼ 1.4 P = 0.0162 treated AR-V7-GFP cells ( 3). D, 80 AR-FL AR- 1.2 Relative splicing of and V7 in pladienolide B–treated cells. 1 60 The position of primers is indicated 0.8 40 in Fig. 5D. E, Relative in vivo tumor 0.6 growth of SF3B2 stably expressing 20 22Rv1 cells with pladienolide B or

Relative splicing 0.4 vehicle (DMSO) treatment 0 0.2 % AR-V7 Depleted population (5 mg/kg) under castration 0 mol/L mol/L mol/L mol/L n ¼ n n F1+R1 F1+R2 conditions ( 5). F, Relative 0 nmol/L4 n 8 n 12 nmol/L16 20 ex.3-ex.4 ex.3-c.ex.3 in vivo tumor growth of SF3B2 or Pladienolide B concentration GFP-stable expressed 22Rv1 cells with pladienolide B treatment (5 mg/kg) under castration E SF3B2-Overexpressed 22Rv1 F PLA-B conditions (n ¼ 5). DMSO * 1,000 * PLA-B GFP * 600 SF3B2 * 800 *

600 400 * * * P < 0.05 400 * P < 0.01 * * * * 200 200 Relative tumor volume (%) Relative tumor volume (%) 0 0 0 7 14 21 28 0 7 14 21 28 Days after initiation of treatment Days after initiation of treatment

activity and disrupts splicing through an interaction with treatment in both 22Rv1 and LNCap95 cells (Fig. 7D; Supple- SF3B3 (12, 34). As SF3B2 was associated with SF3B3 in prostate mentary Fig. S8C). AR-V7-GFP expression was also decreased in cancer cells, we sought to determine whether pladienolide B is a AR-V7-GFP cells after treatment with spliceostatin A, another candidate therapeutic drug for aggressive cancers with high SF3B2 SF3b inhibitor (Supplementary Fig. S8D). Pladienolide B strongly expression. Notably, pladienolide B treatment reduced the inter- inhibited cell proliferation of 22Rv1 cells at a concentration of 4 action of SF3B2-TAP with its associated proteins (Fig. 7A and B). nmol/L, compared with a lesser effect of normal human prostate Given the effect of pladienolide B on SF3B2-associated proteins, cells PNT2 (Supplementary Fig. S8E and S8F). To demonstrate its we examined its effect on AR-V7 splicing (Fig. 7C; Supplementary antitumor effect in vivo, pladienolide B was injected intraperito- Fig. S8B). Pladienolide B decreased AR-V7-GFP expression in neally into hosts bearing 22Rv1 or LNCaP95 tumors with AR-V7-GFP cells in a dose-dependent manner. In addition, AR- stable SF3B2 expression when the tumors reached sizes of approx- V7 expression was signiﬁcantly decreased by pladienolide B imately 100–300 mm3 or approximately 200–400 mm3,

