A Fluorescence In situ Hybridization Screen for E26 Transformation–Specific Aberrations: Identification of DDX5-ETV4 Fusion Protein in Prostate Cancer

E26 transformation–specific (ETS) transcription factors play an important role in cellular proliferation, carcinogenesis, and metastasis (1–3). To date, 27 human ETS family members and over 200 ETS target genes have been identified (3). Translocations involving ETS family genes such as FLI1, ERG, ETV1, ETV4, and ETV6 have been described in hematologic and mesenchymal malignancies including Ewing's sarcomas, leukemias, lymphomas, fibrosarcomas, and mesoblastic nephroma (1, 3, 4) and in rare secretary breast carcinomas (5).

Recently, recurrent gene fusions involving ETS family genes, ERG, ETV1, ETV4, and ETV5, fused to one of five different upstream fusion partners (TMPRSS2, SLC45A3, HERV-K_22q11.23, C15ORF21, and HNRPA2B1) have been identified in a majority of prostate cancers (6–9). The upstream regulatory elements, with their different androgen sensitivity and prostate specificity, are believed to drive aberrant expression of the downstream, ETS genes, contributing to prostate carcinogenesis. Among these gene fusions, the TMPRSS2-ERG fusion is the most prevalent, observed in ∼50% to 70% of hospital-based, prostate-specific antigen (PSA)-screened, prostate cancer cohorts (10–15). Emerging data suggests association of TMPRSS2-ERG fusion subtypes with a more aggressive phenotype in clinically localized prostate cancer as well in androgen-independent metastatic prostate cancer (12, 13, 16–19; reviewed in ref. 20).

Apart from ERG, ETV1, ETV4, or ETV5, aberrations involving other ETS family genes in prostate cancer are still unknown. In this study, we comprehensively surveyed all 27 ETS family genes and the five known 5′ fusion partners for rearrangements using fluorescence in situ hybridization (FISH) analysis on tissue microarray (TMA) from a cohort of localized prostate cancer patients. Whereas confirming the ETS gene rearrangement patterns reported earlier, we noted that >40% of prostate cancers are ETS fusion negative, defining a distinct molecular subtype. In addition, we identified novel 5′ fusion partners of ERG (e.g., SLC45A3), ETV1 (e.g., FLJ35294), and ETV4 (e.g., CANT1 and DDX5). DDX5-ETV4 also represents the first fusion protein identified in prostate cancer.

Study population, clinical data, and TMA construction. A TMA was constructed representing 110 clinically localized prostate cancer patients who underwent radical prostatectomy as a primary therapy between 2004 and 2006 at the University of Michigan Hospital. Three cores (0.6 mm in diameter) were taken from each representative tumor focus and morphology was confirmed by three pathologists (B.H., R.B.S., and R.M.). Detailed clinical, pathologic, and TMA data were maintained on a secure relational database as previously described (21). Patient demographics are shown in Supplementary Table S1. This radical prostatectomy series was part of the University of Michigan Prostate Cancer Specialized Program of Research Excellence Tissue Core. This study was approved by the Institutional Review Board at the University of Michigan Medical School.

FISH and FISH-based screening strategy. A previously validated FISH-based split probe strategy was used to investigate known and novel gene aberrations in prostate cancer (7–9, 21). Interphase FISH was performed as described (7, 21). Bacterial artificial chromosomes (BACs; listed in Supplementary Table S2) were obtained from the BACPAC Resource Center, and probes were prepared as described (7, 9, 21). The integrity and correct localization of all probes were verified by hybridization to metaphase spreads of normal peripheral lymphocytes. Slides were examined using an ImagingZ1 microscope (Carl Zeiss). FISH signals were scored manually (×100 oil immersion) in morphologically intact and nonoverlapping nuclei by three pathologists (B.H., R.B.S., and R.M.), and a minimum of 50 cancer cells from each site were recorded. Cancer sites with very weak or no signals were recorded as insufficiently hybridized. Cases lacking tumor tissue in all three cores were excluded.

A flowchart of FISH-based high-throughput screening strategy is shown in Fig. 1. For detection of ERG, ETV1, ETV4, and ETV5 rearrangements in which breakpoints have been characterized, BACs for split probes were used as previously described (6–9, 21). For the remaining ETS family genes with unknown rearrangement status in prostate cancer, split probes flanking ∼1 Mb regions of interest were used for initial screening on the prostate cancer TMA (Fig. 1A and B). Subsequently, ETS gene–specific probes tightly flanking the gene of interest (∼200 Kb) were used to confirm ETS aberration on tissue sections from prostate cancer cases that were identified on initial screening (Fig. 1C).

