Prostate cancer is the most frequent malignancy in men. Although the majority of patients present with early stage tumors that can be surgically treated in a curative manner, ~20% of the tumors will progress to metastatic and hormone refractory disease, accounting for >250.000 deaths per year worldwide (1). Targeted therapies that would allow for an effective treatment after failure of androgen withdrawal therapy are lacking.

Recent whole genome sequencing studies have shown that the genomic landscape of prostate cancer differs markedly from that of other solid tumor types. Whereas, for example, breast or colon cancer is characterized by high-grade genetic instability and presence of a multitude of mutations, deletions, and amplifications including important therapy target genes such as HER2 and EGFR (2,3), prostate cancers show only comparatively few mutations and almost completely lack amplifications (4–7). In contrast, prostate tumors are typically characterized by translocations, deletions, and gene fusions, the latter of which are recurrently involving androgen-responsive genes and transcription factors of the E-twenty six (ETS) family (8). The most frequent ETS-fusion is caused by interstitial deletion or translocation of a 3.7 Mb genomic segment located between the TMPRSS2 serine protease and the ERG transcription factor at chromosome 21q22. Approximately 50% of prostate cancers carry the TMPRSS2:ERG fusion, which brings ERG under the control of the androgen responsive TMPRSS2 promoter and results in permanent expression of ERG (9). Accordingly, ETS-fusion proteins have been proposed as putative targets for future gene-specific therapies (10).

In a recent study, which was performed in the context of the International Cancer Genome Consortium (11) (ICGC) project on Early-Onset Prostate Cancer, we have carried out integrated genomic analyses, including whole-genome, transcriptome, and DNA methylome sequencing in 11 early onset prostate cancer (EO-PCA) patients and detected a total of 156 individual gene fusions, 140 of which were non-recurrent and unrelated to ETS genes (5). It could be possible that some of these rearrangements result in expressed fusion proteins that could serve as cancer-specific therapy targets, provided that these rearrangements occur at sufficient frequency to justify the efforts of drug development. Accordingly, the aim of the present study was to determine the prevalence of rearrangements of 27 genes by fluorescence in situ hybridization (FISH) analysis in 500 prostate cancer samples in a tissue microarray format.

Rearrangements were detected for 13 (48%) of the 27 tested genes. Recurrent breakage was found for NCKAP5, SH3BGR and TTC3 in 3 tumors each, as well as for ARNTL2 and ENOX1 in 2 cancers each. One rearranged tumor sample was observed for each of VCL, ZNF578, IMMP2L, SLC16A12, PANK1, GPHN, LRP1 and ZHX2. All but four rearrangement were unbalanced, i.e. either the 5′ or the 3′ part of the FISH probe was lost. For ZNF578, SH3BGR, LPR12 and ZHX2 a split signal was found suggesting balanced translocation. Deletions were markedly more frequent than translocations. The most frequently deleted genes were NCKAP5 (7.5%), VCL (6.8%), PANK1 (5.9%), ARNTL2 (5.8%), SLC16A12 (5.6%), SH3BGR (3.0%) and PCNXL2 (1.6%). All detected deletions were heterozygous. No alternations were found for C11orf41, MLLT4, ALDH7A1, EPN1, NR3C1, PACRG, LYRM4, DPF3, FAM154A and WDR67. The number of successfully analyzed samples per target gene, and the frequency and type of rearrangements and deletions for all analyzed genes is summarized in Table III. Representative FISH images are shown in Fig. 1.

The results of the present study demonstrate that most chromosomal rearrangement, including balanced translocations and partial deletions characterized by intragenic breaks, represent very rare events in prostate cancer. The prevalence of breakage events affecting the 27 analyzed genes in this study was usually below 1%.

Based on our data, obtained in a cohort of over 500 tumors, it is not surprising that whole genome sequencing studies on prostate cancer found only few recurrent rearrangements (except TMPRSS2:ERG) in a total of 18 cancers (4,5). Although >250 individual non-ETS gene fusion events (resulting from translocations, inversions and duplications) were identified in these two studies in total, only 16 non-ETS genes in the study by Berger et al (4) and 1 gene in the study by Weischenfeld et al (5) were recurrently hit by structural rearrangements, however, in each case there was a different fusion partner. Only ETS-fusions were highly recurrent in these studies, with 4/7 tumors (4) and 8/11 tumors (5) carrying the TMPRSS2:ERG fusion.

