Identification of a Novel Gene NCRMS on Chromosome 12Q21 with Di

Identi®cation of a novel gene NCRMS on chromosome 12q21 with dierential expression between Rhabdomyosarcoma subtypes

Agnes S Chan1,2, Paul S Thorner1,2, Jeremy A Squire1,3 and Maria Zielenska*,1,2

1Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada M5G 1L5; 2Department of Paediatric Laboratory Medicine, Hospital for Sick Children, Toronto, Ontario, Canada M5G 1X8; 3Department of Oncologic Pathology, Ontario Cancer Institute, Toronto, Ontario, Canada M5G 2M9

Rhabdomyosarcoma (RMS) is a malignant soft tissue detection of signs of skeletal muscle dierentiation. tumor showing varying degrees of skeletal muscle This can include evidence at the light microscopic level, dierentiation. Two major histologic subtypes exist, i.e. cross-striations, or immunohistochemical detection alveolar and embryonal, each with associated molecular of structural proteins such as a-actin, desmin, and genetic changes. We have used Representational Dier- regulatory proteins such as myogenin and MyoD1. ence Analysis (RDA) to compare gene expression Other than this apparent common myogenic lineage, between the two RMS subtypes and have identi®ed the RMS is a heterogeneous tumor with varying histolo- novel gene NCRMS (non-coding RNA in RMS) that has gical and clinical presentations. Two major histological increased expression in the alveolar subtype relative to subtypes are de®ned, namely embryonal (RMS-E) and the embryonal subtype. Multiple alternatively spliced alveolar (RMS-A) with RMS-A being clinically more forms of NCRMS were identi®ed through library aggressive. RMS with mixed embryonal and alveolar screening, RACE, and comparison to human expressed histology occasionally occurs and behaves more like sequence tags (ESTs). Northern blot analysis indicated RMS-A. Despite great improvement over past decades the transcript size to be 1.25 kb in alveolar RMS. There with multi-modal therapy, the 5-year survival rate of was no sequence homology to any of the known genes in RMS remains at about 65 ± 70% with most metastatic GenBank, but extensive homology to ESTs from various cases still fatal. Conventional therapeutic modalities species. Comparison to human genomic sequences maybe approaching their limit in bringing about identi®ed at least 11 exons mapping to chromosomal further improvements in RMS outcome. However region 12q21. Dierential expression of NCRMS was oncogenomic methodologies now oer opportunities noted between various tumor types. Since NCRMS RNA for identifying new classes of tumor-speci®c genes that possesses limited potential for protein coding, yet with may lead to more targeted therapeutic approaches conserved sequences between dierent species, it is likely (reviewed in Merlino and Helman (1999); Dagher and that NCRMS is a functional non-coding RNA. Known Helman (1999) and Anderson et al. (1999b)). genes in its proximity include myogenic regulators Myf5 RMS heterogeneity is also re¯ected on the molecular and Myf6, growth factor Igf1, and another potential level. RMS-E is commonly associated with a loss of dierentially expressed gene (ATP2B1) in RMS isolated heterozygosity (LOH) on chromosomal region 11p15.5 by RDA. suggesting that one or more tumor suppressor genes Oncogene (2002) 21, 3029 ± 3037. DOI: 10.1038/sj/ may map to this region. Putative candidate tumor onc/1205460 suppressor gene(s) in this region include H19, p57KIP2, KVLQT1, GOK, TSSC1, BWR1A and BWR1C Keywords: Rhabdomyosarcoma; non-coding tran- (reviewed in Anderson et al. (1999b)). Comparative scripts; chromosome 12q21; alternative splicing; novel genomic hybridization (CGH) and spectral karyotyping gene (SKY) analysis of RMS genome showed that numeric changes were more common in RMS-E, while structural changes were more prominent in RMS-A Introduction (Pandita et al., 1999). In RMS-A, the most obvious structural changes are the t(2;13) and t(1;13) transloca- Rhabdomyosarcoma (RMS) is the most common tions. The majority