Physical and Transcript Map of the Minimally Deleted Region III on 17P Implicated in the Early Development of Barrett’S Oesophageal Adenocarcinoma

Physical and transcript map of the minimally deleted region III on 17p implicated in the early development of Barrett’s oesophageal adenocarcinoma

Julie R Dunn1,5, Janet M Risk1, Joanne E Langan1, Damian Marlee1, Anthony Ellis2, Fiona Campbell3, Alastair JM Watson2,4, John K Fieldn,1,5

1Molecular Genetics and Oncology Group, Clinical Dental Sciences, The University of Liverpool, Liverpool L69 3BX, UK; 2Department of Gastroenterology, Royal Liverpool University Hospital, Liverpool L69 3BX, UK; 3Department of Pathology, Royal Liverpool University Hospital, Liverpool, UK; 4Department of Medicine University of Liverpool, Royal Liverpool University Hospital, Liverpool, UK; 5Roy Castle International Centre for Lung Cancer Research, Liverpool L3 9TA, UK

Allelic imbalance (AI) studies on chromosome 17 (C17) in Introduction Barrett’s oesophageal adenocarcinoma (BOA) tumours strongly suggest that a minimally deleted region on C17p Chromosome 17 is a cancer gene-rich chromosome, harbours a BOA-associated gene with tumour suppressor containing three known tumour suppressor genes function. This deleted region, designated minimal region (TSGs), p53, BRCA1, NF1, and seven putative TSGs, III (MRIII), lies between the two microsatellite markers including HIC1, DEC1, OVCA1/OVCA2, ELAC2, TOC D17S1852 and D17S954. Computational sequence ana- and DMC1 (Risk et al., 1994; Wales et al., 1995; Phillips lysis techniques, BLAST and NIX, were used to assemble et al., 1996; Schultz et al., 1996; Bruening et al., 1999; a physical map of MRIII, consisting of three overlapping Harada et al., 2001; Tavtigian et al., 2001). We have bacterial artificial chromosome (BAC) clones, 297N7, previously shown that chromosome 17 (C17) is the most 963H4 and 795F17, from the RPCI-11 library. The frequent site of allelic imbalance in Barrett’s oesopha- 270 kb genomic sequence of MRIII was analysed using the geal adenocarcinoma (BOA) tumours (Dolan et al., computational gene prediction methods NIX and TAP to 1998), and that C17 contains 13 common minimally identify putative BOA genes. A transcript map of MRIII deleted regions in BOA tumours (Dunn et al., 1999), has been generated and contains 25 candidate BOA genes, many of which coincide with the location of known or four of which are the named genes MYH3, SCO1, x006 putative TSGs. and MAGOH-LIKE. The other candidates consist of Further evidence for the importance of these 13 seven genes predicted by TAP with associated ESTs minimally deleted regions (MRI-XIII) was provided by identified by NIX, two genes predicted by TAP alone and a more detailed analysis of allelic imbalance on C17 in 12 genes/ESTs (or pairs of ESTs) identified by NIX premalignant tissues taken from numerous sites around alone. No disease-specific mutations were identified in five BOA tumours (Dunn et al., 2000). In that study, we x006 or MAGOH-LIKE, although expression analysis of demonstrated that the majority of the deletions (68%) these genes suggests that they may show alternative observed in the BOA tumours also occurred in splicing or be altered epigenetically or in regulatory histologically defined premalignant tissue adjacent to regions in oesophageal cancer. the tumour, and it was apparent that each oesopha- Oncogene (2003) 22, 4134–4142. doi:10.1038/sj.onc.1206466 gectomy specimen analysed demonstrated a unique pathway of clonally evolving loss of heterozygosity Keywords: chromosome 17; Barrett’s oesophageal ade- (LOH). From these data, we then identified minimally nocarcinoma; transcript map; tumour suppressor gene; deleted region III (MRIII) as the most frequent clonally ONCOGENOMICS gene prediction programmes deleted region in the histologically defined earliest tissue, that is, Barrett’s intestinal metaplasia (BIM) without dysplasia. We suggest that MRIII is the site of an early molecular marker of malignant potential in BOA, which could be used to identify individuals at high risk of developing this disease. BOA has an extremely poor *Correspondence: JK Field, The University of Liverpool, Roy Castle prognosis, with a 5-year survival rate of less than 5% International Centre for Lung Cancer Research, Liverpool L3 9TA, (Rankin, 2000), which is often attributed to the frequent UK; E-mail: [email protected] late presentation with disease. It has been shown that 6Current address: Roy Castle International Centre for Lung Cancer Research early intervention can significantly improve survival Received 24 October 2002; revised 31 January 2003; accepted 3 February (Holscher et al., 1997), thus the early identification of 2003 high-risk patients is desirable. These individuals could Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4135 be placed under increased surveillance and/or interven- D17S954 using NIX (a suite of gene