Main

Prostate cancer is the most frequently diagnosed cancer among men in the United States,1 with one in six men being diagnosed in their lifetime. Localized prostate cancer can be treated by surgical resection or radiation, but recurs in approximately a quarter of patients.2 Advanced prostate cancer is treated by androgen-ablation therapy, though hormone-refractory prostate cancer invariably recurs within 1–2 years.3 Although taxane-based therapies have recently shown promise in prolonging survival,4 new therapeutic targets and molecularly directed therapies for end-stage prostate cancer are urgently needed.

Recently, by examining outlier values of gene expression in human cancers, Tomlins et al5 discovered elevated expression of the ETS family members ERG (v-ets erythroblastosis virus E26 oncogene like) and ETV1 (Ets variant gene 1) to be a common feature of prostate cancer. Elevated expression resulted from chromosomal rearrangement fusing TMPRSS2 (transmembrane protease, serine 2) (chr 21q22.3) to ERG (chr 21q22.2), an apparent intra-chromosomal deletion of ∼3 Mb on chromosome 21, or less frequently to ETV1, an inter-chromosomal rearrangement between chr 21q22.3 and chr 7p21.2.

ETS transcription factors are both positive and negative regulators of gene expression, and are involved in various biological processes including cell proliferation, differentiation, development, transformation and apoptosis.6 In human cancers, ETS genes are frequently found amplified, overexpressed or rearranged, the latter including chimeric fusion products characteristic of Ewing's sarcoma and certain leukemias.6 TMPRSS2, which encodes a serine protease with a transmembrane domain,7 is highly expressed in both normal and neoplastic prostate epithelium,8, 9 and its expression in prostate cancer cells is androgen-regulated.8, 10 In prostate tumors harboring rearrangements, most frequently the non-coding first exon of TMPRSS2 has been found fused to the fourth exon of ERG in the expressed transcript. The rearrangement thereby places ERG under the androgen-regulated transcriptional control of TMPRSS2.5 This finding has immediately suggested an additional possible mechanism for the androgen-dependent growth of therapy-naïve prostate tumors.

Here, we describe the discovery in prostate tumors of novel sequences fused to ERG, which we define as a variant TMPRSS2 isoform-ERG fusion. Our findings further our knowledge of ETS family rearrangements in prostate cancer, and provide for the more accurate molecular diagnosis of this lesion for future molecularly directed therapies.

Materials and methods

Prostate Tumor Specimens

The prostate tumor specimens included in this study, along with detailed clinicopathological annotations and procedures for specimen processing, were described previously.11 Briefly, 54 freshly frozen primary prostate tumor specimens (obtained from radical prostatectomy) and nine therapy-naive metastatic pelvic lymph node specimens (from aborted surgery) were scalpel-dissected such that ≥50% of cells were epithelial, and ≥90% of epithelial cells were neoplastic. Total RNA was extracted using the Trizol (Invitrogen, Carlsbad, CA, USA) method.

RACE

5′ RNA ligase-mediated rapid amplification of cDNA ends (RACE) was performed using the SMART RACE cDNA Amplification kit (Clontech, Mountain View, CA, USA) according to the manufacturer's instructions, using a gene-specific primer from exon 7 of ERG (GCTGCACCCCCTGTGTTTCTAGCATGCATTAACCG). The PCR products were resolved by electrophoresis on a 1% agarose gel, excised and TOPO-TA cloned into pCR4-TOPO (Invitrogen). Purified plasmid DNA from eight independent colonies was bi-directionally sequenced using M13 forward and reverse primers on an ABI 3100 automated capillary DNA sequencer by the Stanford PAN facility.

RT-PCR

RT-PCR was performed as described previously12 to characterize expression of TMPRSS2 isoforms, and the known and variant TMPRSS2-ERG fusions. Briefly, cDNA was synthesized by oligo(dT)-priming using SuperScript First-Strand Synthesis System (Invitrogen). PCR was then carried out using 1/25th the RT reaction product for template, 1 × AmpliTaq PCR buffer, 200 μM dNTPs, 1.5 mM MgCl2, 10 pmol of each gene-specific primer pair (Table S1, Supplementary Information), and 2 U AmpliTaq DNA polymerase in a 50 μl reaction for 35–40 cycles (see Table S1, Supplementary Information for primer-specific annealing/extension conditions), followed by electrophoresis of 1/10th the PCR product on a 1% TAE agarose gel, and imaging using an AlphaImager 2200 (Alpha Innotech, San Leandro, CA, USA).

