Identification of Novel Fusion Genes in Testicular Germ Cell Tumors

Testicular germ cell tumors (TGCT) are the most common cancer in young men, ages 15 to 44 years (1). Although it is a highly treatable cancer type, exemplified by a 10-year net survival rate of 98% in England and Wales (2), the disease affects men in their prime and treatment can lead to substantially increased morbidity, including cardiovascular disease, reduced fertility, and secondary cancers (3). Histologically, there are two main subtypes of TGCTs, seminomas, and nonseminomas. Both are thought to develop from the preinvasive stage termed intratubular germ cell neoplasia (IGCN; also known as carcinoma in situ). Nonseminomas are further divided into the pluripotent embryonal carcinomas (EC) and more differentiated subtypes, with either somatic (teratoma) or extraembryonic differentiation (yolk sac tumors, YST, and choriocarcinomas; ref. 4).

EC cells are highly similar to embryonic stem (ES) cells, derived from the inner cell mass of the blastocyst stage embryo (5). Both cell types exhibit pluripotent characteristics phenotypically and in gene expression profiles (6, 7). Upon extended passaging in vitro, ES cells have been shown to acquire genetic changes similar to those seen in malignant transformation in vivo of TGCTs and EC, including gain of genetic material from chromosomes 12, 17, and X (8). Gain of chromosome arm 12p, often as an isochromosome, i(12p), is found in virtually all cases of TGCT (9, 10). Crucially, despite these similarities, EC cells are malignant in character, whereas ES cells are not. Comparative studies between the two cell types may therefore be useful for characterization of cancer-specific differences in a pluripotent context (5, 11). One such study revealed that several transcription factors located on 12p are overexpressed in EC cells as compared to ES cells (6). Although 12p material is gained in virtually all cases of TGCT, no clear genetic driver for TGCT malignant transformation has been pinpointed (12, 13).

Recently, whole-exome sequencing studies have revealed that the number of nonsynonymous mutations in coding regions of the TGCT genome are few, on a scale similar to that of pediatric cancers (14–16). A number of pediatric cancers with a low mutational load are frequently found to harbor fusion genes with oncogenic properties. Examples are MLL rearrangements in acute lymphoblastic leukemia (17), and subtypes of sarcomas classified by distinct chromosomal translocations (18). In Ewing sarcoma, fusions involving EWSR1 are pathognomonic, while the mutation rate is low, estimated at 0.15/Mb of coding sequence (19). In this study, we have performed RNA sequencing of EC cell lines and their nonmalignant counterpart, ES cell lines. Application of a fusion gene analysis pipeline led to the identification of nine novel fusion genes and transcripts, to our knowledge the first described in TGCT.

Three EC cell lines (2102Ep, 833KE, and NTERA2) and two ES cell lines (H9 and Shef3) were subjected to RNA sequencing. The EC and ES cell lines were established in the lab of Peter W. Andrews, University of Sheffield, where they also were grown and sorted for expression of the pluripotency marker SSEA3 as previously described (11). The extended experimental validation panel consisted of four categories of samples: (i) two additional EC (Tera 1 and NCCIT) and two additional ES (Shef6 and Shef7) cell lines (n = 4); (ii) NTERA2 and 2102Ep cells treated with all-trans retinoic acid (RA) for 0, 3, and 7 days to induce differentiation, as previously described (n = 6; refs. 20, 21); (iii) 36 testicular tissue samples, including six normal testicular parenchyma, six premalignant IGCN, and 24 primary TGCTs, each with only one histologic subtype: EC (n = 8), seminoma (n = 7), choriocarcinoma (n = 1), YST (n = 4), and teratoma (n = 4); and (iv) 20 normal tissues from miscellaneous sites of the body used for exploration of cancer-specificity of the novel transcripts (adipose, bladder, brain, cervix, colon, esophagus, heart, kidney, liver, lung, ovary, placenta, prostate, skeletal muscle, spleen, stomach, testes, thymus, thyroid, and trachea; FirstChoice Human Normal Tissue Total RNA). These were each a pool of RNA from at least three individuals, with the exception of one individual sample from the stomach (Ambion, Applied Biosystems by Life Technologies).

