Article

Novel fusion and chimeric transcripts in ependymal tumors

OLSEN, Thale Kristin, et al.

Abstract

We have previously identified two ALK rearrangements in a subset of ependymal tumors using a combination of cytogenetic data and RNA sequencing. The aim of this study was to perform an unbiased search for fusion transcripts in our entire series of ependymal tumors. Fusion analysis was performed using the FusionCatcher algorithm on 12 RNA-sequenced ependymal tumors. Candidate transcripts were prioritized based on the software's filtering and manual visualization using the BLAST (Basic Local Alignment Search Tool) and BLAT (BLAST-like alignment tool) tools. Genomic and reverse transcriptase PCR with subsequent Sanger sequencing was used to validate the potential fusions. Fluorescent in situ hybridization (FISH) using locus-specific probes was also performed. A total of 841 candidate chimeric transcripts were identified in the 12 tumors, with an average of 49 unique candidate fusions per tumor. After algorithmic and manual filtering, the final list consisted of 24 potential fusion events. Raw RNA-seq read sequences and PCR validation supports two novel fusion genes: a reciprocal fusion involving UQCR10 and C1orf194 in [...]

Reference

OLSEN, Thale Kristin, et al. Novel fusion genes and chimeric transcripts in ependymal tumors. Genes, & Cancer, 2016, vol. 55, no. 12, p. 944-953

DOI : 10.1002/gcc.22392 PMID : 27401149

Available at: http://archive-ouverte.unige.ch/unige:111930

Disclaimer: layout of this document may differ from the published version.

1 / 1 GENES, CHROMOSOMES & CANCER 55:944–953 (2016)

Novel Fusion Genes and Chimeric Transcripts in Ependymal Tumors

Thale Kristin Olsen,1,2,3* Ioannis Panagopoulos,1,2 Ludmila Gorunova,1,2 Francesca Micci,1,2 Kristin Andersen,1 Hege Kilen Andersen,1 Torstein R. Meling,4 Bernt Due-Tønnessen,4 David Scheie,5,6 Sverre Heim,1,2,3 and Petter Brandal1,2,7 1Section for Cancer Cytogenetics,Institute for Cancer Genetics and Informatics,Oslo University Hospital,Norway 2Centre for Cancer Biomedicine,Facultyof Medicine,Universityof Oslo,Norway 3Institute of Clinical Medicine,Facultyof Medicine,Universityof Oslo,Norway 4Departmentof Neurosurgery,Oslo University Hospital,Norway 5Departmentof Pathology,Rigshospitalet,Copenhagen,Denmark 6Departmentof Pathology,Oslo University Hospital,Norway 7Departmentof Oncology,Oslo University HospitalçThe Norwegian Radium Hospital,Norway

We have previously identified two ALK rearrangements in a subset of ependymal tumors using a combination of cytogenet- ic data and RNA sequencing. The aim of this study was to perform an unbiased search for fusion transcripts in our entire series of ependymal tumors. Fusion analysis was performed using the FusionCatcher algorithm on 12 RNA-sequenced ependymal tumors. Candidate transcripts were prioritized based on the software’s filtering and manual visualization using the BLAST (Basic Local Alignment Search Tool) and BLAT (BLAST-like alignment tool) tools. Genomic and reverse tran- scriptase PCR with subsequent Sanger sequencing was used to validate the potential fusions. Fluorescent in situ hybridiza- tion (FISH) using locus-specific probes was also performed. A total of 841 candidate chimeric transcripts were identified in the 12 tumors, with an average of 49 unique candidate fusions per tumor. After algorithmic and manual filtering, the final list consisted of 24 potential fusion events. Raw RNA-seq read sequences and PCR validation supports two novel fusion genes: a reciprocal fusion gene involving UQCR10 and C1orf194 in an adult spinal ependymoma and a TSPAN4-CD151 fusion gene in a pediatric infratentorial anaplastic ependymoma. Our previously reported ALK rearrangements and the RELA and YAP1 fusions found in supratentorial ependymomas were until now the only known fusion genes present in ependymal tumors. The chimeric transcripts presented here are the first to be reported in infratentorial or spinal ependymomas. Fur- ther studies are required to characterize the genomic rearrangements causing these fusion genes, as well as the frequency and functional importance of the fusions. VC 2016 Wiley Periodicals, Inc.

INTRODUCTION (Maher et al., 2009). Several different bioinformat- Fusion genes are known to be of pathogenetic ic tools are currently available for this purpose importance in several human neoplastic entities. (Wang et al., 2013). They are often a result of chromosomal rearrange- ments such as translocations, inversions, or dele- tions. In addition to the pathogenetic, prognostic, Additional Supporting Information may be found in the online version of this article. and diagnostic information they convey, some Supported by: The Norwegian Cancer Society; The Norwegian fusion gene products may also represent promising Children’s Cancer Society; The Radium Hospital Foundation; The Research Council of Norway through its Centers of Excel- targets for novel therapies (Shaw et al., 2013; lence Funding Scheme, project number 179571; RNA sequencing Mertens et al., 2015). was performed by the Norwegian Sequencing Centre, A National Technology Platform Hosted by the University of Oslo/Oslo Uni- Traditionally, fusion genes have been identified versity Hospital and Funded by the Functional Genomics and by cytogenetic methods followed by molecular Infrastructure programs of the Research Council of Norway and the South-Eastern Regional Health authorities. characterization of the chromosomal breakpoints *Correspondence to: Thale Kristin Olsen, Section for Cancer involved. During the last decade, high-throughput Cytogenetics, Institute for Cancer Genetics and Informatics, Oslo University Hospital, P.O. Box 4953 Nydalen, Oslo 0424, Norway. RNA sequencing has become a widely applied, E-mail: [email protected] efficient approach for the detection of gene fusions Received 13 November 2015; Revised 3 July 2016; in cancer, thereby in part omitting the first cytoge- Accepted 3 July 2016 DOI 10.1002/gcc.22392 netic step on the way towards the detection of Published online 28 July 2016 in cancer-specific molecular-level genomic changes Wiley Online Library (wileyonlinelibrary.com).