We further evaluated this TMA for rearrangements of all 5′ partners known at the time this study was started by split probe FISH strategy. For cases rearranged for both ETS gene and known 5′ partners, a previously validated fusion probe strategy was used to confirm potential gene fusions (Fig. 1D; refs. 7, 8, 21). For cases with ETS gene rearrangement only, frozen tissues were obtained to identify 5′ fusion partner by RNA ligase–mediated rapid amplification of cDNA ends (RLM-RACE; Fig. 1D).

Cell line studies. LNCaP, an androgen-sensitive prostate cancer cell line, was maintained in RPMI with 10% FBS. For androgen stimulation experiments, cells were placed for 2 d in phenol red–free RPMI, supplemented with 5% of charcoal-treated FCS before being treated with 1% ethanol or 10 nmol/L R1881 for the following time points 0, 3, 12, 24, and 48 h. Cells were treated for 16 h with androgen and cross-linked for chromatin immunoprecipitation (ChIP) analysis. Total RNA was isolated with Trizol (Invitrogen) according to the manufacturer's instructions.

Quantitative real-time reverse transcriptase PCR. Quantitative real-time reverse transcriptase PCR (QRT-PCR) was carried out according to standard protocols as previously described (7). All oligonucleotide primers were synthesized by Integrated DNA Technologies and are listed in Supplementary Table S3. Samples were normalized by the mRNA level of the housekeeping gene GAPDH. Androgen stimulation reactions were performed in triplicate, and all other reactions were performed in duplicate.

RLM-RACE. RLM-RACE was performed as previously described to identify unknown 5′ partners of aberrant ETS genes in prostate cancer (7). First-strand cDNA was amplified with gene-specific reverse primers ETV1_exon4-5r (7), ETV4_exon7r (8), and 5′ gene racer primers (Invitrogen) using Platinum Taq High Fidelity enzyme (Invitrogen) after the touchdown PCR protocol according to manufacturer's instructions. PCR amplification products were cloned into pCR4-TOPO TA vector (Invitrogen) and sequenced bidirectionally using vector primers as described (7).

ChIP analysis. ChIP experiment was carried out as previously described (22) using 5 μg of antiandrogen receptor (AR) antibody (Upstate) at 16 h after R1881 or ethanol treatment of hormone-deprived cells. The input whole-cell extract DNA and the ChIP-enriched DNA were amplified by ligation-mediated PCR, and subjected to qPCR assessment of target promoters. ChIP enrichment was evaluated as a percentage of input DNA.

Tissue-specific expression. To determine tissue-specific expression of 5′ fusion partners, we interrogated the International Genomics Consortium's expO data set,⁶

Cloning and expression of DDX5-ETV4. Full-length DDX5-ETV4 fusion cDNA was cloned into Gateway cloning system's entry vector pENTR-D-TOPO (Invitrogen) using 5′ RLM-RACE product from case no.85 with 5′ primer containing a Kozak sequence-FLAG tag-start codon-DDX5 NH₂-terminal nucleotide sequence (5′-CACC-ATG-GATTACAAGGATGACGACGATAAG-TCGGGTTATTCGAGTGACCGAGACCGCGGC-3′) and 3′ primer corresponding to the COOH-terminal nucleotides of ETV4, until the stop codon (CTAGTAAGAGTAGCCACCCTTGGGGCCA). Expression constructs were generated by recombination of pENTR-D-Topo-DDX5-ETV4 with pCDNA3.2-DEST vector (Invitrogen), using LR Clonase II (Invitrogen). Human embryonic kidney (HEK) 293 cells were transiently transfected with pCDN3.2-FLAG-DDX5-ETV4 expression construct using FuGENE6 transfection reagent (Roche). Lysates from the transfected cells were resolved by SDS-PAGE and transferred onto Polyvinylidene Difluoride membrane (GE Healthcare). The membrane was incubated for 1 h in blocking buffer [TBS, 0.1% Tween (TBS-T), 5% nonfat dry milk] and incubated overnight at 4°C with rabbit polyclonal antibody against FLAG tag at 1:1,000 dilution (Cell Signaling Technology). After a wash with TBS-T, the blot was incubated with horseradish peroxidase–conjugated secondary antibody and the signals visualized by enhanced chemiluminescence system as described by the manufacturer (GE Healthcare). The blot was reprobed with glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Abcam) for confirmation of equal loading. Prostate tissue lysates were processed similarly for immunoblotting with ETV4 polyclonal antibody (Abnova).

