Novel Research on Fusion Genes and Next-Generation Sequencing

Discovery of Recurrent Translocations in Prostate Cancer

As of 2005, recurrent translocations were only associated with sporadic solid tumors in sarcoma, some thyroid cancers, and pediatric malignancies. Using gene expression arrays and a bioinformatics filtering algorithm called cancer outlier profile analysis (COPA), Chinnaiyan and coworkers first identified recurrent translocations occurring in prostate cancer, finding juxtaposition of the 5′ untranslated region of the prostate-specific gene TMPRSS2 with the proximal coding region of either ERG or ETV1 , members of the ETS family of transcription factors. In general, carcinogenic translocations occur as one of two classes: (1) Juxtaposition of two disparate coding regions that in combination create a chimeric fusion protein having novel function, such as with breakpoint-cluster region ( BCR ) - Abelson kinase ( ABL ) in CML; or (2) juxtaposition of a proto-oncogene with the promoter of a highly expressed gene, which is the nature of recurrent translocations in prostate cancer. The COPA algorithm detects the second class of translocation by virtue of high expression of oncogenes in a particular subset of cancers (outliers). TMPRSS2 expression is driven by the androgen receptor (AR), so translocation results in high expression of any coding sequences fused distal to the TMPRSS2 breakpoint. ERG and ETV1 had been previously identified as elements of recurrent translocations in Ewing sarcomas, and so a similar carcinogenic mechanism was implicated in prostate cancer. The proteins encoded by ETS gene family members are transcription factors with homology to proto-oncogenic sequences within the E26 virus. They have properties similar to classical proto-oncogenes such as Src kinase, in that they have been shown to have transforming capacity when expressed in normal cells. The human genome encodes 27 ETS family members that all bind to a consensus sequence GGA(A/T). The gene family includes ERG and ETV1, which contain the aforementioned 3′ translocations in Ewing sarcoma and have been identified in about half of prostate cancers. Although this family of transcription factors have promiscuous binding to regulatory elements (by virtue of their short DNA response elements), interactions with other transcription-regulating proteins elicits a tissue-specific and biology-specific response.

The ETS transcription factor family members are therefore considered bona fide oncogenes. However, the translocations identified by Tomlins et al. involving ETV1 were identified at a lower-than expected frequency than that predicted by COPA, suggesting unrecognized ETS rearrangements. Subsequent studies identified alternative 5′ translocation partners, in place of TMPRSS2 , which more frequently juxtaposed ETV1 . These translocations had a similar mechanism: the ETV1 locus was fused downstream of one of four genes strongly expressed in prostate. Additionally, rare 3′ translocation partners were identified, which involved other ETS family proto-oncogenes, ETV4 , FLI1 , and ETV5 . In summary, this category of recurrent translocations juxtapose the regulatory element of one of several 5′ partners that are strongly expressed in prostate or prostate cancer, linking them to the majority of the coding sequence of one of at least six ETS transcription factor proto-oncogenes (see Figure 4.1 ). When expressed, the proteins are often truncated at the N-terminus. Tomlins et al. have suggested that ETS rearrangements can be subdivided into one of five classes based on upstream regulatory sequence. The immediate consequence of these translocations is the overexpression of ERG, ETV1, ETV4, ETV5, ELK4, or FLI1 protein, resulting in downstream transcriptional profile changes.

Function of Recurrent Prostate Cancer Translocations

The role of any oncogenic transcription factor will presumably result from alterations in downstream gene expression, so research has focused on determining the targets of the ETS family. Because ERG is the most common TMPRSS2 translocation partner, ERG function in prostate cancer has been the most intensely studied. In particular, identification of ERG transcriptional targets might explain its role in carcinogenesis and suggest a focus for therapeutic intervention. The first effect attributed to overexpression of ETS family genes was increased invasiveness when expressed in RWPE cells, which are immortalized normal prostate epithelial cells. This invasive capacity is mediated by an invasion-associated gene expression program, and is also associated with epithelial–-mesenchymal transition. Although historically proto-oncogenes have been shown to transform model systems, surprisingly ETS overexpression did not lead to transformation in this study. ERG or ETV1 translocations seem to be necessary but not independently sufficient for development of prostate cancer, and it is suspected that additional genetic lesions are required. Supporting this belief, when transgenic mice harboring the prostate-specific overexpression of ERG also have heterozygous loss of the PTEN tumor suppressor, they develop prostate cancer. This is clinically relevant as PTEN loss is frequent in prostate cancer and seems to coincide with ETS gene rearrangements. ETS family overexpression in combination with PTEN loss is therefore a clinically meaningful model of prostate cancer, but knowledge of the oncogenic mechanism is still evolving (explained in more detail later).

