Truncated ETV1, Fused to Novel Tissue-Specific Genes, and Full-Length ETV1 in Prostate Cancer

In this study, we describe the properties of novel ETV1 fusion genes, encoding N-truncated ETV1 (dETV1), and of full-length ETV1, overexpressed in clinical prostate cancer. We detected overexpression of novel ETV1 fusion genes or of full-length ETV1 in 10% of prostate cancers. Novel ETV1 fusion partners included FOXP1, an EST (EST14), and an endogenous retroviral repeat sequence (HERVK17). Like TMPRSS2, EST14 and HERVK17 were prostate-specific and androgen-regulated expressed. This unique expression pattern of most ETV1 fusion partners seems an important determinant in prostate cancer development. In transient reporter assays, full-length ETV1 was a strong transactivator, whereas dETV1 was not. However, several of the biological properties of dETV1 and full-length ETV1 were identical. On stable overexpression, both induced migration and invasion of immortalized nontumorigenic PNT2C2 prostate epithelial cells. In contrast to dETV1, full-length ETV1 also induced anchorage-independent growth of these cells. PNT2C2 cells stably transfected with dETV1 or full-length ETV1 expression constructs showed small differences in induced expression of target genes. Many genes involved in tumor invasion/metastasis, including uPA/uPAR and MMPs, were up-regulated in both cell types. Integrin β3 (ITGB3) was clearly up-regulated by full-length ETV1 but much less by dETV1. Based on the present data and on previous findings, a novel concept of the role of dETV1 and of full-length ETV1 overexpression in prostate cancer is proposed. [Cancer Res 2008;68(18):7541–9]

The ETS transcription factor family is composed of 27 members (1–3). Depending on the cellular context, they can function as transactivators or transrepressors. ETS transcription factors modulate many cellular functions, including proliferation, apoptosis, differentiation, tissue remodeling, migration, invasion, and angiogenesis (1–3). Altered expression or properties of ETS transcription regulators affect the control of these processes.

Recurrent chromosomal rearrangements are well defined in leukemias, lymphomas, and sarcomas (4). These rearrangements result in fusion genes that express oncogenic proteins with altered properties or in overexpression of wild-type oncogenes. In Ewing's sarcoma and in acute myeloid leukemia, gene fusions of members of the ETS gene family have been found. At low frequency, gene fusions have also been described in solid tumors (4). However, recent analyses showed common gene fusions in prostate cancer, all involving members of the ETS transcription factor family (5–9).

In 40% to 70% of clinical prostate cancers, ERG (21q22.1) is directly linked to androgen-regulated, prostate-specific TMPRSS2, which is located 3 Mbp upstream of ERG. At low frequency, fusions of TMPRSS2 to ETV1, ETV4, and ETV5, which map on different chromosomes, have been described (5, 7, 8, 10).

Considering the complexity of fusion genes in hematologic and mesenchymal malignancies, we questioned whether this would also be true for gene fusions in prostate cancer. Here, we describe overexpression of ETV1 in 8 of 84 clinical prostate cancer samples. In four samples, full-length ETV1 is overexpressed, but in the other four samples, we detected novel ETV1 fusion genes, which result in predicted N-truncated ETV1 proteins. Novel fusion partners include FOXP1, an EST (EST14), and an endogenous retroviral sequence (HERVK17, identified in two samples). Like TMPRSS2 (11), both EST14 and HERVK17 are androgen regulated and prostate specific.

Transient reporter assays with full-length ETV1 and N-truncated ETV1 (dETV1) showed that these proteins possess different transcription regulation functions. However, quantitative PCR (QPCR) analysis of prostate epithelial cells with stable overexpression of full-length ETV1 or dETV1 indicated less pronounced differences in expression of candidate target genes. Biological assays showed no significant difference in migration and invasion properties between full-length ETV1 and dETV1-expressing cells. However, full-length ETV1 is capable of inducing anchorage-independent growth, whereas dETV1 is not. We propose a different role of dETV1 and full-length ETV1 in prostate cancer.

