Transcriptome analysis of serous ovarian cancers identifies differentially expressed chromosome 3 genes

Transcriptome Analysis of Serous OvarianCancers Identifies Differentially ExpressedChromosome 3 Genes

Ashley H. Birch,1 Michael C.J. Quinn,1 Ali Filali-Mouhim,2 Diane M. Provencher,2,3

Anne-Marie Mes-Masson,2,4 and Patricia N. Tonin1,5,6*1Department of Human Genetics, McGill University, Montreal, Canada2Centre de Recherche du Centre Hospitalier de l’Universite de Montreal (CHUM), Institut du cancer de Montreal,Montreal, Canada3Division of Gynecologic Oncology, Universite de Montreal, Montreal, Canada4Department of Medicine, Universite de Montreal, Montreal, Canada5The Research Institute of the McGill University Health Centre, Montreal, Canada6Department of Medicine, McGill University, Montreal, Canada

Cytogenetic, molecular genetic and functional analyses have implicated chromosome 3 genes in epithelial ovariancancers (EOC). To further characterize their contribution to EOC, the Affymetrix U133A GeneChip1 was used toperform transcriptome analyses of chromosome 3 genes in primary cultures of normal ovarian surface epithelial

(NOSE) cells (n¼ 14), malignant serous epithelial ovarian tumors (TOV) (n¼ 17), and four EOC cell lines (TOV-81D,TOV-112D, TOV-21G, and OV-90). A two-way comparative analysis of 735 known genes and expressed sequencesidentified 278 differentially expressed genes, where 43 genes were differentially expressed in at least 50% of the TOV

samples. Three genes, RIS1 (at 3p21.31), GBE1 (at 3p12.2), and HEG1 (at 3q21.2), were similarly underexpressed in allTOV samples. Deregulation of the expression of these genes was not associated with loss of heterozygosity (LOH) ofthe genetic loci harboring them. LOH analysis of the RIS1, GBE1, and HEG1 loci was observed at frequencies of 14.3%,

13.7%, and 9.2%, respectively, in a series of 66 malignant TOV samples of the serous subtype. Reduced expressionlevels of RIS1, GBE1, and HEG1 were observed only in the tumorigenic EOC cell lines (TOV-21G, TOV-112D, and OV-90) and did not correlate with LOH. These results combined suggest that RIS1, GBE1, and HEG1, unlike classical tumorsuppressor genes, are not likely to be primary targets of inactivation. This study provides a comprehensive analysis of

chromosome 3 gene expression in NOSE and in EOC samples and identifies chromosome 3 gene candidates for furtherstudy. � 2007 Wiley-Liss, Inc.

Key words: ovarian cancer, chromosome 3, microarray

INTRODUCTION

Numerical and structural aberrations of chromo-some 3 have been frequently documented in epi-thelial ovarian cancer (EOC) through karyotyping,loss of heterozygosity (LOH), and comparative geno-mic hybridization (CGH) techniques [1–5]. Analysesof cytogenetic studies indicated that 3p losses and/orbreaks were associated with early events in ovariancancer [6–9]. Tumor suppressor genes (TSGs) locatedon the 3p arm have been suggested by LOH analyses[10–15], as well as functional analyses involving thetransfer of whole chromosome 3 [16] or chromosome3 fragments [17]. Although a number of promisingcandidates have been identified, such as SEMA3F andSEMA3B (at 3p21.31) [18,19], DRR1/TU3A (at 3p14.2)[20], RASSF1A (at 3p21.31), RAR-b (at 3p24.2), MLH1(at 3p21.3), and DLEC1 (at 3p22.3) [21–23], fewexhibit features of classical TSGs, where inactivationof one allele occurs as a result of LOH andthe second allele is mutated or transcriptionallysilenced through promoter methylation. As pro-posed for the pathogenesis of lung cancer, where

3p deletions are frequent [24–28], a coordinateinactivation of a group of 3p genes by genetic(deletion and/or mutation) and epigenetic (pro-moter methylation and/or haploinsufficiency)mechanisms may be important for the pathogenesisof ovarian cancer.

Chromosomal anomalies of 3q in EOC have alsogarnered attention. However, unlike the 3p arm,recurrent patterns include gains of the 3q arm, as

MOLECULAR CARCINOGENESIS 47:56–65 (2008)

� 2007 WILEY-LISS, INC.

This article contains supplementary material, which may beviewed at the Molecular Carcinogenesis website at http://www.interscience.wiley.com/jpages/0899-1987/suppmat/index.html.

Abbreviations: EOC, epithelial ovarian cancer; LOH, lossof heterozygosity; TSGs, tumor suppressor genes; NOSE, normalovarian surface epithelial.

*Correspondence to: Medical Genetics, Room L10-120, MontrealGeneral Hospital, 1650 Cedar Avenue, Montreal, Quebec,Canada H3G 1A4.

Received 14 February 2007; Revised 4 May 2007; Accepted 7 May2007

DOI 10.1002/mc.20361

detected by array CGH [9,29–32]. Amplification ofthe 3q26 region is frequently reported, resulting inaugmentation of both the PIK3CA and EVI-1 onco-gene transcripts [33,34]. Also found at this locus iseIF-5A2, which has documented oncogenic ability,and is correlated with advanced stages of ovariancancer when overexpressed [35]. Interestingly, thelocus for the RNA template of telomerase (hTR) is alsolocated within the amplified region [36]. Upregulat-ion of its expression has been noted in ovariancarcinomas [37], although a correlation betweenamplification and expression is required to deter-mine whether hTR is a target of the 3q26 amplifica-tion.

Large-scale transcriptome analyses of chromo-some 3 extracted data sets have also revealed differ-ential expression of chromosome 3 genes in fourspontaneously immortalized EOC cell lines (TOV-21G, TOV-81D, TOV-112D, and OV-90) when com-pared with primary cultures of normal ovariansurface epithelium (NOSE) [38]. The tumorigeniccell lines, TOV-21G, TOV-112D, and OV-90, exhib-ited more examples of differentially expressed genesin comparison to TOV-81D, a non-tumorigeniccell line derived from a patient with the mostindolent form of the disease [39]. Phenotypes of theovarian cancer cell lines are also reflected in theirglobal gene expression profiles [40] and in the geneexpression profiles of specific chromosomes (such as17, 22 and X) [38,41–43]. Notable is that OV-90,which is hemizygous for the entire 3p arm, exhibitedthe largest number of underexpressed 3p geneswhen compared with the other EOC cell lines [38].This analysis also revealed examples of similarderegulation of gene expression patterns acrossall tumorigenic EOC cell lines, although theydiffered in their chromosome 3 content. The appa-rent correlation of differential chromosome 3 geneexpression and tumorigenic potential suggests thatthese genes may be implicated in ovarian tumori-genesis or progression. This concept is exemplifiedin the global gene expression analyses of an ovariancancer model originally developed to identify chro-mosome 3p TSGs, where the transfer of normalchromosome 3 fragments into OV-90 resulted inthe suppression of tumorigenicity [17]. Geneticmapping of the transferred fragments identifiedpotential TSG-containing regions. However, expres-sion microarray analyses also identified chromo-some 3 genes deregulated in their expression patternthat did not overlap with the transferred chromoso-mal fragments. Hence, a combination of molecular,cytogenetic and expression array analyses, andmost recently, functional assays, support the con-cept that a number of chromosome 3 genes may playa role in ovarian cancer.

