Research Article

Recurrent Fusion of TMPRSS2 and ETS Transcription Factor Genes in Prostate Cancer

See allHide authors and affiliations

Science  28 Oct 2005:
Vol. 310, Issue 5748, pp. 644-648
DOI: 10.1126/science.1117679


Recurrent chromosomal rearrangements have not been well characterized in common carcinomas. We used a bioinformatics approach to discover candidate oncogenic chromosomal aberrations on the basis of outlier gene expression. Two ETS transcription factors, ERG and ETV1, were identified as outliers in prostate cancer. We identified recurrent gene fusions of the 5′ untranslated region of TMPRSS2 to ERG or ETV1 in prostate cancer tissues with outlier expression. By using fluorescence in situ hybridization, we demonstrated that 23 of 29 prostate cancer samples harbor rearrangements in ERG or ETV1. Cell line experiments suggest that the androgen-responsive promoter elements of TMPRSS2 mediate the overexpression of ETS family members in prostate cancer. These results have implications in the development of carcinomas and the molecular diagnosis and treatment of prostate cancer.

A central aim in cancer research is to identify altered genes that play a causal role in cancer development. Many such genes have been identified through the analysis of recurrent chromosomal rearrangements that are characteristic of leukemias, lymphomas, and sarcomas (1). These rearrangements are of two general types. In the first, the promoter and/or enhancer elements of one gene are aberrantly juxtaposed to a proto-oncogene, thus causing altered expression of an oncogenic protein. This type of rearrangement is exemplified by the apposition of immunoglobulin (IG) and T cell receptor (TCR) genes to MYC, leading to activation of this oncogene in B and T cell malignancies, respectively (2). In the second, the rearrangement fuses two genes, resulting in the production of a fusion protein that may have a new or altered activity. The prototypic example of this translocation is the BCR-ABL gene fusion in chronic myelogenous leukemia (CML) (3, 4). Importantly, this finding led to the development of the promising cancer drug imatinib mesylate (Gleevec) (5). In contrast to leukemias, epithelial tumors (carcinomas) display many nonspecific but few recurrent chromosomal rearrangements (6). This karyotypic complexity is thought to reflect secondary genomic alterations acquired during tumor progression.

We hypothesized that rearrangements and high-level copy number changes that result in marked overexpression of an oncogene should be evident in DNA microarray data but not necessarily by traditional analytical approaches. In the majority of cancer types, heterogeneous patterns of oncogene activation have been observed; thus, traditional analytical methods that search for common activation of genes across a class of cancer samples (e.g., t test or signal-to-noise ratio) will fail to find such oncogene expression profiles. Instead, a method that searches for marked overexpression in a subset of cases is needed. Toward this end, we developed a method termed cancer outlier profile analysis (COPA). COPA seeks to accentuate and identify outlier profiles by applying a simple numerical transformation based on the median and median absolute deviation of a gene expression profile (7) (fig. S1A).

Cancer outlier profile analysis. We applied COPA to the Oncomine database (8), a compendium of 132 gene expression data sets representing 10,486 microarray experiments. COPA correctly identified several outlier profiles for genes in specific cancer types in which a recurrent rearrangement or high-level amplification is known to occur (Table 1 and fig. S1, B and C). We focused our analyses on outlier profiles of known causal cancer genes, as defined by the Cancer Gene Census (9), that ranked in the top 10 outlier profiles in an Oncomine data set (Table 1 and table S1), because we felt these genes would be the most likely to participate in uncharacterized alterations. The general COPA methodology can be applied to any expression data (10).

Table 1.

Cancer outlier profile analysis (COPA). Genes known to undergo causal mutations in cancer that had strong outlier profiles. “X” indicates literature evidence for the acquired pathognomonic translocation. “XX” indicate that samples in the study were characterized for the indicated translocation. “Y” indicates consistent with known amplification. Double asterisks indicate ERG and ETV1 outlier profiles in prostate cancer. A complete listing of genes known to undergo causal mutations ranking in the top 10 of all studies in Oncomine, along with the relevant references, is included as table S1.

