Recurrent gene fusions involving oncogenic ETS transcription factors (including ERG, ETV1, and ETV4) have been identified in a large fraction of prostate cancers. The most common fusions contain the 5′ untranslated region of TMPRSS2 fused to ERG. Recently, we identified additional 5′ partners in ETV1 fusions, including TMPRSS2, SLC45A3, HERV-K_22q11.23, C15ORF21, and HNRPA2B1. Here, we identify ETV5 as the fourth ETS family member involved in recurrent gene rearrangements in prostate cancer. Characterization of two cases with ETV5 outlier expression by RNA ligase–mediated rapid amplification of cDNA ends identified one case with a TMPRSS2:ETV5 fusion and one case with a SLC45A3:ETV5 fusion. We confirmed the presence of these fusions by quantitative PCR and fluorescence in situ hybridization. In vitro recapitulation of ETV5 overexpression induced invasion in RWPE cells, a benign immortalized prostatic epithelial cell line. Expression profiling and an integrative molecular concepts analysis of RWPE-ETV5 cells also revealed the induction of an invasive transcriptional program, consistent with ERG and ETV1 overexpression in RWPE cells, emphasizing the functional redundancy of ETS rearrangements. Together, our results suggest that the family of 5′ partners previously identified in ETV1 gene fusions can fuse with other ETS family members, suggesting numerous rare gene fusion permutations in prostate cancer. [Cancer Res 2008;68(1):73–80]
Whereas gene fusions are common in hematologic and mesenchymal malignancies, until recently, they have not been well defined in common epithelial tumors. To nominate candidate oncogenes from DNA microarray data, we developed a bioinformatics approach [cancer outlier profile analysis (COPA)] to identify genes with marked overexpression in a subset of cancers (1). COPA identified ERG and ETV1 as prominent “outliers” across multiple prostate cancer data sets. Through several molecular techniques, we identified fusions of the 5′ untranslated region (5′-UTR) of TMPRSS2 (21q22) to ERG (21q22) or ETV1 (7p21) in cases that overexpressed the respective ETS family member (1). Subsequently, rare fusions of TMPRSS2 and the ETS family member ETV4 have also been identified (2, 3). Multiple studies have shown that TMPRSS2:ERG fusions are the most predominant subtype of ETS gene fusions (∼50% of prostate cancers), and fusions involving ETV1 or ETV4 are rare (∼1–10% of prostate cancers; refs. 1–7).
More recently, we discovered additional 5′ fusion partners involved in ETV1 gene fusions, including the 5′-UTRs from SLC45A3, HERV-K_22q11.3, C15ORF21, and HNRPA2B1 (4). Importantly, because these 5′ partners are differentially regulated by androgen (androgen-induced, androgen-repressed and androgen insensitive), they define distinct classes of ETS gene rearrangements. To date, these additional 5′ partners have only been identified in ETV1 fusions, and it is unknown if they can fuse with ERG (in rare TMPRSS2:ERG negative cases with ERG outlier expression) or additional ETS family members.
Here we report the discovery of TMPRSS2:ETV5 and SLC45A3:ETV5 gene fusions, identifying a fourth ETS family member, ETV5 (3q27), involved in recurrent gene rearrangements in prostate cancer. This study also shows that the family of 5′ fusion partners we previously identified can fuse to additional ETS family members, suggesting numerous rare ETS gene fusion combinations.
Materials and Methods
Samples and cell lines. Prostate tissues were from the radical prostatectomy series and the Rapid Autopsy Program (8), which are both part of the University of Michigan Prostate Cancer Specialized Program of Research Excellence Tissue Core. Samples were collected with informed consent and prior institutional review board approval. The benign immortalized prostate cell line RWPE was obtained from the American Type Culture Collection. Total RNA was isolated from all samples with Trizol (Invitrogen).
In silico analysis of ETV5 outlier expression. The normalized expression values for ERG, ETV1, ETV4, and ETV5 from four prostate cancer profiling studies [Glinsky et al. (9), Lapointe et al. (10), Vanaja et al. (11), and Yang et al. (GSE8218)] and this study (X. Cao et al.) in the Oncomine database (12) were downloaded, and heat maps were generated using Cluster 3.0 and Java Treeview.
Quantitative PCR. Quantitative PCR was done using SYBR Green dye on an Applied Biosystems 7300 Real-time PCR system (Applied Biosystems) as described (1, 2, 4). Oligonucleotide primers were synthesized by Integrated DNA Technologies and are listed in Supplementary Table S1.