Fluorescence In Situ Hybridization

Fluorescence in situ hybridization (FISH) was carried out on paraffin sections of selected prostate tumor cases arrayed in tissue microarray (TMA) format, as described previously.12 A 6 μM TMA section was baked overnight at 60°C, then deparaffinized in xylene and denatured in 100% ethanol. The TMA was then subjected to sequential pretreatment washes: 0.2 N HCl for 20 min, ddH2O for 10 min, 2 × SSC for 3 min and 1 M NaSCN at 80°C for 30 min. Slides were then protease-treated at 37°C for 10 min, dehydrated in 100% ethanol and air-dried. FISH analysis was performed as described (Rajput et al13) using the following BACs (BACPAC Resources Centre, Children's Hospital Oakland Research Institute, Oakland, CA, USA): RP11-95I21 (5′ ERG), RP11-476D17 (3′ ERG) and RP11-35C4 (telomeric to TMPRSS2). BACs RP11-95I21 and RP11-35C4 were directly labeled by nick translation using Spectrum Green and Orange, respectively (Vysis, Downer's Grove, IL, USA). BAC RP11-476D17 was indirectly labeled using a modified protocol with Cy5 (MetaSystems, Belmont, MA, USA) using the BioPrime DNA labeling system (Invitrogen). FISH signals were visualized on a Zeiss Axioplan epifluorescent microscope, and captured using the ISIS FISH imaging software (MetaSystems, Belmont, MA, USA); signals were counted for 50 tumor cells per specimen.

Results

In an array-based comparative genomic hybridization (array CGH) study of prostate tumors (paper in preparation), we identified a single specimen (PT187) with an apparent single-copy terminal deletion of chr 21q with the break point at ERG (21q22. 2), suggesting a possible inter-chromosomal rearrangement and alternative fusion partner of ERG. To identify the 5′ fusion partner of ERG in this specimen, we carried out RACE, using a 3′ primer within the ERG gene. The resultant PCR products, a predominant band (∼0.9 kb) and a lesser band (∼0.8 kb), were separately gel-purified, cloned and sequenced. The lower molecular-weight (MW) band corresponded to the previously reported TMPRSS2 (exon1)-ERG (exon4) fusion,5 whereas four independent clones from the higher-MW band represented an identical novel sequence (Figure 1a) fused to the fourth exon of ERG. A BLAST search of this sequence to the human expressed sequence tag (EST) database yielded a match with the first exon of three overlapping ESTs (DA872508, DA868984 and DA870830; cloned from prostate tissue), mapping 4 kb upstream (ie distal on chr 21) of exon 1 of TMPRSS2 (Figure 1b).

Figure 1
figure 1

Variant TMPRSS2 isoform and ERG fusion product. (a) RACE identifies novel DNA sequences fused to ERG. Novel sequences derive from a TMPRSS2 isoform with alternate first exon. Junction of TMPRSS2 (isoform 2) and ERG sequences is indicated. RT-PCR primers used to validate expression of TMPRSS2 isoform 2 (black line) and ERG fusion (black/gray line) are indicated. Numbering is from the first nucleotide of RACE product. Predicted start site of translation, an in-frame ATG in exon 4 of ERG, is in bold (b) Genomic structure of ERG (isoform 2),19 TMPRSS2, and TMPRSS2 (isoform 2), and exonic organization of the novel TMPRSS2 (isoform 2,exon1)-ERG (exon4) fusion (dashed connecting line), and the most commonly reported known TMPRSS2 (exon1)-ERG (exon4) fusion (solid connecting line), drawn to scale from genome assembly (http://genome.ucsc./edu). Position of primers is indicated by arrows: RACE/3′-TMPRSS2 (isoform 2)-ERG fusion (black-open), 3′-TMPRSS2-ERG (black-filled), 5′-TMPRSS2 (gray-filled), 5′-TMPRSS2 isoform 2 (gray-open), 3′-TMPRSS2 (black/gray-filled). Mapped ESTs representing additional putative TMPRSS2 variants (see text) are indicated, top. Location of sequences reported by Tomlins et al18 fused to ETV4 is indicated by upwards-pointing arrowhead. (c) RT-PCR analysis of TMPRSS2 (upper panel), TMPRSS2 isoform 2 (middle panel), and control GAPDH (lower panel) expression in four primary prostate tumors (PT) along with no-template control. Molecular sizes (kb) are indicated.

Full size image

The genome annotation (http://genome.ucsc.edu) suggested these sequences to encompass an alternative start site of a variant of TMPRSS2 that included an alternate noncoding exon 1, along with known exons 2 and 3 of TMPRSS2. To define further the novel transcript, we carried out RT-PCR analysis using a 5′ primer within the novel sequences (Figure 1a), and a 3′ primer from the terminal exon (exon 14) of TMPRSS2. We identified an expressed transcript in three of four prostate tumors analyzed (Figure 1c), with a size consistent with its comprising an alternative exon 1 but sharing the known exons 2–14 of TMPRSS2 (Figure 1b). We therefore propose to name the novel non-rearranged transcript variant, TMPRSS2 isoform 2.