Rank % Score Gene Cancer Study Evidence
1 95 20.056 RUNX1T1 Leukemia (View inline) XX
1 95 15.4462 PRO1073 Renal (View inline) X
1 90 12.9581 PBX1 Leukemia (View inline) XX
1 95 10.03795 ETV1 Prostate (View inline) **
1 90 7.4557 WHSC1 Myeloma (View inline) X
1 75 5.4071 ERG Prostate (View inline) **
1 75 4.3628 ERG Prostate (View inline) **
1 75 4.3425 CCND1 Myeloma (View inline) X
1 75 3.4414 ERG Prostate (View inline) **
1 75 3.3875 ERG Prostate (View inline) **
3 95 13.3478 FGFR3 Myeloma (View inline) X
4 75 2.5728 ERBB2 Breast (View inline) Y
6 90 6.6079 ERBB2 Breast (View inline) Y
9 95 17.1698 ETV1 Prostate (View inline) **
9 90 6.60865 SSX1 Sarcoma (View inline) X
9 75 2.2218 ERG Prostate (View inline) **

Outlier profiles for ERG and ETV1 in prostate cancer. In several independent data sets, COPA identified strong outlier profiles in prostate cancer for ERG (21q22.3) and ETV1 (7p21.2) (Table 1), two genes that encode ETS family transcription factors and are involved in oncogenic translocations in Ewing's sarcoma and myeloid leukemias (11, 12). In total, COPA ranked ERG or ETV1 within the top 10 outlier genes in six independent prostate cancer profiling studies.

Fusion of the 5′ activation domain of the Ewing sarcoma breakpoint region 1 (EWSBR1) gene to the highly conserved 3′ DNA binding domain of an ETS family member, such as ERG [t(21;22)] or ETV1 [t(7;22)], is characteristic of Ewing's sarcoma (11, 13, 14). Because translocations involving ETS family members are functionally redundant in oncogenic transformation, only one type of translocation is typically observed in each case of Ewing's sarcoma. We hypothesized that, if ERG and ETV1 are similarly involved in the development of prostate cancer, their outlier profiles should be mutually exclusive—that is, each tumor should overexpress only one of the two genes.

Thus, we examined the joint expression profiles of ERG and ETV1 across several prostate cancer data sets and found that they invariably showed mutually exclusive outlier profiles, consistent with our hypothesis. Exclusive outlier expression of ERG and ETV1 was identified in two large-scale transcriptome studies (15, 16), which profiled grossly dissected prostate tissues with the use of different microarray platforms (Fig. 1). Similar results were obtained in prostate tissue samples obtained by laser capture microdissection (LCM) (fig. S2A). In addition to exclusive outlier expression of either ERG or ETV1 in epithelial cells from prostate cancer or metastatic prostate cancer, ETV1 and ERG were not overexpressed in the precursor lesion prostatic intraepithelial neoplasia (PIN) or adjacent benign epithelia (fig. S2A). The observed exclusive outlier pattern is consistent with other translocations where an activating gene can fuse with multiple partners, such as the fusion of the immunoglobulin heavy chain promoter to CCND1 or FGFR3, t(11,14) or t(4,14), respectively, in specific subsets of multiple myeloma (17) (fig. S2B).

Fig. 1.

COPA of microarray data revealed ETV1 and ERG as outlier genes across multiple prostate cancer gene expression data sets. ETV1 and ERG expression (normalized expression units) are shown from all profiled samples in two-large scale gene expression studies [top data set from (15) and bottom data set from (16)]. Visualization tools incorporated in Oncomine (10) were used to generate graphical displays. Sample classes are indicated according to the color scale. In the data set from (16), prostate cancer samples were classified on the basis of Gleason grade. Scatter plots of ERG and ETV1 expression across all of the profiled samples are shown (right).

Recurrent gene fusion of TMPRSS2 to ERG or ETV1 in prostate cancer. To determine the mechanism responsible for ERG and ETV1 overexpression, we identified prostate cancer cell lines and clinical specimens that overexpressed ERG or ETV1 by using quantitative polymerase chain reaction (QPCR) (Fig. 2A). The LNCaP prostate cancer cell line and two specimens obtained from a patient with hormone-refractory metastatic disease (MET26-RP, residual primary carcinoma in the prostate, and MET26-LN, a lymph node metastasis) overexpressed ETV1.Alymph node metastasis from a second patient (MET28LN) and two prostate cancer cell lines, VCaP and DuCaP, overexpressed ERG. We did not find consistent amplification of ERG or ETV1 in samples with respective transcript overexpression, so we considered the possibility of DNA rearrangements. We measured the expression of ETV1 exons by exon-walking QPCR in samples that displayed ETV1 overexpression. We used five primer pairs spanning ETV1 exons 2 through 7 and found that although LNCaP cells showed essentially uniform overexpression of all measured ETV1 exons, both MET26 specimens showed >90% reduction in the expression of ETV1 exons 2 and 3 compared with exons 4 to 7 (Fig. 2B).