of RMS-A possess one of these malignant soft tissue tumor in children. Found in translocations which involve the in-frame fusion of diverse anatomic sites, diagnosis of RMS is based on PAX3 or PAX7 gene to the FKHR transcriptional activator. Resultant chimeric proteins were found to be more potent transcriptional regulators. The potential gain-of-function was thought to contribute to the more *Correspondence: M Zielenska, Department of Paediatric aggressive behavior of RMS-A (reviewed in Barr Laboratory Medicine, Hospital for Sick Children, Toronto, (1997)). Ontario, Canada M5G 1X8; E-mail: [email protected] Received 11 July 2001; revised 15 November 2001; accepted 11 Recent data are beginning to suggest that LOH at March 2002 11p15.5 is not exclusive to RMS-E (Anderson et al., Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3030 1999a; Visser et al., 1997), and that molecular changes MAGE (Lucas et al., 1998), as well as the isolation of in RMS between, as well as within, subtypes are more herpesvirus-like DNA sequences in AIDS-associated complex and heterogeneous than previously believed. Kaposi's sarcoma (Chang et al., 1994). As RDA The use of genomic screening technologies is increas- requires only minute amount of starting RNA, it is ingly proving to generate new candidate markers suited ideal for screening a rare tumor such as RMS in which for diagnostic and prognostic purposes. In this study, tumor tissue is often limited. we have used Representational Dierence Analysis (RDA) to compare gene expression between RMS-E and RMS-A. RDA allows the isolation of novel genes Results at a manageable scale and oers greater sensitivity than the traditional subtractive hybridization. It is a Difference products isolated by RDA PCR-based subtractive hybridization technique that had been employed successfully in identifying dier- To compare gene expression by RDA, RMS-A and ences between two very similar DNA or cDNA RMS-E cDNA was used as tester in the forward and populations (Hubank and Schatz, 1994; Lisitsyn et reverse subtraction, respectively. After three rounds of al., 1993). Its application in cancer research includes RDA subtraction and ampli®cation, 44 clones were the identi®cation of the tumor suppressor gene PTEN selected for initial dot-blot screening. Of these, 13 (Li et al., 1997), the family of tumor speci®c genes clones showed dierential signal intensity (Figure 1). Sequencing of these 13 clones revealed ®ve unique sequences, designated C1 to C5. C1 was obtained from the forward subtraction suggesting higher expression in RMS-E. C2 to C5 were products of the reverse subtraction and thus were likely to have higher expression in RMS-A. Sequence identities of C1 to C5 were shown in Table 1. Semi-quantitative analysis was performed to con®rm dierential expression in the four samples originally used in RDA (Figure 2). Follow-up analysis was focused on C2 and C3 which showed the greatest dierence in signal strength between RMS-A and RMS-E. As indicated in Table 1, C2 and C3 showed no nucleotide sequence homology to any known gene in GenBank. However, matches to human ESTs and high throughput genomic sequences (HTGS) were found. Homologies to mouse and rat ESTs were also present. Figure 1 Reversed dot-blot array screening of RDA products. After the cloning of RDA products, 44 clones were individually 32 dot-blotted onto duplicate membranes. P-labeled RDA 3rd Confirmation of differential expression of selected RDA dierence products were used as complex probes on each membrane. Membrane on left was probed with cDNA enriched products by semi-quantitative RT ± PCR with RMS-A sequences, while the one on right was probed with To assess relative expression level of C2 and C3 in RMS cDNA enriched with RMS-E sequences. Thirteen clones (under- lined) with dierential signal intensity were detected, of which 12 tumors, a series of 53 RMS tumors were screened with gave stronger signals with RMS-A probe semi-quantitative RT ± PCR with sequence speci®c