In the index prostate tumor specimen on which we carried out RACE, we identified noncoding exon 1 of the variant TMPRSS2 isoform 2 fused to the fourth exon of ERG. To confirm the expression and assess the frequency of the variant TMPRSS2 isoform-ERG fusion in prostate cancer, we used RT-PCR primer sets (Figure 1b; Table S1, Supplementary Information) designed to amplify specifically either the known or variant fusion transcripts. Amplification of GAPDH served as a positive control. Altogether, we surveyed 63 prostate tumors, including 54 primary prostate tumors and nine unmatched (ie from different patients) therapy-naïve lymph node metastases, for which we had previously profiled gene expression by cDNA microarray.11 Exemplary data are shown in Figure 2. Of the 63 specimens, 44 (70%) cases expressed either the known or novel variant TMPRSS2-ERG fusion, and 28 (44%) expressed both. Ten cases (16%) expressed only the known fusion, and notably six cases (10%) expressed only the novel variant isoform fusion (eg Figure 2, specimens PT112 and PL114).

Figure 2
figure 2

Expression of the variant TMPRSS2 (isoform 2)-ERG fusion in prostate cancer. Shown are representative RT-PCR assays scoring the expression of known (top panel) and variant isoform (middle panel) TMPRSS2-ERG fusion transcripts in prostate cancer, along with GAPDH control (lower panel). Molecular sizes (kb) are indicated. PT and PL indicate primary tumor and lymph node metastasis samples, respectively. PCR product sizes are consistent with, and limited DNA sequence analysis confirms that most fusions represent TMPRSS2 (exon1)-ERG (exon4) transcripts (top panel) and TMPRSS2 (isoform 2, exon1)-ERG (exon4) transcripts (middle panel). The higher-MW PCR product from sample PT335 corresponds to fusion of TMPRSS2 (exon1) to ERG (exon3), and a lower-MW PCR product from sample PT316 (not shown) corresponds to fusion of TMPRSS2 (exon2) to ERG (exon5); both variants have been previously reported.15, 17

Full size image

The co-expression of both the known and variant fusions in a subset of specimens might reflect alternative transcriptional start sites from the same rearranged chromosome, or alternatively disparate start sites from two different rearranged chromosomes. To distinguish between these possibilities, we carried out FISH analysis on paraffin sections corresponding to four prostate tumor specimens co-expressing the known and variant fusions. In two of the four specimens analyzed, only a single (rather than multiple) TMPRSS2-ERG chromosomal rearrangement was detected in the majority of cells (Figure 3), indicating that both transcripts could be expressed from the same rearranged locus.

Figure 3
figure 3

FISH analysis of TMPRSS2-ERG rearrangement. Representative FISH images from prostate tumors determined by RT-PCR to express (a) no TMRPSS2-ERG fusion (PT265), and (b, c) both known and novel variant isoform TMPRSS2-ERG fusion transcripts (PT87, PT41). FISH signals in these pseudocolored images are red (just telomeric to TMPRSS2; RP11-35C4), green (5′ ERG; RP11-95I21) and blue (3′ ERG; RP11-476D17). Rearrangement (arrows) is evident by juxtaposition of red and blue signals, typically with loss of green signal. Note that in non-rearranged chromosomes the proximity of green and blue signals can result in an aqua-colored signal.

Full size image

Scoring the known and variant TMPRSS2-ERG fusions together, TMPRSS2-ERG expression showed no statistical association with specimen type (primary vs metastasis), tumor stage or Gleason grade (Table 1). We also found no association with recurrence-free survival time (Figure S1, Supplementay Information), though the number of specimens with clinical follow-up (n=28), and the median clinical follow up time (2 years) were both modest.

Table 1 Histopathological associations of TMPRSS2-ERG fusion
Full size table

Discussion

Our report represents an independent confirmation of the high frequency of TMPRSS2-ERG fusion in prostate cancer. Scored together, the known and/or variant fusion was expressed in 70% of cases assayed. These findings are comparable to the original report by Tomlins et al5 (79% of cases, scored by FISH), and recent confirmations by Soller et al14 (78%, RT-PCR), Yoshimoto et al15 (40%, RT-PCR), Perner et al16 (49%, FISH), and Wang et al17 (59%, RT-PCR). However, it should be noted that our set of cases was selected in part for sufficient scalpel-dissected tumor material for expression profiling (before the availability of robust methods for RNA amplification). Therefore, it is biased towards larger specimens, and our result should not be regarded as a true population estimate.