Fig. 2.

Identification and characterization of TMPRSS2:ETV1 and TMPRSS2:ERG gene fusions in prostate cancer (PCA). (A) Prostate cancer cell lines, hormone refractory metastatic (MET) prostate cancer tissues, and pooled benign prostate tissue (CPP) were analyzed for ERG (solid) and ETV1 (open) mRNA expression by QPCR. Samples with values off the scale are indicated by hatched bars, and the values are given above the graph. (B) Reduced overexpression of ETV1 exons 2 and 3 compared with exons 4 to 7 in MET26 samples. Expression of ETV1 exons 2 to 7 was assessed by QPCR in LNCaP cells and MET26-LN and MET26-RP samples. (C) Schematic of 5′ RLM-RACE revealing fusion of TMPRSS2 with ETV1 in MET26-LN and ERG in MET28-LN. Structures for the TMPRSS2, ERG, and ETV1 genes have their basis in the GenBank reference sequences. The numbers above the exons (indicated by boxes) indicate the last base of each exon. Untranslated regions are shown in corresponding lighter shades. Coding exons not depicted are indicated by hatched boxes. Identified TMPRSS2 fusions are colored and numbered from the original reference sequences. Line graphs show the position and automated DNA sequencing of the fusion points. (D) Validation of TMPRSS2:ETV1 expression using fusion-specific QPCR in MET26-LN and MET26-RP. Expression of ETV1 (solid, right axis) and TMPRSS2:ETV1 (open, left axis) was assessed by QPCR. (E) Validation of TMPRSS2:ERG expression using fusion-specific QPCR in cell lines and PCA specimens. Expression of ERG (solid, right axis) and TMPRSS2:ERG (open, left axis) was assessed by QPCR.

To characterize the complete 5′ ETV1 transcript, we performed 5′ RNA ligase-mediated rapid amplification of cDNA ends (RLM-RACE) on LNCaP cells and MET26-LN. In addition, we also performed RLM-RACE to obtain the complete 5′ transcript of ERG in MET28-LN. Sequencing of the cloned products revealed fusions of the prostate-specific gene TMPRSS2 (18) (21q22.2) with ETV1 in MET26-LN and with ERG in MET28-LN (Fig. 2C). In MET26-LN, two RLM-RACE PCR products were identified. The first product, TMPRSS2:ETV1a, resulted in a fusion of the complete exon 1 of TMPRSS2 with the beginning of exon 4 of ETV1 (Fig. 2C). The second product, TMPRSS2:ETV1b, resulted in a fusion of exons 1 and 2 of TMPRSS2 with the beginning of exon 4 of ETV1 (fig. S3). Both products are consistent with the exon-walking QPCR described above, where MET26-LN showed loss of overexpression in exons 2 and 3. In MET28-LN, a single RLM-RACE PCR product was identified, and sequencing revealed a fusion of the complete exon 1 of TMPRSS2 with the beginning of exon 4 of ERG (TMPRSS2:ERGa) (Fig. 2C).