VC 2016 Wiley Periodicals, Inc. NOVEL CHIMERIC TRANSCRIPTS IN EPENDYMAL TUMORS 945

TABLE 1. Clinical and Pathological Details of the 12 Ependymal Tumor Samples Included in This Study Case Sex/age, WHO Primary tumor/ Extent of PFS/OS Current Case no. in Olsen no. years Location Histology grade recurrence resectiona (years) status et al. (2014)b

1 M/0.8 ST E II or III Primary GTR 6.5c NED 11 2 F/0.9 ST AE/GBM IV Primary STR 3.6c AWD 15 3 F/47 S E II Primary GTR 8.9c NED 5 4 M/53 S E II Primary GTR 7.9c NED 7 5 M/61 IT SE I Primary GTR 7.5c NED 8 6a M/38 ST AE, giant III Primary GTR 1.5 9-1 cell type 6b AE III Recurrent GTR 5.3 AWD 9-2 7 M/75 IT SE I Primary GTR 6.5c NED 10 8 M/42 ST AE III Primary GTR 0.8/1.3 DOD 12 9 F/46 IT SE I Primary GTR 4.9c NED 13 10 F/46 IT E II Primary GTR 1.7 AWD 14 11 F/1 IT AE III Primary STR 1.2 AWD 16

S, spinal; ST, supratentorial; IT, infratentorial; E, ependymoma; AE, anaplastic ependymoma; GBM, glioblastoma multiforme; SE, subependymoma; GTR, gross total resection; STR, subtotal resection; PFS, progression-free survival; OS, overall survival; NED, no evidence of disease; AWD, alive with disease; DOD, dead of disease. aExtent of resection at initial surgery. bAll cases are referred to in Tables 1 and 2 in Olsen et al. (2014). cPatient still alive with no evidence of disease progression.

Ependymal tumors constitute 2% of all primary receptor rearrangements in two supratentorial CNS neoplasms. The annual incidence is about ependymal tumors with considerable astrocytic two to four cases per million inhabitants (Ostrom differentiation (Olsen et al., 2015). The aim of this et al., 2014). Overall 5-year relative survival is study was to extend our search for fusion genes, around 70% (Crocetti et al., 2012). According to now independent of low-resolution genomic data, the new WHO classification, ependymal tumors to the remaining ependymal tumors without ALK are subdivided into subependymomas (WHO rearrangement. grade I), myxopapillary ependymomas (WHO grade I), ependymomas (WHO grade II), anaplas- MATERIALS AND METHODS tic ependymomas (WHO grade III), and RELA fusion-positive ependymomas (Louis et al., 2016). Patients and Tumor Samples It has long been known that spinal ependymo- The present analysis was performed on a series mas harbor recurrent NF2 mutations (Ebert et al., of 12 samples from 11 patients. These were col- 1999). Studies published during the last decade lected between January 2005 and December 2012 have shed considerable light on ependymal tumor at the Department of Neurosurgery, Oslo Univer- biology. The knowledge resulting from this work sity Hospital, Rikshospitalet. The clinical and suggests that ependymomas located in different pathological details of these tumors are described anatomical compartments represent distinct in Table 1. molecular entities (Taylor et al., 2005; Johnson et al., 2010; Pajtler et al., 2015). In the case of DNA and RNA Extraction, High-Throughput infratentorial ependymoma, two clinically differ- Sequencing, and Bioinformatic Analyses ent subgroups have been identified (Witt et al., 2011; Mack et al., 2014). It was recently found DNA was extracted from frozen tumor tissue that the majority of supratentorial ependymomas (Olsen et al., 2014) and RNA extraction and paired- harbor RELA or YAP1 fusions (Parker et al., 2014; end high-throughput sequencing (Olsen et al., Pietsch et al., 2014; Pajtler et al., 2015). 2015) was performed as previously described. We have previously described the chromosomal Three samples of total human brain RNA (Life aberrations detected by karyotyping and compara- Technologies, Carlsbad, CA) were used as normal tive genomic hybridization (CGH) in a series of controls. One sample (Normal3; catalog no. Norwegian ependymal tumor patients (Olsen AM6050) consisted of RNA pooled from 23 adult et al., 2014). Using these chromosomal data com- individuals, whereas the remaining two samples bined with high-throughput RNA sequencing, we (Normal1 and Normal2; catalog no. AM7962) con- demonstrated the presence of ALK tyrosine kinase sisted of RNA from two different adult