5214 Cancer Res; 79(20) October 15, 2019 Cancer Research

SF3B2-Mediated RNA Splicing in Prostate Cancer Progression

respectively, under castrate conditions (Fig. 7E and F; Supple- sequence in introns and promotes transesterification to produce a mentary Fig. S8G and S8H). Notably, pladienolide B treatment lariat structure (42, 43). SF3b, a multiprotein complex with resulted in significant regression of both tumors under castration SF3B1, SF3B2, SF3B3, SF3B4, SF3B5, SF3B6, SF3B7, is a compo- conditions (P < 0.01, Fig. 7E; Supplementary Fig. S8G). Moreover, nent of U2 snRNP (29). SF3B1 (SF3b155) and SF3B6 (p14) form tumors with stable SF3B2 overexpression were more sensitive to a the helix of U2 RNA and pre-mRNA, resulting in the bulged pladienolide B treatment than tumors with GFP expression alone adenosine in the branch point sequence that acts as the nucleo- (P < 0.05, Fig. 7F; P < 0.01, Supplementary Fig. S8H). Therefore, phile (41). SF3B1 (SF3b155) interacts with U2AF2 (U2AF65), the aggressive phenotype conferred by high SF3B2 overexpression which binds to the polypyrimidine tract downstream of the may be reversed by therapeutic targeting, and pladienolide B is a branch point sequence (44, 45). Consistently, cross-linking and candidate therapeutic drug for cancers that are addicted to high immunoprecipitation (CLIP)-seq of U2AF2 revealed that U2AF2 SF3B2 expression. bound to the vast majority of 30 splice sites (30 SS; ref. 46). However, a key role for SF3B2 has not been investigated in detail in previous studies. Discussion SF3B2-associated proteins were not identical to the compo- SF3B2 high expression confers aggressive prostate cancer nents of the SF3B1 complex (Fig. 7). Particularly, SF3B6 was not phenotypes involved in SF3B2-associated proteins, suggesting a distinct func- Cancer cells exhibit not only altered gene expression levels tion from conventional SF3b because SF3B6 binds to the bulged but also splicing of various genes related to cell proliferation adenosine at the branch site (28). Moreover, the global SF3B2- or cell death (4, 35). Mutations in genes encoding splicing binding profile was different from U2AF2 (Fig. 4). Although there factors result in aberrant splicing and cell death in hemato- is a possibility that a difference between PAR-CLIP and CLIP poietic cells (36, 37). Mutations at splice sites also become a causes a different binding pattern, disruption of the SF3B2 motif, cause of aberrant splicing and functional perturbation. How- identified based on SF3B2 PAR-CLIP signals, in cryptic exon 3 in ever, aberrant splicing has been detected in cancer without AR resulted in decreased AR-V7 expression (Fig. 5). The affinity of somatic mutations in either splicing factors or splice sites (35). SF3B2 for RNA may depend on the context of the RNA sequence or Thus, the mechanisms that alter splicing in cancer remain its associated proteins, potentially explaining the poor ability of unclear. SF3B2 PAR-CLIP to detect upstream branch point sites, as PAR- Some splicing variants are known to mediate cancer progres- CLIP efficiently recovers SF3B2-bound RNA and determines the sion (35). Among them, a splicing variant of the AR, AR-V7, has a binding sites at high resolution (13). Some splicing factors, such critical role in castration resistance and anticancer drug sensitiv- as RBFOX2, promote both the inclusion and exclusion of their ity (10). Although an association of U2AF2, Sam68, or CPSF1 target exons depending on their binding position by controlling with AR-V7 expression has been reported (38, 22, 20), it remains the assembly of splicing factors on pre-mRNAs (47). Similarly, unclear whether these splicing factors upregulate AR-V7 expres- SF3B2 depletion or overexpression affected gene expression and sion in patients with prostate cancer and how AR-V7 expression is splicing in a SF3B2 binding–dependent manner (Figs. 4 and 6). upregulated in CRPC. In this study, we now demonstrate that high However, SF3B2 depletion did not affect expression and splicing SF3B2 expression, without a mutation in SF3B2, increased AR-V7 in all genes; as an example, ACTB, a housekeeping gene, was not through direct binding to AR cryptic exon 3 and conferred an affected by SF3B2 depletion (Supplementary Fig. S4C), suggesting aggressive phenotype in prostate cancer by altering gene expres- that SF3B2 is associated more with the regulation of splicing of sion profiles. Furthermore, high SF3B2 expression was associated SF3B2-bound RNA. Therefore, while SF3B1 is thought to have a with poor progression-free survival and overall survival not only central role in SF3b function for general RNA splicing, our data in patients with prostate cancer but also in those with bladder indicate that SF3B2 directly binds not only to introns but also to cancer, AML, lung adenocarcinoma, head and neck squamous cell exons through recognizing a specific RNA motif to promote an carcinoma, and breast cancer (Supplementary Fig. S3), implying inclusion of the target exon. that high SF3B2 expression may be related to other key splice variants that drive cancer progression similar to AR in prostate A compound targeting the SF3B2-containing complex is a cancer. Indeed, in breast cancer, MENAINV, a splicing variant of candidate therapeutic drug for aggressive cancers with high MENA, a member of the ENA/VASP family of actin filament SF3B2 expression elongation factors, promotes haptotaxis on fibronectin gradients, Although some drugs, such as abiraterone or enzalutamide, which is associated with metastasis, and high MENAINV expres- are effective for controlling CRPC by suppressing the andro- sion leads to poor overall survival (39). As another splicing factor gen–AR axis, CRPC eventually acquires resistance to these that leads to aggressive cancer phenotypes at high expression treatments and such resistance may be mediated by modifying levels, SRSF1 (also known as SF2 or ASF) overexpression results in the androgen–AR axis (10). Thus, there is a need to develop an increase in tumor growth along with increases in a splicing novel therapies targeting resistance mechanisms, including AR- variant of S6K1 (40). Thus, high expression of splicing factors, V7–mediated resistance to AR-targeting therapies. The findings including SF3B2, is a cause of the production of splicing variants from this study have therapeutic implications for patients with leading to an aggressive phenotype in cancer. CRPC with high SF3B2 expression. Pladienolide B, a splicing inhibitor, binds to SF3B3 and inhibits SF3B1 activity (34, 48). The molecular mechanisms of SF3B2 for regulating splicing Two clinical trials of E7107, a derivative of pladienolides, The spliceosome, a dynamic megadalton complex consisting of showed that E7107 controlled tumor growth in approximately 5 small nuclear ribonucleoproteins (snRNP; U1, U2, U4/U6, U5), 30% patients, but vision loss occurred in some patients in is assembled on introns and removes the vast majority of introns clinical trials (3/66; refs. 49, 50). The results indicate a poten- from pre-mRNAs (41). U2 snRNP recognizes the branch point tial of splicing inhibitors for cancer treatment, although other

