To assess whether oestrogen-regulated gene (ERG) expression analysis using GeneChip arrays can predict transmembrane protease, serine 2 (TMPRSS2)-ERG fusion. The expression level of the TMPRSS2-ERG gene was studied in various histological grades of prostate cancer and castration-resistant prostate cancer (CPRC).
Patients and methods
GeneChip Affymetrix exon 1.0 ST arrays were used for expression profiling of ERG, erythroblast transformation-specific (ETS) variant gene 1 (ETV1), ETV4 and ETV5 genes in 67 prostate cancer tissue specimens.
Real-time quantitative polymerase chain reaction analysis and in some cases DNA sequencing was used to validate the presence and the expression levels of TMPRSS2-ERG gene fusions.
In our series of patients with prostate cancer over expression of the ERG gene predicted the presence of TMPRSS2-ERG rearrangements in almost all cases.
ETS expression by itself outmatched the diagnostic performance of the ERG exons ratioing allowing equal detection of the less frequent ETS gene fusion transcripts.
The gene fusions were expressed at significantly lower levels in CPRC but occurred more frequently than in primary prostate cancer.
ERG expression analysis using GeneChip arrays appears to be an excellent diagnostic tool for identifying gene rearrangements.
In coming years, measuring expression of the ETS gene family by itself might become a clinically relevant surrogate test to identify patients with fusion-positive prostate cancer.
The variation of gene fusion expression levels, particularly in CPRC, needs to be taken into account when using quantitative molecular diagnosis of prostate cancer.
Prostate cancer constitutes a major healthcare problem. At present, in Europe alone the prostate cancer incidence rate is >15%, with almost 240 000 new cases diagnosed and >85 000 deaths each year . Unfortunately, standard diagnostic test, i.e. serum PSA level, has a low specificity in prostate cancer detection with only a 25–40% positive predictive value within the PSA grey zone range between 4.0 and 10.0 ng/mL .
There is compelling evidence that genomic rearrangements should be considered as initial events in oncogenesis . In recent studies gene alterations involving androgen-regulated transmembrane protease, serine 2 gene (TMPRSS2) and erythroblast transformation-specific (ETS) transcription factor genes have been identified in patients diagnosed with prostate cancer. TMPRSS2 fusion with the ETS family member, oestrogen-regulated gene (ERG), considered as a key oncogene in prostate cancer, constitutes the predominant variant occurring in 40–70% (≈50%) of patients with prostate cancer. Importantly, a recent study showed that the mechanism of fusion could be explained by androgen receptor-mediated recruitment of topoisomerase 2B (TOP2B) and DNA double-strand breakage . Further, recent reports have shown the association of the presence of TMPRSS2-ERG fusion with prostate cancer-specific death  or earlier biochemical recurrence . Considering the high incidence of prostate cancer, TMPRSS2-ERG fusion remains the most common genetic aberration in human malignancies . Both genes are located on chromosome 21; TMPRSS2 at 21q22.3 and ERG at 21q22.2. The predominant mechanism for the gene fusion is the loss of 2.8 Mb of genomic DNA between TMPRSS2 and ERG .
TMPRSS2-ERG rearrangement can be detected in urine after DRE. The presence of the fusion in urine has >90% specificity and 94% positive predictive value for prostate cancer detection . Given the high specificity of this genomic alteration, fusion status may shortly become a clinically relevant biomarker for establishing the presence or absence of prostate cancer . Additionally, targeting gene rearrangement could represent a valuable approach for the development of an effective prostate cancer therapy.
TMPRSS2-ERG translocation occurs almost exclusively in patients with the substantial overexpression of ERG . Therefore, it was hypothesised that the increase in ERG by itself was only good enough to predict the presence of the genomic alteration . In addition, Jhavar et al.  have recently reported that an accurate assessment of ERG expression on the exon level using Affymetrix GeneChip arrays allows identification of TMPRSS2-ERG fusions in human prostate cancer. In the present study, we wanted to confirm the potential utility of the Affymetrix exon 1.0 ST array, as a fast and reliable method of identifying TMPRSS2-ERG fusion transcripts in our collection of 67 prostate cancer specimens from patients with low-grade, high-grade, castration-resistant prostate cancer (CPRC) or metastatic disease. We also assessed whether expression analysis of the ERG oncogene alone and/or ERG exons ratioing might constitute a surrogate marker to predict TMPRSS2-ERG gene fusions.
Patients and Methods
In all, 67 patients with different stages of prostate cancer were selected for this study. A consent form approved by the Institutional Review Board was signed by all study participants. Based on the pathology findings and the case records four different groups of patients with prostate cancer were identified: low-grade (19 patients), high-grade (20), CPRC (21) and metastatic (seven). Low-grade prostate cancer was defined as Gleason score <7 and high-grade as Gleason score ≥7. The metastatic prostate cancer group included patients with lymph node metastases. CPRC referred to patients who had failed conventional hormonal therapy.
