Mapping of deletion breakpoints at the CDKN2A locus in melanoma: detection of MTAP-ANRIL fusion transcripts

Genomic locus at chromosome 9p21 that contains the CDKN2A and CDKN2B tumor suppressor genes is inactivated through mutations, deletions and promoter methylation in multiple human cancers. Additionally, the locus encodes an anti-sense RNA (ANRIL). Both hemizygous and homozygous deletions at the locus targeting multiple genes are fairly common in different cancers. We in this study investigated breakpoints in five melanoma cell lines, derived from metastasized tumors, with previously identified homozygous deletions using array comparative genomic hybridization (aCGH). For breakpoint mapping, we used primer approximation multiplex PCR (PAMP) and inverse PCR techniques. Our results showed that three cell lines carried complex rearrangements. In two other cell lines, with focal deletions of 141 kb and 181 kb, we identified fusion gene products, involving MTAP and ANRIL. We also confirmed the complex rearrangements and focal deletions in DNA from tumor tissues corresponding to three cell lines. The rapid amplification of 3′cDNA ends (3′RACE) carried out on transcripts resulted in identification of three isoforms of MTAP-ANRIL fusion gene. Screening of cDNA from 64 melanoma cell lines resulted in detection of fusion transcripts in 13 (20%) cell lines that involved exons 4-7 of the MTAP and exon 2 or 5 of the ANRIL genes. We also detected fusion transcripts involving MTAP and ANRIL in two of the seven primary melanoma tumors with focal deletion at the locus. The results from the study, besides identifying complex rearrangements involving CDKN2A locus, show frequent occurrence of fusion transcripts involving MTAP and ANRIL genes.


INTRODUCTION
The cyclin-dependent kinase inhibitor 2A/2B (CDKN2A/2B ) locus on chromosome 9p21, a target of frequent inactivation in various human cancers, encodes three tumor suppressors [ 1 -6 ]. The p16 INK4a (p16 henceforth) and p14 ARF (ARF henceforth) are encoded from CDKN2A sequence, shared in alternate reading frames, and CDKN2B encodes p15 INK4b (p15 henceforth). The locus also transcribes a long intergenic noncoding RNA, termed as antisense non-coding RNA in the INK4 locus ( ANRIL ), involved in the repression of CDKN2A and 2B [ 7 , 8 ] . Both p16 and p15 disrupt cell cycle progression through inhibition of retinoblastoma phosphorylation via disruption of CDK4/Cyclin D1 complex. ARF protein stabilizes p53 through binding to MDM2 and inhibits ubiquitin ligase activity [ 9 , 10 ]. Elevated expression of any of the three proteins encoded from the CDKN2A/2B locus results in cell cycle arrest leading to cellular senescence [ 10 ]. Germline mutations specifi c to p16 and to some extent to www.impactjournals.com/oncotarget ARF impart one of the highest genetic risks of melanoma within familial settings [ 10 -12 ].
Somatic inactivation of the CDKN2A gene , a feature common to many cancer types involves point mutations, promoter methylation or deletions at the locus [ 13 -21 ]. Owing to multiple targets, deletions at 9p21 targeting CDKN2A/B genes constitute the predominant alterations at the locus. Previously, we reported deletions at the locus in 78% of cell lines derived from metastasized melanoma tumors [ 22 ]. The mono-allelic and bi-allelic deletions at the locus span from a few hundred kilobases to several megabases, sometimes involving the entire chromosomal arm. Deletions, besides the CDKN2A , also affect neighboring genes including methylthioadenosine phosphorylase ( MTAP ) [ 5 , 23 -25 ]. Despite the fact that deletions at the locus are common occurrences, the resultant breakpoints have remained mostly uncharacterized [ 2 , 26 ]. The large genomic deletions that are often detected by conventional techniques such as fl uorescence in situ hybridization (FISH), array comparative genomic hybridization (aCGH) and multiplex ligation-dependent probe amplifi cation (MLPA) lack details about exact deletion coordinates. Therefore, a precise characterization of deletion breakpoints becomes essential for understanding the complexity of deletions at the 9p21 locus [ 26 , 27 ]. In this study we used PCR based methods for identifi cation of deletion breakpoints in melanoma cell lines followed by validation of those fi ndings in corresponding tumors. We report chromosomal rearrangements in three cell lines and focal deletions in two cell lines. Investigation of two cell lines with focal deletions at the CDKN2A/B locus resulted in detection of MTAP-ANRIL gene fusion at deletion breakpoints , which we validated in corresponding tumors. Further, screening showed the presence of fusion gene transcripts in a total of 13 (20%) of 64 cell lines and two of seven primary tumors that had deletion at the locus.