We generated a comprehensive profile of the rearrangement status of all 27 ETS family genes and all five of the known 5′ fusion partners in prostate cancer using FISH split probe hybridizations on a TMA composed of 110 cases of clinically localized prostate cancers. A flow chart of our systematic ETS FISH screen is described in Fig. 1.

A matrix representation of gene rearrangements for the 27 ETS transcription factors and the five 5′ fusion partners is shown in Fig. 2. ERG, as expected, was the most commonly rearranged ETS gene in prostate cancer, and TMPRSS2 was the most commonly rearranged 5′ fusion partner (7, 9, 21). In the cohort represented on this TMA, ERG was rearranged in 43% (43 of 99) of cases, of which 63% (27 of 43) were fused through deletion of its 5′ end to TMPRSS2, which is similar to previous reports (11, 21). Overall, 98% (42 of 43) of the ERG-positive cases harbored TMPRSS2 as the 5′ fusion partner. Interestingly, one of the ERG-positive cases was negative for TMPRSS2. Upon further inspection of our FISH screen (Fig. 2), this case harbored rearrangement of the 5′ partner SLC45A3 (case no. 102). We proceeded to confirm the genomic fusion of SLC45A3:ERG using a fusion assay with probes 5′ to SLC45A3 and 3′ to ERG (Fig. 3), thus implicating SLC45A3 as a novel 5′ fusion partner of ERG. Previous to this finding, it was thought that ERG exclusively partnered with TMPRSS2 possibly due to their colocalization on chromosome 21. SLC45A3 was previously identified to be the 5′ fusion partner of ETV1 and ETV5 (6, 7).

In contrast to ERG, ETV1 was rearranged in ∼5% (5 of 99) of patients (case nos. 4, 51, 54, 72, and 91; Fig. 2), which is in line with our previous study (21). ETV1 has been shown to be fused to a number of genes including TMPRSS2, SLC45A3, HERV-K_22q11.23, C15ORF21, and HNRPA2B1 (7). In the 110 cases represented in this study, case nos. 4 and 72 harbored rearrangements in ETV1 in addition to rearrangements of SLC45A3 and HNRPA2B1, respectively. We used fusion probe FISH assays to confirm SLC45A3-ETV1 (a probe 5′ to SLC45A3 and 3′ to ETV1) and HNRPA2B1-ETV1 (a probe 5′ to HNRPA2B1 and 3′ to ETV1) fusions in these two cases, respectively. Of note, case no. 72 was the second case of HNRPA2B1-ETV1 reported in the literature thus far (7), suggesting HNRPA2B1:ETV1 fusion is recurrent in prostate cancer. Case no. 4, which harbored SLC45A3-ETV1 fusion, was identified to be the same case as reported earlier (7). In the ETV1-positive case nos. 51, 54, and 91, no rearrangement in the known 5′ partners were identified by FISH. In case no. 54, using RLM-RACE, we identified exons 1 to 4 of ETV1 to be replaced by 263 bp of the 5′ sequence of a prostate EST gene FLJ35294 (17p13.1; Fig. 4A). For ETV1-positive case nos. 51 and 91, we could not perform RACE as frozen tissue was not available, they would presumably harbor uncharacterized 5′ fusion partners.