DNA isolated simultaneously from the cell pellets was tested and authenticated by STR fingerprinting using the AmpFLSTR Identifiler PCR Amplification Kit (Applied Biosystems). Profiles positively matched with those reported in the literature for 2102Ep (7), and obtained from the European Collection of Cell Cultures (for 833KE), ATCC (for NTERA2, NCCIT, and TERA1), the Wisconsin International Stem Cell Bank (H9), and the UK Stem Cell Bank (Shef3, Shef6, and Shef7). The biobank is registered according to Norwegian legislation (no. 953; Biobank Registry of Norway) and the project has been approved by the National Committee for Medical and Health Research Ethics (S-05368 and S-07453b).

Paired-end RNA-sequencing data from the Illumina Human Body Map v2 dataset, consisting of 16 nonmalignant miscellaneous tissue types, was analyzed as an additional source of normal controls (ArrayExpress accession ID E-MTAB-513 and European Nucleotide Archive study accession ID ERP000546).

Library construction was performed using the standard Illumina mRNA library preparation protocol (Illumina Inc), including poly-A mRNA isolation, fragmentation, and gel-based size selection. Shearing to about 250 bp fragments was achieved using the Covaris S2 focused ultrasonicator (Covaris Inc). Paired-end sequencing, 76 bp from each end, was performed according to protocol on a Genome Analyzer IIx (Illumina Inc.).

To identify fusion transcripts specific for EC, we used the fusion detection algorithm deFuse v. 0.6.1 (22) with hg19 sequence reference from UCSC and Ensembl release 69 annotation. To enrich for true positive fusion transcripts specifically expressed in EC cells, several heuristic filtering steps of the initial fusion breakpoint candidates were performed, some adapted from the recommended procedures of the original publication of deFuse (22), namely (i) considering only the nominated fusion breakpoint candidates with a probability score greater than 0.5; (ii) removing breakpoints nominated in EC that were also found in ES cell lines or tissues of the Human Body Map v2 to enrich for malignancy-specific candidates and remove systematic technical artifacts; (iii) removing candidates that had more than five multimapping spanning reads, or a ratio of multimapping spanning reads greater than 25%; and (iv) filtering out candidates nominated due to homologous or repeat sequences by removing candidates that had a deFuse homology score greater than 10 and candidates fulfilling either of the following three criteria: cDNA-adjusted, genome-adjusted, or EST-adjusted percentage of identity greater than 0.1. In an effort to enrich for functionally interesting fusion candidates, we applied four functional filters, of which, candidates needed to pass three in order to proceed: (i) both gene partners are annotated as protein-coding genes; (ii) the number of split reads and spanning reads supporting the fusion breakpoint sequence is greater than or equal to 3 or 5, respectively; (iii) the fusion breakpoint includes 5′UTR or coding parts of at least one of the partner genes; (iv) the distance between the two partner genes is greater than 30 kb. We also used an additional fusion finder algorithm, SOAPfuse v.1.26 (23), and included all fusion breakpoints that were picked up by both deFuse and SOAPfuse independently. For the remaining fusion transcripts, breakpoint alignments were evaluated in the UCSC genome browser and the Integrative Genomic Viewer (IGV). In general, we removed chimeric breakpoint sequences likely to derive from read-through transcripts, and those not aligning to conserved exon to exon boundaries.

Selected fusion transcript candidates were validated with RT-PCR in the RNA-sequenced cell lines and in an extended validation panel. Primers were designed to the fusion transcript breakpoint sequences as detected by deFuse by using the Primer3 web application (24). All primer sequences used in this study are listed in Supplementary Table S1. Briefly, reverse transcription was performed using the high-capacity reverse transcription kit according to protocol (Applied Biosystems by Life Technologies). From 50 ng of starting cDNA template, a PCR protocol was initiated with 15 minutes of HotStarTaq DNA polymerase activation at 95°C, followed by 30 thermal cycles of denaturation for 30 seconds at 95°C, primer annealing for 1 minute at optimal primer melting temperatures (Supplementary Table S1), and extension for 1 minute at 72°C. After the last cycle, a final extension step was performed at 72°C for 10 minutes. The PCR products were separated by electrophoresis at 200 V for 30 minutes on a 2% agarose gel and visualized using ethidium bromide and UV light.