In this study the Affymetrix GeneChip1 arraytechnology was applied to characterize chromosome3 specific transcriptome patterns in primary cultures

ofNOSEsamplesandidentifydifferentiallyexpressedgenes in comparative analyses with malignant serousovarian cancer, the most common histopathologicalsubtype of EOC. We also relate our findings to geneexpression profiles observed in a parallel analysisof the four EOC cell lines (TOV-21G, TOV-81D,TOV-112D, and OV-90) and those modulated as aconsequence of chromosome 3 fragment transferexperiments described in a recent study [17]. Tobegin to address possible mechanisms of inactiva-tion of the underexpressed genes, gene expressionwas analyzed in the context of LOH in the topcandidates.

MATERIALS AND METHODS

Primary Cultures of NOSE Samples and EOC Cell Lines

Primary cultures were derived from NOSEcells from the ovaries of 14 participants with noprior history of ovarian cancer, following prophy-lactic oophorectomy at the Centre de Recherchedu Centre Hospitalier de l’Universite de Montreal(CHUM) Hopital Notre-Dame, and were establishedas described previously [44,45]. EOC cell lineswere derived from a stage IIIc/grade 1–2 papillaryserous adenocarcinoma (TOV-81D), a stage III/grade3 clear cell carcinoma (TOV-21G), a stage IIIc/grade3 endometrioid carcinoma (TOV-112D), and fromthe ascites fluid of a stage IIIc/grade 3 adenocarci-noma (OV-90), all from chemotherapy naıve patients,as described [39]. Cells were cultured in OSE mediumconsisting of 50:50 medium 199:105 (Sigma), with2.5 mg/ml amphotericin B and 50 mg/ml gentamicin[39]. Culture media was supplemented with 15%fetal bovine serum (FBS) for the NOSE cultures and10% FBS for the EOC cell lines.

Ovarian Cancer Samples

Malignant serous tumor (TOV) samples andpatient-matched peripheral blood lymphocytes wereobtained with informed consent from participantsundergoing surgeries performed at CHUM HopitalNotre-Dame, as described previously (Supplemen-tary Table 1) [46]. Tumor grade, histopathologicalsubtype, and disease stage at diagnosis were assignedaccording to the criteria established by the Interna-tional Federation of Gynecology and Obstetrics(FIGO).

Nucleic Acid Extraction

DNA was extracted from EOC cell lines, TOVsamples, and peripheral blood lymphocytes asdescribed previously [45]. Total RNA was extractedwith TRIzolTM reagent (Gibco/BRL, Life Technolo-gies, Inc., Grand Island, NY) from the primarycultures of NOSE samples and the EOC cell lines,grown to 80% confluency in 100 mm petri dishes,and from frozen TOV samples as described previ-ously [43]. The quality of the RNA used for the

SEROUS OVARIAN CANCERS 57

Molecular Carcinogenesis DOI 10.1002/mc

expression microarray analysis was assessed by gelelectrophoresis and with the 2100 Bioanalyzer, usingthe RNA 6000 Nano LabChip kit (Agilent Technol-ogies, Waldbronn, Germany).

Expression Microarray Analysis

Microarray expression analysis was performedusing the Affymetrix GeneChip1 U133A array(Santa Clara, CA). Hybridization and scanning wereconducted at the McGill University and GenomeQuebec Innovation Centre (genomequebec.mcgill.ca). Total RNA was used to prepare biotinyla-ted hybridization target, and hybridization wasperformed as described (genomequebec.mcgill.ca).Microarray experiments were performed once foreach sample.

Gene expression levels were determined fromscanned images using MAS5 software (Affymetrix1

Microarray Suite). The software generates an averagedifference ratio (or raw expression value) fromthe hybridization signals obtained from the 22 probepairs that comprise a probe set. It also generatesa reliability score for each probe set, which reflectsthe level of non-specific binding by comparingthe hybridization signals of the 11 perfect-matchprobes to those of the 11 probes that contain amismatched nucleotide. A high reliability score ofPresent (P call) represents a minimal hybridization tothe mismatch probe set and consistent hybridizationacross all matched probes, where less obvious differ-ences return a borderline score of Marginal (M call) ora low score of Ambiguous (A call). In order toeliminate systematic biases when comparing theexpression values from independently generateddata sets, the raw data was normalized by multi-plying the value for an individual probe set(n¼ 22,216) by 100 and dividing by the mean ofthe raw expression values for the given sample dataset, as described [41,43,47].

Probe sets corresponding to chromosome 3 geneswere identified using the Affymetrix NetAffx BatchQuery tool (affymetrix.com/analysis/index.affx) andthe UniGene Homo sapiens May 2005 database, basedon UniGene Build 184 (www.ncbi.nlm.nih.gov/UniGene/). Additional mapping information wasobtained from University of California Santa Cruz(UCSC) Human Genome Browser database, March2006 (NCBI Build 36.1) assembly (genome.ucsc.edu).Based on these databases, 1,147 probe sets wereidentified that mapped to chromosome 3 genes andESTs. After normalization, the expression valuesbelow 15 were considered to be technical noise andwere reassigned a threshold value of 15, based on themean expression value of the A-calls, in order toavoid overestimating the gene expression differencesthat result from the high variability of low expressionvalues. Probe sets containing A-calls across allsamples were removed from further analysis, as theseexpression values may reflect either absent expres-

sion or poorly designed probe sets. Differentiallyexpressed genes were defined as those which exhib-ited at least a threefold difference in expression in aTOV sample relative to the mean of expression valuesof NOSE samples, where the expression value also felloutside the maximum or minimum range of expres-sion of the 17 NOSE samples.