RNA ligase–mediated rapid amplification of cDNA ends. RNA ligase–mediated rapid amplification of cDNA ends (RLM-RACE) was done using the GeneRacer RLM-RACE kit (Invitrogen) according to the manufacturer's instructions as described (1, 2, 4). To obtain the 5′ end from PCa_ETV5_1, first-strand cDNA was amplified using the GeneRacer 5′ primer and ETV5_exon6_racer-r. For amplification from PCa_ETV5_2, ETV5_exon11-r was used.
Reverse-transcription PCR. To confirm the expression of TMPRSS2:ETV5 fusion transcripts in PCa_ETV5_1, we carried out reverse transcription PCR (RT-PCR) as described (1). cDNA was PCR amplified with Platinum Taq High Fidelity using primers for TMPRSS2:ETV5a/TMPRSS2:ETV5b and TMPRSS2:ETV5c (Supplementary Table S1), and products were resolved by electrophoresis, cloned into pCR4-TOPO, and sequenced as described (1).
Fluorescence in situ hybridization. Interphase fluorescence in situ hybridization (FISH) on formalin-fixed paraffin-embedded tissue sections was done as described (2, 4). Bacterial artificial chromosomes (listed in Supplementary Table S2) were obtained from the BACPAC Resource Center.
In vitro ETV5 overexpression. Full-length human ETV5 cDNA (BC007333) in the Gateway compatible vector pDNR-Dual was obtained from the Harvard Institute of Proteomics (HsCD00003658). Adenoviral and lentiviral constructs were generated by recombination with pAD/CMV/V5 (Invitrogen) and pLenti6/CMV/V5 (Invitrogen), respectively, using LR Clonase II (Invitrogen). Control pAD/LACZ clones were obtained from Invitrogen and Control pLenti6/GUS clones were generated by recombination using a control entry clone (pENTR-GUS, Invitrogen). The University of Michigan Vector Core generated the viruses. The benign immortalized prostate cell line RWPE was infected with adenoviruses expressing ETV5 or LACZ and used 48 h after infection. RWPE cells were also infected with lentiviruses expressing ETV5 or GUS, and stable clones were generated by selection with blasticidin (Inivtrogen). ETV5 expression was confirmed by immunoblotting with a mouse monoclonal anti-ETV5 antibody (Abnova) at 1:500 dilution. Mouse monoclonal anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibody (Abcam) was applied at 1:30,000 dilution for loading control.
Invasion assays. Invasion assays were done as described (4). Equal numbers of the indicated cells were seeded onto basement membrane matrix (extracellular matrix, Chemicon) present in the insert of a 24-well culture plate, with fetal bovine serum added to the lower chamber as a chemoattractant. After 48 h, noninvading cells and extracellular matrix were removed with a cotton swab.
For inhibitor studies, amiloride (20 μmol/L, EMD Biosciences); matrix metalloproteinase (MMP)-3, MMP2/MMP9, MMP8, and the pan-MMP inhibitor GM 6001 (all 10 μmol/L, EMD Biosciences); or vehicle control was added to cells for 24 h before trypsinization and seeding for invasion assays, as described above. For plasminogen activator inhibitor (PAI)-1, cells were trypsinized and treated with the indicated amount of recombinant PAI-1 (EMD Biosciences) for 15 min at indicated concentrations before seeding, as described above.
Expression profiling. Expression profiling of transient RWPE-ETV5 and RWPE-LACZ cells was done using the Agilent Whole Human Genome Oligo Microarray as described with a dye flip hybridization (4). Overexpressed and underexpressed signatures were generated by filtering to include only features with significant differential expression (Plog ratio < 0.01) in both hybridizations after correction for a dye flip. Signatures were loaded into the Oncomine Concepts Map44, 13, 14).
Results and Discussion
In this study, we applied COPA to an expression profiling data set generated as part of an integrative molecular study of prostate cancer progression.5
X Cao et al., in preparation.