To ensure specific amplification of the breakpoint sequences, PCR products from the cell lines that were nominated by RNA sequencing to harbor the individual fusion transcripts were sequenced by Sanger sequencing. PCR products that showed a single-nucleotide band on the agarose gel were sequenced directly from both sides using forward and reverse primers. Prior to sequencing, the PCR products were purified using Illustra ExoStar 1-step cleanup (GE Healthcare). The cycle sequencing reactions were performed using the BigDye Terminator v.1.1 cycle sequencing kit (Applied Biosystems) following the manufacturer's recommendations. The sequencing products were purified using BigDye Xterminator (Applied Biosystems) before being analyzed by capillary electrophoresis using the ABI 3730 DNA Analyzer (Applied Biosystems). The resulting sequences were analyzed using the Sequencing Analysis v.5.3.1 software.

Several fusion transcripts confirmed by regular RT-PCR were recurrent, however with varying nucleotide band intensities between samples, as observed by agarose gel electrophoresis. We performed TaqMan quantitative RT-PCR (qRT-PCR) to quantify the relative expression of these fusion transcripts. Primers and minor groove binding-probes were designed with the Primer Express v.3.0 software (Applied Biosystems) to cross the fusion transcript boundaries (Supplementary Table S1). Two endogenous control assays targeting ACTB and GUSB were analyzed in all samples to normalize for input template amounts. The qRT-PCR reactions were performed in reaction volumes of 10 μL, with 15 ng of template cDNA, TaqMan universal mastermix II with uracil-N-glycosylase (Applied Biosystems) and final primer and probe concentrations of 0.9 and 0.2 μmol/L, respectively. The PCR reactions were run in triplicate on an ABI 7900HT fast real-time PCR system (Applied Biosystems). Expression levels were reported as the median cycle threshold (C_T) of the triplicates and normalized to median C_T values of the endogenous controls. A threshold value at C_T = 35 was set for all assays as positive expression.

Droplet digital PCR (ddPCR) takes advantage of oil/water emulsion, separating PCR reagents and template into thousands of nanoliter-sized droplets. Subsequent thermal cycling by traditional fluorescent PCR specifically amplifies target templates in the droplets. The number of target molecules in a reaction mixture is inferred by counting the number of droplets with and without amplified fluorescent signal. It is also possible to measure two target molecules simultaneously, by multiplexing two PCR assays with different fluorescent dyes (FAM and VIC/HEX). As template molecules distribute randomly into droplets, droplets can be expected to contain one, the other, both, or none of the target molecules by chance in a multiplexed assay. However, if two template targets are located in close proximity on the same DNA molecule and thereby linked, these would distribute together in a nonrandom fashion with a higher number of double-positive droplets than expected by chance.

Here, we performed ddPCR to investigate DNA-level linkage of the partner genes of two recurrent fusion transcripts, as well as two fusion transcripts, that each were expressed only in one EC cell line. As proof of concept, we included duplex linkage assays for the known fusion VTI1A-TCF7L2, which in the NCI-H508 cell line is known to be formed by a genomic deletion (25). To evaluate the integrity of DNA fragments, and as an additional positive control of the ddPCR linkage approach, we used a custom milepost experiment, which measures linkage with assays 1 kb, 10 kb, 50 kb, and 100 kb apart. In all experiments, a reference assay with FAM fluorescence was multiplexed with one of the milepost assays with VIC fluorescence. FAM and VIC assays were also designed for the two partner genes of the fusion transcripts. All assays used in the ddPCR linkage experiments are listed in Supplementary Table S2. As control experiments for the linkage assays we performed fragmentation of genomic DNA with the NspI restriction endonuclease. We ensured that none of the assay's target sequences overlapped with the restriction enzyme target sequence. Each ddPCR experiment was carried out in 22 μL reaction volumes, with final concentrations of 0.9 μmol/L of each primer and 0.25 μmol/L probe, 1× ddPCR supermix (Bio-Rad) and 25 to 50 ng genomic DNA. Droplet generation was performed with 20 μL of the reaction mix, according to the manufacturer's protocol. Droplets were then transferred to a 96-well plate and PCR performed with the following thermal cycling profile: initial enzyme activation at 95°C for 10 minutes, followed by 40 cycles of denaturation at 94°C for 30 seconds and annealing/extension at 60°C for 1 minute. As a final step, the enzyme was deactivated at 98°C for 10 min. The droplets were read using the QX200 droplet reader according to manufacturer's protocol. The data was analyzed using the QuantaSoft software (1.7.4.0917; Bio-Rad). Crosshair gating was used to set a threshold for the four quadrants of droplet populations: double-negative, FAM-positive, VIC-positive, and double-positive. QuantaSoft outputs the concentration in molecules/μL for each of the assays. Additionally, a linkage concentration is calculated based on the ratio of double-positive droplets, given in linked molecules/μL. We calculated percent linkage as the concentration of linked molecules divided by the mean concentration of the individual assays transformed to percentage.