Table 1 Sequence information from GenBank (June 11, 2001) for RDA products C1-5 Name Length Chromosomal location Sequence identity Sequences homologies

C1 216 bp 12q21 ± q23 Human plasma membrane calcium (1 ± 216 bp, 99%) ATPase isoform 1 (ATP2B1) gene C2 341 bp 12q21 Unknown Mouse ESTs (1 ± 136 bp, 95%) (279 ± 341 bp, 87%) Rat ESTs (1 ± 136 bp, 94%) HTGS (1 ± 341 bp, 99+%) C3 371 bp 12q21 Unknown Human ESTs (23 ± 152 bp, 90%) (146 ± 371 bp, 98%) Mouse ESTs (1 ± 63 bp, 87%) (120 ± 240 bp, 80%) (218 ± 285 bp, 92%) Rat ESTs (1 ± 61 bp, 88%) HTGS (1 ± 371 bp, 99+%) C4 422 bp 18q12.3 Unknown KIAA0427 ± gene with protein product, function unknown (1 ± 422 bp, 95%) C5 264 bp mitochondrion Human mitochondrion (1 ± 264 bp, 96%)

Oncogene Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3031 primers. Expression was noted in all samples, including a review, this case was reclassi®ed as a solid form RMS-A. sample of normal muscle. Comparing with RMS-E, For statistical purpose, band intensity of C2, C3 and RMS-A tumors and those with mixed alveolar and PBGD PCR products were measured and relative embryonal histology demonstrated an overall higher expression levels calculated as ratios of C2/PBGD and expression of both C2 and C3 (Figure 3). One case C3/PBGD (Table 2). Dierences in expression level of (denoted by *) in the RMS-E panel was noted to have both C2 and C3 between RMS-A group and RMS-E strong expression of C2 and C3. Upon blinded pathology group of tumors were found to be statistically signi®cant (P50.02, Mann-Whitney Rank Sum statistics).

Isolation of full-length cDNA with various splice forms To determine transcript size, Northern blot was performed with one RMS-A and one RMS-E total RNA. Probing with C3 gave rise to a strong signal of 1.25 kb in RMS-A (Figure 4). Very weak signals of 1.8 ± 2.0 kb were detected in normal brain, testis and liver using C2 as probe with a commercially available poly(A)+ mRNA multiple tissue Northern blot (not shown). To isolate the full-length cDNA, C2 and C3 were used as template in RACE reactions. Only the 3' reaction of C3 yielded a clean band, termed C3-3G. This sequence is of 775 bp, contains a poly(A) tract but no polyadenylation signal. C3-3G showed sequence identity with human ESTs as well as HTGs. There was no match to known gene but homologies to mouse and rat ESTs were present. Comparison to HTGS showed Figure 2 Expression screening of C1 to C5 by RT ± PCR. that C3-3G contained 2 exons (Figure 5). Relative expression levels of C1 to C5 were assessed by RT ± Since similar expression patterns were noted for C2 PCR in the four RMS tumor samples used in RDA. Lanes 1 and and C3 (Figure 3), it was suspected that they could 2 are the RMS-A samples and lanes 3 and 4 are RMS-E samples. Expression of the housekeeping gene PBGD was also determined possibly be dierent segments of the same gene. As for comparison. De®nite dierential expressions between RMS-A RACE was unable to isolate the full-length sequence, and RMS-E were noted for C2, C3 and C4 PCR was performed using one primer from each

Figure 3 Expression screening of C2 and C3 by semi-quantitative RT ± PCR. A series of RMS-E and RMS-A were screened for C2 and C3 expression by semi-quantitative RT ± PCR. C2 and C3 overall expression were higher in RMS-A (lanes 40 ± 52) and in RMS with mixed alveolar-embryonal histology group (lanes 33 ± 38) than in RMS-E (lanes 1 ± 32), though variability between individual cases existed. Level of C2 and C3 were expressed as ratios to the housekeeping gene PBGD for statistical analysis (Table 2). Signi®cant dierences were found between RMS-A and RMS-E tumors for both C2 and C3 expression (Mann-Whitney Rank Sum statistic, P50.02). *This case was re-classi®ed as the solid variant of RMS-A after pathology review. Lane L denoted DNA size ladder, lanes 54 and 55 were RMS with no subtype information, lane 57 was a sample of normal muscle and lane 58 was the negative water control

Oncogene Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3032 Table 2 Relative expression levels of C3 and C2 between RMS-E and RMS-A Band intensity ratio of C3/PGBD Band intensity ratio of C2/PGBD Lane RMS-E (n=32) Lane RMS-A (n=13) Lane RMS-E (n=32) Lane RMS-A (n=13)