www.aacrjournals.org Cancer Res; 79(20) October 15, 2019 5215

Kawamura et al.

pladienolide derivatives or a compound with a different chem- Authors' Contributions ical structure are needed to avoid adverse events. Conception and design: N. Kawamura, K. Nimura Pladienolide B showed strong antiproliferative activity Development of methodology: N. Kawamura, K. Nimura toward aggressive prostate cancer cell tumors with high SF3B2 Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Kawamura, K. Nimura, K. Saga, A. Ishibashi, expression (Fig. 7). The effect of E7107 on the repression of K. Kitamura, H. Nagano, K. Ishida, Y. Kaneda tumor growth has been observed in a limited number of Analysis and interpretation of data (e.g., statistical analysis, biostatistics, patients (49, 50), and in xenografts, the antitumor effect of computational analysis): N. Kawamura, K. Nimura, A. Ishibashi, K. Ishida E7107 is very diverse depending on the cancer cell type (12). Writing, review, and/or revision of the manuscript: K. Nimura, N. Nonomura, These findings suggest that splicing inhibitors show strong Y. Kaneda antiproliferative effects on aggressive tumors with a modified Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Yoshikawa, N. Nonomura, M. Arisawa, splicing pattern by high expression levels of splicing factors. It Y. Kaneda was recently reported that E7107 extends survival in a mouse Study supervision: K. Nimura, Y. Kaneda model of leukemia with a splicing factor SRSF2 mutation Other (provided expert advice from time to time): J. Luo compared with those without a mutation, although it remains unclear how the effect of E7107 is stronger in leukemia with the Acknowledgments SRSF2 mutation (36). Thus, biomarker-based patient selection, We thank Mayuko Okado and Mieko Watanabe for technical assistance and by expression level or mutation of a splicing factor, may be Dr. Yukio Kawahara for technical advice. This work was supported by Platform required to demonstrate efficacy and strong antitumor activity Project for Supporting Drug Discovery and Life Science Research [Basis for Supporting Innovative Drug Discovery and Life Science Research (BINDS)] by a splicing inhibitor. Because pladienolide B shows a strong from AMED under grant number JP19am0101084 and DAICEL, Inc. (to K. antiproliferative effect on prostate cancer with SF3B2 over- Nimura). expression under castrate conditions in vivo,targetingSF3B2 by pharmacologic inhibitors is a viable therapeutic strategy for The costs of publication of this article were defrayed in part by the payment of tumors driven by aberrant splicing. page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Disclosure of Potential Conflicts of Interest N. Kawamura and K. Nimura report applying for patents with DAICEL, Inc. Received December 18, 2018; revised April 24, 2019; accepted August 15, No potential conflicts of interest were disclosed by the other authors. 2019; published first August 20, 2019.