Because our original application of COPA did not identify ETV5 as an outlier in prostate cancer profiling studies (1), we hypothesized that the 17% prevalence in localized prostate cancers in this study was an overestimate of the true prevalence of ETV5 outlier expression. Thus, we examined ETV5 expression across several prostate cancer profiling data sets in the Oncomine database, revealing only two additional cases with ETV5 outlier expression [total, 4 of 269 (∼1.5%); Fig. 2], suggesting that ETV5 outlier expression is indeed rare, similar to ETV4 (∼1.1%; Fig. 2). To confirm both the ETV5 outlier expression in our two samples and the rarity of ETV5 outlier expression, we carried out quantitative PCR on samples from our expression profiling data set as well as 44 additional localized prostate cancers. This analysis confirmed ETV5 outlier expression in PCa_ETV5_1 and PCa_ETV5_2, whereas the additional screen of 44 prostate cancers revealed no further ETV5 outlier cases (Fig. 1B). Together, these results suggest that similar to ETV4, ETV5 outlier expression is rare in prostate cancer.
In previous studies, >95% of cases with ERG or ETV1 outlier expression harbored chromosomal rearrangements (1, 2, 4, 15), suggesting that rearrangements similarly cause ETV5 outlier expression. Previously, we used an exon-walking quantitative PCR strategy that showed loss of outlier expression at the 5′ end of the ETS transcript in cases that harbored gene fusions (1, 2). Using 10 primer pairs across the 13 exons in the ETV5 transcript, we found outlier expression of ETV5 exons 2 to 13 in PCa_ETV5_1 and ETV5 exons 8 to 13 in PCa_ETV5_2 (Fig. 1C). Thus, to identify a potential 5′ partner, we carried out RLM-RACE, which identified three products in PCa_ETV5_1 and one product in PCa_ETV5_2. Sequencing of the PCa_ETV5_1 products showed three different TMPRSS2:ETV5 fusions (TMPRSS2:ETV5a–TMPRSS2:ETV5c; Fig. 3A). The TMPRSS2:ETV5a transcript contained exon 1 of TMPRSS2 fused to exon 2 of ETV5 and TMPRSS2:ETV5b contained exons 1 to 3 of TMPRSS2 fused to exon 2 of ETV5. In the TMPRSS2:ETV5c transcript, sequencing revealed a distinct 1st exon and exons 2 to 3 of TMPRSS2 fused to exon 2 of ETV5. This 1st exon is ∼1.5 kb downstream of the reference sequence 1st exon and overlaps with a reported EST (DA460061; Supplementary Fig. S1).
Sequencing of the PCa_ETV5_2 RLM-RACE product revealed exon 1 of SLC45A3 fused to exon 8 of ETV5 (Fig. 3A). Recently, we identified SLC45A3 as a 5′ fusion partner in ETV1 gene rearrangements (4). Our discovery of the SLC45A3:ETV5 fusion confirms SLC45A3 as another recurrent 5′ partner in ETS gene fusions and suggests that the other 5′ fusion partners may be involved in additional rare ETS rearrangements. Together, RLM-RACE of both cases confirmed the exon-walking quantitative PCR expression patterns and shows the existence of ETV5 gene fusions in prostate cancers with ETV5 outlier expression.
To confirm the presence of the fusion transcripts found in PCa_ETV5_1 and PCa_ETV5_2, we carried out quantitative PCR. As shown in Fig. 3B, quantitative PCR detected TMPRSS2:ETV5b/c exclusively in PCa_ETV5_1 and SLC45A3:ETV5 in PCa_ETV5_2. To confirm the expression of individual TMPRSS2:ETV5 isoforms in PCa_ETV5_1, we carried out RT-PCR using primers specific for TMPRSS2:ETV5a/TMPRSS2:ETV5b and TMPRSS2:ETV5c. Sequencing of amplified products using TMPRSS2:ETV5a/TMPRSS2:ETV5b primers confirmed the expression of both TMPRSS2:ETV5a and TMPRSS2:ETV5b (Supplementary Fig. S2). PCR using primers for TMPRSS2:ETV5c produced two products, which were identified as TMPRSS2:ETV5c and TMPRSS2:ETV5d, another isoform containing the novel 1st exon of TMPRSS2 (found in TMPRSS2:ETV5c) fused to exon 2 of ETV5 (Fig. 3A and Supplementary Fig. S2).