1 0.34 40 0.75 1 0.09 40 0.32 2 0.50 41 2.50 2 0.00 41 0.76 3 0.36 42 0.29 3 0.13 42 0.15 4 excluded 43 1.21 4 excluded 43 0.82 5 4.88 44 2.40 5 2.08 44 1.58 6 0.21 45 1.56 6 0.03 45 1.08 7 0.53 46 6.00 7 0.21 46 3.59 8 0.50 47 1.27 8 0.16 47 0.90 9 0.58 48 0.41 9 0.24 48 0.00 10 0.19 49 1.62 10 0.00 49 1.18 11 excluded 50 0.71 11 excluded 50 0.24 12 0.67 51 5.13 12 0.27 51 3.33 13 excluded 52 0.55 13 excluded 52 0.20 14 1.54 14 0.52 15 1.19 15 0.73 16 excluded 16 excluded 17 0.19 17 0.05 18 0.19 18 0.03 19 0.18 19 0.00 20 0.25 20 0.07 21 excluded 21 excluded 22 0.10 22 0.00 23 0.27 23 0.09 24 1.24 24 0.50 25 0.89 25 0.00 26 0.00 26 0.00 27 0.19 27 0.08 28 0.29 28 0.23 29 0.72 29 0.10 30 0.14 30 0.00 31 0.45 31 0.00 32 0.04 32 0.00

any known gene in GenBank but homologies to mouse and rat ESTs were again noted. The whole sequence matches to two BAC clones on chromosome band 12q21 with over 99% homology. Alignment of C23up with genomic sequences revealed at least seven exons with conserved splice sites (Figure 5). Multiple stop codons were noted in all six frames and the longest possible open reading frame (ORF) was less than 300 bps. However, none of the ORF fell within a strong, favorable context (Kozak, 1996). Using C23up as a probe, a fetal brain phage library was screened yielding three positive clones with a consensus sequence designated as C-con (GenBank Accession #429306). C-con is 1375 bps in length and again showed no homology to known gene in GenBank. Human, mouse, rat, pig and cow ESTs matched dierent segments along the whole sequence. Figure 4 Total RNA Northern blot probed with C3. Using 32P- Complete C-con sequence revealed 100% identity to labeled C3 as probe, a strong signal of 1.25 kb was detected in chromosome 12 HTGS. These HTGS were from the RMS-A only same two BAC clones that matched to C23up, with transcript alignment variation that was suggestive of alternative splicing. Like C23up, C-con also contained sequence. A larger product, named C23up (GenBank seven exons with four exons the same as those in Accession #AF429305), was ampli®ed, puri®ed and C23up. The alignments of C23up and C-con in relation sequenced. C23up is 1103 bps long and contains C2 to each other and to the genomic sequence are shown and C3 sequences at both ends with additional in Figure 5. In comparison to C23up, C-con has one intervening sequence (Figure 5). Like the original C2 less exon (exon 6) of 127 bps. Primers ¯anking exon 6 and C3 sequences, C23up showed sequence identity to were used in PCR and products with (C23up) or human ESTs and HTGS. There was no homology to without (C-con) this exon were found in RMS-A,

Oncogene Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3033

Figure 5 Gene structure of NCRMS. Alignments and relationships of various sequences, C2, C3, C23up, C-con, 3' RACE fragment (C3 ± 3G), ESTs consensus and HTGS sequences, revealed the gene structure of NCRMS. Numbers in boxes are exon numbers

is likely the 3' end of the gene. A closer look at the alignment between these ESTs to C-con revealed that the alignment ceased 6 bps before the end of C-con, at which the polyadenylation signal AATAAA began in these ESTs. The poly-A tracts in these ESTs were found 14 bp downstream. C-con did not have the polyadenylation signal and no poly-A tract was found. Comparison of the EST consensus sequence to HTGS Figure 6 RT ± PCR with primers ¯anking exon 6. The presence of both alternatively spliced transcripts, C23up (with exon 6) and showed that, unlike C-con, these ESTs were not C-con (without exon 6), was demonstrated by RT ± PCR with spliced (Figure 5). Assembling a consensus sequence primers ¯anking exon 6. Products of expected sizes, 373 bps with with C2, C3, C23up and C-con resulted in a sequence exon 6 and 247 bps without, were resulted of about 1.8 ± 1.9 kb which was the size re¯ected from the multiple tissue Northern blot. However, no signi®cant ORF characteristic of a typical gene was RMS-E, as well as normal muscle. Expression was identi®ed. strongest in RMS-A with the product representing C23up more prominent (Figure 6). With C23up and C- Chromosomal location of NCRMS con representing alternatively splice forms, we named this gene NCRMS. The two BAC clones containing the HTGS that match The longest possible ORF in any of the six reading to NCRMS sequence lie side by side on chromosome frames for C-con was less than 300 bps and none was 12q21 between markers D12S1657 and D12S1279 (197 with a favorable context which appeared the same as kbs apart) near stSG58190. This location was con- that of C23up. This suggested that NCRMS was ®rmed in parallel by radiation hybrid mapping of the unlikely to encode a protein product. One end of the C3 sequence. Approximately 16 Mb away towards the C-con sequence matched to multiple human ESTs that centromeric end lie the Myf5, Myf6 and ATP2B1 (C1) belong to the same UniGene cluster. Based on the genes, and about 6 Mb towards the telomeric end lies poly-A tract location of these ESTs, this end of C-con the Igf1 gene (Figure 7).