In our set of specimens, we found no relationship between TMPRSS2-ERG expression and tumor stage, Gleason grade or tumor recurrence. Although the TMPRSS2 (exon 1)-ERG (exon 4) fusion is the most common, other expressed variants have been reported,15, 17 differing with respect to the exon break point sites in TMPRSS2 (exons 1–5) and ERG (exons 2–5). Depending on the exon break point, protein translation is predicted to start either at the native ATG of ERG (exon 3), the native ATG of TMPRSS2 (exon 2), or an in-frame ATG within exon 4 or 5 of ERG; in the latter cases the N terminus of ERG would be truncated. Interestingly, the expression of selected TMPRSS2-ERG break point variants has recently been linked to specific clinicopathological features of prostate tumors.17 Although we had sequenced only a subset of fusion products in our study (see Figure 2 legend), the PCR product sizes of most are consistent with their representing the common TMPRSS2 (exon1)-ERG (exon4) and the novel TMPRSS2 (isoform2, exon1)-ERG (exon4) transcripts, for both of which protein translation is predicted to initiate from an in-frame ATG in ERG exon 4.

Given that TMPRSS2 isoform 2 is expressed in prostate tumors, and that a fraction of prostate tumors harbors rearrangement of the variant TMPRSS2 (isoform 2)-ERG fusion alone (where it is a presumptive pathogenic alteration), it is likely that TMPRSS2 isoform 2 and the resultant ERG fusion are under similar androgen-regulated transcriptional control as is TMPRSS2 (isoform 1) and its fusion. However, further studies are needed to elucidate the roles of the known and variant ERG fusions in the pathogenesis and androgen dependency (and subsequent independency) of prostate cancer.

Interestingly, also mapped on the genome assembly (Figure 1b, top) are additional ESTs that, like the ESTs (DA872508, DA868984 and DA870830) corresponding to TMPRSS2 isoform 2, appear likely to comprise alternative first exons and transcriptional start sites of TMPRSS2 variants. The first exons of these distinct transcripts map both upstream (DB046885; ∼23 kb upstream of TMPRSS2) and downstream (DA865300, DA460061, DA874671, DA627080, DA873416, DR005196 and DB457454) of TMPRSS2 exon 1. It is possible that these putative TMPRSS2 isoforms also serve as androgen-regulated fusion partners for ERG (or ETV1) rearrangements. Indeed, our preliminary data (not shown) identify the additional expression of a DA460061 (exon1)-ERG (exon 4) fusion in the index specimen (PT187). To note, Tomlins et al18 recently reported infrequent cases of prostate cancer harboring novel genomic sequences 8 kb upstream of TMPRSS2 (Figure 1b, arrowhead) fused to ETV4, also an ETS family member. These sequences, which do not match any known EST, are distinct from any of those we report here, but may also result from androgen-regulated expression.

In the index prostate tumor specimen for which we carried out RACE, an apparent terminal deletion identified by array CGH suggested an inter-chromosomal rearrangement. However, RACE-PCR identified only the known and variant isoform TMPRSS2-ERG fusion reported here. A likely explanation is that TMPRSS2-ERG fusion in this specimen results from an unbalanced inter-chromosomal rearrangement between two copies of chr 21. Indeed, a recent study,16 and our own data (unpublished), indicate that only half or less of TMPRSS2-ERG fusion events are associated with intra-chromosomal deletion between the TMPRSS2 and ERG loci on chr 21. Whether other prostate tumor specimens might contain rearrangements fusing ERG to genes other than TMPRSS2, and its isoforms described here, remains to be investigated.

Our findings underscore the pathogenetic relevance of altered ERG expression in prostate cancer, as well as the potential for novel therapeutic opportunities. Importantly, 10% of specimens in our series expressed only the variant isoform TMPRSS2-ERG fusion, which would be missed using published RT-PCR primers from the known TMPRSS2 transcript.5, 14, 15, 17 Our discovery, therefore, has important implications for the accurate detection of TMPRSS2-ERG fusion transcript, which is desirable for defining the molecular and pathological correlates of the fusion in tumor specimens, and may be useful in the future for molecular diagnosis. Specifically, accurate detection of TMPRS22-ERG fusion transcripts in prostate tissue, or in fluids such as urine, prostatic secretions, seminal fluid, ejaculate or blood, may prove useful in the diagnosis of prostate cancer, in prognostication and selection of appropriately aggressive therapies, and possibly in the future in selection of patients for new directed therapies targeting the oncogenic fusion.