Gene expression data was also analyzed usingBioconductor, an open-source software libraryfor the analyses of genomic data [48] based on R,a language and environment for statistical comput-ing and graphics (www.r-project.org). Modifiedt tests were performed with Bioconductor’s limmapackage in order to determine the significance ofthe differential expression, and P values from theresulting comparison were adjusted for multipletesting according to the method of Benjaminiand Hochberg [49]. This method controls the falsediscovery rate, which was set to 0.05 in this analysis.Bioconductor’s genefilter package was used to filterout expression results from probe sets with insuffi-cient variation in gene expression across all testedsamples. In the remaining expression values, a logbase 2 scale of at least 0.5 for the interquartile rangewas required across all tested samples as described[17]. Hierarchical clustering analysis was performedwith R’s cluster package, using the Pearson correla-tion distance.

RT-PCR Analyses

The microarray expression results were validatedby semi-quantitative RT-PCR as previously described[50]. Products were electrophoresed on 1.2% agarosegels, visualized by ethidium bromide staining andcompared with the intensity of 18S cDNA. Primersequences for the analysis of RIS1, HEG1, and 18Sexpression were designed based on the 3’ untrans-lated region of the gene using the Primer3 designsoftware (www.broad.mit.edu/cgi-bin/primer/primer3.cgi/primer3_www.cgi). Primer sequences for the anal-ysis of GBE1 expression were designed to encompassexons 8–12 [17]. Primer sequences and annealing tem-peratures can be found in Supplementary Table 2.

LOH Analysis

A PCR-based assay was used to allelotype sixpolymorphic microsatellite markers. Two markerscorresponded to each of the RIS1 (D3S3624 andD3S3582), GBE1 (D3S1538 and D3S3508) and HEG1(D3S1551 and D3S1269) loci (see SupplementaryTable 2). The genetic markers are described in TheGDB Human Genome Data Base (www.gdb.org). Thepairs of markers that flank RIS1 and HEG1 are bothless than 1 Mb apart. The markers flanking GBE1encompass a 3.5 Mb region, which contains noother known genes or ESTs. The PCR analysis wasperformed using S35dATP radiolabel and annealingtemperatures described in Supplementary Table 2, asreported previously [46,50].

58 BIRCH ET AL.


Allelotyping was performed for DNA samples froma series of 66 malignant serous tumors and matchednormal peripheral blood leukocytes (SupplementaryTable 1). Due to restrictions in DNA availability, only6 (TOV-800EPT, TOV-1007EPT, TOV-1108DT, TOV-1127GT, TOV-1142DT, TOV-1247T) of the 11 tumorsamples examined by both RT-PCR and expressionmicroarray analyses were tested for LOH. LOH wasscored based on the absence or a difference in therelative intensity of signals representing alleles whencomparing products from tumor and matchednormal DNA samples. All samples positive for LOHat individual loci were analyzed twice in independ-ent assays. LOH frequency was established based onallelotyping analyses of the series of 66 malignantserous tumors samples.

RESULTS

Description of Chromosome 3 Gene Coverage

Chromosome 3 spans 199 Mb and representsabout 6.5% of the total DNA in the humangenome (www.ncbi.nlm.nih.gov). Based on theNCBI Build 36.2, this metacentric chromosome isestimated to contain 666 chromosome 3p and803 chromosome 3q genes, although this numberalso includes pseudogenes, microRNAs, and othernon-annotated hypothetical genes. Of these 1,469genes, 1,013 have been annotated as UCSC KnownGenes, corresponding to known coding genes basedon protein data from UniProt and mRNA defined bythe NCBI reference sequence (RefSeq) collection andGenBank (http://genome.ucsc.edu). Using availablegenome databases, expression data from 1,147 probesets, representing 735 chromosome 3 genes andESTs, was extracted from the U133A microarray

analyses of 14 primary cultures of NOSE samples,17 TOV samples and four EOC cell lines (Table 1). Ofthese, 691 were either annotated as Known Genes orRefSeq genes, resulting in the representation ofabout two thirds (68.2%) of the chromosome 3Known and RefSeq genes on the Affymetrix U133AGeneChip1. Based on mapping information, 535probe sets represented 350 3p genes and ESTs, 336 ofwhich are Known or RefSeq 3p genes (71.3% of the471 Known or RefSeq 3p genes). On the 3q arm, 612probe sets represented 385 genes and ESTs, 355 ofwhich are Known or RefSeq 3q genes (65.5% of the542 Known or RefSeq 3q genes). In general, genesrepresented by probe sets that contained at least oneP call were distributed at least once every 4.5 Mb,with the exception of a 10 Mb region encompassingthe centromere, where gene density is relativelysparse.

Two-Way Comparative Analyses of NOSEand TOV Samples

Two-way comparative analyses of MAS5 treateddatasets were performed on expression values gen-erated by the microarray analysis of the 14 NOSE and17 TOV samples. There were 122 genes that exhibitedexpression values with low reliability scores in all ofthe tested samples (Table 1). About 77% of the probesets (n¼883, representing 613 genes) gave expres-sion values with at least one high reliability score inthe samples tested, and these expression values wereused in two-way comparative analyses (Table 1). Inthis group, there were 303 genes that exhibited athreefold (or greater) difference in expression in atleast one TOV sample relative to the mean expressionvalue of the NOSE samples. Within this groupof differentially expressed genes were ATP13A3,

Table 1. Gene Representation and Differential Expression of U133A Microarray Analysis of NOSE and TOV Samples

Description

Number of genes (Probe sets)

All 3p 3q

Genes on chromosome 3 (NCBI) 1,469 666 803Known or RefSeq genes (UCSC) 1,013 471 542Genes or ESTs (probes sets) represented on U133A

array735 (1,147) 350 (535) 385 (612)

Known or RefSeq represented on U133A array 691 (1,079) 336 (504) 355 (575)Genes with at least one P call in the NOSE or TOV

samples613 (883) 292 (402) 321 (481)

Threefold differences in gene expression 303 (379) 128 (151) 175 (228)Threefold differences in genes expression occurring

outside range of NOSE samples in any TOVsample

278 (342) 120 (141) 158 (201)

Threefold differences in gene expression andoutside range of NOSE samples in 50% of TOVsamples

43 (52) 16 (18) 27 (34)

Threefold difference in gene expression and outsiderange of NOSE samples in 100% of TOV samples

3 (4) 2 (2) 1 (2)