References 1. Lee Y, Rio DC. Mechanisms and regulation of alternative pre-mRNA 14. Yokoshi M, Li Q, Yamamoto M, Okada H, Suzuki Y, Kawahara Y. Direct splicing. Annu Rev Biochem 2015;84:291–323. binding of ataxin-2 to distinct elements in 30 UTRs promotes mRNA 2. Danan-Gotthold M, Golan-Gerstl R, Eisenberg E, Meir K, Karni R, Levanon stability and protein expression. Mol Cell 2014;55:186–98. EY. Identification of recurrent regulated alternative splicing events across 15. Nimura K, Ura K, Shiratori H, Ikawa M, Okabe M, Schwartz RJ, et al. A human solid tumors. Nucleic Acids Res 2015;43:5130–44. histone H3 lysine 36 trimethyltransferase links Nkx2-5 to wolf-hirschhorn 3. Dvinge H, Bradley RK. Widespread intron retention diversifies most cancer syndrome. Nature 2009;460:287–91. transcriptomes. Genome Med 2015;7:45. 16. Hu R, Dunn TA, Wei S, Isharwal S, Veltri RW, Humphreys E, et al. Ligand- 4. Dvinge H, Kim E, Abdel-Wahab O, Bradley RK. RNA splicing factors as independent androgen receptor variants derived from splicing of cryptic oncoproteins and tumour suppressors. Nat Rev Cancer 2016;16: exons signify hormone-refractory prostate cancer. Cancer Res 2009;69: 413–30. 16–22. 5. Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, et al. 17. Grasso CS, Wu YM, Robinson DR, Cao X, Dhanasekaran SM, Khan AP, et al. Cancer incidence and mortality worldwide: sources, methods and major The mutational landscape of lethal castration-resistant prostate cancer. patterns in GLOBOCAN 2012. Int J Cancer 2015;136:E359–86. Nature 2012;487:239–43. 6. Beer TM, Armstrong AJ, Rathkopf DE, Loriot Y, Sternberg CN, Higano CS, 18. Carbon S, Ireland A, Mungall CJ, Shu S, Marshall B, Lewis S. AmiGO: et al. Enzalutamide in metastatic prostate cancer before chemotherapy. online access to ontology and annotation data. Bioinformatics 2009;25: N Engl J Med 2014;371:424–33. 288–9. 7. de Bono JS, Logothetis CJ, Molina A, Fizazi K, North S, Chu L, et al. 19. The Cancer Genome Atlas Research Network. The molecular taxonomy of Abiraterone and increased survival in metastatic prostate cancer. N Engl J primary prostate cancer. Cell 2015;163:1011–25. Med 2011;364:1995–2005. 20. Etten JLV, Nyquist M, Li Y, Yang R, Ho Y, Johnson R, et al. Targeting a single 8. Ryan CJ, Smith MR, de Bono JS, Molina A, Logothetis CJ, de Souza P, et al. alternative polyadenylation site coordinately blocks expression of andro- Abiraterone in metastatic prostate cancer without previous chemotherapy. gen receptor mRNA splice variants in prostate cancer. Cancer Res 2017;77: N Engl J Med 2013;368:138–48. 5228–35. 9. Scher HI, Fizazi K, Saad F, Taplin M-E, Sternberg CN, Miller K, et al. 21. Li Y, Alsagabi M, Fan D, Bova GS, Tewfik AH, Dehm SM. Intragenic Increased survival with enzalutamide in prostate cancer after chemother- rearrangement and altered RNA splicing of the androgen receptor in a apy. N Engl J Med 2012;367:1187–97. cell-based model of prostate cancer progression. Cancer Res 2011;71: 10. Antonarakis ES, Armstrong AJ, Dehm SM, Luo J. Androgen receptor variant- 2108–17. driven prostate cancer: clinical implications and therapeutic targeting. 22. Liu LL, Xie N, Sun S, Plymate S, Mostaghel E, Dong X. Mechanisms of the Prostate Cancer Prostatic Dis 2016;19:231–41. androgen receptor splicing in prostate cancer cells. Oncogene 2014;33: 11. Lu C, Luo J. Decoding the androgen receptor splice variants. Transl Androl 3140–50. Urol 2013;2:178–86. 23. Hu R, Lu C, Mostaghel EA, Yegnasubramanian S, Gurel M, Tannahill C, 12. Mizui Y, Sakai T, Iwata M, Uenaka T, Okamoto K, Shimizu H, et al. et al. Distinct transcriptional programs mediated by the ligand-dependent Pladienolides, new substances from culture of streptomyces platensis full-length androgen receptor and its splice variants in castration-resistant mer-11107. J Antibiot 2004;57:188–96. prostate cancer. Cancer Res 2012;72:3457–62. 13. Hafner M, Landthaler M, Burger L, Khorshid M, Hausser J, Berninger P, et al. 24. Nelson PS, Clegg N, Arnold H, Ferguson C, Bonham M, White J, et al. The Transcriptome-wide identification of RNA-binding protein and MicroRNA program of androgen-responsive genes in neoplastic prostate epithelium. target sites by PAR-CLIP. Cell 2010;141:129–41. Proc Natl Acad Sci U S A 2002;99:11890–5.

