Loss of Heterozygosity at Chromosome SegmentXq25-26.1 in Advanced Human Ovarian Carcinomas

Chan Choi,1 Sunghee Cho,1 Izumi Horikawa,1 Andrew Berchuck,2 Nancy Wang,3 Edward Cedrone,3Sang Woo Jhung,4 Jung Bin Lee,5 Judith Kerr,6 Georgia Chenevix-Trench,6 Seongyeon Kim,7 J. Carl Barrett,1*and Minoru Koi1*1Laboratory of Molecular Carcinogenesis, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina2Department of Obstetrics and Gynecology, Duke University School of Medicine, Durham, North Carolina3Department of Pediatrics, University of Rochester School of Medicine and Dentistry, Rochester, New York4Department of Pathology, Chonnam University Medical School, Kwangju, Korea5Department of Forensic Medicine, Seoul University Medical School, Seoul, Korea6Queensland Institute of Medical Research, Brisbane, Queensland, Australia7Department of Statistics, North Carolina State University, Raleigh, North Carolina

To determine whether a tumor suppressor gene of importance to epithelial ovarian cancer resides on the X chromosome, weexamined loss of heterozygosity (LOH) in 123 epithelial ovarian cancer cases. In 54 such cases, we examined LOH at 26 loci onthe human X chromosome. In eight cases, we examined LOH in 14 loci and in 61 cases we examined LOH in 13 loci. MatchedDNA samples from tumors and corresponding normal tissues were analyzed by polymerase chain reaction (PCR) amplificationof microsatellite markers. Frequent losses were found in epithelial carcinomas at the Xq25-26.1 region, including DXS1206(34.5% loss in informative cases), DXS1047 (27.7%), HPRT (24.1%), and DXS1062 (33.3%). The minimum overlapping region ofLOH was approximately 5 megabases (Mb), flanked by DXS1206 (Xq25) and HPRT (Xq26.1). The methylation status of theremaining allele of the androgen receptor gene in the tumors exhibiting LOH at the Xq25-26.1 region suggested that the losswas exclusively in the inactive X chromosome. We next determined whether a significant relationship exists between Xq LOHand other parameters, including histologic grade and/or clinical stage of the tumors and LOH at TP53. The Xq LOH had asignificant association with grade 2 to 3 tumors at stages II to IV. Sixteen of 18 cases that showed Xq LOH revealed LOH at theTP53 locus, and 45% of tumors exhibiting LOH at TP53 showed Xq LOH. These results suggest that there may be a tumorsuppressor gene or genes which escape inactivation of the X chromosome at Xq25-26.1, and that the loss of the gene(s) atXq25-26.1 is frequently accompanied by loss of the TP53 or loss of another gene on chromosome 17. These losses maycontribute to the progression from a well-differentiated to a more poorly differentiated state or to metastatic aggressiveness.Genes Chromosomes Cancer 20:234–242, 1997. r 1997 Wiley-Liss, Inc.†

INTRODUCTION

The development of cancer is associated withmultiple genetic alterations in oncogenes and tu-mor suppressor genes. In ovarian cancer, thesechanges include amplification of ERBB-2 (Slamonet al., 1989), atypical expression of EGF and FMS(Borrensen, 1992; Scwab and Amler, 1990), andmutations in KRAS (Teneriello et al., 1993; Chene-vix-Trench et al., 1997) and TP53 (Teneriello et al.,1993; Mazars et al., 1991; Milner et al., 1993).Cytogenetic and molecular genetic studies suggestthat changes in other oncogenes and tumor suppres-sor genes may be involved in ovarian carcinogen-esis. Chromosome analysis has identified frequentbreakpoints in 1p, 1q, 3p, 6q, 11p, and 19p, loss ofan X chromosome, and gain of chromosomes 7 and12 (Pejovic et al., 1989, 1991, 1992; Gallion et al.,1990; Jenkins et al., 1993; Thompson et al., 1994).Loss of heterozygosity (LOH), suggesting the pres-ence of a tumor suppressor gene, has been reportedfor many chromosome arms in ovarian cancers,

including 3p, 4p, 5q, 7p, 8q, 11p, 13q, 14q, 15q, 22q,and Xp, as well as for both arms of 6, 9, 12, 17, and19 (Chenevix-Trench et al., 1994, 1997; Lee et al.,1990; Sato et al., 1991; Zheng et al., 1991; Van-damme et al., 1992; Yang-Feng et al., 1992, 1993;Cliby et al., 1993; Osborne and Leech, 1994).These studies suggest that ovarian carcinoma cellsundergo multiple genetic changes during tumorprogression (Yang-Feng et al., 1993). However, therole of these genetic abnormalities in the etiologyand progression of this disease remains poorlyunderstood.