RESULTS
The pair wise copy-number data from 44 metastatic melanoma cell lines and corresponding peripheral blood mononuclear cells from a previous study was used to select cell lines with homozygous deletions (HD) [ 22 ]. 12 of the 44 cell lines carried HD at the 9p21 locus that ranged between 139 kb to 5.6 Mb (Supplementary Figure  S1). Three cell lines  with deletions less than 500 kb and two cell lines (MaMel-103a and MaMel-08a) with more than 2 Mb deletions were selected for mapping of breakpoints using primer approximation multiplex PCR (PAMP) or inverse PCR techniques (Supplementary Figure S2). The results from three cell lines, MaMel-30, MaMel-19 and MaMel-08a, were validated in the corresponding metastasized tumors; however, tumor tissues corresponding to MaMel-95 and MaMel-103a were not available.

Breakpoint cloning with PAMP and inverse PCR
The deletion breakpoints in MaMel-30, MaMel-95 and MaMel-103a were mapped using PAMP. However, due to the presence of complex rearrangements, the deletions in MaMel-19 and MaMel-08a could not be mapped with PAMP. The deletion points in MaMel-19 and MaMel-08a at telomeric and centromeric ends were independently cloned using inverse PCR. The general mapping of breakpoints was initiated by validation of single nucleotide polymorphisms (SNPs) at the deletion junctions by PCR and the aberrations in SNP calls were used to determine the correct deletion borders ( Figure 1 , Supplementary  Table S1). The deletion coordinates described in this study are according to Human Genome Variation Society (HGVS) nomenclature version 2.0 (last-probe-present_fi rst-probedeleted)_(last-probe-deleted_fi rst-probe-present) or (genomic -end-position-last-positive-probe_genomic-start-position-fi rst -negative-probe)_(genomic-end-position-last-negative-probe _genomic-start-position-fi rst-positive-probe) [ 28 ].
The aCGH data for MaMel-95 indicated a deletion of 152 kb ( Figure 1 , Supplementary Table S1). Sequencing of the amplifi ed product from PAMP confi rmed a deletion of 181 kb with coordinates at hg19 chr9:g. (21,855,742_22,036,449) Table S1). The sequence at the deletion junction revealed 4 bp nucleotide overlap between the fusion ends (Supplementary Table S2). The deletion on telomeric side occurred within intron 6 of the MTAP (Ensembl Transcript: ENST00000380172 ; intron 7 of Ensembl Transcript: ENST00000580900 ) and within intron 3 of the ANRIL gene (Ensembl Transcript: ENST00000428597 ), indicating a possible gene fusion between MTAP and ANRIL ( Figure 2B ).
For MaMel-19 cell line, the aCGH indicated a deletion of 289 kb ( Figure 1 , Supplementary Table S1). Sequencing of the inverse PCR amplicons from both ends showed a deletion and an inversion of 320 kb and 1.1 Mb, respectively (Supplementary Table S1). The deletion included MTAP and CDKN2A genes (Chr9:21,674,438-21,994,653). The 1.1 Mb (chr9:21,994,653-23,090,311) chromosomal inversion involved CDKN2B, ANRIL and DMRTA1 ( Figure 2D ). Southern hybridization of DNA from the cell line showed a single 5.6 kb fragment corresponding to CDKN2B compared to a control DNA that showed 4 kb and 5.6 kb fragments containing CDKN2A and CDKN2B , respectively ( Figure 2D ). The deletion and inversion in the DNA from tumor tissue corresponding to the MaMel-19 cell line was confi rmmed with primers specifi c to the identifi ed break points ( Figure 2D , Supplementary Table S3).
The aCGH data for MaMel-08a indicated a deletion of 3.4 Mb ( Figure 1 , Supplementary Table  S1). Inverse PCR showed non-reciprocal translocations with chromosomes 6 and 15 with a deletion of 3.2 Mb (Supplementary Table S1, Figure 2E ). The deletion on the telomeric side featured a translocation involving the reverse strand of chromosome 15q12 locus (t(9;15) (p21.3;-q12)). A 4 nucleotide overlap was detected between the fusion sequences at telomeric end (Supplementary Table S2). The translocated locus corresponded to a no-gene region located 7 kb downstream of pseudogene interferon, nu 1, pseudogene ( IFNNP1 ) on chromosome 9 (Ensembl Transcript: ENST00000429219 ) and intron 23 of oculocutaneous albinism II ( OCA2 ) gene on chromosome 15 (Ensembl Transcript: ENST00000354638 ). On the  Table S2). The translocated locus corresponded to a no-gene region located 301 kb downstream of IZUMO family member 3 (IZUMO3) on chromosome 9 (Ensembl Transcript: ENST00000604921 ) and intron 2 of family with sequence similarity 83, member B ( FAM83B )(Ensembl Transcript: ENSG00000168143 ) on chromosome 6 ( Figure  2E ). Using primers specifi c to the identifi ed breakpoints, we confi rmed the translocations on both telomeric and centromeric ends in DNA from tumor tissue corresponding to the MaMel-08a cell line ( Figure 2E ).