Tissue samples. Primary prostate cancer samples were obtained by radical prostatectomy, regional lymph node metastases were collected during surgery, and recurrences were obtained by transurethral resection. Samples were snap frozen and stored in liquid nitrogen. Use of the samples for research purposes was approved by the Erasmus MC Medical Ethics Committee according to the Medical Research Involving Human Subjects Act (MEC-2004-261).

H&E-stained tissue sections were histologically evaluated by two pathologists (T.H. van der Kwast and A. van Leenders). All samples contained at least 70% tumor cells.

BALB/c mouse prostate tissues were collected at different developmental stages (16.5 and 18.5 embryonal days and postnatal days 3, 9, 15, and 50).

RNA and DNA preparations. RNA from clinical prostate cancer specimens was isolated from frozen tissue sections using RNA-Bee (Campro Scientific). DNA was isolated from frozen sections using the DNeasy DNA Extraction kit (Qiagen). RNAs from the prostate cancer cell lines LNCaP and DuCaP cultured in the presence of 10⁻⁹ mol/L of the synthetic androgen R1881, or in the absence of hormone, from PNT2C2 cells overexpressing full-length ETV1 or dETV1, and from mouse prostates of different developmental stages were isolated using the RNeasy RNA Extraction kit (Qiagen).

Breakpoint mapping. Positions of fusion points were mapped by standard long-range PCR on 200 ng genomic DNA in the presence of 0.5 μmol/L of each forward (fusion partner) and reverse (ETV1) primer with Taq polymerase and ProofStart DNA polymerase (Qiagen). For primers, see Supplementary Table S1. PCR products were separated on a 1% agarose gel. Specific amplified fragments were isolated and sequenced in an ABI 3100 genetic analyzer (Applied Biosystems).

mRNA expression. mRNA expression was analyzed by reverse transcription-PCR (RT-PCR) or by QPCR. cDNA was prepared with Moloney murine leukemia virus reverse transcriptase (Invitrogen Life Technologies) and an oligo(dT)₁₂ primer. cDNAs of 16 different tissues were purchased from Clontech.

RT-PCR products were analyzed over a 1.5% agarose gel. QPCR was done in Power SYBR Green PCR Master Mix (25 μL) containing 0.33 μmol/L forward and reverse primer in an ABI Prism 7700 Sequence Detection System (Applied Biosystems). Amplified products were quantified relative to porphobilinogen deaminase (PBGD; human RNAs) or hypoxanthine guanine phosphoribosyl transferase 1 (Hprt; mouse RNAs) by the standard curve method (Applied Biosystems). For primers, see Supplementary Table S1.

RNA ligase-mediated rapid amplification of cDNA ends. 5′-RNA ligase-mediated rapid amplification of cDNA ends (RLM-RACE) was performed using the GeneRacer kit (Invitrogen). cDNA was amplified with Taq polymerase (Qiagen) using the GeneRacer 5′-primer and a gene-specific primer (ETV1 exon 6 reverse). RACE PCR products were analyzed on a 1.5% agarose gel, and bands were excised, purified, and sequenced.

Interphase fluorescence in situ hybridization. Interphase fluorescence in situ hybridization (FISH) was done on 5-μm frozen tissue sections as described previously (5). Bacterial artificial chromosome (BAC) clones RP11-79G16 (ETV1), RP11-154H23 (FOXP1), RP11-460G19 (EST14), RP11-1099M24, and RP11-1B5 (both flanking HERVK17; see Fig. 2A) were purchased from BacPac Resources. Specificity of BACs was confirmed on metaphase chromosome spreads. BAC DNA clones were either Spectrum Orange or Spectrum Green labeled using a Nick Translation Reagent kit (Vysis). Tissue sections were counterstained with 4′,6-diamidino-2-phenylindole in antifade solution (Vector Laboratories). Images of the three fluorochromes were collected on an epifluorescence microscope (Leica DM) equipped with appropriate filter sets (Leica) and a charge-coupled device cooled camera (Photometrics).