5216 Cancer Res; 79(20) October 15, 2019 Cancer Research

SF3B2-Mediated RNA Splicing in Prostate Cancer Progression

25. Corcoran DL, Georgiev S, Mukherjee N, Gottwein E, Skalsky RL, Keene JD, 38. Stockley J, Markert E, Zhou Y, Robson CN, Elliott DJ, Lindberg J, et al. The et al. PARalyzer: definition of RNA binding sites from PAR-CLIP short-read RNA-binding protein Sam68 regulates expression and transcription func- sequence data. Genome Biol 2011;12:R79. tion of the androgen receptor splice variant AR-V7. Sci Rep 2015;5: 26. Georgiev S, Boyle AP, Jayasurya K, Ding X, Mukherjee S, Ohler U. Evidence- srep13426. ranked motif identification. Genome Biol 2010;11:R19. 39. Oudin MJ, Jonas O, Kosciuk T, Broye LC, Guido BC, Wyckoff J, et al. Tumor 27. Cretu C, Schmitzova J, Ponce-Salvatierra A, Dybkov O, De Laurentiis EI, cell–driven extracellular matrix remodeling drives haptotaxis during met- Sharma K, et al. Molecular architecture of SF3b and structural conse- astatic progression. Cancer Discov 2016;6:516–31. quences of its cancer-related mutations. Mol Cell 2016;64:307–19. 40. Karni R, de Stanchina E, Lowe SW, Sinha R, Mu D, Krainer AR. The gene 28. Will CL, Schneider C, MacMillan AM, Katopodis NF, Neubauer G, Wilm M, encoding the splicing factor SF2/ASF is a proto-oncogene. Nat Struct Mol et al. A novel U2 and U11/U12 snRNP protein that associates with the Biol 2007;14:185–93. premRNA branch site. EMBO J 2001;20:4536–46. 41. Wahl MC, Will CL, Luhrmann€ R. The spliceosome: design principles of a 29. Will CL, Urlaub H, Achsel T, Gentzel M, Wilm M, Luhrmann€ R. Charac- dynamic RNP machine. Cell 2009;136:701–18. terization of novel SF3b and 17S U2 snRNP proteins, including a human 42. Berglund JA, Rosbash M, Schultz SC. Crystal structure of a model branch- Prp5p homologue and an SF3b DEADbox protein. EMBO J 2002;21: point–U2 snRNA duplex containing bulged adenosines. RNA 2001;7: 4978–88. 682–91. 30. Guo Z, Yang X, Sun F, Jiang R, Linn DE, Chen H, et al. A novel androgen 43. MacMillan AM, Query CC, Allerson CR, Chen S, Verdine GL, Sharp PA. receptor splice variant is up-regulated during prostate cancer progression Dynamic association of proteins with the pre-mRNA branch region. and promotes androgen depletion–resistant growth. Cancer Res 2009;69: Genes Dev 1994;8:3008–20. 2305–13. 44. Valcarcel J, Gaur RK, Singh R, Green MR. Interaction of U2AF65 RS region 31. Hafeez BB, Adhami VM, Asim M, Siddiqui IA, Bhat KM, Zhong W, et al. with Pre-mRNA of branch point and promotion base pairing with U2 Targeted knockdown of notch1 inhibits invasion of human prostate cancer snRNA. Science 1996;273:1706–9. cells concomitant with inhibition of matrix metalloproteinase-9 and 45. Singh R, Valcarcel J, Green MR. Distinct binding specificities and functions urokinase plasminogen activator. Clin Cancer Res 2009;15:452–9. of higher eukaryotic polypyrimidine tract-binding proteins. Science 1995; 32. Stoyanova T, Riedinger M, Lin S, Faltermeier CM, Smith BA, Zhang KX, et al. 268:1173–6. Activation of Notch1 synergizes with multiple pathways in promoting 46. Shao C, Yang B, Wu T, Huang J, Tang P, Zhou Y, et al. Mechanisms for U2AF castration-resistant prostate cancer. Proc Natl Acad Sci 2016;113: to define 30 splice sites and regulate alternative splicing in the human E6457–66. genome. Nat Struct Mol Biol 2014;21:997–1005. 