To understand the molecular mechanisms of thisdisease, it is necessary to isolate the genes underly-ing these chromosomal abnormalities. Among af-fected chromosomes, the X chromosome has fre-

*Correspondence to: Minoru Koi and J. Carl Barrett, Laboratory ofMolecular Carcinogenesis, National Institute of EnvironmentalHealth Sciences, P.O. Box 12233, Research Triangle Park, NC 27709.E-mail: Koi@niehs.nih.gov

Received 11 September 1996; Accepted 25 April 1997

GENES, CHROMOSOMES & CANCER 20:234–242 (1997)

r 1997 Wiley-Liss, Inc. †This article is a US Government work and, assuch, is in the public domain in the United States of America.

quently been lost in ovarian carcinoma (Pejovic etal., 1992; Thompson et al., 1994). Previous studiesof ovarian tumors have shown a significant LOH atXp21 (Yang-Feng et al., 1992, 1993). Recently,specific LOH was found at Xq near the androgenreceptor (AR) gene on the inactive X chromosomein low-malignant-potential (LMP) tumors. It hasbeen suggested that the gene that escapes inactiva-tion of the X chromosome contributes to this typeof tumor (Cheng et al., 1996). Furthermore, thepresence of gene(s) that cause a senescence-likecell growth arrest in rodent and human tumor cellshas been demonstrated through X chromosometransfer experiments (Klein et al., 1991; Wang et al.,1992). These observations prompted us to map aputative tumor suppressor gene or genes on the Xchromosome. To this end, we initiated studies todefine the precise chromosomal region of LOH onthe X chromosome in ovarian epithelial tumors. Inthis study, we analyzed LOH in at least 13 loci ofthe X chromosome in 123 tumor samples. We alsoanalyzed the association between Xq LOH andother parameters, including histologic grade, stage,tissue type, and LOH at the TP53 locus.

MATERIALS AND METHODS

Patient Materials

Ovarian tumors and corresponding normal tissuewere obtained from 123 patients, including 39 casesfrom North America, 15 from South Korea, and 69from Australia. These included the following histo-logic type of tumors: 81 serous, 13 mucinous, 11endometrioid, 8 clear cell, 3 undifferentiated, 1Brenner, and 6 mixed or of unknown histology.There were two benign tumors, nine LMP, and 112carcinomas. Clinical stages of tumors were deter-mined according to the International Federation ofGynecology and Obstetrics (FIGO) classification.Of the 90 carcinoma cases in which the stage wasknown, 11 were stage I, 11 were stage II, 58 werestage III, and 10 were stage IV. The grade of thetumors was also determined according to the FIGOclassification. Among 89 carcinoma cases in whichthe grade was known, 9 were grade 1, 27 were grade2, 5 were grade 2/3, and 48 were grade 3.

Dissection of Tumor Tissues and DNA Extraction

The Australian tumors were prepared in one oftwo ways. They were either dissected free fromnecrotic and connective tissue, and mechanicallydispersed. Dead and red cells were removed onFicoll-paque and the tumor cells pelleted andfrozen before storage at -70°C (Chenevix-Trench et

al., 1992). This method was used for two benign,nine LMP, and six stage I carcinomas. Alternatively,the tumors were disaggregated with collagenaseand the tumor cells allowed to settle overnight bygravity. Using these methods, the nontumor con-tamination was less than 5%. In 25 cases from NorthAmerica, tumor tissues were obtained at the time ofsurgery and were frozen at -80°C, and sequential5-µm sections were cut for hematoxylin and eosinstaining. Hematoxylin and eosin stained sectionswere examined, and tumor specimens with morethan 50% tumor cell content were subjected toDNA isolation. In 15 cases from Korea, tumortissues were isolated from formalin-fixed paraffin-embedded tissue blocks and subjected to DNAisolation. In 14 cases from North America, includ-ing four cases of stage 1 carcinomas, no procedurewas performed to evaluate the tumor content ofsurgically removed tumor tissue. Genomic DNAwas extracted and purified either from blood orfrozen tissues (108 North American and Australiancases) or from formalin-fixed paraffin-embeddedtissue blocks (15 Korean cases) (Chenevix-Trenchet al., 1994; Ignar-Trowbridge et al., 1992).