Genes, Chromosomes & Cancer DOI 10.1002/gcc 946 OLSEN ET AL. individuals. Sequencing was performed by the Fluorescence In Situ Hybridization Norwegian Sequencing Centre (http://www. Bacterial artificial (BAC) probes sequencing.uio.no/) using the HiSeq 2000 plat- for the UQCR10 (clone RP11-772E21), C1orf194 form (Illumina, San Diego, CA) with a strand- (clone RP11-695F11), DPYSL2 (clone RP11- specific, paired-end TruSeq library protocol. Read 1G11), and GFAP (clone RP11-643P22) genes length for the control samples was 125 base pairs. were retrieved from the Human 32K BAC Re- The presence of fusion gene transcripts was Array library (BacPac Resource Center, Oakland, analyzed using the algorithm FusionCatcher (Nic- CA) (Osoegawa et al., 2001). Two probes overlap- orici et al., 2014) with Ensembl release 75 as refer- ping the CLU locus (clones RP11-810P7 and ence. RNA-seq reads were also aligned using RP11-16P20; BacPac Resource Center, https://bac- TopHat version 2.0.9, as described by Trapnell pac.chori.org) were used. The clones were select- et al. (2012). Gene expression in ependymal and ed based on their mapping to the chromosomal normal samples was quantified using htseq-count sites of the putative gene fusions. Clones were (Anders et al., 2015) and DESeq2 (Anders and grown in selective media according to the manu- Huber, 2010). Candidate fusion transcripts from facturer’s protocol and BAC DNA was extracted FusionCatcher were manually assessed using the using the High Pure Plasmid isolation kit (Roche, National Center for Biotechnology Information Basel, Switzerland). The DNA probes were (NCBI) Basic Local Alignment Search Tool labeled by nick translation using fluorescein iso- (BLAST; http://www.ncbi.nlm.nih.gov/blast) and thiocyanate and Texas Red (PerkinElmer, Wal- the University of California Santa Cruz (UCSC) tham, MA) as described previously (Nyquist et al., BLAT-like alignment tool (BLAT; http://genome. 2011). Interphase FISH was performed on fresh ucsc.edu/cgi-bin-hgBlat). slides using frozen tumor cell suspension and the slides were counterstained using 40,6-diamidino-2- Fusion Validation by PCR and Direct Sanger phenylindole (DAPI; Vector Laboratories, Burlin- Sequencing game, USA). CytoVision software (Leica biosys- cDNA was synthesized from extracted RNA by tems, Wetzlar, Germany) was used to analyze the reverse transcription using the iScript Advanced FISH images. cDNA Synthesis Kit for RT-qPCR (Bio-Rad, Her- cules, USA). PCR primers were designed using Primer3 (http://primer3.ut.ee/) and synthesized by Ethics Statement Life Technologies. All primer sequences used are This study has been approved by the regional listed in Supporting Information Table 1. The 25 health research ethics committee (project number ml PCR mix consisted of ExTaq premix (TaKaRa, S-06046). Written informed consent was obtained Otsu-shi, Japan), 0.4 mM forward and reverse prim- from all patients or from their parents. One patient er, and 2 ml cDNA. PCR was run on the C1000 was included postmortem following particular per- Thermal Cycler (Bio-Rad) with an initial denatur- mission from the Norwegian Directorate of ation at 948C for 30 sec, followed by 35 cycles (7 Health. sec at 988C, 30 sec at 588C, and 2 min at 728C) and a final extension of 5 min at 728C before cooling. RESULTS Long-range PCR was performed using the LA PCR kit v2.1 (TaKaRa) with thermal cycles as rec- On average, 96 million read pairs were obtained ommended by the manufacturer. per tumor sample (range 83–114 million). Among Three microliter of the PCR products were the normal control samples, 234 million read pairs stained with GelRed (Biotium, Hayward, CA) and were obtained from the pooled RNA, whereas 122 run on a 1% agarose gel electrophoresis. PCR million read pairs were obtained from each of the products were purified using the PCR cleanup kit two individual RNA samples. (Macherey-Nagel, Duren,€ Germany) and direct Fusion analysis of the 12 tumor samples yielded Sanger sequencing of purified PCR products was a total of 841 candidate fusion transcripts. There performed via the GATC Biotech Light Run was on average 49 unique candidate fusions per Sequencing service (http://www.gatc-biotech.com/ tumor sample (range 34–74). 120 fusion transcripts lightrun). The Sanger sequences and chromato- were detected in the three control samples. Thus, grams were then analyzed using the NCBI there was a total of 961 fusion candidates present BLAST and UCSC BLAT tools. in the 15 samples analyzed.

Genes, Chromosomes & Cancer DOI 10.1002/gcc NOVEL CHIMERIC TRANSCRIPTS IN EPENDYMAL TUMORS 947

TABLE 2. Validated Chimeric Transcripts After Fusion-Specific PCR and Sanger Sequencing 50 gene 50 gene 30 gene Fusion breakpoint breakpoint breakpoint 30 gene breakpoint ID Sample 50 gene 30 gene (chr:nucleotide) (description) (chr:nucleotide) (description)

F02 4 C1orf194 UQCR10 1:109650634 Exon 2 (mid-exon) 22:30163372 50 UTR F02r 4 UQCR10 C1orf194 22:30165800 30 UTR 1:109649744 Exon 2 (boundary) F05 11 TSPAN4 CD151 11:850367 Exon 2 (boundary) 11:836063 Exon 3 (boundary) F12 7 DIS3L CLU 15:66621611* Exon 14 (mid-exon) 8:27462849 Exon 5 (boundary) F15 5 DPYSL2 GFAP 8:26514806* 30 UTR 17:42983136* 30 UTR F19 9 CLU AGT 8:27462642 Exon 4 (mid-exon) 1:230846601 Exon 2 (boundary) N04 Normal3 CLOCK NMU 4:56342120 Exon 7 (boundary) 4:56496627 Exon 2 (boundary)

Transcript breakpoints are described according to GRCh37/hg19 annotation. Asterisks (*) indicate that the breakpoint could not be precisely iden- tified due to similarities in sequence between the two genes.

Putative fusion transcripts that were tagged by of the PCR products (samples 4, 5, 7, and 11; FusionCatcher as “no product” (tumor Table 2) involving a total of 10 different genes. samples: n 5 305; control samples: n 5 36) or One of the fusions found in the normal controls, “banned” (because of their known presence in CLOCK-NMU, was also validated (N04; Table 2). healthy individuals; tumor samples: n 5 39; control In all remaining cases, PCR and Sanger sequenc- samples: n 5 2) were excluded. The remaining 579 ing results were negative. putative fusion transcripts were then manually Two of these chimeric transcripts (F12 and F19; visualized using the BLAST and BLAT tools with Table 2) were found in subependymomas and the putative fusion sequence provided by involved exon 5 of the CLU gene. There was also FusionCatcher. a TSPAN4-CD151 fusion in a pediatric infratento- In the cases where BLAST/BLAT reported the rial ependymoma (fusion no. F05), a reciprocal putative fusion sequence as one continuous tran- fusion involving UQCR10 and C1orf194 in an adult script, the fusion candidate was considered a read spinal ependymoma (F02 and F02r), and a through and disregarded (tumor samples: n 5 401; DPYSL2-GFAP fusion in a subependymoma control samples: n 5 71). Fusion sequences that (F15). In the case of F12, FusionCatcher originally mapped to not only one or two, but several geno- reported this as a fusion involving CAPS and CLU. mic regions (unspecific sequences; tumor samples: However, PCR and Sanger sequencing revealed n 5 48; control samples: n 5 2), and masked repeat the fusion to be complex, involving mainly materi- sequences (tumor samples: n 5 19; control sam- al from the DIS3L and CLU genes, separated by a ples: n 5 4) were also removed. The final list of short sequence which was found to belong to the candidate fusions consisted of 30 unique fusion CAPS gene. gene candidates (tumor samples: n 5 26; control To examine how the reported chimeric tran- samples: n 5 4; Supporting Information Table 2). scripts related to the expression levels of the genes Two of these were the ALK rearrangements involved, we quantified gene expression for all the described in our previous paper (Olsen et al., genes involved in a PCR-verified fusion (Table 2), 2015) and thus removed from the final candidate using all reads which were properly aligned to the list. reference by the TopHat algorithm. Results are In this final list (Supporting Information Table shown in Supporting Material. There were no 2), candidate fusions were found in 8 of 12 tumor cases of obvious over- or underexpression in any of samples (not in samples 6a, 6b, 8, and 10). the fusion-positive samples. Twenty-six different genes were involved in the We also performed FISH using custom-made 24 putative fusions seen in the tumor samples. BAC probes to look for genomic rearrangements Notably, the GFAP gene was involved in 11 differ- that could explain the presence of the chimeric ent candidate fusion events (46%) and CLU in transcripts. Cell suspension from short-term tumor eight (33%). CHI3L1, CST3, and AGT were each cultures was available for FISH analysis in cases 4, involved in two different putative fusions. In one 5, and 7. Sample 4 was tested for the UQCR10- case, two reciprocal transcript variants were found C1orf194 fusion, sample 5 was tested for the (UQCR10-C1orf194 and C1orf194-UQCR10). DPYSL2-GFAP fusion and splitting of the CLU Six of the 24 candidate tumor fusions were vali- gene (in two separate experiments), whereas sam- dated by RT-PCR and direct Sanger sequencing ple 7 was tested for splitting of the CLU gene.