3′RACE of MTAP-ANRIL fusion transcript and expression in E.coli system
In MaMel-30 and MaMel-95 cell lines, the cloning of the breakpoints showed partial homozygous deletion of MTAP and ANRIL genes and resultant putative gene fusion. Using a combination of rapid amplifi cation of 3′ cDNA ends (3′RACE) and MTAP gene specifi c primers, we detected two transcripts in MaMel-30 and a single transcript in MaMel-95.
One transcript detected in both cell lines had resulted from fusion of fi rst six exons of MTAP gene with ANRIL from exon 5 onwards. In MaMel-30 cell line, the common transcript had 1467 nucleotides including 113 bp 5′untranslated region (UTR), 702 bp coding region and 652 bp 3′UTR and polyadenylation sequences (Supplementary Figure S3A). The transcript in MaMel-95 had 1425 nucleotides including 113 bp 5′UTR, 702 bp coding region and 610 bp 3′UTR and polyadenylation sequences. The sequence was identical in both the cell lines for 5′UTR and coding regions. However, two transcripts differed in length at 3ʼUTR. The ATG start codon was detected at position 114 and TGA stop codon was located at position 813 (Supplementary Figure S3B). The putative translation product for both transcripts encoded 233 amino acids, of which 230 amino acids were from MTAP and three amino acids from ANRIL .
The second transcript identifi ed in MaMel-30 included fi rst seven exons of MTAP , followed by ANRIL gene, exon 5 onwards. The transcript was 1590 bp long with 113 bp 5′UTR, 825 bp coding region and 652 bp 3′UTR and polyadenylation sequences. The ATG start codon was detected at position 114 and TGA stop codon at position 936 (Supplementary Figure S3C). The transcript putatively encoded 274 amino acids, of which 271 amino acids were from MTAP and three amino acids from ANRIL gene.
The two transcripts from ATG until the end of polyadenylation sequence were cloned into pET-20b and expressed under T7 promoter using IPTG induction. The resulting protein fractions upon detection with MTAP antibody showed corresponding protein bands for both the transcripts (data not shown).