The use of patient materials was approved by the local ethical committee of the Radboud University Nijmegen Medical Centre. At radical prostatectomy, TURP or lymph node dissection specimens were snap frozen in liquid nitrogen. Tumour prostatic tissues were selected for purity of cancer cells and processed by step sectioning.
RNA Isolation and Reverse Transcriptase Reaction
Total RNA was extracted from fresh-frozen tissue sections or cell cultures using Trizol reagent (Invitrogen). RNA quantity and purity were determined on a NanoDrop spectrophotometer and BioAnalyzer. In all, 2 μg total RNA was DNase-I-treated and cDNA was synthesised using random primers and SuperScript II-(Invitrogen) reverse transcriptase (RT, Invitrogen). Upon quantification of the RT-reaction, 10 ng of cDNA was used for PCR analysis. Gene expression profiling was performed using a microarray technique (GeneChip, Affymetrix).
Expression Profiling Using Affymetrix GeneChip Exon 1.0 ST Arrays
Expression profiles were determined using Affymetrix 1.0 Human Exon ST arrays according to the manufacturer's instructions. The Affymetrix GeneChip Human Exon 1.0 ST array consists of ≈5.4 million probes and enables measurement of gene expression on both the gene and exon level. Each gene is represented by >20 probes generally covering the entire length of the transcript . The Affymetrix GeneChip Whole Transcript Sense Target Labelling Assay was used to generate amplified and biotinylated sense-strand DNA targets from the entire expressed genome (1.5 μg total RNA) without bias. Manufacturer's instructions were followed for the hybridisation, washing, and scanning steps. Arrays were hybridised by rotating them at 60 rpm in the Affymetrix Gene Chip hybridisation oven at 45 °C for 16 h. After hybridisation, the arrays were washed in the Affymetrix GeneChip Fluidics station FS 450. The arrays were scanned using the Affymetrix Gene Chip scanner 3000 7G system. Gene- and exon-level expression signal estimates were derived from cell intensity files (CEL) generated from Affymetrix GeneChip Exon 1.0 ST arrays using the multiarray analysis algorithm  implemented from the Affymetrix Power Tools software. Gene-level estimates were obtained using the ‘core’ metaprobe list annotation release 21. Exon-level data were filtered to include only those probe sets in the core metaprobe list. To analyse ERG expression level, 25 core probes are used and distributed along the exons 2–11. All probe sets (25) that mapped within ERG were found and assigned to the appropriate exon identified from Ensembl. For each probe set within an exon, the log2 expression level was calculated. Where more than one value was available for a single exon, the median value was taken as the representative ratio for that exon. Affymetrix GeneChip Exon 1.0 ST allows detection of exons 1–13 of ETS variant gene 1 (ETV1), exons 2–13 of ETV4 and exons 1–13 of ETV5.
Detection of TMPRSS2-ERG Fusions by Real-time PCR
Fluorescence based real-time PCR assays were designed and optimised specifically for TMPRSS2-ERG gene fusion transcripts and hypoxanthine-guanine phosphoribosyltransferase (HPRT). Primer pairs and TaqMan probes were designed (Table 1; TIB Molbiol, Berlin). The TMPRSS2-ERG assay detects the three most common gene fusions identified in prostate cancer . The primers were located in exon 1 of TMPRSS2 and exon 4 of ERG. The probe was located in exon 4 of ERG and labelled with two fluorochromes: a 5′-end reporter fluorescent dye and a quencher dye at the 3′-end. Blunt-ended PCR products were cloned in the pCR-Blunt cloning vector (Invitrogen). Calibration curves with a wide linear dynamic range (10–1.106 copies) were generated using serial dilutions of the plasmids. The amplification efficiency of the primer pair was determined using the calibration curve and was >1.85. Control samples with known template concentrations were used as a reference. In all, 2 μL of each cDNA sample were amplified in a 20 μL PCR reaction containing 10 pmol of each primer, 2 pmol of TaqMan probe and 1 x Probe Master mix (Roche). The following amplification conditions were used: 95 °C for 10 min followed by 50 cycles at 95 °C for 10 s, 60 °C for 30 s and cooling at 40 °C for 55 s (Light Cycler LC480, Roche). The crossing point (Cp) values were determined using the Light Cycler 480 SW 1·5 software (Roche). The Cp values of the samples were converted to concentrations by extrapolation in the generated calibration curve. Real-time PCR experiments were performed twice for each sample, and classified as positive for TMPRSS2-ERG whenever they contained ≥10 copies in a single experiment. A fluorescent signal of <1 (465–510 nm) was classified as background signal. The assay performance of the real-time PCR experiments were evaluated during in-study validation. The reference control samples had an inter- and intra-assay variation <30%.