Genes, Chromosomes & Cancer DOI 10.1002/gcc 948 OLSEN ET AL.

Figure 1. Raw RNA-seq reads containing the fusion breakpoints for (A) F02; C1orf194- UQCR10 (n 5 62), (B) F02r; UQCR10-C1orf194 (n 5 3), and (C) F05; TSPAN4-CD151 (n 5 20). [Color figure can be viewed at wileyonlinelibrary.com]

Although CLU fusions had not been detected in files using the “grep” UNIX command and a sample 5, this tumor was also analyzed using sequence of 20 nucleotides; 10 from each fusion FISH because this, too, was a subependymoma. partner. This was found for fusions F02, F02r, and The two other subependymomas of the series dis- F05 (Table 2; reads shown in Fig. 1). played CLU fusions and we wanted to investigate We performed fusion-specific PCR on genomic whether this aberration was present in all sube- tumor DNA for fusions F02, F02r, and F05. For pendymomas of our series. fusion F02, we were able to amplify a fragment con- The number of nuclei available for analysis taining nucleotides 109650635-109650756 from chro- ranged from 50 to 74. In sample 4, 23 of 53 nuclei mosome 1 and 30163373-30163426 from chromosome (43%) displayed three green (C1orf194) and two 22 (Fig. 2). This corresponds to a fusion between red (UQCR10) signals. The remaining 30 nuclei intron 1 of C1orf194 (breakpoint in chr1:109650756) displayed two green and two red signals. However, and the 50 UTR of UQCR10 (breakpoint in no fusion (i.e., yellow) signals were seen. In sam- chr2:30163373); as originally reported by the Fusion- ple 5, 74 of 74 nuclei displayed two yellow (CLU) Catcher algorithm (Supporting Information Table 2). signals and 52 of 52 nuclei displayed two red For fusion F02r, we were able to amplify a frag- (DPYSL2) and two green (GFAP) signals. In sam- ment of approximately 3 kb (Fig. 3). The 50 end of ple 7, 50 of 50 nuclei displayed two yellow (CLU) this fragment contained a 200 bp sequence from signals. Thus, we were not able to detect any the 30 UTR of UQCR10, whereas the 30 end con- genomic rearrangement that could explain the tained a 757 bp segment from intron 2 and exon 3 presence of the chimeric transcripts. Due to lack of C1orf194. We performed Sanger sequencing of material, we were not able to examine by FISH using primers located closer to the putative break- for the C1orf194-UQCR10 or AGT-CLU fusions. point (Supporting Information Table 1); however, We then looked for raw RNA-seq reads covering we were not able to characterize the genomic the fusion breakpoints by searching the raw .fastq breakpoint further.

Genes, Chromosomes & Cancer DOI 10.1002/gcc NOVEL CHIMERIC TRANSCRIPTS IN EPENDYMAL TUMORS 949

Figure 2. Sanger chromatograms of the fusion breakpoint between C1orf194 and UQCR10 (F02) on the (A) RNA and (B) DNA levels. The chromatogram at the top shows the reverse transcriptase PCR fragment, whereas the chromatogram at the bottom shows the genomic PCR fragment. [Color figure can be viewed at wileyonlinelibrary.com]

For fusion F05, we were able to amplify a frag- translocation. UQCR10-C1orf194 and C1orf194- ment of approximately 10 kb (Fig. 3) using long- UQCR10 do not have identical breakpoints; how- range PCR. The 50 end of this fragment contained ever, seemingly balanced translocations may be an approximately 1 kb long sequence from intron unbalanced when investigated at the nucleotide 1ofTSPAN4. The 30end contained an approxi- level (Mertens et al., 2015). The tumor in ques- mately 350 bp long sequence from intron 2 of tion, a grade II spinal ependymoma, has been CD151 (Fig. 3). As for the UQCR10-C1orf194 investigated by G-banding (case 7 in our previous fusion, subsequent rounds of Sanger sequencing paper (Olsen et al., 2014)) but did not display any with primers closer to the putative breakpoint structural aberrations corresponding to these gene were not successful, as we did not obtain chroma- fusions. The karyotype was near-triploid, with tograms of sufficient quality for sequence analysis. numerical aberrations only. Metaphase CGH anal- ysis of the tumor did not reveal any focal gains or DISCUSSION losses in the UQCR10 and C1orf194 loci. Still, a We performed high-throughput RNA sequenc- substantial number of RNA-seq reads spanned the ing to detect novel fusion genes in a small, hetero- fusion breakpoints (Fig. 1). This discrepancy geneous series of ependymal tumors. Six of 24 might be due to representative tumor cells not (25%) tumor-specific candidate fusion transcripts being able to divide in culture prior to G-banding detected were also detected by reverse transcrip- analysis. It is also possible that the genomic rear- tase PCR. In three of these transcripts, we found rangement causing the fusion was too small to be RNA-seq reads containing the breakpoint; two detected by the low resolution level of reciprocal fusions involving C1orf194 and UQCR10 karyotyping. in an adult ependymoma of the spinal cord, and a The genomic PCR results also support TSPAN4-CD151 fusion in a pediatric infratentorial TSPAN4-CD151 as a fusion gene caused by a geno- ependymoma. Using genomic PCR, we were able mic rearrangement. TSPAN4 and CD151 were pre- to validate the C1orf194-UQCR10 fusion on the viously reported to be involved in cancer-related DNA level, with a breakpoint identical to the one gene fusions (Asmann et al., 2011; Lapuk et al., provided by FusionCatcher. We were not able to 2012; Seo et al., 2012; Yoshihara et al., 2015). characterize its corresponding reciprocal variant, Lapuk et al. (2012) detected a WDTC1-CD151 UQCR10-C1orf194, precisely. However, our geno- fusion in a prostate adenocarcinoma in which exon mic PCR results indicate that this fusion also is 2ofCD151 constituted the 30 moiety of the chime- present at the DNA level. ra. This fusion was validated using PCR analysis This reciprocal fusion is most likely caused by a of tumor cDNA followed by Sanger sequencing. genomic rearrangement such as a t(1;22) In sample 11 of our series, a pediatric infratentorial