prediction and tion, to improve the poor prognosis of this cancer. BLAST analysis programmes, http://www.hgmp.mrc. The MRIII region is defined by four BOA tumours ac.uk/Registered/Webapp/nix/) (Williams, 1998) and and is limited by the microsatellite markers D17S1852 TAP (a programme that predicts alternatively spliced and D17S954 in Dunn et al. (1999). The overall products of the same gene, http://sapiens.wustl.edu/ frequency of LOH at MRIII in BOA is 75% and the Bzkan/TAP) (Kan et al., 2001). defining markers both map to 11.573 Mb on chromo- A total of 25 candidate BOA genes were identified some 17p. MRIII is distinct from p53 (Dunn et al., within MRIII, including two characterized genes (SCO1 2000), and multiplex PCR analysis of D17S1852 and and MYH3), one gene with a high degree of homology D17S954 with chromosome 12 markers has shown that to a characterized Drosophila gene (MAGOH-LIKE) the allelic imbalance observed at these sites is because of (Zhao et al., 1998) and one gene whose predicted protein loss of genetic information and not amplification (JR product has been investigated with relation to myelo- Dunn, unpublished). This paper describes the physical dysplasia (x006) (Soenen et al., 1998) (Figure 2 and and transcript mapping of MRIII, which was under- Table 1). The remaining 21 candidate genes consist of taken in order to provide a scaffold from which to seven genes predicted by TAP with associated ESTs identify candidate BOA genes. The coding regions of identified by NIX, two genes predicted by TAP alone two candidate genes were identified and analysed for and 12 genes/ESTs (or pairs of ESTs) identified by NIX mutations in BOA tumours. alone (Table 1). A total of 19 genes were identified on the forward DNA strand and 10 on the reverse DNA strand. In five instances (genes 1, 6, 18, 20 and 21), a putative gene was identified in the same position on both Results DNA strands. Of the 12 genes predicted by NIX alone, alignment of Physical map of MRIII the EST sequences with other NIX predicted gene The Whitehead Institute Chromosome 17 Sequencing features (exons, promoters, CpG islands) was possible in and Mapping web pages (http://www.genome.wi.mit.e- 10/12 cases, thus lending more weight to their status as du/seq/mapping.html) were searched by BLAST for genuine genes (data not shown). Only genes 7, 17 and 25 clones containing D17S1852 and D17S954. Three over- did not align with any other features. Genes 1, 6, 20, 21, lapping sequenced clones from the RPCI-11 BAC 22 and 24 were also aligned to Unigene clusters, and library (Osoegawa et al., 2001) were identified, and genes 3 and 21 to a poorly predicted protein. Genes 11 these spanned MRIII. 297N7 (Accession No. and 4 were predicted by TAP alone. Gene 11, predicted AC002347) contained D17S1852, 795F17 (AC005284) by TAP on the forward DNA strand, aligns with two contained D17S954 and an analysis of the sequence of ESTs identified by NIX, which were aligned to other these two BACs indicated that 963H4 (Accession No. transcripts on the reverse DNA strand using Sequench- AC015908) bridged the gap between the two clones ert, thus these NIX ESTs were assigned by the authors (Figure 1). to gene x006 (Table 1), and it is possible that the TAP- predicted Gene 11 is actually x006. Gene 4 (forward DNA strand) lies close to the start of the MYH3 gene Transcript map of MRIII predicted by NIX, and since the MYH3 gene was too large to input into the TAP programme in a single To identify candidate BOA genes within MRIII, we sequence fragment, it is thought that Gene 4 is actually annotated the genomic sequence between D17S1852 and at the start of MYH3. Therefore, we suggest that 22/25 of the predicted genes listed in Table 1 are genuine candidates, and it is possible that Genes 17, 7 and 25 are misassigned ESTs and should be treated with caution (although not disregarded) as prospective candidates. In three out of the four named genes, there were discrepancies in both the size of gene and the number of exons predicted by the two methods (NIX and TAP), with TAP missing 50 exons that were defined by NIX using cDNA sequence in two cases. Thus, MAGOH- LIKE was predicted by NIX to be 14.76 kb with nine exons, but to be only 10.874 kb with four exons by TAP (Table 1 and Figure 3). Similarly, MYH3 (a characterized gene) was predicted by NIX to be 26 kb with 36 exons, but TAP predicted a smaller (13.15 kb) gene with fewer (16) exons. In this case, TAP missed both 50 and 30 exons. The murine MYH3 has 40 exons so it is not clear Figure 1 Physical map of minimal region III (MRIII). Physical map of MRIII based on MS marker mapping and sequence whether the NIX prediction of this gene is also missing analysis by NIX. ’, MS marker mapped by PCR and electronic some exons. Gene x006, however, was predicted by TAP PCR (names in boxes); ——, sequenced BAC clone (names in bold) to be 8.9 kb with three exons, but NIX predicted a