5′-FAM-TGG TCC TCA CTC ACA ACT GAT AAG GCT TCC-BBQ-3′
5′-CTC AAC TTT AAC TGG AAA GAA TGT C-3′
5′-TCC TTT TCA CCA GCA AGC T-3′
5′-LC610-TTG CTT TCC TTG GTC AGG CAG TAT AAT C-BBQ-3′
Expression Profiling of ERG Genomic and Exon Level
The Affymetrix exon 1.0 ST arrays were used to assess the expression of ERG gene and its exons. There was ERG over expression in 38 (57%) of 67 prostate cancer tissue specimens (Table 2, Fig. 1). The number of samples over expressing ERG was similar in the low- (58%) and high-grade (55%) prostate cancer and CPRC (52%) groups. The highest number of overexpressing ERG samples was found in patients with metastatic prostate cancer (five of seven patients). Additionally, for each cancer specimen we calculated exon 4 to exon 2 (ex4/ex2) and exon 6 to 2 (ex6/ex2) relative ratios of the ERG oncogene. In all but one case ERG over expression coexisted with the elevated exons ratios (Table 2, Fig. 2A,B).
Table 2. ERG overexpression and identification of TMPRSS2-ERG fusions
Prostate cancer specimens, n
ERG, n (%) or n/N
ERG (ex4/ex2 ratio) or ERG (ex6/ex2), n (%) or n/N
TMPRSS2-ERG expression (qPCR analysis), n (%) or n/N
ERG Expression Profile and TMPRSS2-ERG Fusion Status
Considering that TMPRSS2-ERG fusion occurs in >90% of cases in patients with ERG elevation  and that the exons of ERG that become fused to TMPRSS2 are expressed at a very high levels , we investigated the association between ERG over expression/ERG (ex4/ex2 and ex6/ex2) ratios and the presence of the rearrangement. In patients with primary prostate cancer (low- and high grade) an almost perfect correlation was found between ERG alteration assessed with GeneChip arrays and the expression of fusion transcript in quantitative (q)PCR analyses, i.e. only in one patient with the substantial over expression of ERG and its increased exons ratios was there no expression of fusion transcript. However, a discrepancy between array analysis and qPCR was found in patients with CPRC (see next paragraph).
TMPRSS2-ERG Gene Expression Levels in Primary and CPRC
A qPCR analysis with specific probes for the detection of TMPRSS2-ERG fusion transcripts was used to assess TMPRSS2-ERG gene expression level. There was increased TMPRSS2-ERG expression in 41 (61%) of 67 prostate cancer tissue specimens (Table 2). There was a significantly (P < 0.001) higher expression of TMPRSS2-ERG in primary prostate cancer than in CPRC (Fig. 3). Although the expression of TMPRSS2-ERG in patients with CRCP was low its presence was detected much more frequently than in primary prostate cancer, in 71% and 54% of cases, respectively (Fig. 4, Table 2). In four (19%) patients with CPRC with low ERG expression on exon array, qPCR revealed the presence of TMPRSS2-ERG transcript.
TMPRSS2-ERG Fusions and Gleason Score
In our series of patients, there was no correlation between the percentage of specimens with the rearrangements (P = 0.729; r2 = 0.026) and the Gleason score (Fig. 5).
TMPRSS2 and ETV Genes: ETV1, ETV4 and ETV5
The fusions of TMPRSS2 with other than ERG members of ETS transcription factors occur with much lower frequency. ETV1, ETV4 and ETV5 genes altogether account for at most 10% of the rearrangements [2007,2008]. Furthermore, the outlier profiles of ETS genes are mutually exclusive meaning each tumour could over express only one of those genes. Indeed, we observed a unique and substantial over expression of ETV1, ETV4 and ETV5 in four (6%), two (3%) and two (3%) of the 67 tissue specimens, respectively. Despite their relevant over expression we were unable to identify any 5'-fusion partners for the ETV genes in our cohort of patients with prostate cancer. Importantly, the over expression of ETV1 and ETV5 was seen only in aggressive prostate cancer phenotypes: CPRC (three patients) or high-grade prostate cancer (three).
The discovery of the BCR-ABL fusion protein together with the implementation of the translocation-specific treatment has revolutionised the management of chronic myeloid leukaemia (CML) and significantly increased the survival of patients with CML . Likewise, the identification of prostate cancer-specific fusion of the 5'-untranslated region of TMPRSS2 with the 3'-exons of the ETS transcription factors might not only improve diagnosis but also might affect the treatment of prostate cancer. However, the detection of gene fusions using traditional approaches: RT-PCR and/or fluorescence in situ hybridization (FISH) is technically demanding and difficult to apply in clinic. Therefore, we wanted to identify fusion-positive patients in a fast and reliable way using a modern array technology. In the present study, we showed that GeneChip array of ERG expression provides an almost perfect prediction of the TMPRSS2-ERG rearrangement in patients with primary prostate cancer.