Genes, Chromosomes & Cancer DOI 10.1002/gcc 950 OLSEN ET AL.

Figure 3. Genomic PCR indicating the presence of genomic rear- to intron 2 and exon 3 of C1orf194 (lower UCSC Genome Browser rangements leading to the formation of UQCR10-C1orf194 (F02r) and image). (B) Gel electrophoresis of long-range PCR fragment to the left. TSPAN4-CD151 (F05) fusion genes. (A) Gel electrophoresis of nested Lane M1: 1 kb DNA ladder; lane M2: lambda-HindIII marker; lane 2: an genomic PCR is shown to the left. Lane M: 1 kb DNA ladder; the approx. 10,000 bp fragment. The sequence of the 50 end mapped to strongest bands represent 6000, 3000, and 1000 bp fragments. Lane 1: intron 1 of TSPAN4 (upper UCSC Genome Browser image) whereas An approximately 3000 bp fragment. The sequence of the 50 end of the sequence of the 30 end mapped to intron 2 of CD151 (lower this fragment mapped to the 50 UTR of UQCR10 (upper UCSC UCSC Genome Browser image). [Color figure can be viewed at Genome Browser image) whereas the sequence of the 30 end mapped wileyonlinelibrary.com] anaplastic ependymoma, exon 3 of TSPAN4 (50) system. For example, the FGFR3-TACC3 fusion in was fused to exon 3 (30), the first coding exon of glioblastoma is caused by a 70 kb tandem duplica- CD151. The putative fusion transcript involves the tion on 4p16.3 (Parker et al., 2013). first coding exon of TSPAN4 fused to the entire are evolutionarily conserved, widely expressed coding sequence of CD151. transmembrane . Interestingly, CD151 is TSPAN4 and CD151 are neighboring genes only reported to contribute to several cellular processes 4 kb apart on the plus strand of . implicated in cancer, such as tumor initiation and They both encode members of the progression, angiogenesis, and metastasis (Hemler, protein superfamily. TSPAN4 is normally located 2014; Sadej et al., 2014). To investigate further if downstream of CD151, but in our case, TSPAN4 the chimeric TSPAN4-CD151 transcript is caused was the 50 partner of the chimeric transcript. Such by a tandem duplication, one might analyze the an arrangement could potentially be caused by a tumor using array CGH (aCGH). This approach tandem duplication. This has previously been can reveal copy number gains of the duplicated described in other tumors of the central nervous region as demonstrated in the FGFR3-TACC3

Genes, Chromosomes & Cancer DOI 10.1002/gcc NOVEL CHIMERIC TRANSCRIPTS IN EPENDYMAL TUMORS 951 study by Parker et al. (2013). Another approach least 200 interphase nuclei be counted in FISH could be to perform whole-genome sequencing analyses of hematological disorders (Wolff et al., (WGS) and subsequent structural variation analy- 2007). In addition, the FISH was performed on sis (Nakagawa et al., 2015). short-term cultured tumor material. It is possible Several of the genes involved in the remaining that the parenchymal tumor cells did not thrive RT-PCR validated chimeric transcripts (F12, F15, and divide under in vitro conditions. Stromal cells and F19; Table 2) involved highly expressed present in the original sample may have genes, such as GFAP and CLU (Supporting Materi- “outgrown” the neoplastic cells and skewed the al). Such highly expressed genes may generate results in favor of normal FISH signals, reflecting artifactual chimeric transcripts during RNA-seq a growth selection difference that may be particu- sample preparation (Sboner et al., 2010). This, and larly significant in low-grade CNS neoplasms (Ful- the fact that no raw RNA-seq reads contained the ler and Perry, 2002). Backing this suspicion is the putative breakpoints of fusions F12, F15, and F19, circumstance that early karyotypic studies of oligo- led us to believe that these transcripts may repre- dendrogliomas reported very low frequencies of sent false positives. It is known that other artifacts 1p/19q deletions (Griffin et al., 1992; Thiel et al., may occur during cDNA synthesis. For example, 1992) whereas later studies, which did not depend template switching can generate false chimeras on tumor cultures, reported the frequency of this during reverse transcription and/or amplification aberration to be 60–70% (Reifenberger et al., (Ozsolak and Milos, 2011; Zaphiropoulos, 2011). 1994; Bello et al., 1995). Interphase FISH on Emerging high-throughput sequencing technolo- formalin-fixed, paraffin embedded tumor material, gies, such as direct RNA sequencing without tissue slides or extracted nuclei, is not affected by cDNA synthesis, might reduce the problem of the ability of neoplastic cells to divide in vitro and technical artifacts during RNA sequencing (Ozso- could in addition be performed on larger numbers lak et al., 2009; Ozsolak and Milos, 2011). of nuclei as well as on nuclei from morphologically Another scenario to consider is the possibility distinct tissue areas. that the chimeric transcripts may have resulted We found a CLOCK-NMU chimeric transcript in from posttranscriptional events, not from genomic one of our normal control samples (Table 2). In rearrangements. Trans-splicing, the joining of two this case, exon 7 of CLOCK was fused to exon 2 of separate RNA molecules, is known to occur fre- NMU. These two genes are located approximately quently in trypanosomes and nematodes. It has 50 kb apart on the same strand of chromosome 4 also been suggested in mammals, including and are both involved in the regulation of circadi- humans (Gingeras, 2009; Zaphiropoulos, 2011; an rhythm (Takahashi et al., 2008; Martinez and Jividen and Li, 2014). The biological role of this O’Driscoll, 2015). The CLOCK-NMU fusion has mechanism is not known (Gingeras, 2009; Zaphir- previously been detected in a lung adenocarcino- opoulos, 2011). Some have suggested that trans- ma (Yoshihara et al., 2015). Due to the close prox- splicing events may precede the formation of imity of these two genes, this might represent a genomic rearrangements in cancer (Li et al., 2008; read through fusion, something that typically Zaphiropoulos, 2011). A recent study reported occurs when two adjacent genes are spliced recurrent chimeric transcripts in chronic lympho- together (Nacu et al., 2011; Jividen and Li, 2014). cytic leukemia. Since no corresponding genomic However, as the CLOCK and NMU genes are sepa- rearrangement could be found by FISH, WGS, or rated by another gene (PDCL2), a read through Southern blotting, the authors suggested that the transcription event is perhaps less likely. Other observed chimeras were the result of trans-splicing possible explanations could be cDNA-generated (Velusamy et al., 2013). However, this finding has artifacts or trans-splicing events as previously dis- later been debated (Vandepoele et al., 2015). cussed. It is also possible that the fusion is caused The FISH analyses performed in this study by a germline microdeletion without major detri- were inconclusive. Only three of six fusion tran- mental biological consequences. scripts could be directly tested for using this meth- In this study, we have described the detection od (UQCR10-C1orf194, CAPS-DIS3L-CLU, and of two cancer-specific fusion genes in a heteroge- DPYSL2-GFAP). The numbers of nuclei available neous series of ependymal tumors; a reciprocal for analysis was insufficiently low in all cases UQCR10-C1orf194 fusion in an adult ependymoma (range 50–74). To detect putative abnormalities in of the spinal cord and a TSPAN4-CD151 fusion in small clones, higher numbers of nuclei are needed. a pediatric infratentorial anaplastic ependymo- It is, for example, generally recommended that at moma. The genomic rearrangements causing