CLDN11, COL8A1, DCBLD2, GBE1, HEG1, LXN, andRIS1, all of which exhibited a threefold difference ofexpression in all 17 TOV samples. Interestingly, all ofthese genes exhibited underexpression relative to theNOSE samples. Of these 303 genes displaying at leastone 3-fold difference of expression in the TOVsamples, 278 genes (169 overexpressed genes and116 underexpressed genes) exhibited expressionvalues in any of the TOV samples that also felloutside either the maximum or minimum value ofexpression of NOSE samples (Supplementary Table3). Using both of these criteria to define differentialexpression, there were 43 genes (22 overexpressedand 21 underexpressed genes) that exhibited differ-ential expression in at least 50% of TOV samples.Showing differential expression in all 17 TOV

samples are three of the eight genes previouslymentioned (RIS1, GBE1, and HEG1), all of whichwere underexpressed (Table 2). Although the other40 genes in this category show differential expres-sion in fewer than 17 TOV samples, many exhibitedfold changes in the same direction across all TOVsamples, including those that showed less than athreefold difference from the mean of the NOSEsamples. Fourteen of the 22 overexpressed genes(63.6%) and 16 of the 21 underexpressed genes(76.2%) exhibited changes in a consistent directionacross all TOV samples, regardless of whether theyfulfilled the criteria for differential expression(Table 2). In the 278 genes displaying differentialexpression in at least one TOV sample, the patterns offold change among the non-differentially expressed

Table 2. Genes Differentially Expressed at Least Threefold in at Least 50% of Serous TOV Samples

60 BIRCH ET AL.


samples were less consistent. Thirty-four of 169overexpressed genes (20.1%) and 50 of 116 under-expressed genes (43.1%) showed consistently pos-itive or negative fold changes across all 17 TOVsamples.

Hierarchal cluster analysis of the expression valuesrepresenting the 278 differentially expressed genesplaced the NOSE and TOV samples into two separatebranches in the resulting dendrogram (Supplemen-tary Fig. 1a). Interestingly, this cluster analysisplaced TOV-81D, the non-tumorigenic EOC cell line,with the NOSE samples, whereas the tumorigenicEOC cell lines (TOV-21G, TOV-112D and OV-90)were placed within the branch that contained theTOV samples. Although this analysis was generatedusing a filtered list which contained differentiallyexpressed genes, the segregation of NOSE from TOVsamples is also observed in a hierarchal clusteranalysis of all chromosome 3 probe sets displayingat least one P call (n¼ 883) (Supplementary Fig. 1b).

A modified t-test based on R programming lan-guage was used to evaluate the significance ofexpression values of the 613 genes with at leastone high reliability score. This analysis identified 330genes (188 overexpressed and 142 underexpressedgenes in TOV samples) with significantly differentexpression values (P<0.05) between the TOV andthe NOSE samples (Supplementary Table 4). Acomparison of both the fold-difference comparisonand this statistical method identified 210 differ-entially expressed genes in common (SupplementaryTable 4). All of the genes showing differentialexpression in at least 25% of the TOV samples wereidentified by the statistical method, and RIS1, GBE1,and HEG1 ranked as highly significantly different inexpression (P<1.73E�11) between the NOSE andTOV samples.

RT-PCR Analyses of RIS1, GBE1, and HEG1

RT-PCR analyses of RIS1, GBE1, and HEG1 inprimary cultures of NOSE and TOV samples wereperformed to further characterize the expression ofthe top candidates identified by expression micro-array analyses. The expression levels detectable bythe semi-quantitative RT-PCR method closely reflectthe gene expression levels reported from the expres-sion microarray analyses (Fig. 1). Also included in thecomparative RT-PCR analyses are the EOC cell lines.Notable is the strong expression of all three genes inthe non-tumorigenic cell line, TOV-81D, relative tothose of the tumorigenic cell lines, TOV-21G, TOV-112D, and OV-90. This observation is consistent withexpression microarray analyses of these genes ascompared with the NOSE samples (Fig. 2).

LOH Analysis of RIS1, GBE1, and HEG1 Loci

Genes exhibiting decreased expression could becandidate TSGs. As LOH has often been observed asan initial event in the inactivation of classical TSGs,

LOH frequencies of RIS1, GBE1, and HEG1 wereinvestigated. Two polymorphic microsatellite repeatmarkers representing each locus were analyzed forLOH in a subset of the TOV samples examined byexpression microarray analyses, the EOC cell lines,and a series of 66 independently ascertained serousmalignant ovarian cancer samples. The majority ofthe microarrayed samples exhibited no evidence ofLOH at any of the tested loci (Fig. 1). The LOH of OV-90 was observed for RIS1 and GBE1 but not withHEG1, consistent with previous reports showingreduction to hemizygosity of the entire 3p arm [17](Fig. 1). These data suggest that the decreasedexpression may not be associated with LOH ofloci harboring the differentially expressed genes.The LOH frequencies for RIS1, GBE1, and HEG1 lociin the series of 66 malignant ovarian cancer sampleswere 14.3%, 13.7%, and 9.2%, respectively. We haveobserved frequencies of LOH as high as 39% for agiven marker on the 3p arm in unselected malignantovarian cancer samples [13]. Classical TSGs display-ing loss of expression are often biallelically inacti-vated, with LOH occurring in almost every case. Thelow frequency of LOH observed at these loci suggeststhat these genes are not primary targets of inactiva-tion.

DISCUSSION

Expression microarray analyses have identified anumber of differentially expressed genes in malig-nant serous ovarian cancer samples that differ intheir expression profiles from primary cultures ofNOSE samples. The biological significance of thesefindings is evident in the observation thatthe hierarchical cluster analyses not only groupedthe TOV samples separately from the NOSE samples,but that the non-tumorigenic EOC cell line wasgrouped with the NOSE samples and the tumorigenicEOC cell lines generally fell in the same major branchas the TOV samples. These results point to tran-scriptome changes associated with malignant andtumorigenic phenotypes. This is evident by GeneOntology (GO) analysis of the differentiallyexpressed genes, which indicated that these genesare associated with various cellular functions such asapoptosis, cell proliferation, regulation of growth,the cell cycle, cellular adhesion or signal trans-duction, all general processes known to be alteredin tumor formation. A review of the literature hasalso revealed that at least 41 of the differentiallyexpressed genes have been previously studied in thecontext of ovarian cancer, and 137 have beenimplicated in a general cancer context (Supplemen-tary Table 3). These genes include underexpressedputative TSGs such as RASSF1A, RAR-b [21,28,51,52]and DOC [53], and overexpressed genes, such CP,TRAIL, and RHOA [54–59] and EVI-1 [34,60]. Fur-thermore, some of the differentially expressed genesidentified in the present study were also found to be



differentially expressed between serous LMP andmalignant TOV samples in an independent Affyme-trix GeneChip1 analysis from our group [61]. Suchgenes include those that exhibit overexpressionin the malignant tumors, such as ATR, SEMA3B,PRKCD, TM4SF1, and UMPS, and those that displayreduced expression in the malignant tumors, namelySMARCC1 and PDIR [61]. Although questions havebeen raised about the use of different models of NOSEcells incomparative analyses with malignant ovarian

cancer derived tissue samples [62], all of the exam-ples describe data consistent with that observed inthe present study, suggesting that primary cultures ofNOSE samples could reveal transcriptome patternsassociated with malignant disease. Taken together,these observations imply that some of the differ-entially expressed genes may be also be associatedwith tumorigenic potential and warrant furtherinvestigation.