Screening for fusion transcripts in additional melanoma cell lines and primary melanoma tumors
In order to detect the MTAP-ANRIL fusion gene transcript, we screened 174 metastatic melanoma cell lines for the deletion at the locus. Initial MLPA analysis identifi ed deletions at the 9p21 locus in 100 (58%) cell lines. Out of those 100 cell lines, 49 ( Table 1 ).
Additionally, we used data from 60 primary melanoma tumors that had been previously screened for deletions at the CDKN2A/B locus [ 29 ]. Of 60 primary tumors, seven showed focal deletions at the CDKN2A/B locus that extended into MTAP in fi ve tumors. Screening of cDNA from seven tumors with deletions at the CDKN2A/B locus showed fusion transcripts in two tumors. In one tumor, we detected fusion transcript similar to the one detected in MaMel-95 cell line involving exon 6 of MTAP and exon 5 of ANRIL with latter contributing three amino-acid residues ( Figure 4 , Table 1 ). The second tumor showed three fusion transcripts. One of those fusion transcripts (transcript 1) involved intron 4 of MTAP and exon 6 of ANRIL , with stop codon located within intron 4 of MTAP itself ( Figure 4 , Table 1 ). The second transcript (transcript 2) from the same tumor showed fusion involving exon 6 of MTAP and exon 6 of ANRIL ; however, the end of the transcript remains to be determined ( Figure 4 , Table 1 ). The transcript 3 involved fusion between intron 6 of MTAP and exon 6 of ANRIL with stop codon within intron 6 of MTAP ( Figure 4 , Table 1 ).

DISCUSSION
In this study we fi ne mapped deletions at the 9p21 locus that harbors CDKN2A/B complex to clone the resultant breakpoints in melanoma cell lines derived from metastasized tumors. We used a previously generated aCGH data to identify the extent of the deletions in the investigated cell lines to map the breakpoint architectures. Our results showed that in fi ve cell lines the deletions ranged from 141 kb to 3 Mb and involved complex rearrangements including translocations, inversions, insertions and gene fusions. The availability of tumor tissues corresponding to the cell lines allowed us to confi rm the breakpoints identifi ed in cell lines through use of specifi c primers, which ruled out the results being cell culture artifacts. Two of the cell lines showed gene fusions involving parts of the MTAP and non-coding ANRIL genes resulting from the focal deletion of The cloning of breakpoints at the CDKN2A locus did not reveal any signifi cant sequence homologies or an overlap at the junctions. Our observations were in accordance with previous reports in other tumors indicating that double strand breaks in 9p21 locus occur randomly and are regulated by unknown factors [ 5 , 30 ]. Three cell lines with chromosomal rearrangements showed two separate deletions within the 9p chromosome, with one deletion at the CDKN2A locus and the other in a region approximately 700 kb centromeric to CDKN2A . This was evident from the sequencing data from MaMel-103a cell line that showed retention of a 97 bp fragment at the deletion junction, which appeared as an insertion. Similarly in MaMel-19 deletion at the CDKN2A locus accompanied an inversion of 1.1 Mb gene region. In MaMel-08a, deletions in the 9p21 locus resulted in nonreciprocal translocations with chromosome 6 on one end and with chromosome 15 on the other end of the deletion. This observation of dual deletions on chromosome 9p arm was in line with a previous study in cancer cell lines, which showed that all deletions in chromosome 9p arm depend, in some way, on deletions at CDKN2A locus [ 31 ].
In addition to the chromosomal rearrangements, the focal deletions at 9p21 locus resulted in fusion of parts of MTAP with ANRIL , a non-coding gene. The full length sequence of fusion transcript detected in two cell lines (MaMel-30 and MaMel-95) revealed three splice variants, two of those in MaMel-30 and one in MaMel-95; however, MaMel-30 ✓ Ex: exon number in the respective gene, HD: Homozygous deletion, LOH: Loss of heterozygosity. The sequence data for the full length and partial transcripts have been submitted to the GenBank databases under accession numbers KT386339-KT386352. www.impactjournals.com/oncotarget A comparison of the fusion junction in MaMel-30 at genomic and cDNA levels indicated that the splicing in terminal exons is regulated by factors within 3′ region of MTAP gene [ 32 , 33 ]. It can be speculated that the putative protein from MTAP-ANRIL fusion transcripts is analogous to truncated MTAP protein, as ANRIL contributed only three amino acid residues. The activity of MTAP in methionine salvage pathway is dependent on amino acid residues in exons 6 and 7; transcripts without exon 7 lack in enzymatic activity [ 32 ]. That implies that MTAP-ANRIL fusion transcripts without exons 7-8 detected in the present study in 11 of 13 cell lines and in two primary tumors might be enzymatically inactive. Prior to our report, a similar MTAP-ANRIL fusion gene was reported in a glioma tumor resulting from fusion of MTAP exon 4 with exon 2 of ANRIL and the resultant transcript contained 27 amino acids from the non-coding gene [ 34 ]. In contrast all the fusion transcripts detected in this study, contained mainly three amino acids from the ANRIL gene with the exception of UKRV-Mel-22 cell line that showed a transcript similar to that shown in a previous study [ 34 ]. One of the three incompletely characterized fusion transcripts in a primary tumor contained additional amino acid resides from exon 6 of the ANRIL gene.
The deletions at the 9p21 locus confer growth advantage to tumor cells due to occurrence of three critical tumor suppressor genes within a distance of 50 kb that regulate retinoblastoma and p53 pathways [ 35 , 36 ]. Evidence from mice models and human tumors point out that CDKN2A represents the 'weakest links' of retinoblastoma and p53 pathways in melanoma development [ 37 ]. The frequency of cancer specifi c homozygous deletion at CDKN2A is higher than at any other locus within human genome [ 31 , 36 ]. Cytogenetic aberrations such as translocations and gene fusions may be of greater importance in epithelial tumorigenesis [ 38 -40 ]. The Cancer Genome Atlas (TCGA) Network reported 224 gene fusions in 333 cutaneous melanomas with a recurrent GRM8-CNTNAP2 fusion in two tumors. Out of the 224 gene fusions, only one was shown to involve the fusion of CDKN2A locus with NPHP4 on chr.1p36 [ 13 ]. In addition, recurrent gene fusions have been discovered in a variety of epithelial tumors, such as prostate, lung, stomach and colorectal cancers and some of the fusion genes have been confi rmed to be ideal molecular targets for cancer therapy, as the EML4-ALK fusion gene in non-small-cell lung cancer [ 41 -47 ]. Whether, the MTAP-ANRIL fusion transcripts detected in this study exert an oncogenic infl uence, if any, on tumor, remains to be determined. Nevertheless, our results in this study show that in melanoma cell lines and primary tumors with focal deletion at the CDKN2A/B locus, the transcription of MTAP-ANRIL fusion gene is a frequent occurrence.