Validation of TMPRSS2:ERG and TMPRSS2:ETV1 gene fusions in prostate cancer. On the basis of these results, we designed QPCR primer pairs with forward primers in TMPRSS2 and reverse primers in exon 4 of ERG or ETV1. We performed SYBR Green (Molecular Probes, Eugene, OR) QPCR with the use of both primer pairs across a panel of samples from 42 cases of clinically localized prostate cancer and metastatic prostate cancer and depict representative results (Fig. 2, D and E). These 42 cases were selected on the basis of previous cDNA microarray or QPCR results indicating overexpression of ERG or ETV1. We were limited to samples with remaining material, and thus this cohort does not represent a random sampling. In addition to QPCR, we also performed standard reverse transcription PCR (RT-PCR) with the same primers used for QPCR, or with a different forward primer in TMPRSS2 and reverse primers in exon 6 of ERG and exon 7 of ETV1 on a subset of the samples with or without fusions as determined by using QPCR (fig. S4, A and B). Electrophoresis of QPCR products and sequencing of cloned RT-PCR products from MET-26RP and MET-26LN revealed the presence of both TMPRSS2:ETV1a and TMPRSS2:ETV1b. The molecular evidence for TMPRSS2:ERG and TMPRSS2:ETV1 fusions in cases and cell lines overexpressing the respective ETS family member are summarized (fig. S5). From QPCR melt curve analysis and gel electrophoresis of QPCR and RT-PCR products, PCA4 produced a larger amplicon than TMPRSS2:ERGa. Subsequent RLM-RACE analysis and sequencing of the RT-PCR product confirmed a fusion of the complete exon 1 of TMPRSS2 with the beginning of exon 2 of ERG (TMPRSS2:ERGb) (fig. S3). Evidence for the TMPRSS2:ERG and TMPRSS2:ETV1 fusions were only found in cases that overexpressed ERG or ETV1, respectively, by QPCR or DNA microarray. These results are also in agreement with the exclusive expression observed in our outlier analysis.

Genomic confirmation of TMPRSS2:ETV1 translocation and ERG rearrangement. We used interphase fluorescence in situ hybridization (FISH) to validate the rearrangements at the chromosomal level on formalin-fixed paraffin-embedded (FFPE) specimens from the two cases initially used for RLM-RACE, MET26 and MET28 (Fig. 3). With the use of probes for TMPRSS2 and ETV1, normal peripheral lymphocytes (NPLs) demonstrated a pair of red and a pair of green signals (Fig. 3A). However, MET26 showed fusion of one pair of signals, indicative of probe overlap (Fig. 3B) and consistent with the expression of the TMPRSS2:ETV1. Because of the proximity of TMPRSS2 to ERG on chromosome 21, ∼3 megabases (fig. S6A), we used probes spanning the 5′ and 3′ regions of the ERG locus to assay for gene rearrangements. By using these probes, we observed a pair of yellow signals in NPLs (Fig. 3C); however, in MET28, one pair of probes split into separate green and red signals, indicative of a rearrangement at the ERG locus (Fig. 3D) and consistent with the expression of TMPRSS2:ERG. We next performed both individual FISH analyses described above on serial tissue microarrays containing cores from 13 cases of localized prostate cancer and 16 cases of metastatic prostate cancer (Fig. 3E). Of 29 cases, 23 (79.3%) showed evidence of TMPRSS2:ETV1 fusion (7 cases) or ERG rearrangement (16 cases).

Fig. 3.

Interphase FISH on FFPE tissue sections confirms TMPRSS2:ETV1 gene fusion and ERG gene rearrangement. (A and B) NPLs showed two ETV1 (red) and two TMPRSS2 (green) signals, whereas MET26 showed fusion of the signals as indicated by the yellow signal (yellow arrowhead). (C and D) For detection of ERG gene rearrangements, we used a split-signal approach, with two probes spanning the ERG locus. NPLs showed two yellow signals, indicating overlap of the 5′ (green signal) and 3′ (red signal) regions of ERG, whereas MET28 shows a rearrangement of ERG as indicated by the split signal of the 5′ and 3′ probes (red and green arrows). Scale bars for all images are 2.5 μm. (E) Matrix representation of FISH results using the same probes as (A) to (D) on an independent tissue microarray containing cores from clinically localized (PCA) and metastatic (MET) prostate cancer. Cores positive for TMPRSS2:ETV1 probe fusion or split-signal ERG probes are indicated by colored cells. All negative findings are indicated by gray cells. The number of positive cases for each feature is indicated to the right of the matrix.

As additional confirmation of the ERG rearrangement, we performed FISH on metaphase spreads of VCaP cells, which express the TMPRSS2:ERGa transcript. This assay revealed co-localization of 5′ TMPRSS2 and 3′ ERG probes with splitting of the 5′ and 3′ ERG signals, supporting the molecular results (fig. S6). In addition, Southern blotting using a probe in the intron between exons 1 and 2 of TMPRSS2 revealed a unique band in VCaP cells, consistent with a rearrangement at this locus (fig. S7).