RNA sequencing of the three EC (2102Ep, 833KE, and NTERA2) and two ES (H9 and Shef3) cell lines generated a total of 199 million pairs of 76 bp sequencing reads that passed filtering (Supplementary Table S3).

Fusion transcript analyses of the RNA-seq data resulted in an initial list of 1210 unique fusion breakpoints with a probability score above 0.5. Subsequent heuristic filtering nominated nine fusion transcripts that were considered strong enough for further experimental validation (Fig. 1). Briefly, 283 candidate fusions were first removed as they were also detected in ES cell lines or normal human tissues (i.e., external data from the Illumina Human Body Map v2 data set). Further technical filtering of fusion breakpoints with a high ratio of multimapping spanning reads and breakpoints with a high degree of homology or breakpoint sequence identity, removed 621 and 130 candidates respectively. To enrich for functionally important breakpoints, we removed breakpoints that did not pass at least three out of four functional filters. As an additional step, we used the SOAPfuse fusion finder to identify a list of 81 potential EC fusion breakpoint candidates. Of these, only 11 overlapped with the initial list of EC specific breakpoints generated by deFuse, where five of these were already kept through the filtering process. The remaining six overlapping candidates were retrieved for evaluation in the final candidate list. After filtering steps, a list of 65 unique candidate fusion breakpoints remained, and was manually curated by viewing alignments in IGV and the UCSC genome browser. Fusion transcripts likely to be generated by polymerase read-through were filtered out, except for a read-through between CLEC6A and CLEC4D located on chromosome arm 12p found to be recurrent in all three EC cell lines. Fusion candidates in which the breakpoint did not match intact conserved exon–exon boundaries were filtered out, except for a breakpoint, ETV6-RP11-434C1.1, also located on chromosome arm 12p, which was not strictly a fusion transcript but a transcript produced from an unannotated alternative promoter. This alternative promoter of ETV6 was specially considered based on the known oncogenic relevance of the ETS family of transcription factors (26). The final list of transcripts selected for experimental validation consisted of two interchromosomal and seven intrachromosomal fusion transcript candidates (Fig. 2; Table 1). Of the seven intrachromosomal fusions, five included breakpoints with both partner genes located on chromosome arm 12p.

We performed RT-PCR to validate the presence of the nine nominated fusion transcript breakpoints in the EC and ES cell lines investigated, and for further clinical evaluation in a series of IGCN and TGCTs. All nine nominated fusion transcripts were confirmed by RT-PCR spanning the breakpoint. Successful Sanger sequences were produced from eight of these, confirming breakpoint sequences between the two gene pairs, and all found to use intact exon–exon boundaries (Supplementary Fig. S1). The eight fusion transcripts as well as the alternative promoter usage of ETV6 all had intact open reading frames (ORFs), theoretically encoding functional proteins. The ORFs of five out of nine encode N-terminally truncated proteins of the downstream partner gene, whereas three out of nine encode the full-length coding sequence of the downstream partner. None of the fusion transcripts encode potential hybrid proteins with in-frame coding sequence from both intact partner proteins. However, one of the fusion transcripts, PPP6R3-DPP3, encodes an out-of-frame ORF encoding 198 amino acids.