The roles of RIS1, GBE1, and HEG1, which were allsignificantly underexpressed in TOV samples, havenot yet been elucidated in ovarian cancer. RIS1 (at3p21.31), which codes for the transmembraneprotein 158 (TMEM158), was identified as a geneupregulated during Ras-induced senescencein human diploid fibroblasts transduced with aRasV12-containing retrovirus [63]. RIS1 expressionwas observed upregulated in a non-small cell lungcancer line after transfection of the gene TLSC1 andsubsequent suppression of tumorigenicity [64].Although RIS1 has been shown to exhibit a highfrequency of LOH in breast tumors, LOH did notcorrelate with gene expression, and deregulation ofgene expression was not attributable to alteration ofpromoter activity by methylation [65]. In a geneexpression analysis of ductal carcinoma in situ ofbreast cancer, RIS1 expression was completely absentin tumor cells and yet present in the immediatelyadjacent stromal cells, leading the authorsto hypothesize that the characteristics of the proteinthat plays a role in senescence may also cause it to beemitted by stromal cells as an anti-proliferativesignal [66]. Overexpression of RIS1 was observed inWilms tumors harboring somatic CTNNB1 muta-tions, suggesting an association between the RAS andWNT signaling pathways [67]. RIS1 has recently beenidentified as a target in the mutator pathway ofcolorectal cancers exhibiting microsatellite instabil-ity, where mutations have been observed in the CGN

Figure 1. Semi-quantitative RT-PCR and LOH analysis of RIS1, GBE1, and HEG1. RT-PCR analysis of NOSEsamples, TOV samples and the EOC cell lines. The expression of 18S cDNA is also shown for comparison. Results ofLOH analysis of TOV samples and EOC cell lines is shown, where LOH status is indicated as positive (þ), negative(�), not informative (NI) or not done (ND).

Figure 2. Expression microarray analysis of RIS1, GBE1, and HEG1in NOSE samples and the EOC cell lines. Graphical representations ofnormalized U133A GeneChip1 expression values generated byMAS5 treated data are shown for the NOSE samples, the TOVsamples and both the tumorigenic EOC cell lines (TOV-112D, TOV-21G, and OV-90) and the non-tumorigenic EOC cell line (TOV-81D).Supplementary Figure 1. Hierarchical clustering analysis of NOSE andTOV samples and EOC cell lines using differentially expressed probesets. The analysis was performed on the 342 probe sets (representing278 genes) exhibiting differential expression in two-way analysis ofNOSE (n¼14) and TOV (n¼17) samples, using R’s cluster packagewith the Pearson correlation distance (1-correlation) (Panel A). Theanalysis was performed on the 883 probe sets (representing 613genes) corresponding to chromosome 3 that displayed at least one Pcall, regardless of differential expression between the NOSE and TOVsamples. The analysis was performed using R’s cluster package withthe Pearson correlation distance (1-correlation) (Panel B).

62 BIRCH ET AL.


codon repeat tract that encodes 14 alanines [68]. Inorder to deduce the molecular function of the gene,a homozygous RIS1-null mouse knockout was cre-ated recently [69]. Although there were no overtchanges in embryonic development, adult physiol-ogy, or frequency of tumor formation, followingchemical induction, further studies are warranted asa modest (although not statistically significant)increase in tumor size was observed in the RIS1-nullmice.

GBE1 (at 3p12.2) and HEG1 (at 3q21.2) have notbeen previously studied in the context of humancancers. GBE1 (at 3p12.2) codes for glucan (1,4-alpha-), branching enzyme 1 and has been associatedwith glycogen storage disease type IV. Recently,overexpression of GBE1 was noted in breast cancercell lines displaying full or partial promoter methyl-ation of CDH1, which codes for E-cadherin, whencompared to breast cancers with wildtype or non-sense mutations in this gene [70]. HEG1 (at 3q21.2) isa homolog of the zebrafish heart of glass gene thatencodes a protein with calcium-binding domainswhich is suspected to regulate concentric heartgrowth [71]. Interestingly, the human gene bearssimilarity to mucin 13, encoded by the adjacentgene, MUC13 (at 3q21.2), and contains similartransmembrane and epidermal growth factor-likedomains [72]. Mucins, heavily glycosylated proteinsthat are either secreted or found in the surfacemembranes of epithelial cells, play an important rolein cancers through aberrant signaling and celladhesion, as well as through their protection of thetumor cells [73]. In contrast to our findings withHEG1, mucins such as mucins 1, 4, and 16 (the latterwhich is also known as CA-125 and used a serummarker for monitoring recurrence of ovarian cancer),are commonly overexpressed in ovarian tumors[74,75]. While HEG1 has not been studiedin humans, it was recently found to be upregulatedwhen STAT3 is constitutively expressed in lungepithelial cells, indicating a key role in both tumordevelopment and wound healing pathways [76].

Interestingly, RIS1, GBE1, and HEG1 are differ-entially expressed in the EOC cell lines. The lowestlevels of expression occurred in all three tumorigenicEOC cell lines (TOV-21G, TOV-112D, and OV-90). Incontrast, the non-tumorigenic EOC cell line (TOV-81D) displayed expression levels in the range of theNOSE samples, indicating the gene expressionpattern of TOV-81D is most consistent with theprimary cultures of NOSE samples. Low or absentgene expression was not concordant with LOH inthe tumor samples tested or in the EOC celllines. However, RIS1 expression was shown to beupregulated in derivatives of OV-90 EOC cellline hybrids rendered non-tumorigenic throughtransfer of normal chromosome 3 fragments [17].The upregulation was due to transcriptional mod-ification of the endogenous RIS1, as none of the non-

tumorigenic OV-90 hybrids acquired copies of RIS1through chromosome 3 fragment transfer [17]. Incontrast to RIS1, GBE1 was transferred within a 3p12-pcen fragment in the derivation of all non-tumori-genic OV-90 hybrids [17]. Gene expression wasdetectable by expression microarray analyses in allnon-tumorigenic OV-90 hybrids, although it was notdifferentially expressed in all hybrids relative to theparental OV-90 tumorigenic cell line. HEG1 was nottransferred, nor was its expression altered, in any ofthe non-tumorigenic OV-90 hybrids [17]. Thesecombined results suggest that RIS1, GBE1, andHEG1 may play roles in ovarian cancer, but areunlikely targeted for inactivation as observed withclassical TSGs [77] and are instead transcriptionallymodified by an upstream factor.