Melanoma cell lines and DNA/RNA extraction
174 cell lines included in this study were derived from metastasized tumors of 134 melanoma patients. The melanoma patients were recruited according to the eligibility criteria that included histologically confi rmed melanoma of skin, mucosa or with unknown primaries. The cell lines derived from tumor biopsies were maintained in RPMI 1640 (LT) media supplemented with 10% fetal calf serum, 5 mM L-glutamine, 100 U/ml penicillin, and 100 lg/ml streptomycin at 37°C in a humidifi ed 5% CO2 atmosphere, as described previously [ 22 ]. DNA and RNA were extracted from cell lines using Qiagen All Prep mini kit (Qiagen, Germany). Additionally, DNA samples from three tumor tissues, corresponding to three cell lines were also included in the study. The ethical approval for the study was granted by Ethics Commission of the Clinical Medicine Faculty of University of Heidelberg, Germany.

Analysis of array-based comparative genomic hybridization (aCGH)
The aCGH was performed previously on genomic DNA from 44 cell lines and corresponding peripheral blood mononuclear cells (PBMCs) as described previously [ 22 ]. The aCGH data was taken as a basis for selection of the cell lines with HD. A copy number (CN) =0 at the position of a SNP was considered as HD and CN=1 as hemizygous deletions, CN=2 as diploid and CN>2 as amplifi cation. Previously generated CEL fi les for 44 paired melanoma and blood tissues were accessible from the Gene Expression Omnibus database at NCBI (http://www.ncbi. nlm.nih.gov/geo/) under accession number GSE17534.