Construction of expression plasmids. cDNAs of full-length ETV1 and the different ETV1 fusion transcripts were PCR amplified and cloned into pGEMT-easy (Promega). For primers, see Supplementary Table S1. Inserts were sequence verified and subsequently cloned into the Not1 site in the pcDNA3 expression vector (Invitrogen). Similarly, full-length ETV1 cDNA and dETV1 cDNA were integrated in the expression vector pWPXLd (provided by Didier Trono, University of Geneva, Geneva, Switzerland).

Reporter assays. LNCaP prostate tumor cells and immortalized nontumorigenic PNT2C2 prostate epithelial cells (provided by Norman Maitland, University of York, York, United Kingdom; ref. 12) were grown in DMEM supplemented with 5% FCS and antibiotics. Cells were cotransfected with full-length ETV1 or truncated ETV1 expression constructs and the ETS reporter PALx8-TK-Luc (provided by Boh Wasylyk, Institute of Genetics and Molecular and Cellular Biology, Strasbourg, France) essentially as described (13). Cells were harvested after 24 h and luciferase activity was measured in a LUMAC 2500 Biocounter.

Western blot analysis. For Western blot analysis, LNCaP cells were transfected with pcDNA3-ETV1 or pcDNA3-dETV1 expression construct or empty vector. PNT2C2 cells were transfected with pWPXLd-ETV1, pWPXLd-dETV1 expression vectors, or control pWPXLd-GFP. Cells were harvested after 48 h. Western blot analysis was carried out using standard procedure with ER81 (COOH terminus; Santa Cruz Biotechnology) and β-actin loading control (Sigma) antibodies. Bands were visualized by chemiluminescence (Pierce).

Infection with ETV1 lentivirus. To obtain lentiviruses, 293T cells were cotransfected with pWPXLd-ETV1, pWPXLd-dETV1, or pWPXLd-GFP (control) and pPAX2 and pMD2.G (Didier Trono) using the calcium phosphate precipitation method. PNT2C2 cells were infected with lentiviruses expressing either full-length ETV1 or dETV1, or with control virus. Pools of infected cells were propagated and used in the biological assays as described below.

Migration and invasion assays. Migration and invasion assays of PNT2C2-ETV1, PNT2C2-dETV1, and control PNT2C2-GFP cells (1 × 10⁵ per well) were performed according to the instructions of the manufacturer of the Transwells (Chemicon). The migration assay was stopped after 24 h of incubation; the invasion assay was terminated after 48 h.

Proliferation assay. Equal amounts of PNT2C2-ETV1, PNT2C2-dETV1, and control PNT2C2-GFP cells were seeded in T25 culture flasks. At days 0, 2, 4, 6, and 8, thiazolyl blue tetrazolium bromide dissolved in PBS (MTT reagent; AppliChem) was added, and after 4 h, cells were harvested. Cells were suspended in DMSO-Sörensen buffer and absorbance at 570 nm was measured.

Soft agar assay. A bottom layer of 0.6% low-melting agarose in normal culture medium was prepared in six-well culture plates. On top, a layer of 0.3% agarose containing 1 × 10⁴ cells (PNT2C2-ETV1, PNT2C2-dETV1, or control PNT2C2-GFP) was plated. At day 14, cells were stained with crystal violet and numbers of colonies in representative microscope fields were counted.

We investigated 84 clinical prostate cancer samples (49 primary tumors, 11 lymph node metastases, and 24 recurrent tumors) for ETV1 overexpression. In eight samples, divided over each clinical subgroup, ETV1 overexpression was found (Fig. 1A); however, QPCR failed to detect TMPRSS2-ETV1 fusion transcripts (Fig. 1B). QPCR with two ETV1 primer sets, one amplifying an exon 1 to 2 fragment and a second set amplifying an exon 11 to 12 fragment, showed in four samples (37, 89, 308, and 247) an ∼1:1 signal ratio, indicative for full-length ETV1 expression (Fig. 1C). However, in four other samples (32, 104, 116, and 342), a high exon 11 to 12 to exon 1 to 2 ratio, indicative for gene fusion, was detected. 5′-RACE followed by sequencing revealed that the four tumor samples with equal signal intensities for the two amplified ETV1 fragments indeed overexpressed full-length ETV1. Novel fusion genes were present in samples with high ETV1 exon 11 to 12 to exon 1 to 2 ratios. These novel ETV1 fusion partners were FOXP1, a gene encoding a spliced EST (here denoted EST14), and an endogenous retroviral repeat sequence (denoted HERVK17; two samples). All fusion transcripts were confirmed by RT-PCR (Fig. 1D).