Next, to confirm these fusions at the genomic level, we carried out FISH using split signal assays around the 5′ partners (TMPRSS2 and SLC45A3) as well as fusion assays (5′ TMPRSS2 or SLC45A3 and 3′ ETV5; Fig. 3C). In PCa_ETV5_1, multiple hybridizations failed to produce interpretable signals for TMPRSS2, ETV5, or ERG probes (positive control) because the only tumor cells in the tissue section were present in a small focus at the extreme edge. Review of all blocks from the case failed to yield more informative sections. However, in PCa_ETV5_2, we confirmed rearrangements in SLC45A3 and ETV5 and fusion of the 5′ SLC45A3 and 3′ ETV5 signals (Fig. 3C). Together, quantitative PCR and RT-PCR validated the presence of TMPRSS2:ETV5 and SLC45A3:ETV5 in PCa_ETV5_1 and PCa_ETV5_2, respectively, and FISH confirmed the fusion of the SLC45A3 and ETV5 genomic loci in PCa_ETV5_2.
In previous studies, we and others have shown that ERG and ETV1 mediate invasiveness in prostate cancer cell lines harboring ETS rearrangements and benign prostate cell lines ectopically overexpressing ETS family members (4, 16).6
S. A. Tomlins et al., submitted for publication.
To investigate the transcriptional program regulated by ETV5 overexpression, we profiled transient RWPE-ETV5 cells and analyzed the expression signatures using the Oncomine Concepts Map,4 a tool for analyzing associations between >20,000 biologically related gene sets by disproportionate overlap (13, 14). Previously, Oncomine Concepts Map analysis has identified enrichment of concepts related to invasion in our “overexpressed in RWPE-ERG or RWPE-ETV1” signatures, consistent with the phenotypes of these cells (4).6 We identified 420 features overexpressed in transient RWPE-ETV5 compared with RWPE-LACZ cells.
Whereas we observed enrichment of our “overexpressed in RWPE-ERG or RWPE-ETV1” concepts in our “overexpressed in RWPE-ETV5” signature (Fig. 4B), we unexpectedly observed more significant enrichment with our “underexpressed in RWPE-ERG or RWPE-ETV1” signatures. For example, our underexpressed and overexpressed in RWPE-ERG (transient) signatures were both enriched in our overexpressed in RWPE-ETV5 signature (OR, 7.97 and 2.45; P = 1.3e−25 and 5e−4, respectively). Because distinct subsets of genes overexpressed in RWPE-ETV5 cells were both overexpressed and underexpressed in RWPE-ERG and RWPE-ETV1 cells, this suggests that ETV5, ERG, and ETV1 differentially regulate a common set of target genes when overexpressed in benign prostate cells.
Importantly, Oncomine Concepts Map analysis identified a network of invasion related concepts that shared enrichment with our overexpressed in RWPE-ETV5, RWPE-ERG, and RWPE-ETV1 concepts, such as “up-regulated genes in HMLHT (or MCF-10A) breast cells expressing STAT3-C” (OR, 5.38; P = 9.9e−5). In HMLHT and MCF-10A cells, ectopic expression of STAT3-C resulted in increased invasion in a MMP9-dependent manner without effecting proliferation (17). Importantly, we observed overexpression of MMP9 and PLAU in RWPE-ETV5 cells, as we have previously shown that ERG-mediated invasion in RWPE cells is inhibited by amiloride, a specific inhibitor of PLAU, and PAI-1, a PLAU and PLAT inhibitor.6 In this study, we show that amiloride, PAI-1, and a MMP2/MMP9 inhibitor inhibit ETV5-mediated invasion (Fig. 4C). Together, the overlapping target genes of ERG, ETV1, and ETV5 support the functional redundancy of ETS rearrangements in prostate cancer.
ETV1, ETV4, and ETV5 compose the PEA3 subfamily of ETS genes. Members of this subfamily share a highly conserved ETS binding domain (∼95% amino acid homology) and are almost 50% identical along the full protein (18). In this study, COPA identified ETV5 as an outlier in our prostate cancer expression profiling microarray data set. In this data set, the incidence of ETV5 overexpression was ∼17% (2 of 12 localized prostate cancers), but an in silico screen of 269 prostate cancer cases and quantitative PCR analysis of an additional 44 prostate cancers revealed ETV5 overexpression in ∼1% (4 of 313) of cases, similar to ETV4. Our work here shows that all PEA3 subfamily members are involved in rare ETS rearrangements in prostate cancer (only 1–8% of cases), with the increased prevalence of ERG fusions possibly due to the proximity of ERG and TMPRSS2 on chromosome 21.