Primer approximation multiplex PCR (PAMP)
Initially at the each deletion junction at least three SNPs with CN=0 and ≥1 as shown by aCGH data were validated by PCR. This validation was used to determine a minimum distance of 10-20 kb that was used for designing of primers for primer approximation multiplex PCR www.impactjournals.com/oncotarget (PAMP). In those cases, where the distance between SNPs with CN=0 and ≥1 at the deletion border was more than 20-30 kb, we mapped additional SNPs that were present on the array. This validation was used for confi rmation of the CN status determined from aCGH data.
PAMP was carried out on DNA from the selected cell lines with HD at the 9p21 locus using a modifi cation of a previously described method [ 48 ]. A series of 9-10 forward and reverse primers (24 nt long, GC>40% and Tm>67° C) on either side of the deletion loci were designed using Primer 3 software. Specifi city of the primers was determined using UCSC BLAT tool. Distance between any two forward or reverse primers was set to a maximum of 1 kb, assuming that the fi nal amplicon will be less than 2 kb long in case primer pair nearest to the deletion junction results in amplifi cation [ 48 ]. For mapping of the deleted locus, nine forward and nine reverse primers were used. The forward/reverse primers were split into three groups with three primers pairs in each group. Multiplex PCR was performed using all nine combinations with the forward and reverse primer sets (Supplementary Figure S5). In the fi nal step, fresh genomic DNA was used for amplifi cation over the breakpoints using specifi c primers identifi ed in the previous step and the PCR products were subjected to Sanger sequencing (Supplementary Figure S5, Supplementary Table S3).
The PAMP reaction was carried out in a total volume of 10 μl containing 10 ng of template DNA, 0.15-0.25 μM of each forward and reverse primer, 5x HF reaction buffer, 0.11 mM of each dNTP, 0.4 U of Phusion High-fi delity DNA polymerase (New England Biolabs). Sequencing reactions were carried out using BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Warrington, UK) and run on an ABI Prism 3130XL Genetic Analyzer.

Inverse PCR
Breakpoints in MaMel-19 and MaMel-08a cell lines were fi ne mapped by inverse PCR. Using the aCGH data as the basis, the breakpoints from the both the ends were mapped independently and narrowed down to approximately 1 kb by validating SNPs within and around the breakpoint. Based on the known sequence within the region, the restriction sites were determined and primers were designed in the known region adjacent to the unknown sequence. After digestion and ligation, the linear DNA was transformed into circular so that the used primers could amplify the unknown region as in a standard PCR (Supplementary Figure S4). For MaMel-19, Msp1 (New England Biolabs, Germany) and HpyCH4IV (New England Biolabs, Germany) restriction enzymes were used to digest genomic DNA for mapping of the breakpoint on telomeric and centromeric sides, respectively. For MaMel-08a, mix of SpeI, AvrII, NheI (New England Biolabs, Germany) restriction enzymes was used for the digestion. 500ng DNA was digested with corresponding restriction enzymes at 37°C for 16 hours, and the digested DNA was purifi ed through spin columns (Fermentas, Germany) and subjected to ligation reaction with T4 DNA ligase (New England Biolabs, Germany) at 16°C for 8 h in 80 μl reaction volume. Inverse PCR was performed using LongAmpTaq 2x MasterMix (New England Biolabs, Germany) and specifi c primers (Supplementary Table S3). Sequencing was performed using BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Warrington, United Kingdom) and fi nal products were run on an ABI Prism 3130XL Genetic Analyzer.