In contrast to ERG and TMPRSS2, ETV1 and its three novel fusion partners all map to different chromosomes: ETV1 is located on 7p, and FOXP1, EST14, and HERVK17 on 3p, 14q, and 17p, respectively (Fig. 2A). Chromosomal rearrangements in the four samples with fusion transcripts were confirmed by interphase FISH with specific BAC probes (Fig. 2A and B). HERVK17 did not function as a retrotransposon (14) because in both samples 104 and 116 split signal FISH with flanking BACs showed separation of genomic fragments proximal and distal to one HERVK17 copy (Fig. 2C). Their appropriate orientations (Fig. 2A) allow fusion of ETV1 to FOXP1 and EST14 by standard recurrent chromosomal translocations. The HERVK17-ETV1 fusion can most likely be explained by the integration of an ETV1 genomic segment into the fusion chromosome (17p).

For fusions of HERVK17 and EST14 to ETV1, we precisely mapped the fusion point by long-range PCR followed by sequencing (Fig. 2D; Supplementary Fig. S1). The breakpoints in ETV1 are in intron 4 (104 and 342) and in intron 5 (116); the latter breakpoint is in an Alu repeat (Fig. 2D; Supplementary Fig. S1). The breakpoint in EST14 is in its only intron; however, the genomic alteration turned out to be more complex (Fig. 2D). Additionally to the fusion to ETV1, a 117-kbp deletion from EST14 to C14orf25 (intron 4), removing FOXA1, was found. Both the interstitial deletion and the fusion to ETV1 involve a LINE retroviral repeat, pointing to a role of this sequence in genomic instability (14, 15). HERVK17 is a defective retroviral sequence (Fig. 2A and D). The two breakpoints in HERVK17 mapped within 1 kbp from each other, both in a HERVK sequence flanking the 3′-LTR.

To increase our knowledge on properties and expression of fusion genes, we studied in detail the composition of fusion transcripts and the regulation of expression of the various ETV1 fusion partners, including TMPRSS2. Figure 3A schematically summarizes the major ETV1 fusion transcripts found by RT-PCR and sequencing. Note that part of the fusion transcripts lacks ETV1 exon 5 due to alternative splicing. Depending on the transcript, the ATG start codon is provided by ETV1 exon 6 or by the fusion partner. In all cases, the stop codon is in ETV1 exon 12.

In the FOXP1-ETV1 fusion transcript, part (154 bp) of FOXP1 exon 11 is coupled to the ETV1 sequence (Fig. 3A; Supplementary Fig. S2). The same part of this exon is present at the 5′-end of mRNA encoding the FOXP1C isoform (16). Unfortunately, the complex structure of the FOXP1 locus, including several different first exons and extensive alternative splicing, prevented accurate detailed analyses of its expression.

EST14 maps between MIPOL1 and FOXA1 on 14q21.1. The EST14-ETV1 fusion transcript contains part of the known exon 1 of this two exon EST (UniGene Hs.229997), linked to ETV1 (Fig. 3A; Supplementary Fig. S3). EST14 is weakly androgen regulated in LNCaP and DuCaP prostate cancer cells (Fig. 3B). EST14 expression is highest in the prostate (Fig. 3C). Similarly, the flanking gene FOXA1, but not MIPOL1, is preferentially expressed in the prostate, indicating a common control region (Supplementary Fig. S4). Expression of the mouse Est14 ortholog (mEst14; Mm.387080) is also prostate specific (data not shown). Like expression of mTmprss2, expression of mEst14 increased during mouse prostate development and is highest in the adult prostate (Fig. 3D).