Five of the fusion transcripts were only expressed in the originally nominated EC cell lines, and were thus considered private fusion events. The remaining four candidates were however found to be recurrently expressed in TGCT (Supplementary Table S4), and crucially, not in normal testicular parenchyma. The four recurrent candidates included the read-through between CLEC6A and CLEC4D, alternative promoter usage of ETV6, and two fusion transcripts both involving the first two exons of RCC1 (Regulator of chromosome condensation 1) as an upstream partner gene connected to HENMT1 and ABHD12B, located 80 Mbp apart on chromosome 1 and on chromosome 14 respectively. Expression of these recurrent transcripts was variable, with both strongly positive and weaker bands detected by agarose gel electrophoresis. For more accurate assessment of expression, we used qRT-PCR to quantify the expression level of each fusion transcript. All custom TaqMan assays were found to have efficiencies between 80% and 90%. Here, a total of 73 tissue samples and cell lines were tested. None of the recurrent transcripts were expressed in normal testicular parenchyma (n = 6). The read-through between CLEC6A and CLEC4D was found to be expressed in all subtypes of TGCT, as well as premalignant IGCN (6/6) and ES cell lines (3/3). However, only two of the four teratoma tissue samples showed expression of the read-through. The read-through was only detected in 1 (the placenta) of 20 normal tissues from the human body (Fig. 3A). Alternative promoter usage of the ETV6 gene was predominantly detected in EC tissue samples and cell lines (75% and 92%, respectively; 6/8 and 12/13). Alternative promoter usage was also observed in ES cell lines (100%; 3/3), seminoma (14%; 1/7), Cc (100%; 1/1), and YST (50%; 2/4). None of the 20 included normal tissues expressed the alternative promoter of ETV6 (Fig. 3B). The intrachromosomal fusion transcript RCC1-HENMT1 was widely expressed in all subtypes of TGCT, IGCN, ES cell lines, and in 6/20 normal tissue types (spleen, esophagus, trachea, thyroid, thymus, and skeletal muscle; Fig. 3C). By contrast, the interchromosomal fusion transcript RCC1-ABHD12B was found expressed predominantly in EC tissue samples and cell lines (100%) and in seminomas (86%; 6/7). RCC1-ABHD12B was further detected in 67% (4/6) IGCN, in one of four teratomas and in one of three ES cell lines. None of the tested tissue samples from sites of the human body showed expression of this fusion transcript (Fig. 3D).

Total RNA isolated from NTERA2 and 2102Ep cell lines treated with RA for 0, 3, and 7 days were included in the validation panel. The NTERA2 cell line has previously been shown to differentiate in culture when treated with morphogens such as RA (21). 2102Ep, on the other hand, does not differentiate upon in vitro treatment with RA and remains pluripotent (27). Intriguingly, qRT-PCR analyses of both fusion transcripts involving RCC1 revealed that expression decreased upon treatment with RA in NTERA2, but not in 2102Ep (Fig. 4). The measured ΔΔC_T was −3.9 and −3.0 between 0 and 7 days of RA treatment for RCC1-ABHD12B and RCC1-HENMT1 respectively. Additionally, expression of the ETV6 transcript involving the alternative promoter was silenced after 7 days of RA treatment (C_T > 35; ΔΔC_T of −6.8). Expression of the CLEC6A-CLEC4D read-through did not change significantly upon RA treatment in NTERA2. The expression of all fusion transcripts remained unchanged in the 2102Ep cell line upon treatment with RA (Fig. 4), in line with this cell line's previously reported nonresponsiveness to RA treatment.

Because the two fusion transcripts involving RCC1 were recurrently expressed and the two partner genes were not located close to RCC1, the possibility of genome-level rearrangements as a mechanism resulting in gene fusion was tested by ddPCR linkage analysis. For another two fusion genes EPT1-GUCY1A3 and PPP6R3-DPP3, expressed in the 833KE and NTERA2 cell lines, respectively, DNA copy number data indicated a shift in the vicinity of all four genetic loci (data not shown).