In addition to RIS1, a number of genes shown hereto be differentially expressed in at least one tumorsample were also shown deregulated in OV-90 EOCcell line hybrids rendered non-tumorigenic throughtransfer of normal chromosome 3 fragments [17]. Thegenes RAFTLIN and FSTL1, found in this study to beunder-expressed in TOV samples relative to NOSEsamples, were upregulated in hybrids of OV-90 wheretumorigenicity was lost. Additionally, several genesshowing overexpression in TOV samples relative toNOSE samples, namely CP, EVI-1, HGD, CSTA, ZIC1,and BCHE, were expressed at reduced levels inthe hybrids displaying loss of tumorigenicity. Thesegenes warrant further investigation in ovarian cancer,as their differential expression was associated withalteration in tumorigenic phenotype.

The present study has identified chromosome 3genes differentially expressed in malignant ovariancancer. Although some of these differentiallyexpressed genes have been previously identified inthe context of ovarian cancer, this study has identi-fied a number of genes that are deserving of futureresearch. Relating gene expression profiles with anin vitro model of ovarian cancer defined by four wellcharacterized EOC lines that are distinguishable bytheir tumorigenic potential, and a recently modifiedEOC cell line model (OV-90), provided a frameworkfor further prioritizing these genes for furtheranalyses.

ACKNOWLEDGMENTS

We thank Suzanna Arcand, Zhen Shen, LisePortelance, Manon de Ladurantaye, Marise Roy andStephanie Lepage, for their helpful expertise. A.H.B.is a recipient of a graduate scholarship from theDepartment of Medicine and the Research Instituteof the McGill University Health Centre. The ovariantumor bank was supported by the Banque de tissus etde donnees of the Reseau de recherche sur le cancerof the Fonds de la Recherche en Sante du Quebec.This work was supported by grants from the GenomeCanada and Quebec and Canadian Institutesof Health Research to P.N.T, D.P. and A.-M.M.-M.



REFERENCES

1. Bernardini M, Weberpals J, Squire JA. The use of cytogeneticsin understanding ovarian cancer. Biomed Pharmacother2004;58:17–23.

2. Mertens F, Johansson B, Hoglund M, Mitelman F. Chromo-somal imbalance maps of malignant solid tumors: Acytogenetic survey of 3185 neoplasms. Cancer Res 1997;57:2765–2780.

3. Pejovic T, Heim S, Mandahl N, et al. Chromosome aberra-tions in 35 primary ovarian carcinomas. Genes Chromo-somes Cancer 1992;4:58–68.

4. Whang-PengJ,KnutsenT,DouglassEC,etal.Cytogeneticstudiesin ovarian cancer. Cancer Genet Cytogenet 1984;11:91–106.

5. Yonescu R, Currie JL, Hedrick L, Campbell J, Griffin CA.Chromosome abnormalities in primary endometrioid ovariancarcinoma. Cancer Genet Cytogenet 1996;87:167–171.

6. Hoglund M, Gisselsson D, Mandahl N, et al. Multivariateanalyses of genomic imbalances in solid tumors revealdistinct and converging pathways of karyotypic evolution.Genes Chromosomes Cancer 2001;31:156–171.

7. Simon R, Desper R, Papadimitriou CH, et al. Chromosomeabnormalities in ovarian adenocarcinoma: III. Using break-point data to infer and test mathematical models foroncogenesis. Genes Chromosomes Cancer 2000;28:106–120.

8. Taetle R, Aickin M, Panda L, et al. Chromosome abnormal-ities in ovarian adenocarcinoma: II. Prognostic impact ofnonrandom chromosome abnormalities in 244 cases. GenesChromosomes Cancer 1999;25:46–52.

9. Taetle R, Aickin M, Yang JM, et al. Chromosome abnormal-ities in ovarian adenocarcinoma: I. Nonrandom chromosomeabnormalities from 244 cases. Genes Chromosomes Cancer1999;25:290–300.

10. Dodson MK, Hartmann LC, Cliby WA, et al. Comparison ofloss of heterozygosity patterns in invasive low-gradeand high-grade epithelial ovarian carcinomas. Cancer Res1993;53:4456–4460.

11. Ehlen T, Dubeau L. Loss of heterozygosity on chromosomalsegments 3p, 6q and 11p in human ovarian carcinomas.Oncogene 1990;5:219–223.

12. Fullwood P, Marchini S, Rader JS, et al. Detailed genetic andphysical mapping of tumor suppressor loci on chromosome3p in ovarian cancer. Cancer Res 1999;59:4662–4667.

13. Lounis H, Mes-Masson AM, Dion F, et al. Mapping ofchromosome 3p deletions in human epithelial ovariantumors. Oncogene 1998;17:2359–2365.

14. Petursdottir TE, Thorsteinsdottir U, Jonasson JG, et al.Interstitial deletions including chromosome 3 commoneliminated region 1 (C3CER1) prevail in human solid tumorsfrom 10 different tissues. Genes Chromosomes Cancer2004;41:232–242.

15. Zheng JP, Robinson WR, Ehlen T, Yu MC, Dubeau L.Distinction of low grade from high grade human ovariancarcinomas on the basis of losses of heterozygosity onchromosomes 3, 6,and 11 and HER-2/neu gene amplifica-tion. Cancer Res 1991;51:4045–4051.

16. Rimessi P, Gualandi F, Morelli C, et al. Transfer of humanchromosome 3 to an ovarian carcinoma cell line identifiesthree regions on 3p involved in ovarian cancer. Oncogene1994;9:3467–3474.

17. Cody NA, Ouellet V, Manderson EN, et al. Transfer ofchromosome 3 fragments suppresses tumorigenicity of anovarian cancer cell line monoallelic for chromosome 3p.Oncogene 2007;26:618–632.

18. Tse C, Xiang RH, Bracht T, Naylor SL. Human Semaphorin 3B(SEMA3B) located at chromosome 3p21.3 suppresses tumorformation in an adenocarcinoma cell line. Cancer Res2002;62:542–546.