Fusion of TMPRSS2 and ERG results in androgen regulation of ERG. TMPRSS2 is expressed in normal and neoplastic prostate tissue and is strongly induced by androgen in androgen-sensitive prostate cell lines (1820). To investigate whether the TMPRSS2:ERG fusion results in the androgen regulation of ERG, we assessed the expression of ERG by QPCR in androgen-treated VCaP cells, which express TMPRSS2:ERGa, and LNCaP cells, which do not express a fusion transcript. Both VCaP and LNCaP respond to androgen stimulation with increased expression of PSA, which is expressed at a similar amount in both cells and is sensitive to the androgen receptor antagonists bicalutamide and flutamide (Fig. 4A). However, in addition to expressing ∼2000-fold more ERG than LnCAP cells, only VCaP cells responded to androgen stimulation with increased ERG expression sensitive to bicalutamide and flutamide (Fig. 4B). A similar increase in ERG expression upon androgen stimulation was observed in DuCaP cells, which express TMPRSS2:ERGa, whereas RWPE, PC3, and PC3 cells expressing the human androgen receptor express low concentrations of ERG that are not androgen-responsive (fig. S8). These results suggest that the fusion with TMPRSS2 may explain the aberrant expression of ERG or ETV1 in specific subsets of prostate cancer.

Fig. 4.

Androgen regulation of ERG in VCaP prostate cancer cells carrying the TMPRSS2:ERG fusion. (A) PSA expression relative to GAPDH in androgen-sensitive LNCaP (open) and VCaP (solid) cells was assessed by QPCR. (B) ERG (exon 5 to 6) expression relative to GAPDH in LNCaP (open, left axis) and VCaP (solid, right axis) cells. Cell lines were incubated with vehicle or 10 μM of the androgen receptor antagonists bicalutamide or flutamide for 2 hours before treatment for 24 hours with 0.5 nM of the synthetic androgen R1881 or vehicle as indicated. Relative PSA or ERG for each sample was normalized to the amount in the LNCaP control.

Conclusions. The existence of recurring gene fusions of TMPRSS2 to the oncogenic ETS family members ERG and ETV1 may have important implications for understanding prostate cancer tumorigenesis and developing novel diagnostics and targeted therapeutics. Several lines of evidence suggest that these rearrangements occur in the majority of prostate cancer samples and drive ETS family member expression. Across three independent microarray data sets, ERG or ETV1 was markedly overexpressed in 95 of 167 (57%) prostate cancer cases, whereas overexpression was never observed across 54 benign prostate tissue samples. Furthermore, a recent study reported that ERG was the most commonly overexpressed oncogene by QPCR in prostate cancer, with 72.0% of cases overexpressing ERG (21). By using a combination of assays, we found evidence of fusion with TMPRSS2 in 20 of 22 (>90%) cases that overexpressed ERG or ETV1, suggesting that the fusion is the most likely cause for the overexpression. FISH analysis on a set of 29 prostate cancer cases selected independently of any knowledge of ERG or ETV1 expression indicates that 23 of 29 (79%) had TMPRSS2:ETV1 fusions or ERG rearrangement. It is possible that this cohort is not representative of all prostate cancer samples and that this may be an overestimate of the prevalence of TMPRSS2 fusions with ETS family members, because our split-signal approach can detect additional rearrangements involving ERG. However, the reported frequencies of ERG or ETV1 overexpression in prostate cancer with our fusion transcript and FISH results suggest that TMPRSS2 fusions with ETV1 or ERG occur in the majority of prostate cancer cases. Coupled with the high incidence of prostate cancer [an estimated 232,090 new cases will be diagnosed in the United States in 2005 (22)], the TMPRSS2 fusion with ETS family members is likely to be the most common rearrangement yet identified in human malignancies and the only rearrangement present in the majority of one of the most prevalent carcinomas.

Future efforts will be directed at characterizing the expressed protein products, including the effects of N-terminal truncation of ERG and ETV1, and identifying downstream targets and the functional role of the fusions in prostate cancer development. Importantly, the existence of TMPRSS2 fusions with ETS family members in prostate cancer suggests that causal gene rearrangements may exist in common epithelial cancers but may be masked by the multiple nonspecific chromosomal rearrangements that occur during tumor progression.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Tables S1 and S2

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article