ETV4 rearrangements were present in 5% (5 of 100) of patients in our cohort (case nos. 40, 46, 53, 64, and 85; Fig. 2). Case nos. 64 and 85 had rearrangement through translocation (split), whereas deletion of the probe 5′ to ETV4 was identified in case nos. 40 and 46. In contrast, case no. 53 revealed deletion of the probe 3′ to ETV4 (the significance of which is unclear). Of note, case nos. 40 and 64 also showed an aberration for TMPRSS2 when screened by TMPRSS2 split probe. These two cases were confirmed to harbor TMPRSS2-ETV4 fusion using a fusion probe assay (using probes 5′ to TMPRSS2 and 3′ to ETV4). Interestingly, in case no. 40, instead of split signals, a 5′ end deletion was detected using probes 5′ and 3′ to ETV4, suggesting that the TMPRSS2-ETV4 fusion in this case occurred through an unbalanced translocation. RLM-RACE further revealed that exon 1 of ETV4 was replaced with exons 1 to 2 of TMPRSS2. Thus, this case is different from the previously reported TMPRSS2-ETV4 fusion case where a novel upstream exon of TMPRSS2 was involved in the fusion (8). ETV4 was rearranged in case nos. 46 and 85, whereas no genetic aberration in the known 5′ partners was observed by FISH. To characterize the ETV4 transcript in these two cases, we performed RLM-RACE. Interestingly, in case no. 46, we identified exon 5 of ETV4 fused to exon 1a of the CANT1 gene located on 17q25.3, identical to the fusion sequence reported before (Fig. 4A; ref. 23). In case no. 85, another novel 5′ partner gene, DEAD (Asp-Glu-Ala-Asp) box polypeptide 5 (DDX5), was identified by RLM-RACE (Fig. 4A). Sequence analysis of DDX5-ETV4 revealed that the fusion transcript is composed of exons 1 to 3 of DDX5, fused in frame to exons 5 to 13 of ETV4 (Fig. 4A). Furthermore, we confirmed genomic fusions of FLJ35294-ETV1, CANT1-ETV4, and DDX5-ETV4 by fusion probe FISH assay (Fig. 5).

No case with ETV5 rearrangement was identified in the present cohort, consistent with our previous report suggesting that this gene fusion is relatively rare (6). Among the remaining 23 ETS transcription factors, a small proportion of prostate cancer cases revealed aberrant 5′ or 3′ deletions for ETS genes using our FISH screening strategy (Fig. 2). We found single cases of deletions in the 5′ end of ETV3, ELF1, and SPIC, and the 3′ end of ETV3, ELF2, and ELK3. Although FISH screening with a ∼1 Mb split probe approach yielded split signals for the ELK3 locus in case no.12 and ELF4 locus in case no.21, further FISH analysis by narrowing down the region identified the breakpoint location 220 Kb downstream of ELF4 gene and 140 Kb upstream of ELK3 gene. Of note, BACs used to detect ELK4 rearrangement were the same as used to detect SLC45A3 aberration because of proximity (∼20 Kb) of the two genes. Split signals were observed in four cases (nos. 4, 39, 67, and 102). Of these, case no. 4 and no. 102 were identified to harbor SLC45A3:ETV1 and SLC45A3:ERG fusions, respectively. None of these cases harbored fusions between ELK4 and SLC45A3 (data not shown). Thus, many of these singleton deletions are likely nonspecific, nonrecurrent lesions.

Furthermore, we screened our 110 patient TMA for all of the known 5′ partners of ETS aberrations in prostate cancer and found TMPRSS2 to be rearranged in 47% of cases (48 of 102). SLC45A3 was rearranged in 2% (2 of 89), followed by HNRPA2B1 in 1% (1 of 99), and C15ORF21 in 1% (1 of 88) of cases. HERV-K_22q11.23 rearrangement was not identified in this cohort. Notably, as shown in Fig. 2, FISH analysis revealed several cases with 5′ partner rearrangements without ETS gene aberrations, indicating these ones may harbor non-ETS gene fusion partners.

Next, we sought to characterize the 5′ fusion partners of ETV1 and ETV4 identified in this study. By QRT-PCR, ETV1 or ETV4 overexpression was confirmed in case nos. 54, 46, and 85, which harbored FLJ35294-ETV1, CANT1-ETV4, and DDX5-ETV4 fusions, respectively (Fig. 6A). As it is possible that aberrant ETS expression was driven by the regulatory elements of FLJ35294, CANT1, and DDX5, we explored their tissue specificities and androgen regulation in prostate cancer. Similar to TMPRSS2, SLC45A3, FLJ35294, and CANT1 showed marked overexpression in prostate cancer compared with other cancer types using the expOdata set of International Genomic Consortium accessed in Oncomine [Fig. 4B, P = 2.8 × 10⁻⁷ (FLJ35294); P = 6.8 × 10⁻⁷ (CANT1); ref. 7]. By contrast, DDX5 exhibited high expression in all tumor types (Fig. 4B), suggesting that it may have a housekeeping function rather than prostate specificity.