DNA Primers and PCR Condition

Each of the matched pairs of normal and tumorDNAs were subjected to PCR analysis with 26microsatellite markers on the X chromosome:DXS987, DXS996, DXS 999 (Weissenbach et al.,1992), DXS 237 (Gedeon et al., 1992), KAL (Bou-loux et al., 1991), DXS 207 (Oudet et al., 1993), DXS989, DXS1047, DXS1053, DXS1062, DXS1193,DXS1195, DXS1206, DXS1226, DXS1227, DXS1229,(Gyapay et al., 1994), DXS6680, DXS6679 (Carv-alho et al., 1994), DMD (Clemens et al., 1991),DXS228, PFC (Coleman et al., 1991), DXS7 (Mooreet al., 1992), MAOA (Hinds et al., 1992), HPRT, ARA(Edwards et al., 1992), and DXS458 (Weber et al.,1990). All of the markers except HPRT (tetra-nucleotide repeats) and AR (tri-nucleotide repeats)contain dinucleotide repeats. LOH at the TP53locus was also determined by PCR analysis (Futrealet al., 1991). PCR reactions were carried out asdescribed below with the inclusion of one [g- 32P]ATP-end-labeled primer. The PCR was performedin 20 µl volumes of a mixture containing 1 X PCRbuffer [10 mM Tris (pH 8.0), 50 mM KCl, 1.5 mMMgCl2], 1 µM each of unlabeled and labeledprimer, 20 ng of template DNA, 0.5 units of TaqDNA polymerase, and 200 µM of each deoxynucleo-side triphosphate. The reactions were cycled 30times; each cycle consisted of 1 min at 94°C, 30 secat 44-55°C, 1 min at 72°C, and 7 min at 72°C for

235CHROMOSOME ARM Xq LOH IN OVARIAN CARCINOMA

final elongation in a 480 thermal cycler (PerkinElmer Cetus, Emeryville, CA). The PCR productswere denatured and separated on a 6% polyacryl-amide gel containing 7 M urea for 2–3 h at roomtemperature. After electrophoresis, the gel wasdried and exposed to X-ray film for 12-16 h.

LOH Analysis

In cases where a particular marker was heterozy-gous in normal tissue DNA, LOH was assessed inthe corresponding tumor sample by three indepen-dent observers. When the results were not visuallyobvious, the intensity of signals was measured byphosphorimage analysis (ImageQuanty, MolecularDynamics, Sunnyvale, CA). The ratio of the signalintensity between the two alleles in the tumorDNA was compared with the ratio of signal inten-sity in corresponding normal tissue DNA. LOH wasdefined in any case where the value of the tumorallele ratio, as defined above, was #50% of thevalue of the normal allele ratio. The cases whereLOH was excluded by the above definition, butwhere there was a difference in the ratio betweentumor and normal tissues, may be due to differen-tial amplification, contamination of normal tissue,or heterogeneity in the tumor tissue. In this study,these cases were categorized as having allelic imbal-ance (AI).

Analysis of X Chromosome Inactivation

The methylation status near the polymorphictriplet repeat in the AR gene at Xq21 was examined(Allen et al., 1992). One hundred nanograms ofgenomic DNA from the patient’s normal or tumortissues were incubated at 37°C for 16 h in 10 µl ofthe appropriate buffer with or without 10 units ofmethylation-sensitive restriction endonucleaseHpaII (New England Biolabs, MA). After heatinactivation at 95°C for 10 min, 1 µl of this reactionsolution was added to the PCR reaction to amplifythe region containing both the triplet repeat andthe HpaII sites. The PCR primers and conditionswere described previously (Allen et al., 1992). ThePCR products were denatured and separated on a6% polyacrylamide gel containing 7 M urea. Afterelectrophoresis, the gel was dried and exposed toX-ray film for 16 h.