Oncogene Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4136

Figure 2 Transcript map of MRIII showing candidate BOA gene predictions by NIX and TAP (data from Table 1) (diagram is drawn a approximately to scale). Rectangle: named gene; rhombus: putative gene; diamond: EST; black vertical lines: scale in kb; heavy black horizontal line: genomic DNA. Candidate genes on the forward strand are above the genomic line, those on the reverse strand are below the genomic line. Predictions coincident on both strands span the genomic line. Gene symbols, where known, are written in italics. The prediction program used for each analysis is shown on the right

smaller gene (6.31 kb) with seven exons. In this case, additional from TAP analysis) and x006 (seven from TAP predicted an extra 50 exon. SCO1 was predicted by NIX analysis) were examined for disease-specific muta- both NIX and TAP to be 5.7 kb with four exons. tions using SSCP and sequencing, respectively. Muta- tion analysis of the MAGOH-LIKE exons was Identification of MAGOH-LIKE and x006 as potential performed by SSCP because all of the exons were small candidate BOA genes and it was the most rapid method. x006, however, had larger exons and was therefore analysed by sequencing. Public domain websites contained some information The tumours used for SSCP analysis of MAGOH- about the function, expression pattern and conservation LIKE were selected as described in Materials and between the species of genes MYH3, SCO1, MAGOH- Methods, and included a tumour that defined MRIII LIKE and x006. All four genes are both highly (i.e. showed LOH at MRIII bounded by retention of conserved and widely expressed in a variety of human heterozygosity on either side). Tumours defining MRIII tissues, and were thus all considered to be good have probably ‘lost’ this second allele at MRIII candidates as a BOA TSG. This is because TSGs following a hypothesized mutation in the first allele identified as being involved in the development of cancer (Knudson’s ‘two-hit’ hypothesis for tumour suppressor by LOH analysis often play a role in the control of cell gene inactivation). By using 10 pairs of normal/tumour growth and death. They are, therefore, usually critical to samples for the SSCP analysis, it was assumed that any the cell and are often expressed in many cell types. causative mutations would be identified. This is because MYH3 and SCO1, however, do not have a cancer- the overall frequency of LOH at MRIII in our earlier related function. MYH3 is a fast skeletal muscle gene study was 75% and, in addition, we have found LOH at and SCO1 is involved in copper transport in the MRIII in Barrett’s intestinal metaplasia in 4/5 samples respiratory chain and these genes were thus listed for tested (Dunn et al., 1999, 2000). The two tumours used further investigation at a later date. Conversely, in the sequence analysis of x006 were selected because MAGOH-LIKE and x006 both have a putative cancer- they were the most likely tumours to show a mutation, related function. The human chromosome 1 homologue since they only showed LOH at MRIII (i.e. defined of MAGOH-LIKE has been associated with increased MRIII). Thus the LOH must have originated in MRIII cell proliferation in mouse cell lines (Zhao et al., 1998), and the remaining copy of the cancer-causing gene and the x006 protein is linked to myelodysplasia located in this region would contain a mutation (the first (Soenen et al., 1998). Therefore, the genes that were ‘hit’ or Knudson’s ‘two-hit’ hypothesis). OE33 BOA cell considered to have the highest priority in the search for line DNA was also analysed for mutations in both the BOA genes were MAGOH-LIKE and x006, which genes. were subjected to expression and mutational analyses. No consistent disease-specific alterations were observed in either of the genes’ coding regions. Mutation and expression analysis of MAGOH-LIKE and x006 Expression analysis In order to assess the expression Mutation analysis All of the predicted exons of patterns of the candidate BOA genes MAGOH-LIKE MAGOH-LIKE (nine from NIX analysis, plus 1 and x006 in normal and diseased oesophagus (BOA and

Oncogene Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4137 Table 1 List of candidate BOA genes within MRIII, identiﬁed by NIX and TAP Position (kb) Prediction method Gene/EST Size of gene/EST No. of exons

D17S1852 (limit) 1 1.2 (F and R) NIX EST 466 bp 2 3–3.8 (R) TAP Gene 825 bp 1 NIX EST 3 13.2–13.7 (R) TAP Gene 450 bp 1 NIX EST 4 4.4–30.74 (F) NIX MYH3 26 kb 36 6.85–20 (F) TAP MYH3 13.15 kb 16 6.5–13.9 (F) TAP MYH3 (ALT) 7 kb 4 5 31.2–32.47 (F) TAP Gene 1.3 kb 2 6 34.73 (F and R) NIX EST 300 bp 7 37.45 (R) NIX EST 333 bp 8 48.7–54.7 (F) TAP Gene 6 kb 2 NIX 4 ESTs 9 56.7–63 (F) TAP Gene 6.3 kb 2 57–57.6 (F) TAP Gene (ALT) 0.6 kb 2 NIX 2 ESTs 10 67.4–73.1 (F) NIX SCO1 5.7 kb 4 67.2–73 (F) TAP SCO1 5.7 kb 4 11 80.24–80.88 (F) TAP Gene 640 bp 1 12 79.7–86.01 (R) NIX x006 6.31 kb 7 73–81.9 (R) TAP x006 8.9 kb 3 80.64–81.9 (R) TAP x006 (ALT) 1.69 kb 2 13 86.5–86.57 (F) TAP Gene 75 bp 1 87.34(F) NIX EST 200 bp 14 88.69–89.9 (F) TAP Gene 1.31 kb 1 89.7 (F) NIX 2 ESTs 650 bp 15 90.35–105.11 (F) NIX MAGOH-LIKE 14.76 kb 9 TAP MAGOH-LIKE 10.875 kb 4 90.9–101.74 (F) TAP MAGOH-LIKE 9.6 kb 4 90.9–100.5 (F) (ALT) 16 106 (R) NIX 2 ESTs 465 bp 17 107 (F) NIX EST 325 bp 18 116.9–117.35 (F) TAP Gene 450 bp 1 NIX EST 270 bp 117.2 (R) 19 136.09 (F) NIX EST 333 bp 20 209.24–223.87 (F and R) NIX Gene with 3 ESTs 14.6 kb 8 21 245.15 (F and R) NIX EST 399 bp 22 247.15 (F) NIX EST 195 bp 23 260.6 (F) NIX EST 195 bp 24 265.4 (F) NIX EST 399 bp 25 269.9 (F) NIX EST 333 bp D17S954 (limit)