Oncogene Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3034 3G), PCR (C23up) and cDNA library screening (C- Differential expression of NCRMS in various tumors con) which all indicated alternative splicing of a single Expression of NCRMS was further assessed in various gene we termed NCRMS. Sequence analysis of tumor types using the primers ¯anking exon 6. Both NCRMS indicated that all these sequences had no splice forms, C23up and C-con showed dierential homology to any known gene deposited in GenBank as expression between tumor types (Figure 8). Most of June 11, 2001. However, they all showed 99+ per tumor types showed weaker expression level than cent identity to human genomic sequences. Introns and RMS except synovial sarcoma and neuroblastoma that exons were identi®ed in these sequences with conserved showed comparable expression level of NCRMS as splice sites. Alternative splicing of NCRMS was RMS. especially evident between C23up and C-con. With clear intron and exon structure, it is unlikely that these are pseudogenes. Discussion Though multiple stop codons were present and thus no signi®cant ORF was found in all frames, functions To broaden our knowledge in the tumorigenetic of these sequences were implicated through their pathways of RMS, we began our study in comparing signi®cant homologies to various ESTs within and gene expression between the two major subtypes in across species. As no typical ORF was found, it would RMS. By RDA, we isolated two sequences, C2 and C3, be dicult to de®ne the full-length transcript. How- both with higher expression in most RMS-A when ever, based on comparison with transcript size compared to RMS-E. Based on these two sequences, determined by Northern Blot, it is likely that the full- other related sequences were isolated by RACE (C3- length sequence has been obtained. The limited potential available for protein coding suggests that NCRMS is an example of non-coding RNA (ncRNA) (Erdmann et al., 2001). Non-coding RNA are mRNA transcripts that function merely as RNA and are not thought to encode a protein product. These RNA species do not conform to the well-accepted genetic paradigm of `function after translation', and it is probable that many ncRNA went unrecognized for many years so there is not an extensive literature on these intriguing molecules. In addition, most gene annotating softwares are designed to examine ORFs, the protein coding region, when de®ning a gene so that awareness of this class of RNA is limited. More data is available from two relatively well-studied ncRNA, XIST and H19. XIST was expressed exclusively from the inactivated X chromosome in human and is known to directly participate in the dosage-compensation process be- Figure 7 Chromosomal location of C2 and C3. C2 and C3 tween male and female. XIST transcripts were found to mapped to a region on chromosome 12q with a neighboring gene- arrangement comparable to genes on chromosome 11p that have physically coat the X chromosome to prevent expres- been implicated in Wilms' tumor and in RMS sion of target genes. Its function is also in cooperation

Figure 8 Expression screening of C23up and C-con expression in various tumor types by RT ± PCR. RT ± PCR products of C23up or C-con were visualized by gel electrophoresis. Visual scores of 1+ to 4+ were given for each sample showing clear PCR products. An average score was calculated for each tumor type for comparison (the bottom row). Comparable expression was noted in synovial sarcoma and neuroblastoma as that of RMS. A. Synovial sarcoma, B. Leimyosarcoma, C. Fibrosarcoma, D. Breast cancer, E. Neuroblastoma, F. Wilms tumor, G. Ewings sarcoma, H. Osteosarcoma, I, Lymphoma, J. Rhabdomyosarcoma, K. Colon cancer, L. positive control (Rhabdomyosarcoma)