Genes, Chromosomes & Cancer DOI 10.1002/gcc 952 OLSEN ET AL. these fusion genes are not precisely characterized. Ellison DW. 2016. The 2016 World Health Organization classi- fication of tumors of the central nervous system: A summary. Such characterization could potentially be Acta Neuropathol 131:803–820. achieved by WGS, which enables structural vari- Mack SC, Witt H, Piro RM, Gu L, Zuyderduyn S, Stutz€ AM, Wang X, Gallo M, Garzia L, Zayne K, Zhang X, Ramaswamy ant and copy number analysis at base-pair V, J€ager N, Jones DTW, Sill M, Pugh TJ, Ryzhova M, Wani resolution. KM, Shih DJH, Head R, Remke M, Bailey SD, Zichner T, Faria CC, Barszczyk M, Stark S, Seker-Cin H, Hutter S, Johann P, Bender S, Hovestadt V, Tzaridis T, Dubuc AM, ACKNOWLEDGMENTS Northcott PA, Peacock J, Bertrand KC, Agnihotri S, Cavalli FMG, Clarke I, Nethery-Brokx K, Creasy CL, Verma SK, The authors thank Lisbeth Haugom for technical Koster J, Wu X, Yao Y, Milde T, Sin-Chan P, Zuccaro J, Lau L, Pereira S, Castelo-Branco P, Hirst M, Marra MA, Roberts assistance and Mads Motrøen for help with R/ SS, Fults D, Massimi L, Cho YJ, Van Meter T, Grajkowska W, Bioconductor analyses. Lach B, Kulozik AE, von Deimling A, Witt O, Scherer SW, Fan X, Muraszko KM, Kool M, Pomeroy SL, Gupta N, Phillips J, Huang A, Tabori U, Hawkins C, Malkin D, Kongkham PN, Weiss WA, Jabado N, Rutka JT, Bouffet E, Korbel JO, Lupien REFERENCES M, Aldape KD, Bader GD, Eils R, Lichter P, Dirks PB, Pfister SM, Korshunov A, Taylor MD. 2014. Epigenomic alterations Anders S, Huber W. 2010. DESeq: Differential expression analy- define lethal CIMP-positive ependymomas of infancy. Nature sis for sequence count data. Genome Biol 11:R106. 506:445–450. Anders S, Pyl PT, Huber W. 2015. HTSeq—A Python framework Maher CA, Kumar-Sinha C, Cao X, Kalyana-Sundaram S, Han B, to work with high-throughput sequencing data. Bioinformatics Jing X, Sam L, Barrette T, Palanisamy N, Chinnaiyan AM. 31:166–169. 2009. Transcriptome sequencing to detect gene fusions in can- Asmann YW, Hossain A, Necela BM, Middha S, Kalari KR, Sun cer. Nature 458:97–101. Z, Chai HS, Williamson DW, Radisky D, Schroth GP, Kocher Martinez VG, O’Driscoll L. 2015. Neuromedin U: A multifunc- JP, Perez EA, Thompson EA. 2011. A novel bioinformatics tional neuropeptide with pleiotropic roles. Clin Chem 61:471– pipeline for identification and characterization of fusion tran- 482. scripts in breast cancer and normal cell lines. Nucleic Acids Res Mertens F, Johansson B, Fioretos T, Mitelman F. 2015. The 39:e100. emerging complexity of gene fusions in cancer. Nat Rev Cancer Bello MJ, Leone PE, Vaquero J, de Campos JM, Kusak ME, 15:371–381. Sarasa JL, Pestana~ A, Rey JA. 1995. Allelic loss at 1p and 19q Nacu S, Yuan W, Kan Z, Bhatt D, Rivers CS, Stinson J, Peters frequently occurs in association and may represent early onco- BA, Modrusan Z, Jung K, Seshagiri S, Wu TD. 2011. Deep genic events in oligodendroglial tumors. Int J Cancer 64:207– RNA sequencing analysis of readthrough gene fusions in human 210. prostate adenocarcinoma and reference samples. BMC Med Crocetti E, Trama A, Stiller C, Caldarella A, Soffietti R, Jaal J, Genomics 4:11. Weber DC, Ricardi U, Slowinski J Brandes A, RARECARE Nakagawa H, Wardell CP, Furuta M, Taniguchi H, Fujimoto A. Working Group. 