Southern blot
Ten μg of genomic DNA extracted from cells lines using AllPrep mini kit (Qiagen, Germany) was restriction digested overnight with EcoR I (New England Biolabs, Germany) at 37°C. The digestion product was subjected to electrophoresis on a 1% agarose gel for 12 hours at 30 volt. After staining with Gel-red, the gel was soaked for 30 minutes in alkali denaturation solution (0.4N NaOH and 1M NaCl) and another 30 minutes in renaturation solution (0.5 M Tris-Cl, pH 7.2 and 1 M NaCl), then transferred onto a Hybond N+ membrane (GE-Healthcare Bio-Sciences, Germany) in 6xSCC buffer for 16-18 hours. The gel re-stained with Gel-red to confi rm the complete transfer of DNA onto the membrane. The exon 2 of CDKN2B was PCR amplifi ed and used as probe [ 49 ]. The probe was labeled with α-32p-dCTP using random primers and purifi ed through column (Roche, Mannheim, Germany). The DNA immobilized membrane was prehybridized with 100 ng salmon sperm DNA for 20 minutes at 68°C in a 10 ml QuikHyb Hybridization Solution (Agilent, Germany) and then hybridized with labeled probe for 1 hour at 68°C in roller bottles. After washing twice for 15 minutes at room temperature with a 2x SSC buffer and 0.1% SDS wash solution and once for 30 minutes at 60°C with 0.1x SSC buffer and 0.1% SDS wash solution, the membrane was exposed to a photographic fi lm overnight at -80°C.

Rapid amplifi cation of 3′cDNA ends (3′ RACE)
The full-length cDNA sequence of MTAP-ANRIL fusion gene was determined using 3′RACE (SMART RACE cDNA amplifi cation kit, Clontech) and MTAP gene specifi c primers (Supplementary Table S3). For 3′RACE, an internal primer was designed within exon 4 of MTAP gene on the basis of results from PAMP.
The cDNA was prepared using 0.5 -1 μg of total RNA using SMARTer RACE cDNA Amplifi cation Kit (Clontech Laboratories Inc, USA). 3′RACE was performed on cDNA using internal forward primer MTAP-EX4-F (Supplementary Table S3) and Universal Primer Mix from Advantage 2 PCR Kit (Clontech Laboratories Inc, USA). The reactions were performed in 50 μl volume, with thermal conditions of 94°C for initial 2 minutes followed by 35 cycles of 94°C for 30 seconds, 68°C for 30 seconds, 72°C for 4 minutes, and fi nal elongation time for 10 minutes. The reaction products were cloned into pCR 2.1-TOPO vectors (TOPO TA cloning kit, Invitrogen, Germany) and sequenced using vector primers. Based on the sequences, two different primers were designed specifi c to each fusion transcript (exon 6 of MTAP and exon 5 of ANRIL , F65; exon 7 of MTAP and exon 5 of ANRIL , F75; Supplementary Table S3). A nested 3′RACE was performed on the previously amplifi ed product using fusion-specifi c internal primers (F65 or F75) and nested universal primer. Amplifi ed products purifi ed from agarose gels were cloned into pCR 2.1-TOPO vectors (TOPO TA cloning kit, Invitrogen, Germany). Finally, the full length of fusion transcripts were amplifi ed using forward primer MTAP-EX1-F and corresponding reverse primers (Supplementary Table S3) and confi rmed by DNA sequencing.

Molecular cloning of MTAP-ANRIL fusion gene
A full length MTAP-ANRIL cDNA was ligated with linearized PCR 2.1-TOPO vector (TOPO TA cloning, Invitrogen, USA). Plasmid DNA was used as a template and the insert sequence was re-amplifi ed using forward primer with NdeI restriction site (Fwd-primer_NdeI, Supplementary Table S3) and M13 as reverse primer. The amplicon was gel purifi ed (PureLink Quick Gel Extraction and PCR Purifi cation Combo Kit, Invitrogen, USA). The gel purifi ed product and circular pET20b(+) vector were digested overnight with respective restriction enzymes (NdeI and EcoRI in MaMel95; NdeI and BamHI in MaMel30). Following the inactivation of the enzymes and column purifi cation (MinElute Reaction Cleanup Kit, Qiagen, Germany), the digested product was incubated overnight at 16°C using T4 ligase (T4 DNA Ligase, New England Biolabs, Germany). The ligation mix was transformed into DH-5α (Z-Competent Cells, HiSS Diagnostics Germany) and plasmid DNA sequenced to confi rm the position and orientation of the cloned fragment.