The sequence position (kb) from the microsatellite marker D17S1852, and the DNA strand containing the predicted gene (F, forward; R, reverse) is indicated. NIX and TAP predictions are based on the HGP sequence database information as at January 2002 oesophageal squamous cell carcinoma (OSCC)), and in normal oesophageal RNA, BOA cell line RNA and other human tissues, three ESTs from each gene were OSCC RNA. Expression data for other human tissue analysed by RT–PCR and SAGEmap virtual Northern types from RT–PCR and SAGEmap analyses indicated analysis (http://www.ncbi.nlm.nih.gov/SAGE). The that both genes are expressed in at least seven other same ESTs were used for both analyses in order to human tissue types, thus both genes are conﬁrmed as obtain a direct comparison between the SAGEmap data BOA candidate genes because of their wide tissue and our RT–PCR results. Validation of DNase I expression. treatment was by PCR using intronic primers. Unfortu- Furthermore, the RT–PCR data indicated that both nately, there was no BOA tumour tissue RNA available MAGOH-LIKE and x006 appear to be differently for this analysis. In addition, there is no commercially expressed in BOA and OSCC compared to normal available oesophageal adenocarcinoma RNA, thus the oesophagus, which suggests that they may be important only available veriﬁed BOA cell line OE33 (ECACC Ref genes in the development of oesophageal cancer (BOA No. 96070808) was used as the source of BOA RNA. and OSCC) (Figure 4). EST HS846185, which aligns The RT–PCR results indicated that the MAGOH- with exon 1 of MAGOH-LIKE, demonstrated an LIKE and x006 genes are expressed in normal and additional 587 bp transcript in both BOA and OSCC diseased (BOA and OSCC) oesophagus (Figure 4, compared to normal oesophagus, which only expressed Table 2). All three ESTs from each gene were expressed the expected 251 bp transcript (Figure 4a data for BOA

Oncogene Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4138

Figure 3 Figure showing the relative positions of candidate BOA gene predictions on BAC clone 297N7 as predicted by NIX and TAP (data from Table 1) (diagram is drawn approximately to scale). Gene names are given in italics. The number of predicted exons for each gene is shown, and alternative transcripts predicted by TAP are shown directly above the initial gene prediction. Rectangle: gene (shown above the line if on the forward strand and below the line if on the reverse strand); black vertical lines within genes: exon boundaries; black horizontal line: BAC 297N7. The prediction program used in each analysis is shown on the right. The positions of the EST primers used for expression analysis of MAGOH-LIKE and x006 are shown directly below their positions as predicted by NIX

oesophageal adenocarcinoma cell line, OE33, in order to determine if genomic alterations could be respon- sible for the altered expression of these genes in this cell line. However, no mutations were identiﬁed in OE33, suggesting that the MAGOH-LIKE and x006 genes are altered epigenetically in oesophageal cancer.

Discussion Figure 4 RT–PCR results for MAGOH-LIKE (a and b) and x006 (c and d) ESTs. (a) HS846185; (b) HS025225; (c) AW368449; (d) Our previous studies based on allelic imbalance data HS387329. N: normal oesophagus; BOA: Barrett’s oesophageal suggest the presence of a human Barrett’s oesophageal adenocarcinoma; OSCC: oesophageal squamous cell carcinoma adenocarcinoma (BOA) tumour suppressor gene (TSG) in a minimally deleted region of chromosome 17p not shown). MAGOH-L1KE exon 7 EST, HS025225, (C17p) centromeric to the p53 TSG (Dunn et al., 1999, appeared to be downregulated in BOA compared to 2000), and multiplex PCR has confirmed that the allelic normal oesophagus and OSCC (Figure 4b), although imbalance in this region is because of loss of hetero- quantitative RT–PCR data would be necessary to zygosity (LOH) rather than amplification of the region confirm this. Expression of MAGOH-LIKE EST (Dunn, unpublished). In this study, we have assembled AW104858 (exon 8) was not altered in BOA and OSCC physical and transcript maps of the deleted region compared with normal oesophagus (data not shown). (minimal region III–‘MRIII’) from which the candidate Similarly, x006 EST AW368449 (exon 1) demonstrated BOA TSG can be identified. The region is 270 kb in two additional transcription products in both BOA and length. OSCC compared to normal oesophagus (Figure 4c) and The MRIII region has also been implicated in the EST HS387329 (located between exons 6 and 7) development of oesophageal squamous cell carcinoma appeared to be downregulated in BOA and OSCC (OSCC) (Huang et al., 2000). The minimally deleted compared to normal oesophagus (Figure 4d). Expres- region in Huang et al.’s study however is limited by the sion of x006 EST AI004998 (located between exons 6 two MS markers D17S804 (distal end at 10.641 Mb) and and 7) was not altered in BOA and OSCC compared D17S799 (proximal end at 12.895 Mb), and is thus with normal oesophagus (data not shown). Only the 42 Mb in size, more than four times the size of MRIII. MAGOH-LIKE exon 7 EST, HS025225, was identified LOH frequencies of 88% at D17S804 and 90% at in SAGEmap libraries, while two ESTs from the x006 D17S799 were observed. The LOH frequencies for the gene were represented (AW368449 (exon 1) and MS markers D17S954 and D17S1852 were not reported HS387329 (between exons 6 and 7)). in this paper. In the light of the RT–PCR results, mutation analysis LOH analysis has been used successfully in the of MAGOH-LIKE and x006 was undertaken in Barrett’s isolation of several candidate TSGs, most recently