HERVK17 maps between ALOX15B and ALOX12B on 17p13.1 (Fig. 2A). The 5′-LTR of this HERV has promoter activity, and several unspliced transcripts from this promoter are known (Hs.336697; transcript 1 in Fig. 2D). HERVK17-ETV1 fusion transcripts are composed of 264 bp of the retroviral transcript linked to either ETV1 exon 5 or 6 (Fig. 3A; Supplementary Fig. S5). In wild-type retroviruses, the same splice donor site, as used here for HERVK17-ETV1 transcripts, is used to remove the gag sequence and produces mRNAs encoding pol and env. We detected by RT-PCR in prostate cells a novel HERVK17 transcript, starting in the 5′-LTR (exon 1), followed by a second exon in the HERV sequence downstream of the 3′-LTR (transcript 2, Fig. 2D). The breakpoints are in the intron of this novel transcript. Expression of HERVK17 is strongly androgen regulated and even more prostate specific than TMPRSS2 and EST14 (Fig. 3B and C). HERVK17 flanking ALOX15B, but not ALOX12B, is also preferentially expressed in the prostate (Supplementary Fig. S4).

Figures 3A and 4A summarize the different open reading frames and predicted translated proteins of the ETV1 fusion transcripts. Wild-type ETV1 is composed of 12 exons, with the start codon in exon 1 and the stop in exon 12 (17). The DNA-binding domain of ETV1, ETS domain, is located in the COOH-terminal half of the protein. In the NH₂-terminal region, amino acids 42 to 73 are the acidic transactivation domain (TAD). In fusion transcripts, translation is predicted to start at an internal ATG in ETV1 exon 6 (amino acid 132) or at an in-frame ATG from the fusion partner followed by ETV1 fragments of different sizes (amino acids 80–477 or shorter; Figs. 3A and 4A). So, all fusion transcripts predict production of ETV1 lacking the NH₂-terminal TAD.

To test the functional properties of ETV1 in prostate cells, expression constructs of full-length ETV1 and truncated ETV1 (dETV1) were generated. The correct size of the proteins was verified by Western blotting in transiently transfected immortalized nontumorigenic PNT2C2 prostate epithelial cells with low endogenous ETV1 expression and in an LNCaP prostate cancer subline without ETV1 expression (Fig. 4B and C). Transient transfection of PNT2C2 cells with ETV1 expression constructs and an ETS reporter gene clearly showed that full-length ETV1 functioned as a transactivator, whereas dETV1 was not or hardly active (Fig. 4B; see also ref. 18). Similar results were obtained in LNCaP cells (Fig. 4C) and in 3T3 cells (data not shown). Other dETV1 fusion transcripts (Fig. 3A) gave identical results in this assay (data not shown). In a competition assay, dETV1 diminished the activity of full-length ETV1 (Fig. 4D). So, full-length ETV1 and N-truncated ETV1 possess different transcription regulation functions, suggesting that prostate cancers overexpressing full-length ETV1 and those expressing N-truncated ETV1 are not identical. dETV1 might compete with full-length ETV1 for the ETS-binding sites in the reporter construct or form heterodimers with full-length ETV1, thereby weakening the much stronger transactivation of full-length ETV1 (see also ref. 18).

Next, we compared the properties of full-length ETV1 and dETV1 in various in vitro biological assays. First, lentiviruses expressing either full-length ETV1 or dETV1 were generated, and PNT2C2 cells were infected with these viruses. Pools of stable-transfected cells overexpressing ETV1 or dETV1, PNT2C2-ETV1, and PNT2C2-dETV1, respectively, were propagated and ETV1 protein expression was verified by Western blotting (Fig. 5A). PNT2C2 cells infected with a GFP lentivirus were used as controls. Overexpression of full-length ETV1 or truncated ETV1 had no effect on proliferation of PNT2C2 cells as determined in a standard MTT assay (data not shown). Compared with infected control cells (Fig. 5B and C) and uninfected parental cells (data not shown), both PNT2C2-ETV1 and PNT2C2-dETV1 showed increased migration and invasion (Fig. 5B and C). So, we did not observe a significant difference between both PNT2C2 sublines in these assays. However, overexpression of full-length ETV1 strongly stimulated anchorage-independent growth of PNT2C2 cells, whereas dETV1 had no effect (Fig. 5D).