Protein expression and western blot analysis
Plasmid DNA samples with pET20b-short, pET20b-long and empty pET20b vector were transformed separately into BL21(DE3) competent E.coli (New England Biolabs, Germany). The bacterial culture was subsequently induced for gene expression using IPTG (isopropyl b-D-1-thiogalactopyranoside) to a fi nal concentration of 1 mM for 4 h at 37°C. The cells were harvested by centrifugation at room temperature, resuspended in 1X SDS loading dye, denatured by boiling for 3 minutes and centrifuged to remove any insoluble materials. Expression of recombinant protein was analyzed by 15% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and the protein bands were visualized with Coomassie staining solution. The approximate molecular weight was estimated by using prestained protein molecular weight standards.

Multiplex ligation-based probe amplifi cation (MLPA)
Deletions at the CDKN2A locus in 174 cell lines were screened with MLPA ME024A kit (MRC, Holland). The ME024-MLPA mix contained 33 probes, out of which 23 probes were specifi c for the CDKN2A/2B gene (including two probes for CDKN2B-AS1 ). The MIR31 , MTAP and PAX5 genes are covered by two probes each. Four more probes targeted the region between the MIR31 gene and the 9p telomere. The probes for the sequences further centromeric to the CDKN2A/2B locus were not included. The kit also contained 12 reference probes, located on stable genomic regions in most tumor types. Briefl y, 75-100 ng of genomic DNA was subjected to 16 h of incubation with probe mix followed by ligation reaction, followed by multiplex PCT. Fragment analysis was performed on a capillary sequencer (ABIPrism 3130xl Genetic Analyzer). The results were analyzed using Coffalyser software (MRC-Hollanf); threshold to defi ne deletion was set at the suggested delta value of 0.3).

Reverse transcription PCR (RT-PCR)
The reverse transcription reaction was carried out using 100 ng of total RNA, 100 pmol (0.5 μg) of Oligo(dT) primer, 4 μl of 5X reaction buffer, 0.5 μl Ribolock RNase inhibitor, 2 μl dNTP mix, 2 μl (40 U) M-MuLV reverse transcriptase in a reaction volume of 20 μl with RNAse-free water. The reaction mix was incubated at 37°C for 1 hour and heat inactivated at 70°C for 10 min. cDNA was amplifi ed using 10pmol of gene specifi c forward and reverse primers (MTAP-EX4-F,ANRIL-EX6-R; Supplementary Table S3) in a 20μl reaction volume, with 1μl reverse transcription product, 2μl 10 x reaction buffer, 0.20 mM of each dNTP, 0.4U Taq DNA polymerase (Genaxxon, Germany) and 1.5 mM of MgCl 2 under the following condition: 2 minutes at 94°C, 40 cycles of 30 seconds at 94°C, 30 seconds at 58°C, 120 seconds at 72°C, and a fi nal extension at 72°C for 10 minutes. A nested PCR was carried out using 10 pmol of forward primer MTAP-EX4-F2 and reverse primer ANRIL-EX6-R2 (Supplementary Table S3). www.impactjournals.com/oncotarget DNA sequencing PCR products were sequenced bi-directionally with both forward and reverse primers, using Sanger chain termination method. The PCR amplicons were treated with ExoSapIT (Amersham Biosciences, Uppsala, Sweden) for 30 min at 37°C, followed by heat inactivation at 85°C for 15 min. Sequencing reaction were carried out with either forward or reverse primers, using BigDye Terminator Cycle Kit (Applied Biosystems, Warritong, UK). The reaction products after purifi cation were analyzed on an ABI prism 3130xl Genetic analyzer (Applied Biosystems). The sequencing data were analyzed using the Sequencing Analysis 5.2 (Applied Biosystems) software.

ACKNOWLEDGMENTS
This work was supported by TRANSCAN through German Ministry of Education and Science under grant number 01KT1511.

CONFLICTS OF INTEREST
No, there is no confl ict of interest that I should disclose, having read the above statement.

ACCESSION NUMBERS
The full length sequences of MTAP-ANRIL fusion gene have been submitted to the GenBank database under accession numbers KT386339-KT386341. The other 11 partial sequences encompassing the fusion junction were also submitted to the GenBank database under accession numbers KT386342-KT386352.