We and others have shown TMPRSS2 to be a 5′ fusion partner for all ETS genes with known rearrangements (1–6, 19–21). Consistent with previous TMPRSS2 fusions to ERG, ETV1, or ETV4, where several isoforms have been reported per case (1, 6, 20, 22), here we identified four TMPRSS2:ETV5 isoforms in PCa_ETV5_1. In most ETS fusions, the 5′ partner does not contribute to the coding sequence of the fusion transcript, suggesting the production of a truncated ETS protein and the importance of the promoter region of the 5′ partner in driving ETS gene outlier expression. In TMPRSS2:ETV5a, exon 1 of TMPRSS2 is fused to ETV5 at exon 2, allowing for production of the full-length ETV5 protein (reported start codons of TMPRSS2 and ETV5 are in exons 2 and 2, respectively; see Fig. 3A). Alternatively, TMPRSS2:ETV5b and TMPRSS2:ETV5c transcripts contain the start codon for TMPRSS2 in frame with the untranslated and coding sequence of ETV5, possibly resulting in the production of a fusion protein. Importantly, expression of TMPRSS2:ERG isoforms that encode a fusion production has been associated with more aggressive prostate cancers (20). Additional ETV5 fusion positive cancers will need to be characterized to determine if chimeric protein coding ETV5 fusions are similarly aggressive.
Additionally, the TMPRSS2:ETV5c/d transcripts contained a distinct 1st exon compared with the 1st exon of the reported TMPRSS2 reference sequence. This 1st exon in our TMPRSS2:ETV5c/TMPRSS2:ETV5d transcripts is ∼1.5 kb downstream of the reference sequence 1st exon and overlaps with the EST DA460061, which was originally found in a tongue tumor library (Supplementary Fig. S1). Distinct sequences of TMPRSS2 involved in ETS gene fusions have been found by other studies. Recently, Lapointe et al. showed that 34 of 63 (54%) cases expressed a novel 1st exon of TMPRSS2 that is ∼4 kb upstream of the reference sequence and overlaps with three ESTs (DA872508, DA868984, and DA870830; ref. 6; Supplementary Fig. S1). In addition, Lapointe et al. (6) report the same 1st exon in TMPRSS2 that we found in our TMPRSS2:ETV5c transcript in a single TMPRSS2:ERG case. Additionally, the TMPRSS2:ETV4 rearrangement showed two fusion transcripts that each contained a novel 1st exon of TMPRSS2 that is ∼8 kb upstream of the reported reference sequence (ref. 2, Supplementary Fig. S1). Together, these results support our hypothesis that the dysregulation of ETS family members is driven by androgen response elements within the upstream sequence of TMPRSS2 and other previously identified 5′ partners. Whereas we were unable to show the TMPRSS2:ETV5 fusion at the genomic level using FISH due to lack of tissue, the presence of multiple splice variants in other TMPRSS2:ETS gene fusions confirmed by FISH suggests that the transcripts are produced as the result of a chromosomal rearrangement rather than trans-splicing.
We recently discovered novel 5′ fusion partners to ETV1, including SLC45A3, HERV-K_22q11.23, C15ORF21, and HNRPA2B1. In this study, by applying COPA to a prostate cancer profiling data set, we identified rare ETV5 gene fusions with two different 5′ partners, TMPRSS2 and SLC45A3. Ectopic overexpression of ETV5 in benign prostate cells induced invasion and an invasive transcriptional program that is analogous to that seen with the overexpression of ERG and ETV1. Collectively, this study identifies novel ETS gene fusions involving the third member of the PEA3 subfamily, ETV5, and shows that the family of 5′ partners can fuse to multiple ETS family members, supporting the existence of multiple rare ETS gene fusions.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
B.E. Helgeson, S.A. Tomlins, and N. Shah contributed equally to this work.
Grant support: Department of Defense grant PC040517 (no. W81XWH-06-1-0224; R. Mehra); NIH Prostate Specialized Program of Research Excellence grants P50CA69568 and R01 CA102872; Early Detection Research Network grant UO1 CA111275-01; the Prostate Cancer Foundation; a sponsored research agreement from Gen-Probe, Inc.; Clinical Translational Research Award from the Burroughs Welcome Foundation (A.M. Chinnaiyan); and a Rackham Predoctoral Fellowship (S.A. Tomlins). S.A. Tomlins is a Fellow of the Medical Scientist Training Program.
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 S. Dhanasekaran, J. Siddiqui, M. LeBlanc, L. Wang, A. Menon, and B. Han for technical assistance, and the University of Michigan Vector Core for virus generation.