In the present cohort all but one patient with a high level of ERG expression had rearrangement, whereas those with low levels of ERG expression did not. The identification of the translocation is also possible using ERG exons ratioing assuming that the expression of the ERG exons that become fused to TMPRSS2 is elevated . It has been reported that most of the prostate cancer specimens that exhibit the translocation harbour increased expression of ERG exons 4–11 relative to the expression of exons 2 and 3  and the predominant translocation includes TMPRSS2 exon 1 fused with ERG exon 4 (TMPRSS2 ex1-ERG ex4) . The increased relative expression ratios of ex4/ex2 and/or ex6/ex2 occurred in all but one patient with ERG elevation. The only limitation of ERG exon ratioing compared with ERG overexpression alone is its inability to identify the infrequent fusion transcripts containing ERG exon 2. This is due to the fact that Exon 1.0 ST array is lacking probe sets for ERG exon 1 and in patients overexpressing ERG exon 2 and/or overexpressing all of the assessed exons of ERG (2–11), the differential expression ratio between the exons does not allow prediction of fusions. The DNA sequencing showed in one case that the discrepancy between ERG ex4/ex2 ratio and ERG elevation was due to the rare fusion variants of TMPRSS2 exon 1 and/or TMPRSS2 exon 2 with ERG exon 2 (TMPRSS2 ex1/ex2-ERG ex2), i.e. the fusion-positive patient with a substantial overexpression of ERG had a low ex4/ex2 ratio due to the elevated expression of both exons. Therefore, in patients containing the TMPRSS2-ERG exon 2 translocation ERG exons ratioing (ex4/ex2) is unable, contrary to ERG expression by itself, to identify the gene fusion (Supplementary Fig. 1). Importantly, the combined analysis of ERG expression together with ERG exons ratioing allows not only the prediction of fusion but also to relatively well assign the ERG exons involved in the translocation.
Interestingly, in CPRC group we observed a difficult to explain more frequent but significantly lower than in primary prostate cancer expression of TMPRSS2-ERG (Fig. 3). Due to this relatively low expression of fusion transcripts the array analysis did not allow prediction of the translocations in four of 21 CPRC tumours. There was a slightly lower expression of TMPRSS2-ERG in patients with lymph node metastases compared with those with primary prostate cancer. Therefore, prostate cancer progression or/and androgen-deprivation therapy might lead to the variations in the level of fusions expression.
A previous report, although performed on small series of patients, is in agreement with our present observations. Jhavar et al.  also used Affymetrix GeneChip arrays and identified TMPRSS2-ERG fusion transcripts in most of the radical prostatectomy specimens by calculating the ERG expression together with the ratio of its exons 4–11 relative to exons 2 and 3. According to the Authors, Affymetrix Exon 1.0 ST arrays due to the analysis of gene expression at the exon-level contrary to Affymetrix U133A human gene arrays allows accurate discrimination between fusion-negative and -positive samples based on the ERG expression only. The incidence of samples with an intermediate level of ERG expression on the Affymetrix U133A human gene arrays impedes the clear distinction between the presence and absence of the fusion. In addition, Jhavar et al. claimed the mechanism of the fusion involves the deletion of exon 8. Contrary to this, we did not find lower expression of ERG exon 8 compared with exons 4–7 and 9–11 in any of our prostate cancer specimens, which does not support the theory of the splicing out of ERG exon 8 in each case of TMPRSS2-ERG translocation. The discrepancy between studies might be related to the slight difference in methodology. Jhavar et al. evaluated the ratio of ERG gene expression of prostate cancer tissue specimen to that reported in the non-neoplastic prostate epithelium. We were analysing only the expression of the ERG gene in cancer tissue.
In conclusion, we were able to show that analysis of ERG expression on Exon 1.0 ST arrays allows prediction, in a fast and reliable way, of TMPRSS2–ERG fusions in patients with prostate cancer. Considering the high accuracy of the technique, GeneChip arrays of ERG expression might shortly become a clinically relevant tool in identifying fusion-positive patients. The variation of gene fusion expression levels in the present study, particularly in CPRC, demands further investigations.
Maciej Salagierski was supported by the European Urological Scholarship Programme grant: S-02-2008 from the European Association of Urology (EAU) for performing the project: ‘New targets for molecular diagnosis of Prostate Cancer’ in the Department of Urology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands.
Conflict of Interest
Jack A. Schalken: Employer owns and has license to third party. Maciej Salagierski, Frank P. Smit and Sander Jannink: no conflicts of interest declared.