Oncogene Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4139 Table 2 RT–PCR and SAGEmap virtual Northern expression results for three ESTs from each of the MAGOH-LIKE and x006 candidate genes RT–PCR SAGEmapa

EST ID Gene (exon)b Oes N BOA* OSCC Liv N Prst N Brn N Brst N Col* T Stom N Tissue

HS846185 MAG(1) + +(2) +(2) + + + À + À NF HS025225 MAG(7) + +k + ÀÀÀ+ + + Brn(N) AW104858 MAG(8) + + + ÀÀÀÀ + À NF AW368449 x006 (1) + +(3) +(3) ÀÀÀ+ + + Kid(N) Ov(N) Brst(T) Brn(T) Panc(T) Prst(T) HS387329 x006 (between 6 and 7) + +k +k À + ÀÀ+ À NF AI004998 x006 (between 6 and 7) + + + + + + À + À Col(T) Brst(T)

+: positive expression; À: negative expression; number in brackets indicates the number of different-sized transcripts observed when greater than one; k: decreased band intensity compared to normal oesophagus RNA; MAG: MAGOH-LIKE; OesN: normal oesophagus; BOA: Barrett’s oesophageal adenocarcinoma; OSCC: oesophageal squamous cell carcinoma tissue; Liv(N): normal liver; Prst(N): normal prostate; Brn(N): normal brain; Brst(N): normal breast; Col(T): colon tumour; Stom(N): normal stomach; *: RNA extracted from cell lines. atissues positive for the SAGEtag from the given EST are listed. NF: EST not found in SAGEmap database. bnumber in brackets after gene name indicates the exon to which the EST aligns. Brn (T): brain tumour; Kid (N): normal kidney; Ov (N): Normal ovary; Panc (T): pancreatic tumour; Prst (T): prostate tumour; Brst (T): breast tumour. including an ovarian cancer TSG ‘Ov/Br septin’on and should be treated with caution (although not C17q25 (Russell et al., 2000), an oesophageal cancer disregarded) as prospective candidates. ESTs are une- TSG ‘DEC1’ on C9q32 (Nishiwaki et al., 2000), dited single-pass sequencing reads and are therefore ‘LOH11CR2A’ (C11q23) implicated in breast and lung prone to error; at best there is a 97% accuracy rate carcinomas (Monaco et al., 1997), and a renal cell (Hillier et al., 1996). carcinoma TSG ‘TU3A’ on C3p14.3–14.2 (Yamato et al., By using more than one method of gene finding (NIX 1999). Each of these studies was conducted in a manner suite of programmes and TAP) and the Sequenchert similar to that used to localize a BOA TSG in this sequence alignment programme, we believe that the project. Firstly, common region of allelic loss was candidate genes identified in this study are more likely to identified in the tumours and the limiting MS markers be genuine than if they had been identified using a single determined; secondly, a physical contig of genomic method. It is recognized however that before further segments was constructed to span the region of interest; investigation of candidate genes is undertaken (e.g. and finally the candidate TSG was identified by mutation analysis), it would be prudent to supplement sequence analysis of the region. This is currently the current information with laboratory-generated data considered the most appropriate approach for identify- such as RT–PCR, thus eliminating false exon predic- ing novel TSGs in cancer research, when using genetic tions made by NIX or TAP. instability as the starting point. The preliminary RT–PCR expression analysis results Computational methods of gene identification have reported here provide evidence for alternative splicing in been used successfully in the identification of G-protein- both MAGOH-LIKE and x006 in oesophageal cancer coupled receptor genes in two separate studies (March- (BOA and OSCC) as alternative transcripts for MA- ese et al., 1999; Wittenberger and Chica Schaller, 2001), GOH-LIKE exon 1 and x006 exon 1 are observed in both of which utilized expressed sequence tag (EST) both BOA and OSCC. In addition, the TAP predictions database searches. It is envisaged that many more for MAGOH-LIKE and x006 lend further support to human genes will be identified using such methods as the these results. TAP has predicted an alternatively spliced completion of the Human Genome Sequence progresses. MAGOH-LIKE gene that had lost exon 1 and gained Certainly, given the information we have accumulated another exon between exons 2 and 3. It is noteworthy leading up to a fully sequenced candidate BOA gene that since PCR primers for MAGOH-LIKE EST exon 1 region, computational analyses have been the most gave a PCR product, it is most likely that exon 1 has rapid and effective method of identification of candidate been missed by the TAP programme. This is probably genes for further investigation. because TAP only uses EST data, whereas NIX uses In all, 42% (11/26) of the candidate genes identified in cDNA data for their predictions. Thus, TAP may miss MRIII were predicted by both NIX and TAP. Since the some 50 ends of predicted genes. TAP program utilizes WU2BLASTN (Gish, 1996–2000) Some apparent discrepancies were observed in the to search the EST database ‘dbEST’ (i.e. a different RT–PCR data for both MAGOH-LIKE and x006. method to NIX), we believe that these gene predictions Specifically, out of the three ESTs used for expression are more robust than those that were predicted by one analysis of these genes, only one EST appeared to be method alone. We suggest that 22/25 of the predicted downregulated in each case. Possible explanations for genes listed in Table 1 are genuine candidates, and it is these discrepancies are as follows. Firstly, the ESTs that possible that genes 17, 7 and 25 are mis-assigned ESTs are apparently downregulated could represent alterna-