19. Xiang R, Davalos AR, Hensel CH, Zhou XJ, Tse C. Naylor SL.Semaphorin 3F gene from human 3p21. 3 suppresses tumorformation in nude mice. Cancer Res 2002;62:2637–2643.

20. Wang L, Darling J, Zhang JS, et al. Loss of expression ofthe DRR 1 gene at chromosomal segment 3p21.1 in renalcell carcinoma. Genes Chromosomes Cancer 2000;27:1–10.

21. Agathanggelou A, Honorio S, Macartney DP, et al. Methyl-ation associated inactivation of RASSF1A from region 3p21.3in lung, breast and ovarian tumours. Oncogene 2001;20:1509–1518.

22. Imura M, Yamashita S, Cai LY, et al. Methylation andexpression analysis of 15 genes and three normally-methylated genes in 13 Ovarian cancer cell lines. CancerLett 2006;241:213–220.

23. Kwong J, Lee JY, Wong KK, et al. Candidate tumor-suppressor gene DLEC1 is frequently downregulatedby promoter hypermethylation and histone hypoacetylationin human epithelial ovarian cancer. Neoplasia 2006;8:268–278.

24. Dallol A, Forgacs E, Martinez A, et al. Tumour specificpromoter region methylation of the human homologue ofthe Drosophila Roundabout gene DUTT1 (ROBO1) in humancancers. Oncogene 2002;21:3020–3028.

25. Huebner K. Tumor suppressors on 3p: A neoclassic quartet.Proc Natl Acad Sci USA 2001;98:14763–14765.

26. Imreh S, Klein G, Zabarovsky ER. Search for unknown tumor-antagonizing genes. Genes Chromosomes Cancer2003;38:307–321.

27. Xian J, Clark KJ, Fordham R, Pannell R, Rabbitts TH, RabbittsPH. Inadequate lung development and bronchial hyperplasiain mice with a targeted deletion in the Dutt1/Robo1 gene.Proc Natl Acad Sci USA 2001;98:15062–15066.

28. Zabarovsky ER, Lerman MI, Minna JD. Tumor suppressorgenes on chromosome 3p involved in the pathogenesisof lung and other cancers. Oncogene 2002;21:6915–6935.

29. Dent J, Hall GD, Wilkinson N, et al. Cytogenetic alterations inovarian clear cell carcinoma detected by comparativegenomic hybridisation. Br J Cancer 2003;88:1578–1583.

30. Kiechle M, Jacobsen A, Schwarz-Boeger U, Hedderich J,Pfisterer J, Arnold N. Comparative genomic hybridizationdetects genetic imbalances in primary ovarian carcinomasas correlated with grade of differentiation. Cancer 2001;91:534–540.

31. Patael-Karasik Y, Daniely M, Gotlieb WH, et al. Comparativegenomic hybridization in inherited and sporadic ovariantumors in Israel. Cancer Genet Cytogenet 2000;121:26–32.

32. Staebler A, Heselmeyer-Haddad K, Bell K, et al. Micro-papillary serous carcinoma of the ovary has distinct patternsof chromosomal imbalances by comparative geno-mic hybridization compared with atypical proliferativeserous tumors and serous carcinomas. Hum Pathol 2002;33:47–59.

33. Shayesteh L, Lu Y, Kuo WL, et al. PIK3CA is implicated as anoncogene in ovarian cancer. Nat Genet 1999;21:99–102.

34. Sunde JS, Donninger H, Wu K, et al. Expression profilingidentifies altered expression of genes that contribute to theinhibition of transforming growth factor-{beta} signaling inovarian cancer. Cancer Res 2006;66:8404–8412.

35. Guan XY, Fung JM, Ma NF, et al. Oncogenic role of eIF-5A2 inthe development of ovarian cancer. Cancer Res 2004;64:4197–4200.

36. Sugita M, Tanaka N, Davidson S, et al. Molecular definition ofa small amplification domain within 3q26 in tumors of cervix,ovary, and lung. Cancer Genet Cytogenet 2000;117:9–18.

37. Park TW, Riethdorf S, Riethdorf L, Loning T, Janicke F.Differential telomerase activity, expression of the telomerasecatalytic sub-unit and telomerase-RNA in ovarian tumors. IntJ Cancer 1999;84:426–431.

38. Manderson EN, Mes-Masson AM, Novak J, et al. Expressionprofiles of 290 ESTs mapped to chromosome 3 in humanepithelial ovarian cancer cell lines using DNA expressionoligonucleotide microarrays. Genome Res 2002;12:112–121.

64 BIRCH ET AL.


39. Provencher DM, Lounis H, Champoux L, et al. Character-ization of four novel epithelial ovarian cancer cell lines. InVitro Cell Dev Biol Anim 2000;36:357–361.

40. Tonin PN, Hudson TJ, Rodier F, et al. Microarray analysis ofgene expression mirrors the biology of an ovarian cancermodel. Oncogene 2001;20:6617–6626.

41. Arcand SL, Mes-Masson AM, Provencher D, Hudson TJ,Tonin PN. Gene expression microarray analysis and genomedatabases facilitate the characterization of a chromosome 22derived homogeneously staining region. Mol Carcinog 2004;41:17–38.

42. Benoit MH, Hudson TJ, Maire G, et al. Global analysis ofchromosome X gene expression in primary cultures ofnormal ovarian surface epithelial cells and epithelial ovariancancer cell lines. Int J Oncol 2007;30:5–17.

43. Presneau N, Mes-Masson AM, Ge B, Provencher D, HudsonTJ, Tonin PN. Patterns of expression of chromosome 17 genesin primary cultures of normal ovarian surface epithelia andepithelial ovarian cancer cell lines. Oncogene 2003;22:1568–1579.

44. Kruk PA, Maines-Bandiera SL, Auersperg N. A simplifiedmethod to culture human ovarian surface epithelium. LabInvest 1990;63:132–136.

45. Lounis H, Provencher D, Godbout C, Fink D, Milot MJ, Mes-Masson AM. Primary cultures of normal and tumoral humanovarian epithelium: A powerful tool for basic molecularstudies. Exp Cell Res 1994;215:303–309.

46. Dion F, Mes-Masson AM, Seymour RJ, Provencher D, ToninPN. Allelotyping defines minimal imbalance at chromosomalregion 17q25 in non-serous epithelial ovarian cancers.Oncogene 2000;19:1466–1472.