QPCR experiments were done to assess the expression of endogenous ETS target genes that are presumed to mediate migration, invasion, or anchorage-independent growth (1–3). Genes studied encoded matrix metalloproteinases (MMP), uPA, uPAR, and integrins (Fig. 6).³

First, we confirmed overexpression of ETV1 and dETV1 mRNA in the PNT2C2-ETV1 and PNT2C2-dETV1 lines, respectively (Fig. 6A). Next, we studied expression of members of the MMP family. Highest induction was detected for MMP1 (Fig. 6B) and MMP7 (data not shown). MMP3 was also induced, but MMP9 was not (data not shown). In none of the experiments, a significant difference between PNT2C2-ETV1 and PNT2C2-dETV1 was detected. Similarly, expression of uPAR was induced in both cell types, and the effect of ETV1 and dETV1 on uPA expression was limited (Fig. 6C). ITGAV and ITGB3 expression were induced in PNT2C2-ETV1 and PNT2C2-dETV1 cells (Fig. 6D). However, particularly the expression of ITGB3 was strongly stimulated in PNT2C2-ETV1 cells but much less in cells overexpressing dETV1. Neither ETV1 nor dETV1 affected the expression of ITGB1 and ITGB5 mRNA (data not shown).

The results presented in this study reveal several important aspects of prostate cancer. First, it is increasingly becoming clear that the prostate-specific and androgen-regulated expression of many ETV1 fusion partners, as shown here for EST14 and HERVK17, and previously for TMPRSS2 (11), is an important determinant in fusion gene selection. In agreement with this observation, recently, for three other ETV1 fusion partners, SLC45A3, HERVK22q11.23, and C15orf21, preferential expression in prostate cancer has been documented (19). SLC45A3 and HERVK22q11.23 are up-regulated by androgens, but C15orf21 is down-regulated. Similarly, prostate-specific and androgen-regulated KLK2 and CANT1 are novel ETV4 fusion partners (20). An explanation for this remarkable selection would be a colocalization of prostate-specific genes in specific chromosome territories or prostate-specific transcription factories, favoring their selective interactions with partner oncogenes (21–23). Alternatively, it can be postulated that certain regions in the genome, involved in regulation of prostate-specific gene expression, are preferentially unstable at the shift in cellular programming from proliferation (DNA replication) to differentiation (expression of prostate-specific genes) during development or at tissue renewal (24). These unstable regions could be targets for gene fusions and, if fused to oncogenes, be involved in tumor development.

In the LNCaP prostate cancer cell line, the complete ETV1 locus is translocated from chromosome 7 to chromosome 14 and integrated into the last intron of MIPOL1 (19). Remarkably, MIPOL1 is a directly flanking gene of EST14, which was found as ETV1 fusion partner in this study. In addition, evidence exists that in the MDA-PC2A cell line, which shows a (7;14) chromosomal translocation, ETV1 is linked to the same chromosome 14 region (19). Combined, these findings indicate that chromosome 14 contains a small region that favors integrations and rearrangements of full-length ETV1 or ETV1 fusion genes not only in cell lines but also in clinical prostate cancer. It remains to be established whether overexpression of full-length ETV1 in other clinical samples is the result of genomic rearrangement of the complete ETV1 locus. Elucidation of the chromatin structure of the chromosome 14 genomic region might shed more light on the mechanism of gene rearrangement in prostate cancer.

A further aspect of this study concerns the role of repeat sequences in prostate cancer. From the eight ETV1 fusion partners (refs. 8, 19 and this study), two are members of the HERVK subfamily of endogenous retroviral repeat sequences: HERVK17 and HERVK22q11.23. Previously, only one similar gene fusion has been reported, HERVK19-FGFR1, in a myeloproliferative disorder (25). The finding of a role of common repeats, encoding apparently insignificant defective transcripts, in a frequent disease such as prostate cancer urges to reconsider the role of such repeats in disease. In this regard, the HERVK retroviral subfamily is of particular interest because many members possess active promoters (26, 27).