Oncogene Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4140 tively spliced exons that are under-represented in data at the Whitehead Institute website (http://www.genome. tumour tissue. Secondly, the two ESTs that show no wi.mit.edu) using the sequences of D17S1852 and D17S954. changes in expression in BOA or OSCC could actually Clones containing these markers were purchased from be downregulated, but the change is too small to see in Research Genetics and screened by PCR to verify the marker these experiments, and, conversely, it is possible that the content. All PCR primers (including sequencing primers) used in this study were designed using the Primer 3 design ESTs that do show downregulation in BOA and/or programme (http://www.genome.wi.mit.edu) and purchased OSCC are not actually downregulated. Quantitative from MWG Biotech Ltd. RT–PCR data would determine if either of these alternatives were the case. Finally, it is possible that Computational sequence analysis: transcript mapping the sequence data at the site of the ESTs with unchanged expression levels (or indeed at the site of the apparently 20 kb segments (with 5 kb overlap) of the sequence of MRIII downregulated ESTs) may be poor and has led to were analysed using the NIX and TAP programmes in order to predict genes and alternative splice products, respectively. inaccuracies in RT–PCR amplification. Since NIX produces results for all predicted features, ranging The results in this study highlight the enormous from poor to excellent predictions, we employed set criteria for potential of computational gene prediction methods the selection of candidate genes predicted by NIX for further coupled with laboratory-based methods. The bioinfor- investigation as follows. matic approach provided an important sieve mechanism by which we are able to prioritize our analysis to two 1. Only ‘excellent’ predictions were noted. BLAST alignments candidate genes. The RT–PCR analysis suggests that for these predictions were also visually checked to verify if a both the MAGOH-LIKE and x006 genes have an significant proportion (B95%) of the matched sequence alternative form that is important in oesophageal cancer was aligning to genomic sequence. In addition, all development (BOA and OSCC), but is not transcribed transcript alignments were verified using Sequenchert in the normal oesophagus. If these results are confirmed, (Gene Codes Corporation), using the following parameters: and alternative gene products are made in oesophageal minimum 85% bp match and minimum 20 bp overlap. cancer, it will be important to ascertain whether these 2. Exons predicted by fewer than two exon prediction products are also found in other human cancers. program, which did not align with any other predicted The major objective to this study was to identify the features in NIX, and which were not predicted by TAP are gene that could be considered to be an early marker of not included in Table 1. malignant potential specific to BOA. Even though there Expression analysis of candidate genes was compelling evidence that both MAGOH-LIKE and Culture of OE33 and OE47 cell lines and RNA extraction Cells x006 were candidate genes based on the alternative were grown to 85% confluence in RPMI 1640 with 2 mm splicing of exon 1, the results presented here indicate glutamine and 10% FBS in 550 ml tissue culture flasks, before that both forms of each gene are found in both BOA detachment with 0.25% trypsin/EDTA and subsequent cen- and OSCC. In order to ascertain the frequency of trifugation at 100 g for 10 min to pellet the cells. Cell pellets occurrence of the alteration in BOA compared to were snap frozen on dry ice before storage at À801C prior to OSCC, a larger data set will be required. The likelihood RNA extraction. Each cell pellet contained B3 Â 107 cells. that either MAGOH-LIKE or x006 is involved in Total RNA was extracted using the Qiagen Rneasy mini kit, oesophageal cancer development is further strengthened and quantified using a GeneQuant II Spectrophotometer. by the apparent downregulation of 30 ESTs in BOA and OSCC, respectively. Quantitative RT–PCR analysis will RT–PCR RT–PCR reactions were carried out using the t be required to expand on these results. RetroScript first-strand synthesis RT–PCR kit (Ambion). Total RNA was purchased from Clontech (normal RNA from In summary, the expression analyses of MAGOH- stomach, liver, prostate and brain) and Invitrogen (normal LIKE and x006 by RT–PCR suggests that both genes oesophagus, squamous cell tumour oesophagus and normal remain good candidates for a tumour suppressor/ breast), and isolated from OE33 (BOA) and OE47 (colon cancer-related gene, albeit perhaps not specific for adenocarcinoma). Cell line RNA samples were first treated BOA, and both genes warrant further investigation in with DNase enzyme (1 U) (Gibco BRL) to remove contam- BOA and other cancers. inating DNA. First strand cDNA was synthesized using 50 mm. In conclusion, we have developed a transcript map of random decamers and 1–2 mg total RNA, according to kit MRIII, a region of C17p that we believe contains an instructions. Candidate gene transcripts (ESTs) were amplified early molecular marker of malignant potential in BOA. by PCR (1 cycle of 941C for 5 min; 30 cycles of 941C for 30 s, 1 1 1 This transcript map is the first of its kind in BOA, and 59 C for 30 s, 72 C for 30 s; 1 cycle of 72 C for 5 min) using Bioline Taq polymersase (1 U). EST primers were designed as will be the foundation for the identification of the BOA described previously (Dunn et al., 2000). marker using a combination of computational- and laboratory-based methods. SAGEmap virtual Northern analysis In order to determine the breadth of expression of MAGOH-LIKE and x006 in a wider variety of other human tissues (cancer and normal), three EST sequences each from the MAGOH-LIKE and x006 Materials and methods candidate genes were analysed using the virtual Northern tool on the US National Centre for Biotechnology Information Identification of BAC clones spanning MRIII (NCBI) Serial Analysis of Gene Expression (SAGE) map BAC clones overlapping MRIII were identified by BLAST website (http://www.ncbi.nlm.nih.gov/SAGE/SAGEtag.cgi). analysis of the chromosome 17 physical mapping and sequence In brief, the EST sequence obtained from the NIX output