Little is known about the prevalence of individual gene rearrangements (except TMPRRS2:ERG) in prostate cancer. Two studies performed by Reid et al (14) and us analyzed breakage of the PTEN tumor suppressor, and reported 7% (13/187) (14) and 3% (162/5,404) (5) of PTEN breakage, which was typically (3 out of 4 affected cases) associated with deletions of the second PTEN allele. In addition, we have previously studied breakage of the 3p13 tumor suppressor FOXP1 (15) and found 1.2% of rearrangements. These data suggest that rearrangements are infrequent even for genes with a key role including PTEN. The 0.2–1% of rearrangements found for half of the genes analyzed in the present study fit well to these numbers.

The selection of the 27 genes analyzed in this study was based on the findings of our International Cancer Genome (ICGC) project, where we employed the paired end deep sequencing strategy (16) to specifically identify gene breakages, translocations and gene fusions. In the present study we found a total of 140 non-ETS gene rearrangements. For the present study, we randomly selected genes that were potentially involved in non-ETS fusions between protein-coding genes or gene inactivation by translocation or gene breakage (5). Such genes are candidates for a dual tumor relevant function, including a putative tumor suppressor function based on inactivation by gene breakage, as well as a putative oncogenic in case of expressed fusion genes.

In this study, deletion of the analyzed region was more frequent than rearrangement. This fits well with the known relevance of many of the analyzed genes, which were located at chromosomal regions that are frequently deleted on prostate cancer, including for example PANK1, VCL and SLC16A12 (10q22-q23, deleted in 20–30%) (17–19), NCKAP5 (2q21, deleted in 10–30%) (17–19), or ARNTL2 (12p11-p12, deleted in 15–60%) (17,19), explaining the markedly higher frequency of deletions as compared to rearrangements. The deletion frequencies observed in the present study were markedly lower than in these studies, which can be explained by the fact that we did not use a deletion-specific FISH assay including a combination of a locus-specific and a centromere reference probe. With the break-apart probe used in this study, we only called absolute deletions showing unequivocal loss of one red-green signal pair but missed relative deletions, which frequently occur in aneuploid cancers.

Several of the genes analyzed in this study, including NCKAP5:MGAT5, C11orf41:RAG1, SH3BGR:RIPK4, FAM154A:IRAK3 and CCNT1:PANK1, were involved in gene fusions leading to overexpression of the fusion partner according to our previous study (5). Such fusion genes may represent suitable targets for new gene specific therapies, since they are specific for the cancer cells. However, the vast majority of gene breakages detected in this study were unbalanced, with loss of either the 3′ or the 5′ fraction of the gene, suggesting a partial deletion of these genes. Only 4 genes, ZNF587, SH3BGR, LRP12 and ZHX2, showed balanced rearrangements that might have led to gene fusions. These findings suggest that intragenic breaks may in most cases indicate a deletion break point located inside a coding gene, while formation of a specific rearrangement with a possible functional fusion gene seems to be a comparatively rare event.

We manufactured break-apart probe assays to detect rearrangements of the 27 candidate genes in a tissue microarray format. The use of our tissue microarray format in combination with FISH enables a fast and cheap analysis of gene rearrangements to detect common recurrent gene changes. Break-apart assays are capable of detecting all types of rearrangements of a probed gene, including translocation, (partial) deletion and inversion, and are thus optimally suited to estimate the prevalence of rearrangements for a given gene. We selected a cut-off level of ≥60% affected tumor cell nuclei for the detection of rearrangements in order to avoid false-positive findings due to truncated cell nuclei in 4 μm tissue sections. This cut-off was based on our previous studies analyzing breakage of ERG (20) and PTEN (5,17). Using this threshold we found a high (>95%) correlation between ERG breakage by FISH and ERG expression be immunohistochemistry (20), supporting the validity of our approach to screen for recurrent gene rearrangements.

In summary, the present study shows that a multitude of genes can be affected by chromosomal rearrangements in prostate cancer, but the frequency of specific rearrangements is typically in the range of 1% or less. In most cases, these rearrangements will result in gross deletions inactivating the affected gene. True translocations, potentially resulting in fusion genes, are comparatively rare.