Oncogene Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3035 with tsix which is an anti-sense transcript of XIST t(2;13) translocation, were selected for subtractions in RDA. (Brown et al., 1992). H19 is a developmentally RDA was carried out as previously described (Lisitsyn et al., regulated gene that maps to the imprinted chromoso- 1993) with slight modi®cations as outlined below. Unless mal region 11p15.5 with Igf2. It was found to regulate otherwise stated, manufacturers' protocols were followed Igf2 expression in a complex manner involving an closely. Additional frozen tissues or RNA of various tumor types imprinting mechanism and dierential allelic methyla- were obtained from dierent research laboratories in Toronto tion (Srivastava et al., 2000; Thorvaldsen et al., 1998). for subsequent dierential screening and that included four Since these ncRNA are involved in regulating the cases of breast cancer; three cases each of synovial sarcoma, expression of nearby genes, they are also referred as leiomyosarcoma, ®brosarcoma, neuroblastoma, Ewings sar- riboregulators (Erdmann et al., 2001). This group of coma, osteosarcoma, colon cancer; two cases of Wilm's RNA may contain poly-A tract or repeats, involve tumor and a case of lymphoma. alternative splicing, and they can be associated with Unless otherwise stated, manufacturers' protocols were disease or imprinting. Since there is currently no followed closely. common characteristic to ncRNA, it is dicult to postulate a precise function to the NCRMS transcript. Synthesis of first and second strand cDNA The chromosomal location of NCRMS at 12q21 may cDNA was synthesized from each tumor RNA individually. be important in the etiology of this muscle tumor. Two Five mg of RNA was ®rst treated with DNaseI (GIBCO ± myogenic genes, Myf5 and Myf6 are located on BRL 18068-015), then reverse transcribed with Superscripts II chromosome 12q21, about 16 Mb centromeric from (GIBCO ± BRL 18064-014). cDNA-RNA hybrid was puri®ed NCRMS. In addition, the human plasma membrane with QIAquick PCR-puri®cation column (QIAGEN 28104). Ca(2+)-ATPase gene (ATP2B1) (RDA clone C1) maps Second strand cDNA was synthesized by incubating over- adjacent to Myf5. 6 Mb telomeric from NCRMS is the night the above cDNA ± RNA hybrid in RNase H (4 U), E. insulin-linked growth factor 1 gene (Igf1). IGF1 coli DNA Ligase (20 U), E. coli T4 DNA Polymerase (8 U), protein, a structural homolog of IGF2, was shown to and dNTPs (160 mM) in a total volume of 150 ml. Resultant stimulate skeletal muscle hypertrophy (Musaro et al., double-stranded cDNA was gel puri®ed with QIAEX II (QIAGEN 20021). cDNA of the same RMS subtype were 1999). Since gene expression dysregulation is suggested pooled after completing the 2nd strand synthesis. in this chromosomal region associated with myogenesis, a complex mechanism in gene regulations is likely involved. Representational difference analysis In summary, our data suggest that the chromosomal RDA was carried out as previously described using SAU 3A I region 12q21 may be important in myogenesis and the (NEB 169CL) and the set of Bam oligo adaptors with three development of RMS. We have identi®ed a novel gene rounds of hybridizations and PCR ampli®cations (Lisitsyn NCRMS in this region which could act as a ncRNA. A and Wigler, 1995). In the forward subtraction, RMS-E possible alternative pathway in RMS tumorigenesis cDNA was used as tester for the identi®cation of genes with may involve gene regulation by NCRMS. Comparable higher expression in RMS-E; while in the reverse subtraction, RMS-A cDNA was used as tester to identify genes expression level of NCRMS in synovial sarcoma and with higher expression in RMS-A. neuroblastoma may suggest a common pathway utilized by these tumors. Further investigations are needed to con®rm this hypothesis and to study the Analysis of RDA products complex interaction of genes that are involved in RMS RDA third dierence products from both subtractions were tumorigenesis. resolved on agarose gel. Selected bands or smears, when bands were too close to be separated, were gel puri®ed and cloned with pUC19 plasmid (GIBCO ± BRL 15364-011) and DH-5a competent cells (GIBCO ± BRL 18265-017). For initial screening, PCR products of inserts from 44 selected Materials and methods clones were dot-blotted on duplicated nylon membranes. Membranes were then probed with 32P-labeled RDA 3rd Tumor samples dierence products, one from each direction of subtraction. RMS patient tumor tissues were obtained from the frozen Clones that showed dierential signal intensity were tumor bank in the department of Paediatric Laboratory sequenced for semi-quantitative PCR analysis. Medicine (Division of Pathology) in the Hospital for Sick Children. Specimens were obtained after informed consent Semi-quantitative PCR analysis was received from all parents/guardians of children at diagnosis. A total of 32 RMS-E, 13 RMS-A, six RMS with First strand cDNA was generated from total RNA as mixed alveolar and embryonal histology, two histologically described above. Using 1/10 of reversed transcription product unclassi®ed RMS, and one normal muscle sample were as template, PCR were performed for each sample with gene included. Cryosections of selected samples were reviewed and speci®c primers for C2 and C3, as well as primers for the those with at least 90% tumor without extensive necrosis housekeeping gene Porphobilinogen Deaminase (PBGD). were selected. Total RNA was extracted with TRIzol Reagent Primers used are as follows: PBGD 5'-CAT GTC TGG (GIBCO BRL 15596). Only good quality RNA showing TAA CGG CAA TGC GGC TGC-3' and 5'-GAA CTC strong and discrete bands of the 18S and 28S ribosomal RNA CAG ATG CGG GAA CTT TC-3' with product size of upon gel electrophoresis was used. Four RMS samples, two 490 bps; C2 5'-CAC GGA GTC AAC AGA AAA-3' and 5'- RMS-E and two RMS-A, one with t(1;13) and the other with TGT GCT GAG TGG AGT AAC-3' with product size