Oncogene Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4141 was pasted into the SAGEmap virtual Northern query electrophoresis on polyacrylamide gels made with Protogel sequence box. The SAGEmap database was then automati- (30% acrylamide with 37.5 : acrylamide : bis acrylamide ratio, cally searched for matching SAGEmap library ‘TAGs’, and a National Diagnostics) under the following eight running list of the tissue libraries in which the EST was identified was conditions. given. 1. 8 and 10% gel at 201C, Mutation analysis of MAGOH-LIKE and x006 2. 8 and 10% gel at 51C, 3. 8 and 10% gel with 5% glycerol at 201C, SSCP All 10 of the predicted exons for the candidate gene 4. 8 and 10% gel with 5% glycerol at 51C. MAGOH-LIKE were examined for mutations using the single- strand conformational polymorphism (SSCP) technique. Gels were then silver stained and analysed for polymorph- Primers were designed in the flanking introns and all SSCP isms. PCR products were o200 bp. DNA (200 ng) was used for each template. DNA was Sequencing All of the seven exons of candidate gene x006 derived from the BOA cell line, OE33, and the following BOA were analysed for mutations by sequencing on an LiCor normal and tumour paired tissues: automated sequencer using an Epicentre SequiTherm EX- CELtII DNA Sequencing kit-LC. Sequencing DNA templates DNA pair 1 – defined MRIII at both proximal and distal were derived using eLONGases Enzyme mix (Gibco BRL) limits, from B200 ng DNA from two BOA normal/tumour pairs that DNA pair 2 – showed LOH at MRIII, but did not define showed LOH at MRIII and the BOA cell line. DNA (150 ng) the region, from BAC 297N7 was used as a positive control. Templates DNA pair 3 – showed no LOH at MRIII. were amplified using primers (10 pmol/ml) designed to cover DNA samples from the blood of six unaffected individuals 800–1000 bp of sequence with the exon in the middle. were used as controls. Any exons demonstrating a polymorph- Sequencing primers were designed immediately outside the ism were subjected to a more rigorous SSCP analysis using a exon sequence and were 50labelled with IRD700. Templates further 10 BOA DNA normal and tumour pairs. DNA was were verified for sequencing by the presence of a single, obtained as described in Dunn et al. (1999). appropriately sized product on agarose gel electrophoresis, PCR primers were designed to fall immediately outside each then cleaned using the QIAquickt PCR purification kit exon such that the whole exon was amplified in each case. (Qiagen). Each exon was sequenced twice in each direction Verification of the amplification of a single, appropriately sized using a different PCR template. Sequence traces were analysed product was by agarose gel electrophoresis. SSCP DNA using the LiCor base-calling software, by eye, and using the templates were prepared immediately prior to running SSCP Sequenchert programme. gels as follows: 2 ml of PCR product and 10 ml of SSCP loading buffer (80% formamide, 100 mm NaOH, 10 mm EDTA, 0.05% Acknowledgements bromophenol blue and 0.05% xylene cyanol) were denatured We thank Dr Lakis Liloglou for advice on SSCP analysis. This at 951C for 3 min, then snap cooled on iced water. A volume of study was supported by a grant from the North West Cancer 1–4 ml of each denatured DNA sample was subjected to Research Fund UK (Grant CR536).