By QRT-PCR, expression of endogenous FLJ35294 (11.6-fold at 48 hours; P = 0.0006) and CANT1 (3.8-fold at 24 hours; P = 0.0014) were significantly induced by treatment with synthetic androgen R1881 in the LNCaP prostate cancer cell line (Fig. 6B). Furthermore, ChIP analysis showed AR occupancy of the promoter region in the FLJ35294 (9-fold enrichment; P = 0.005) and CANT1 (40-fold enrichment; P = 0.0014) genes (Fig. 6C). By contrast, the expression of DDX5 was not affected by R1881 stimulation (Fig. 6B) and AR was not recruited to the promoter region of DDX5 (Fig. 6C). Therefore, the FLJ35294-ETV1 and CANT1-ETV4 fusions can be categorized as class II gene fusions in prostate cancer, which include rearrangements involving fusions from prostate-specific androgen-induced 5′ partner genes (21), whereas DDX5-ETV4 may represent a class IV gene fusion, in which nontissue-specific promoter elements drive ETS gene expression.

Like the previously reported prostate cancer gene fusions, FLJ35294-ETV1 and CANT1-ETV4 fusions contain no predicted translated sequence (exons) from the 5′ partner. However, unlike almost all previously reported gene fusions in prostate cancer, the DDX5-ETV4 fusion transcript codes for a putative fusion protein, with an NH₂-terminal 102 amino acid stretch of DDX5 protein fused to 416 COOH-terminal amino acids of ETV4 (Fig. 6D). We confirmed that a fusion protein is expressed by this chimeric transcript by cloning the full-length cDNA from case no. 85 into an expression vector encoding an in-frame NH₂-terminal FLAG epitope tag. Upon transient transfection into HEK 293 cells, we detected the expression of the DDX5-ETV4 fusion protein across independently derived clones (Fig. 6D). To assess whether case no. 85 expresses the DDX5-ETV4 fusion protein, we subjected tissue extracts to immunoblot analysis using an antibody to the COOH-terminal end of ETV4 (Fig. 6D). As expected, only the DDX5:ETV4-positive case expressed the aberrant 57-kDa fusion protein. To our knowledge, this is the first report of characterization and expression of a fusion protein in prostate cancer.

In this study, we used a systematic and comprehensive approach to screen for recurrent gene rearrangements in the ETS family of oncogenic transcription factors. Although we used this approach to delineate ETS aberrations in prostate cancer, one could use this FISH screen in other epithelial tumors or sarcomas. Gene rearrangements, amplifications, and deletions of the ETS family are all captured by this assay. If the FISH evaluation processes were automated, one could envision scaling up this approach to cover all of the genes in the genome, which would help in discovery of recurrent gene fusions in cancer (which often are difficult to detect by array comparative genomic hybridization based technologies).

Combining the comprehensive FISH analyses carried out in this study with the published literature, we note that only ERG, ETV1, ETV4, and ETV5 gene fusions have been identified in prostate cancer. Notably, in this study, 44% (47 of 106) of prostate cancer cases do not harbor any ETS aberration, indicating a substantial fraction of prostate cancers can be clearly designated as being “ETS gene fusion negative.” This observation will stimulate research into identifying recurrent non-ETS aberrations or “nongene fusion” mechanisms in prostate cancers. This observation is also consistent with our previous work and that of others, which indicate that ETS fusion–positive and ETS fusion–negative cancers have distinct transcriptional signatures across profiling studies (7, 24). Interestingly, using a meta-Cancer Outlier Profile Analysis strategy, we have recently identified SPINK1 overexpression in a subset of ETS rearrangement–negative prostate cancers (∼11% of total cases; ref. 25).

In addition to ETS gene fusions, loss of PTEN activity is another common molecular aberration in prostate cancer (26–28). Yoshimoto and colleagues (29) have suggested that TMPRSS2-ERG fusion may be accompanied by PTEN deletion, and TMPRSS2-ERG fusion and PTEN deletion together are a predictor of PSA recurrence in prostate cancer. These findings suggest that these two leading genetic events may be concomitant in a subset of cases, and these aberrations may cooperate to promote prostate cancer progression.