Oncogene Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3036 155 bps; C3 5'-GCA CCT TGG AAA GAA TAG-3' and 5'- sequences. Sequence-speci®c primers were designed to match CCA GGG TTA TCT CAC AAC-3' with product size with the special end primers provided with the kit for use in 126 bps. 50 ml reactions were set up containing a ®nal 1 mM PCR ampli®cation. The sample used in RACE was a case of of each primer, 200 mM dNTPs, 2.5 mM MgCl2, 1X Perkin RMS-A that showed strong expression of target sequences. Elmer PCR buer II, and 2.5 U of Taq Polymerase Gold PCR reactions were optimized to amplify a single product. (Perkin Elmer). PCR was performed with initial denaturiza- Nested PCR was performed when primary PCR failed to tion and activation at 958C for 10 min, followed by 35 cycles result in a single product. PCR products were resolved in of 958C for 30 s, 528C for 45 s, and 728C for 45 s. A ®nal agarose gel and puri®ed with QIAEX II (QIAGEN 20021), extension was done for 10 min at 728C. PCR products were T-tailed cloned into pUC19 plasmid and transformed into visualized by gel electrophoresis with ethidium bromide. Gel DH5a competent cells. Positive clones were sequenced and images were digitally captured over UV. Intensities of all analysed. bands were measured by software NIH Image 1.60. Relative expression level of C2 and C3 in each sample was calculated Phage library screening as ratios of C2/PBGD and C3/PBGD. Dierences in C2 and C3 expression between RMS-A and RMS-E groups of A commercially available fetal brain phage library (Strata- tumors were assessed with Mann ± Whitney Rank Sum gene 936206) was screened with 32P-labeled C23up PCR statistic. product according to standard protocol (Sambrook et al., 1989). A total of three rounds of screening were performed. Individual positive phage clones were converted back to Northern blot analysis plasmid, cloned according to ZAP-cDNA synthesis kit Ten mg of total RNA from one RMS-A and one RMS-E (Stratagene 200403) and sequenced for further analysis. tumor were resolved on formaldehyde agarose gel and alkaline transferred to a positively charged nylon membrane. NCRMS expression screening C3 PCR product was gel-puri®ed and labeled with 32Pby random priming (Boehringer Mannheim 1004 760). Probing For PCR ¯anking exon 6, 20 ml reactions were set up was done at a speci®c activity of 1.566106 c.p.m. in 10 ml of containing a ®nal 1 mM of each primers, 200 mM dNTPs, Hybrisol (Oncor S4040) in hybridization oven overnight at 2.5 mM MgCl2, 1X Perkin Elmer PCR buer II, and 1 U of 428C. After hybridization, membrane was washed four times: Taq Polymerase Gold. PCR templates were 1 mlofRT once in 26SSC+1% SDS, once in 0.26SSC+0.1% SDS, product (1st strand cDNA). Primers ¯anking exon 6 are: 5'- and twice in 0.16SSC+0.1% SDS at 688C. The membrane CCG ATT ATT ACC AAA GAC AAT G-3' and 5'-TCC was then allowed to expose overnight at 7808C to autorad GTG TCC CTT GTG ATC TCT G-3'. PCR product size is ®lm. 373 bps with exon 6 (C23up) and 247 bps without exon 6 (C- Multiple tissues Northern blot was obtained from OriGene con). PCR condition used were as follows: Initial denaturiza- (HB-2010) with 12 dierent normal human Poly(A)+ tion at 958C for 10 min; 40 step cycles of 958C for 35 s, 558C mRNA. Two dierent ways of probe-labeling were done. for 40 s, 728C for 1 min; ®nal extension 728C for 10 min. Strippable probes were PCR generated and labeled with 32P PCR products were visualized by gel electrophoresis in 1% by Strip-EZ PCR (Ambion 1475). Regular probe was labeled agarose gel containing ethidium bromide. by random priming as above with 32P. Overnight hybridization was carried out as described above but in UltraHyb (Ambion 8670). Probing and washing procedures were also the same as above. Signals were detected with phosphor screen. Acknowledgments We would like to thank Dr Johanna Rommens for Rapid amplification of cDNA ends (RACE) providing the fetal brain library, Dr Irene Andulis, Dr SMART RACE cDNA ampli®cation kit (Clontech K1811-1) Bharati Bapat, Dr Suzanne Kamel-Reid and Nicole was used to amplify both the 5' and 3' ends of target Fabricus for providing RNA or tumor tissues.