References

Bruening W, Prowse AH, Schultz DC, Holgado Madruga M, Li G, Giffen C and Taylor PR. (2000). Carcinogenesis, 21, Wong A and Godwin AK. (1999). Cancer Res., 59, 4973– 2019–2026. 4983. Kan Z, Rouchka E, Gish Wand States D. (2001). Genome Res, Dolan K, Garde J, Gosney JR, Sissons M, Wright T, 11, 875–888. Kingsnorth AN, Walker SJ, Sutton R, Meltzer SJ and Field Marchese A, Sawzdargo M, Nguyen T, Cheng R, Heng HH, JK. (1998). Br. J. Cancer, 78, 950–957. Nowak T, Im DS, Lynch KR, George SR and O’Dowd BF. Dunn J, Garde J, Dolan K, Gosney JR, Sutton R, Meltzer SJ (1999). Genomics, 56, 12–21. and Field JK. (1999). Oncogene, 18, 987–993. Monaco C, Negrini M, Sozzi G, Veronese ML, Vorechovsky I, Dunn JR, Garde J, Dolan K, Gosney JR, Oates BC, Watson Godwin AK and Croce CM. (1997). Genomics, 46, 217–222. AJM, Fielding P and Field JK. (2000). Clin. Cancer Res., 6, Nishiwaki T, Daigo Y, Kawasoe T and Nakamura Y. (2000). 4033–4042. Gene Chromosome Cancer, 27, 169–176. Gish W. (1996–2000). WU-BLASTN2. http://blast.wustl.edu/. Osoegawa K, Mammoser AG, Wu C, Frengen E, Zeng C, Harada H, Nagai H, Tsuneizumi M, Mikami I, Sugano S and Catanese JJ and de Jong PJ. (2001). Genome Res., 11, 483–496. Emi M. (2001). J. Hum. Genet., 46, 90–95. Phillips NJ, Ziegler MR, Radford DM, Fair KL, Steinbrueck Hillier LD, Lennon G, Becker M, Bonaldo MF, Chiapelli B, T, Xynos FP and Donis-Keller H. (1996). Cancer Res., 56, Chissoe S, Dietrich N, DuBuque T, Favello A, Gish W, 606–611. Hawkins M, Hultman M, Kucaba T, Lacy M, Le M, Le N, Rankin SC. (2000). Oesophageal Cancer: staging. Practical Mardis E, Moore B, Morris M, Parsons J, Prange C, Rifkin Management of Oesophage Disease. Adam A, Mason R and L, Rohlﬁng T, Schellenberg K and Marra M. (1996). Owen WJ (eds) Oxford: Isis Medical Media. Genome Res., 6, 807–828. Risk JM, Whittaker A, Fryer A, Ellis A, Shaw JM, Field EA, Holscher AH, Bollschweiler E, Schneider PM and Siewert JR. Friedmann PS, Bishop DT, Bodmer J, Leigh IM and Field (1997). Br. J. Surg., 84, 1470–1473. JK. (1994). Nat. Genet., 8, 319–321. Huang J, Hu N, Goldstein AM, Emmert-Buck MR, Tang ZZ, Russell SE, McIlhatton MA, Burrows JF, Donaghy PG, Roth MJ, Wang QH, Dawsey SM, Han XY, Ding T, Chanduloy S, Petty EM, Kalikin LM, Church SW, McIlroy

Oncogene Transcript map and candidate Barrett’s TSGs on C17P JR Dunn et al 4142 S, Harkin DP, Keilty GW, Cranston AN, Weissenbach J, Thomas A, Tranchant M, Woodland AM, Labrie F, Hickey I and Johnston PG. (2000). Cancer Res., 60, Skolnick MH, Neuhausen S, Rommens J and Cannon- 4729–4734. Albright LA. (2001). Nat. Genet., 27, 172–180. Schultz DC, Vanderveer L, Berman DB, Hamilton TC, Wong Wales MM, Biel MA, el Deiry W, Nelkin BD, Issa JP, AJ and Godwin AK. (1996). Cancer Res., 56, 1997–2002. Cavenee WK, Kuerbitz SJ and Baylin SB. (1995). Nat. Med., Soenen V, Preudhomme C, Roumier C, Daudignon A, Lai JL 1, 570–577. and Fenaux P. (1998). Blood, 91, 1008–1015. Williams GW. (1998). URL: http://www.hgmp.mrc.ac.uk/ Tavtigian SV, Simard J, Teng DH, Abtin V, Baumgard M, NIX/. Beck A, Camp NJ, Carillo AR, Chen Y, Dayananth P, Wittenberger T and Chica Schaller H. (2001). J. Mol. Biol., Desrochers M, Dumont M, Farnham JM, Frank D, Frye C, 307, 799–813. Ghaffar S, Gupte JS, Hu R, Iliev D, Janecki T, Kort EN, Yamato T, Orikasa K, Fukushige S, Orikasa S and Horii A. Laity KE, Leavitt A, Leblanc G, McArthur-Morrison J, (1999). Cytogenet. Cell Genet., 87, 291–295. Pederson A, Penn B, Peterson KT, Reid JE, Richards S, Zhao XF, Colaizzo-Anas T, Nowak NJ, Shows TB, Elliott Schroede M, Smith R, Snyder SC, Swedlund B, Swensen J, RWand Aplan PD. (1998). Genomics, 47, 319–322.

Oncogene