In transient reporter assays, full-length ETV1 is a strong transactivator, whereas dETV1 is not or hardly active (Fig. 4; ref. 18). However, dETV1 can stimulate expression of endogenous target genes (Fig. 6). So, it seems that the acidic NH₂-terminal region, which functions as a dominant TAD in transient transfections, is less important for activation of endogenous ETV1 target genes. This finding implies that ETV1 possesses additional TADs that remain to be defined in more detail. Moreover, it is of high interest to identify genes that are preferentially regulated by ETV1 and that are responsible for anchorage-independent growth of prostate cells overexpressing full-length ETV1. In this regard, ITGB3 is an attractive candidate.

The large family of ETS transcription factors displays a wide variety of biological activities, including cellular proliferation, apoptosis, differentiation, tissue remodeling, migration, invasion, and angiogenesis (2, 28). The effects might depend on the cellular context and on the expression levels of the individual ETS factors. In most tumor types, a role of overexpressed wild-type ETS factors, including ETV1, has been described (2, 28). Although clinical prostate cancers can overexpress full-length ETV1 (this study), most prostate cancers show overexpression of N-truncated ETS transcription regulators. It is tempting to speculate that the combination of prostate specificity and protein truncation is a unique prerequisite for initial oncogenic properties of a weaker, more specific ETS in prostate cancer.

Previously, the biological and molecular effects of overexpression of full-length ETV1 in prostate cancer, as shown here, have not been studied. However, recently, the effects of dERG, as expressed from TMPRSS2-ERG (29, 30), and of dETV1, as expressed from TMPRSS2-ETV1 (19), on cell growth, migration, and invasion have been described. Some differences in proliferation and migration were found (this study and ref. 29), which might be due to differences between dERG and dETV1 or to the different cellular context. In agreement with our findings, in all studies, the fusion proteins were able to stimulate invasion of the target cells. Like shown here, stimulation of invasive growth correlated with up-regulation of genes known to stimulate tumor invasion and metastasis. The unique differences between full-length ETV1 and dETV1 warrant further in-depth investigation of the mechanism of prostate cancer growth.

We propose that overexpression of truncated ETV1 or other members of the ETS transcription family is most important in earlier stages of prostate cancer, whereas overexpression of wild-type ETS transcription factors combined with down-regulation of fusion gene expression comes into play at late stages of the disease (Supplementary Fig. S6). This hypothesis is supported by several observations: (a) the soft agar growth of PNT2C2-ETV1 cells indicates that full-length ETV1 is more oncogenic than dETV1 (Fig. 5); (b) in prostate cancer xenografts, overexpression of fusion genes is detected in hormone-dependent samples, whereas in hormone-independent xenografts fusion gene expression is shut off, and overexpression of a full-length ETS factor is turned on (5); and (c) our limited clinical data indicate that full-length ETV1 is expressed in the two recurrent tumors (Fig. 1A). Moreover, the two patients with overexpression of full-length ETV1 in primary tumors had a remarkable short survival time (50 and 19 months) compared with the three patients with primary tumors with ETV1 fusion gene expression (99, 141, and >172 months, respectively) and with patients with TMPRSS2-ERG fusion or without gene fusion (data not shown). Obviously, these clinical data should be validated in a larger patient cohort.

In conclusion, the data presented in this study show that ETS genes play a pivotal role in prostate cancer, probably affecting many stages of tumor growth. Investigation of the mechanism of gene fusion and translocation, and the function of the various truncated and full-length ETS transcription factors will contribute to a much broader knowledge of prostate tumor development and to the identification of novel therapeutic targets.

No potential conflicts of interest were disclosed.

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 Arno van Leenders for pathology, Wilma Teubel for collection of clinical samples, Wytske van Weerden for xenograft tissues, Anieta Siewerts for RNA isolation, Boh Wasylyk for the ETS reporter, Norman Maitland for PNT2C2 cells, Didier Trono for cloning vectors, and Erik Jan Dubbink for critical reading of the manuscript.