2012. Epidemiology of glial and non-glial brain 2015. Cancer whole-genome sequencing: Present and future. tumours in Europe. Eur J Cancer 48:1532–1542. Oncogene 34:5943–5950. Ebert C, von Haken M, Meyer-Puttlitz B, Wiestler OD, Nicorici D, Satalan M, Edgren H, Kangaspeska S, Murumagi A, Reifenberger G, Pietsch T, von Deimling A. 1999. Molecular Kallioniemi O, Virtanen S, Kilkku O. 2014. FusionCatcher—a genetic analysis of ependymal tumors. NF2 mutations and chro- tool for finding somatic fusion genes in paired-end RNA- mosome 22q loss occur preferentially in intramedullary spinal sequencing data. bioRxiv. doi: 10.1101/011650. ependymomas. Am J Pathol 155:627–632. Nyquist KB, Thorsen J, Zeller B, Haaland A, Trøen G, Heim S, Fuller CE, Perry A. 2002. Fluorescence in situ hybridization Micci F. 2011. Identification of the TAF15-ZNF384 fusion (FISH) in diagnostic and investigative neuropathology. Brain gene in two new cases of acute lymphoblastic leukemia with a Pathol 12:67–86. t(12;17)(p13;q12). Cancer Genet 204:147–152. Gingeras TR. 2009. Implications of chimaeric non-co-linear tran- Olsen TK, Gorunova L, Meling TR, Micci F, Scheie D, Due- scripts. Nature 461:206–211. Tønnessen B, Heim S, Brandal P. 2014. Genomic characteriza- Griffin CA, Long PP, Carson BS, Brem H. 1992. Chromosome tion of ependymomas reveals 6q loss as the most common aber- abnormalities in low-grade central nervous system tumors. Can- ration. Oncol Rep 32:483–490. cer Genet Cytogenet 60:67–73. Olsen TK, Panagopoulos I, Meling TR, Micci F, Gorunova L, Hemler ME. 2014. Tetraspanin proteins promote multiple cancer Thorsen J, Due-Tønnessen B, Scheie D, Lund-Iversen M, stages. Nat Rev Cancer 14:49–60. Krossnes B, Saxhaug C, Heim S, Brandal P. 2015. Fusion genes Jividen K, Li H. 2014. Chimeric RNAs generated by intergenic with ALK as recurrent partner in ependymoma-like gliomas: A splicing in normal and cancer cells. Genes Chromosomes Can- new brain tumor entity? Neuro Oncol 17:1365–1373. cer 53:963–971. Osoegawa K, Mammoser AG, Wu C, Frengen E, Zeng C, Johnson RA, Wright KD, Poppleton H, Mohankumar KM, Catanese JJ, de Jong PJ. 2001. A bacterial artificial chromosome Finkelstein D, Pounds SB, Rand V, Leary SES, White E, Eden library for sequencing the complete . Genome C, Hogg T, Northcott P, Mack S, Neale G, Wang Y-D, Coyle Res 11:483–496. B, Atkinson J, DeWire M, Kranenburg TA, Gillespie Y, Allen Ostrom QT, Gittleman H, Liao P, Rouse C, Chen Y, Dowling J, JC, Merchant T, Boop FA, Sanford RA, Gajjar A, Ellison DW, Wolinsky Y, Kruchko C, Barnholtz-Sloan J. 2014. CBTRUS sta- Taylor MD, Grundy RG, Gilbertson RJ. 2010. Cross-species tistical report: Primary brain and central nervous system tumors genomics matches driver mutations and cell compartments to diagnosed in the United States in 2007–2011. Neuro Oncol 16: model ependymoma. Nature 466:632–636. iv1–iv63. Lapuk AV, Wu C, Wyatt AW, McPherson A, McConeghy BJ, Ozsolak F, Milos PM. 2011. RNA sequencing: Advances, chal- Brahmbhatt S, Mo F, Zoubeidi A, Anderson S, Bell RH, lenges and opportunities. Nat Rev Genet 12:87–98. Haegert A, Shukin R, Wang Y, Fazli L, Hurtado-Coll A, Jones Ozsolak F, Platt AR, Jones DR, Reifenberger JG, Sass LE, EC, Hach F, Hormozdiari F, Hajirasouliha I, Boutros PC, McInerney P, Thompson JF, Bowers J, Jarosz M, Milos PM. Bristow RG, Zhao Y, Marra MA, Fanjul A, Maher CA, 2009. Direct RNA sequencing. Nature 461:814–818. Chinnaiyan AM, Rubin MA, Beltran H, Sahinalp SC, Gleave Pajtler KW, Witt H, Sill M, Jones DTW, Hovestadt V, Kratochwil ME, Volik SV, Collins CC. 2012. From sequence to molecular F, Wani K, Tatevossian R, Punchihewa C, Johann P, Reimand pathology, and a mechanism driving the neuroendocrine pheno- J, Warnatz HJ, Ryzhova M, Mack S, Ramaswamy V, Capper D, type in prostate cancer. J Pathol 227:286–297. Schweizer L, Sieber L, Wittmann A, Huang Z, van Sluis P, Li H, Wang J, Mor G, Sklar J. 2008. A neoplastic gene fusion Volckmann R, Koster J, Versteeg R, Fults D, Toledano H, mimics trans-splicing of RNAs in normal human cells. Science Avigad S, Hoffman LM, Donson AM, Foreman N, Hewer E, 321:1357–1361. Zitterbart K, Gilbert M, Armstrong TS, Gupta N, Allen JC, Louis DN, Perry A, Reifenberger G, von Deimling A, Figarella- Karajannis MA, Zagzag D, Hasselblatt M, Kulozik AE, Witt O, Branger D, Cavenee WK, Ohgaki H, Wiestler OD, Kleihues P, Collins VP, von Hoff K, Rutkowski S, Pietsch T, Bader G,