Among the known 5′ fusion partners, we have shown the observed prevalence of chromosomal rearrangements in prostate cancer to be TMPRSS2 >> SLC45A3 > HNRPA2B1, CANT1, FLJ35294, C15ORF21, and HERV-K_22q11.2. Recently, SLC45A3 was identified as a 5′ fusion partner of ETV1 and ETV5 (6, 7). Our current findings confirmed SLC45A3 as a recurrent 5′ partner occurring in ∼2% of prostate cancer cases. More importantly, this is the first study that shows that SLC45A3 can serve as an alternative 5′ partner of ERG, as all previously reported ERG fusions involved TMPRSS2 as the 5′ partner. Thus, these findings suggest that other androgen-regulated genes can also serve as 5′ fusion partners of ERG. This study also identified cases with aberrations in 5′ fusion partners not associated with fusion to the ETS family genes, suggesting the possibility that other novel non-ETS genes are involved in chromosomal rearrangements in prostate cancer.

The ETS gene fusions that we have identified thus far are all mutually exclusive from one another and thus are presumably “drivers” for their respective tumor, although some of them are relatively rare in our cohort. One limitation of our current study is that our surgical case series may not be representative of general prostate cancer patient population. In a recent study, Demichelis and colleagues (17) observed a significant lower frequency of TMPRSS2-ERG fusions in a Swedish population–based prostate cancer cohort. Therefore, additional studies with larger population-based cohort may further define prevalence of ETS gene fusions in prostate cancer.

In a previous study, we identified several 5′ fusion partners involved in ETS fusions that are differentially regulated by androgen (androgen induced, androgen repressed, and androgen insensitive) that define distinct classes of ETS gene rearrangements (7). In the present study, we identified two prostate-specific, androgen-induced genes FLJ35294 and CANT1 partnering ETV1 and ETV4, repectively. Although this study was in preparation, Hermans and colleagues (23) reported KLK2 and CANT1 genes as novel fusion partners of ETV4 in prostate cancer. Interestingly, the CANT1-ETV4 fusion transcript in our study contained the same alternative 1st exon of CANT1 as described by Hermans and colleagues (23), which they reported was prostate specific. These observations also suggest that CANT1-ETV4 fusion is a recurrent aberration in prostate cancer.

To date, almost all of the gene fusions identified in prostate cancer have been chimeric mRNA transcripts with no associated fusion protein. Here, we report the first fusion protein in prostate cancer composed of the 102 NH₂-terminal amino acids of DDX5, DEAD (Asp-Glu-Ala-Asp) box polypeptide 5, fused to the COOH-terminal 419 amino acids of ETV4. This fusion protein seems to be expressed at high levels and presumably activates an ETS oncogenic program. How the NH₂-terminal 102 amino acids of DDX5, which seem to contain neither the DEAD nor the helicase domain, might affect ETV4 is unclear and would be best addressed in future studies. Interestingly, DDX5 is known to interact with a large number of signaling molecules including Steroid receptor coactivator, estrogen receptor α, GSK3b, etc.⁸

Additional sequence analysis of the genomic region of FLJ35294 shown in this study indicates the presence of an endogenous retroviral element (HERVK17p13.1). Hence this discovery has identified the second member of the HERVK class fusions in prostate cancer.

The University of Michigan has filed a patent on ETS gene rearrangements in prostate cancer, on which R. Mehra, S.A. Tomlins, and A.M. Chinnaiyan are coinventors, and the diagnostic field of use has been licensed to Gen-Probe Incorporated. Gen-Probe has not played a role in the design and conduct of the study, nor in the collection, analysis, or interpretation of the data, and had no involvement in the preparation, review, or approval of the article. A.M. Chinnaiyan serves as a consultant to Gen-Probe, Inc. The other authors disclosed no potential conflicts of interest.

Grant support: Department of Defense PC040517 (R. Mehra, W81XWH-06-1-0224), Department of Defense PC060266 (J. Yu), the NIH (Prostate Specialized Programs of Research Excellence P50CA69568, R01 CA102872), the Early Detection Research Network (UO1 CA111275-01), and the Prostate Cancer Foundation. A.M. Chinnaiyan is supported by a Clinical Translational Research Award from the Burroughs Welcome Foundation. S.A. Tomlins is supported by a Rackham Predoctoral Fellowship and is a Fellow of the Medical Scientist Training Program. S.A. Tomlins is supported by the GPC Biotech Young Investigator Award from the Prostate Cancer Foundation.

Fusion sequences identified in this article are deposited in Gene Bank, under the accession numbers EU693077, EU693078, and EU693079.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Jill Granger for editorial review of this article and Michele LeBlanc and Julie Kim for technical support.