Rather than identifying targets by examining gene expression changes after introduction of ERG, the ETS cistrome (the gene regulatory regions to which ETS directly binds) can be directly characterized using chromatin immunoprecipitation (ChIP). ChIP was used to show that ETS family binding sites often overlap AR transcriptional targets. Unexpectedly, sequence analysis of AR targets identified using ChIP often did not contain canonical 15-base pair androgen response elements (AREs). Instead, AR targets contained a partial six-base pair ARE, which was also associated with an ETS response element in 70% of the targets, suggesting coregulation of gene expression. Additionally, androgen treatment of LNCaP prostate cancer cells induced redistribution of ETS1 , as well as increased ETS1 binding to the AR cistrome. AR and ETS1 interact in immunoprecipitation experiments, and so although ETS1 has not been formally shown to be involved in recurrent translocations in prostate cancer, the potential that other ETS family members associated with translocations could substitute for ETS1 is intriguing.

The finding of ETS transcription factors and AR target co-occupancy was confirmed and further explored using next-generation sequencing (NGS) of ChIP targets. The ETS response element was the second most enriched motif after the ARE in the AR cistrome: 40% of the AR binding sites had an ARE, 29% had an ETS response element, and this overlapping pattern was recapitulated across cell lines and found in a human prostate cancer sample. Furthermore, ERG and AR binding sites also overlap in a mouse model of prostate-specific ERG overexpression in a PTEN heterozygous context. ERG and AR proteins directly interact, but unexpectedly, ERG binds the AR promoter to downregulate AR expression in vitro ; this leads to low AR-induced gene expression in tumors harboring ERG translocations, which seems counterintuitive. Lastly, ERG-mediated AR target repression seems to be mediated by methylation of histone H3-K27, an enzymatic reaction carried out by EZH2 and Polycomb Repressive Complex 2 (PRC2).

Epidemiology/Biomarker Studies

When initially described, the TMPRSS2–ERG and TMPRSS2–ETV1 fusions were identified in 23 of 29 samples (79%) analyzed independent of any knowledge regarding the more general involvement of ETS family gene expression or translocations. However, this value underestimates the rate of translocations involving the ETS family in prostate cancer and overestimates the frequency of TMPRSS2-ERG translocations. Because multiple 5′ and 3′ translocations have been identified, which result in overexpression of an ETS family member, any study looking at just one pair of translocation partners will underestimate the overall rate of translocation. TMPRSS2-ERG , however, are the most common translocations and account for 90% of all prostate cancer oncogenic translocations and occur in about 50% of all prostate cancer specimens examined to date, although this was mostly evaluated in prostate-specific antigen (PSA)-screened populations.

Because of the ubiquity of these translocations and the evolving debate around the benefit of PSA-based prostate cancer screening, ETS-associated translocations have been proposed as a molecular marker for prostate cancer. The most compelling data for TMPRSS2-ERG as a biomarker examined urine from >1300 men with prostate cancer and correlated it with pathological indicators of disease aggressiveness. Subanalysis of >1100 men undergoing prostate biopsy had urinary TMPRSS2-ERG measurement prior to prostate biopsy. Presence of a positive signal correlated strongly with risk of cancer diagnosis, number of positive cores, total cancer involvement of all cores, and clinically significant cancer as defined by Epstein criteria. The relationship of fusion transcript presence and pathologic surrogates of disease aggressiveness was further examined in a cohort of patients who underwent radical prostatectomy, and the presence of TMPRSS2-ERG correlated with clinically significant disease (defined as tumor size >1 cm ³ , Gleason score >6, or nonorgan confined disease). This is provocative preliminary work, and requires greater delineation of thresholds and prospective validation before it can be considered for routine application.