References

Anderson J, Gordon A, McManus A, Shipley J and Kozak M. (1996). Mamm. Genome, 7, 563 ± 574. Pritchard-Jones K. (1999a). Neoplasia, 1, 340 ± 348. Li J, Yen C, Liaw D, Podsypanina K, Bose S, Wang SI, Puc Anderson J, Gordon A, Pritchard-Jones K and Shipley J. J, Miliaresis C, Rodgers L, McCombie R, Bigner SH, (1999b). Genes Chromosomes Cancer, 26, 275 ± 285. Giovanella BC, Ittmann M, Tycko B, Hibshoosh H, Barr FG. (1997). Int. J. Biochem. Cell Biol., 29, 1449 ± 1461. Wigler MH and Parsons R. (1997). Science, 275, 1943 ± Brown CJ, Hendrich BD, Rupert JL, Lafreniere RG, Xing Y, 1947. Lawrence J and Willard HF. (1992). Cell, 71, 527 ± 542. Lisitsyn N, Lisitsyn N and Wigler M. (1993). Science, 259, Chang Y, Cesarman E, Pessin MS, Lee F, Culpepper J, 946 ± 951. Knowles DM and Moore PS. (1994). Science, 266, 1865 ± Lisitsyn N and Wigler M. (1995). Methods Enzymol., 254, 1869. 291 ± 304. Dagher R and Helman L. (1999). Oncologist, 4, 34 ± 44. Lucas S, De Smet C, Arden KC, Viars CS, Lethe B, Lurquin Erdmann VA, Barciszewska MZ, Szymanski M, Hochberg C and Boon T. (1998). Cancer Res., 58, 743 ± 752. A, de Groot N and Barciszewski J. (2001). Nucleic Acids Merlino G and Helman LJ. (1999). Oncogene, 18, 5340 ± Res., 29, 189 ± 193. 5348. Hubank M and Schatz DG. (1994). Nucleic Acids Res., 22, Musaro A, McCullagh KJ, Naya FJ, Olson EN and 5640 ± 5648. Rosenthal N. (1999). Nature, 400, 581 ± 585.

Oncogene Non-coding transcript NCRMS on 12q21 in Rhabdomyosarcoma AS Chan et al 3037 Pandita A, Zielenska M, Thorner P, Bayani J, Godbout R, Thorvaldsen JL, Duran KL and Bartolomei MS. (1998). Greenberg M and Squire JA. (1999). Neoplasia, 1, 262 ± Genes Dev., 12, 3693 ± 3702. 275. Visser M, Sijmons C, Bras J, Arceci RJ, Godfried M, Sambrook Joseph, Fritsch EF and Maniatis Tom. (1989). Valentijn LJ, Voute PA and Baas F. (1997). Oncogene, 15, Molecular cloning: a laboratory manual. Cold Spring 1309 ± 1314. Harbor Laboratory Press: New York. Srivastava M, Hsieh S, Grinberg A, Williams-Simons L, Huang SP and Pfeifer K. (2000). Genes Dev., 14, 1186 ± 1195.

Oncogene