To establish the integrity of DNA to be included in the linkage analyses, and as proof of concept, we performed a ddPCR milepost experiment. Here, multiplexed fluorescent TaqMan assays measured linkage 1 kb, 10 kb, 50 kb, and 100 kb apart. DNA isolated by the AllPrep method from the NTERA2 cell line showed highest integrity, with 90% linkage at 1 kb, 52% at 10 kb, and 11% at 50 kb (Supplementary Fig. S2). No evidence of linkage was seen at 100 kb. DNA from the 833KE cell line and DNA isolated by phenol–chloroform from NTERA2 showed slightly lower linkage scores, indicating more highly degraded DNA (Supplementary Fig. S2). After DNA fragmentation with the NspI restriction endonuclease, all linkage was substantially reduced, but some background levels remained (0–3.5%; Supplementary Fig. S2). Fusion gene linkage analysis of VTI1A-TCF7L2, previously reported to be caused by a deletion in the NCI-H508 cell line, showed that 18.5% of molecules are linked and contain both fusion partner targets (Fig. 5). The interchromosomal fusion EPT1-GUCY1A3 and the intrachromosomal fusion PPP6R3-DPP3 showed evidence of DNA-level linkage, with 15% and 18% linkage rates respectively (Fig. 5). However, we found no evidence for DNA-level linkage for the recurrently expressed fusions RCC1-ABHD12B and RCC1-HENMT (Fig. 5). As a control experiment, we found that the DNA-level linkage in all tested samples was lost upon digesting the DNA with the restriction enzyme NspI.

In this study, we have identified the presence of fusion genes and transcripts in TGCT. Nine novel fusion genes and transcripts are reported, of which RCC1-HENMT1, RCC1-ABHD12B, CLEC6A-CLEC4D, and alternative promoter usage of ETV6 are recurrently expressed in a significant number of clinical TGCT samples. Whereas these were detected only on the transcript level, two DNA-level fusions were identified, EPT1-GUCY1A3 and PPP6R3-DPP3, although these were privately expressed by individual EC cell lines.

The nonsynonymous mutation rate in TGCTs has recently been found to be low, on a scale similar to pediatric cancers (14–16). The fact that few genes are recurrently mutated indicates that other molecular mechanisms are responsible for the development of TGCT. The genomes of TGCTs are generally aneuploid, with several recurrent gains and losses of chromosomal material (10, 28). Structural rearrangements in these aneuploid tumors are known from cytogenetic banding analyses, and i(12p) is found in the majority of TGCT (9, 29, 30). Cases without i(12p) often have gain of parts of 12p material, and/or extra copies of the whole of chromosome 12 (31). Gain or loss of chromosomal material not only leads to increase or loss of gene copies and subsequent expression changes, but can also introduce genomic rearrangements that form fusion genes. The nine novel fusion genes and transcripts found in this study, all consisting of intact ORFs that potentially encode full-length or N-terminally truncated proteins, suggest that fusion genes play an important role and may be drivers for the malignant development of TGCTs. CLEC6A and CLEC4D, found to be involved in a read-through expressed in a high number of clinical TGCT samples, and ETV6 with a novel alternative promoter, are all located on chromosome arm 12p. In addition, all genes involved in the CD9-ANO2, TSPAN9-FOXJ2, and TSPAN9-GUCY2C fusion transcripts, expressed in the 2102Ep EC cell line, are located on 12p. These findings indicate that 12p is a dynamic region in TGCTs, and that gain of 12p material may be associated with expression of recurrent fusion transcripts and transcript variants.