Genes, Chromosomes & Cancer DOI 10.1002/gcc NOVEL CHIMERIC TRANSCRIPTS IN EPENDYMAL TUMORS 953

Yaspo ML, von Deimling A, Lichter P, Taylor MD, Gilbertson Takahashi JS, Hong HK, Ko CH, McDearmon EL. 2008. The R, Ellison DW, Aldape K, Korshunov A, Kool M, Pfister SM. genetics of mammalian circadian order and disorder: Implica- 2015. Molecular classification of ependymal tumors across all tions for physiology and disease. Nat Rev Genet 9:764–775. CNS compartments, histopathological grades, and age groups. Taylor MD, Poppleton H, Fuller C, Su X, Liu Y, Jensen P, Cancer Cell 27:728–743. Magdaleno S, Dalton J, Calabrese C, Board J, Macdonald T, Parker BC, Annala MJ, Cogdell DE, Granberg KJ, Sun Y, Ji P, Li Rutka J, Guha A, Gajjar A, Curran T, Gilbertson RJ. 2005. X, Gumin J, Zheng H, Hu L, Yli-Harja O, Haapasalo H, Radial glia cells are candidate stem cells of ependymoma. Can- Visakorpi T, Liu X, Liu CG, Sawaya R, Fuller GN, Chen K, cer Cell 8:323–335. Lang FF, Nykter M, Zhang W. 2013. The tumorigenic Thiel G, Losanowa T, Kintzel D, Nisch G, Martin H, Vorpahl K, FGFR3-TACC3 gene fusion escapes miR-99a regulation in Witkowski R. 1992. Karyotypes in 90 human gliomas. Cancer glioblastoma. J Clin Invest 123:855–865. Genet Cytogenet 58:109–120. Parker M, Mohankumar KM, Punchihewa C, Weinlich R, Dalton Trapnell C, Roberts A, Goff L, Pertea G, Kim D, Kelley DR, JD, Li Y, Lee R, Tatevossian RG, Phoenix TN, Pimentel H, Salzberg SL, Rinn JL, Pachter L. 2012. Differen- Thiruvenkatam R, White E, Tang B, Orisme W, Gupta K, tial gene and transcript expression analysis of RNA-seq experi- Rusch M, Chen X, Li Y, Nagahawhatte P, Hedlund E, ments with TopHat and Cufflinks. Nat Protoc 7:562–578. Finkelstein D, Wu G, Shurtleff S, Easton J, Boggs K, Yergeau Vandepoele K, Philippe J, Denys B. 2015. The YPEL5-PPP1CB D, Vadodaria B, Mulder HL, Becksford J, Gupta P, Huether R, fusion transcript is detected in different hematological malig- Ma J, Song G, Gajjar A, Merchant T, Boop F, Smith AA, Ding nancies and in normal samples. Leuk Res Rep 4:51–54. L, Lu C, Ochoa K, Zhao D, Fulton RS, Fulton LL, Mardis Velusamy T, Palanisamy N, Kalyana-Sundaram S, Sahasrabuddhe ER, Wilson RK, Downing JR, Green DR, Zhang J, Ellison AA, Maher CA, Robinson DR, Bahler DW, Cornell TT, Wilson DW, Gilbertson RJ. 2014. C11orf95–RELA fusions drive onco- TE, Lim MS, Chinnaiyan AM, Elenitoba-Johnson KSJ. 2013. genic NF-jB signalling in ependymoma. Nature 506:451–455. Recurrent reciprocal RNA chimera involving YPEL5 and Pietsch T, Wohlers I, Goschzik T, Dreschmann V, Denkhaus D, PPP1CB in chronic lymphocytic leukemia. Proc Natl Acad Sci Dorner€ E, Rahmann S, Klein-Hitpass L. 2014. Supratentorial USA 110:3035–3040. ependymomas of childhood carry C11orf95-RELA fusions lead- Wang Q, Xia J, Jia P, Pao W, Zhao Z. 2013. Application of next ing to pathological activation of the NF-jB signaling pathway. generation sequencing to human gene fusion detection: Com- Acta Neuropathol 127:609–611. putational tools, features and perspectives. Brief Bioinform 14: Reifenberger J, Reifenberger G, Liu L, James CD, Wechsler W, 506–519. Collins VP. 1994. Molecular genetic analysis of oligodendroglial Witt H, Mack SC, Ryzhova M, Bender S, Sill M, Isserlin R, tumors shows preferential allelic deletions on 19q and 1p. Am J Benner A, Hielscher T, Milde T, Remke M, Jones DT, Pathol 145:1175–1190. Northcott PA, Garzia L, Bertrand KC, Wittmann A, Yao Y, Sadej R, Grudowska A, Turczyk L, Kordek R, Romanska HM. Roberts SS, Massimi L, Van Meter T, Weiss WA, Gupta N, 2014. CD151 in cancer progression and metastasis: A complex Grajkowska W, Lach B, Cho YJ, von Deimling A, Kulozik AE, scenario. Lab Invest 94:41–51. Witt O, Bader GD, Hawkins CE, Tabori U, Guha A, Rutka JT, Sboner A, Habegger L, Pflueger D, Terry S, Chen DZ, Rozowsky Lichter P, Korshunov A, Taylor MD, Pfister SM. 2011. Delin- JS, Tewari AK, Kitabayashi N, Moss BJ, Chee MS, Demichelis eation of two clinically and molecularly distinct subgroups of F, Rubin MA, Gerstein MB. 2010. FusionSeq: A modular posterior fossa ependymoma. Cancer Cell 20:143–157. framework for finding gene fusions by analyzing paired-end Wolff DJ, Bagg A, Cooley LD, Dewald GW, Hirsch BA, Jacky RNA-sequencing data. Genome Biol 11:R104. PB, Rao KW, Rao PN. 2007. Guidance for fluorescence in situ Seo JS, Ju YS, Lee WC, Shin JY, Lee JK, Bleazard T, Lee J, Jung hybridization testing in hematologic disorders. J Mol Diagn 9: YJ, Kim JO, Shin JY, Yu SB, Kim J, Lee ER, Kang CH, Park 134–143. IK, Rhee H, Lee SH, Kim JI, Kang JH, Kim YT. 2012. The Yoshihara K, Wang Q, Torres-Garcia W, Zheng S, Vegesna R, transcriptional landscape and mutational profile of lung adeno- Kim H, Verhaak RGW. 2015. The landscape and therapeutic carcinoma. Genome Res 22:2109–2119. relevance of cancer-associated transcript fusions. Oncogene 34: Shaw AT, Hsu PP, Awad MM, Engelman JA. 2013. Tyrosine 4845–4854. kinase gene rearrangements in epithelial malignancies. Nat Rev Zaphiropoulos PG. 2011. Trans-splicing in higher eukaryotes: Cancer 13:772–787. Implications for cancer development? Front Genet 2:92.

Genes, Chromosomes & Cancer DOI 10.1002/gcc