33. Bonnal S, Vigevani L, Valcarcel J. The spliceosome as a target of novel 47. Misra A, Ou J, Zhu LJ, Green MR. Global promotion of alternative internal antitumour drugs. Nat Rev Drug Discov 2012;11:847–59. exon usage by mRNA 30 end formation factors. Mol Cell 2015;21: 34. Kotake Y, Sagane K, Owa T, Mimori-Kiyosue Y, Shimizu H, Uesugi M, et al. 997–1005. Splicing factor SF3b as a target of the antitumor natural product pladie- 48. Effenberger KA, Urabe VK, Prichard BE, Ghosh AK, Jurica MS. Interchange- nolide. Nat Chem Biol 2007;3:570–5. able SF3B1 inhibitors interfere with pre-mRNA splicing at multiple stages. 35. Sveen A, Kilpinen S, Ruusulehto A, Lothe RA, Skotheim RI. Aberrant RNA RNA 2016;22:350–9. splicing in cancer; expression changes and driver mutations of splicing 49. Eskens FA, Ramos FJ, Burger H, O'Brien JP, Piera A, de Jonge MJ, et al. Phase factor genes. Oncogene 2016;35:2413–27. I pharmacokinetic and pharmacodynamic study of the first-in-class spli- 36. Lee SC-W, Dvinge H, Kim E, Cho H, Micol JB, Chung YR, et al. ceosome inhibitor E7107 in patients with advanced solid tumors. Modulation of splicing catalysis for therapeutic targeting of leukemia Clin Cancer Res 2013;19:6296–304. with mutations in genes encoding spliceosomal proteins. Nat Med 50. Hong DS, Kurzrock R, Naing A, Wheler JJ, Falchook GS, Schiffman JS, et al. 2016;22:672–8. A phase I, open-label, single-arm, dose-escalation study of E7107, a 37. Shirai CL, Ley JN, White BS, Kim S, Tibbitts J, Shao J, et al. Mutant U2AF1 precursor messenger ribonucleic acid (pre-mRNA) splicesome inhibitor expression alters hematopoiesis and pre-mRNA splicing in vivo. administered intravenously on days 1 and 8 every 21 days to patients with Cancer Cell 2015;27:631–43. solid tumors. Invest New Drugs 2014;32:436–44.

www.aacrjournals.org Cancer Res; 79(20) October 15, 2019 5217

SF3B2-Mediated RNA Splicing Drives Human Prostate Cancer Progression

Norihiko Kawamura, Keisuke Nimura, Kotaro Saga, et al.

Cancer Res 2019;79:5204-5217. Published OnlineFirst August 20, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-18-3965

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2019/08/20/0008-5472.CAN-18-3965.DC1

Visual A diagrammatic summary of the major findings and biological implications: Overview http://cancerres.aacrjournals.org/content/79/20/5204/F1.large.jpg

Cited articles This article cites 50 articles, 17 of which you can access for free at: http://cancerres.aacrjournals.org/content/79/20/5204.full#ref-list-1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/79/20/5204. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.