Both fusion transcripts that involved RCC1 and the alternative promoter usage of ETV6 are overrepresented in undifferentiated histologic subtypes of TGCT, and show substantially reduced expression in the NTERA2 EC cell line upon RA-induced differentiation in vitro. This indicates that they are all associated with the pluripotent phenotype. ETV6 (ets variant 6) encodes a transcription factor of the ETS family, a family of transcription factors that are frequently involved as fusion gene partners in cancer. ETV6 itself has repeatedly been reported as a fusion gene partner in hematologic malignancies (32). In fact, a chromosomal rearrangement (4;12)(q11-q12;p13) found in cases of acute myeloid leukemia fuses CHIC2 (exon 1–3) and ETV6 (exon 2–8) using the same breakpoint of ETV6 as the one identified here connected to the newly identified alternative promoter sequence (33, 34). RCC1 (Regulator of chromosome condensation 1) has important functions in the cell cycle and acts as a guanine nucleotide exchange factor (GEF) for the RAS homologue RAN (35). However, the fusion transcripts involving RCC1 only includes the two first noncoding exons in the 5′UTR of RCC1 connected to either HENMT1 or ABHD12B located 80 Mb downstream on chromosome 1 and on chromosome 14, respectively. The long genomic distance between the partner genes suggests that these fusion transcripts are not expressed as a result of a read-through mechanism (36). To investigate if the fusion transcripts involving RCC1 were caused by genomic rearrangements, we applied linkage analysis using multiplexed fluorescent PCR assays with ddPCR. To our knowledge, this approach has not been used previously to detect rearrangements of genes resulting in fusion genes. However, linkage analysis with ddPCR has been proven successful in showing the arrangement of the Killer-cell immunoglobulin-like receptor gene complex and in chromosomal phasing (37, 38). We found no indication of DNA-level linkage for the partner genes of the RCC1 fusion transcripts. However, the partner genes of EPT1-GUCY1A3 and PPP6R3-DPP3, which also had indications of chromosomal breakpoints from DNA copy number data, were found to be linked at the DNA-level, indicating bona-fide genomic rearrangements in their respective cell lines, 833KE and NTERA2. The absence of linked partner genes of the RCC1 fusions, implies that these fusion transcripts are expressed as the result of posttranscriptional mechanisms, such as trans-splicing (39). However, we cannot rule out chromosomal rearrangements as an underlying cause of the RCC1 fusion transcripts, as only a proportion of the input DNA in our assay consisted of DNA fragments longer than 50 kb, and none more than 100 kb in length. A chromosomal rearrangement resulting in a fusion gene may include intronic regions that are longer than these DNA fragments. Such rearrangements will be missed by ddPCR linkage analysis. For optimal sensitivity, linkage analysis should be carried out on DNA samples isolated by protocols that maintain long DNA molecules intact.

Expression of the interchromosomal RCC1-ABHD12B fusion transcript and transcripts involving the alternative promoter of ETV6 was not detected in normal testis or other normal tissue samples from 20 different human organs. Also, expression of CLEC6A-CLEC4D was only observed in normal tissue from the placenta, indicating that it may be specifically expressed in adult male TGCT. These molecules are therefore highly specific for TGCTs in a stemness setting, and could prove to have important roles for TGCT malignant transformation, as well as biomarkers for TGCT disease. Diagnosis of TGCT through sensitive detection of these molecules in excreted body fluids such as seminal fluid or serum could have clinical potential (40).

In conclusion, to our knowledge, we present here the first fusion genes to be described in TGCT, including recurrent expression of RCC1 involving fusions and alternative promoter usage of the ETV6 gene, both associated with the pluripotency phenotype. These transcript variants may be important drivers of malignancy, and could potentially serve as diagnostic markers in the clinic.

No potential conflicts of interest were disclosed.

Conception and design: A.M. Hoff, S. Alagaratnam, P.W. Andrews, R.I. Skotheim

Development of methodology: A.M. Hoff, J. Bruun, R.I. Skotheim

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.M. Hoff, S. Alagaratnam, J. Bruun, P.W. Andrews, R.A. Lothe, R.I. Skotheim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.M. Hoff, S. Alagaratnam, S. Zhao, R.A. Lothe, R.I. Skotheim

Writing, review, and/or revision of the manuscript: A.M. Hoff, S. Alagaratnam, S. Zhao, J. Bruun, R.A. Lothe, R.I. Skotheim

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.M. Hoff, S. Alagaratnam, S. Zhao, R.A. Lothe, R.I. Skotheim

Study supervision: R.I. Skotheim

The study was funded by grants from the Norwegian Cancer Society (A.M. Hoff was financed as PhD student from the PR-2007-0166 grant to R.I. Skotheim), NorStore (for storage of computer files; project NS9013K granted to R.I. Skotheim), the Medical Research Council UK (grant number G0700785 to P.W. Andrews), the Research Council of Norway through its Centers of Excellence funding scheme (project number 179571 granted to R.A. Lothe), and from the KG Jebsen Foundation (granted to R.A. Lothe).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.