47. Novak JP, Sladek R, Hudson TJ. Characterization of variabilityin large-scale gene expression data: Implications for studydesign. Genomics 2002;79:104–113.

48. Gentleman RC, Carey VJ, Bates DM, et al. Bioconductor:Open software development for computational biology andbioinformatics. Genome Biol 2004;5:R80.

49. Benjamini Y, Yekutieli D. Quantitative trait Loci analysis usingthe false discovery rate. Genetics 2005;171:783–790.

50. Presneau N, Dewar K, Forgetta V, Provencher D, Mes-Masson AM, Tonin PN. Loss of heterozygosity and tran-scriptome analyses of a 1.2 Mb candidate ovarian cancertumor suppressor locus region at 17q25.1-q25.2. MolCarcinog 2005;43:141–154.

51. Grunt TW, Puckmair K, Tomek K, Kainz B, Gaiger A. An EGFreceptor inhibitor induces RAR-beta expression in breast andovarian cancer cells. Biochem Biophys Res Commun 2005;329:1253–1259.

52. Makarla PB, Saboorian MH, Ashfaq R, et al. Promoterhypermethylation profile of ovarian epithelial neoplasms.Clin Cancer Res 2005;11:5365–5369.

53. Mok SC, Wong KK, Chan RK, et al. Molecular cloning ofdifferentially expressed genes in human epithelial ovariancancer. Gynecol Oncol 1994;52:247–252.

54. Chan A, Wong F, Arumanayagam M. Serum ultrafiltrablecopper, total copper and caeruloplasmin concentrations ingynaecological carcinomas. Ann Clin Biochem 1993;30:545–549.

55. Hough CD, Cho KR, Zonderman AB, Schwartz DR, Morin PJ.Coordinately up-regulated genes in ovarian cancer. CancerRes 2001;61:3869–3876.

56. Lancaster JM, Sayer R, Blanchette C, et al. High expression oftumor necrosis factor-related apoptosis-inducing ligand isassociated with favorable ovarian cancer survival. ClinCancer Res 2003;9:762–766.

57. Lee CM, Lo HW, Shao RP, et al. Selective activation ofceruloplasmin promoter in ovarian tumors: Potential use forgene therapy. Cancer Res 2004;64:1788–1793.

58. Pulay T, Csomor S, Kovacs I, Szarka G, Somos P. Changes ofserum proteins in patients with malignant gynaecologicaltumours. Acta Chir Hung 1988;29:305–314.

59. Horiuchi A, Imai T, Wang C, et al. Up-regulation of smallGTPases, RhoA and RhoC, is associated with tumorprogression in ovarian carcinoma. Lab Invest 2003;83:861–870.

60. Brooks DJ, Woodward S, Thompson FH, et al. Expression ofthe zinc finger gene EVI-1 in ovarian and other cancers. Br JCancer 1996;74:1518–1525.

61. Ouellet V, Provencher DM, Maugard CM, et al. Discrim-ination between serous low malignant potential and invasiveepithelial ovarian tumors using molecular profiling. Onco-gene 2005;24:4672–4687.

62. Zorn KK, Jazaeri AA, Awtrey CS, et al. Choice of normalovarian control influences determination of differentiallyexpressed genes in ovarian cancer expression profilingstudies. Clin Cancer Res 2003;9:4811–4818.

63. Barradas M, Gonos ES, Zebedee Z, et al. Identification of acandidate tumor-suppressor gene specifically activatedduring Ras-induced senescence. Exp Cell Res 2002;273:127–137.

64. Sussan TE, Pletcher MT, Murakami Y, Reeves RH. Tumorsuppressor in lung cancer 1 (TSLC1) alters tumorigenicgrowth properties and gene expression. Mol Cancer 2005;4:28.

65. Silva J, Silva JM, Barradas M, et al. Analysis of the candidatetumor suppressor Ris-1 in primary human breast carcinomas.Mutat Res 2006;594:78–85.

66. Verlinden I, Janssens J, Raus J, Michiels L. Microdissection andSAGE as a combined tool to reveal gene expression in ductalcarcinoma in situ of the breast. Mol Carcinog 2004;41:197–206.

67. Zirn B, Samans B, Wittmann S, et al. Target genes of theWNT/beta-catenin pathway in Wilms tumors. Genes Chro-mosomes Cancer 2006;45:565–574.

68. Iglesias D, Fernandez-Peralta AM, Nejda N, et al. RIS1, a genewith trinucleotide repeats, is a target in the mutator pathwayof colorectal carcinogenesis. Cancer Genet Cytogenet 2006;167:138–144.

69. Nieto M, Barradas M, Criado LM, Flores JM, Serrano M, LlanoE. Normal cellular senescence and cancer susceptibility inmice genetically deficient in Ras-induced senescence-1(Ris1). Oncogene 2007;26:1673–1680.

70. Lombaerts M, van Wezel T, Philippo K, et al. E-cadherintranscriptional downregulation by promoter methylation butnot mutation is related to epithelial-to-mesenchymal tran-sition in breast cancer cell lines. Br J Cancer 2006;94:661–671.

71. Mably JD, Mohideen MA, Burns CG, Chen JN, Fishman MC.Heart of glass regulates the concentric growth of the heart inzebrafish. Curr Biol 2003;13:2138–2147.

72. Lang T, Hansson GC, Samuelsson T. An inventory of mucingenes in the chicken genome shows that the mucin domainof Muc13 is encoded by multiple exons and that ovomucin ispart of a locus of related gel-forming mucins. BMC Genomics2006;7:197.

73. Hollingsworth MA, Swanson BJ. Mucins in cancer: Protectionand control of the cell surface. Nat Rev Cancer 2004;4:45–60.

74. Chauhan SC, Singh AP, Ruiz F, et al. Aberrant expression ofMUC4 in ovarian carcinoma: Diagnostic significance aloneand in combination with MUC1 and MUC16 (CA125). ModPathol 2006;19:1386–1394.

75. Taylor-Papadimitriou J, Burchell J, Miles DW. Dalziel M.MUC1 and cancer. Biochim Biophys Acta 1999;1455:301–313.

76. Haura EB, Zheng Z, Song L, Cantor A, Bepler G. Activatedepidermal growth factor receptor-Stat-3 signaling promotestumor survival in vivo in non-small cell lung cancer. ClinCancer Res 2005;11:8288–8294.

77. Presneau N, Manderson EN, Tonin PN. The quest for a tumorsuppressor gene phenotype. Curr Mol Med 2003;3:605–629.



Transcriptome analysis of serous ovarian cancers identifies differentially expressed chromosome 3 genes

Documents