Statistical Analysis

We used the chi-square test to determine pos-sible differences among proportions of LOH at 24gene loci. We also used the test to determinewhether a significant relationship existed betweenXq LOH and TP53 LOH. We used the Cochran-

Armitage trend test (Agresti, 1990) to show theassociation of LOH with grade and stage.

RESULTS

Table 1 summarizes the locus symbols, theircytogenetic positions (Willard et al., 1994), and thepercentage of LOH at each locus in 123 cases fromNorth America, South Korea, and Australia. In theAustralian tumors, LOH at some of the loci hasbeen reported before (Chenevix-Trench et al.,1997). The percentage of LOH was determinedusing the ratio between the number of cases withLOH and the number of informative cases. Lossesof DXS1206 (34.5%, P , 0.001), DXS1047 (27.7%,P 5 0.002)and HPRT (24.1%, P 5 0.046) weresignificantly higher than losses of other gene loci,suggesting the presence of a tumor suppressor genein this region. Figure 1 illustrates a common regionof LOH in 37 carcinomas (33% of carcinomasexamined) that showed LOH at one of the loci onthe X chromosome. Among these tumors, 32 (28.6%)showed LOH at several loci on Xq containing theXq25-26.1 region. Tumors D35, A24, and A56defined the HPRT locus as the distal breakpoint.Five tumors (D39, A30, A47, A23, and D35) re-tained heterozygosity at the DXS458 locus, suggest-ing that a specific LOH localizes between DXS458(Xq21.3) and HPRT (Xq26.1) in ovarian carcinomas.Tumor D35 further defined the minimum overlap-ping region of LOH between DXS1206 (excluded)and HPRT (excluded). The predicted physicaldistance between these two markers has beenestimated to be approximately 5 megabases (Mb)(Willard et al., 1994).

Twenty-two carcinomas (19.8%) showed LOH atthe loci on the Xp region. Tumors A21 and A70define the second minimum overlapping regionbetween DMD and DXS 6679 at the Xp21 region,suggesting the presence of another tumor suppres-sor locus in this region (Fig. 1). However, frequencyof LOH at any locus in this region was not statisti-cally significant (Table 1).

Figure 2 shows examples of PCR analyses ofLOH at DXS1206, DXS1047, and HPRT. TumorK10 showed LOH at DXS1206, DXS1047, andHPRT. Tumor A24 showed LOH at DXS1206 andDXS1047. Although weak PCR signals correspond-ing to the lost alleles were visible in both loci, theratio of the signal intensity between the two allelesin the tumor was less than 50% of the value of thenormal allele ratio. This may be due to contamina-tion of non-neoplastic stromal cells with tumortissue or to heterogeneity within a tumor tissue.

236 CHOI ET AL.

Tumor D35 had LOH at DXS1047, but retainedheterozygosity at DXS1206 and HPRT.

We analyzed nine cases of LMP tumors and twocases of benign tumors from Australia for severalloci on the X chromosome where a previous study(Cheng et al., 1996) found a specific loss at Xq nearthe AR locus in these tumors. All of the LMPtumors were informative at AR locus and twotumors showed LOH at this locus but not at theother loci examined (Fig. 3). Two benign tumorsshowed no LOH on these loci. These resultsconfirm the previous finding by Cheng et al. (1996)that some LMP tumors show LOH at Xq near ARlocus (Xp11.21-Xq22). Our results further suggestthat the region of LOH found in carcinomas (Xq25-26.1) is different from that found in LMP tumors(Xp21.2-Xq21.3) (Fig. 3).

To determine whether the Xq loss is in active X,inactive X, or both, we analyzed the methylationstatus of an HpaII site 5’ to the (CAG) repeat in theremaining AR gene for the carcinomas that showedLOH at the Xq25-26.1 and AR loci. This region onthe active X chromosome is unmethylated; thus,the HpaII site of this locus is sensitive to HpaII

digestion. In contrast, the same site on the inactiveX is methylated and resistant to HpaII digestion. Ineach of six carcinomas examined (case N1, N23,D12, A10, A24, and A68), a remaining allele con-tained an HpaII-sensitive site, suggesting that theremaining chromosome is the active X chromosome(Fig. 4).

We next analyzed the relationship between XqLOH and other parameters, including histologicgrade, clinical stage, and tumor type (serous vs.nonserous). As shown in Table 2, Xq LOH wasassociated with a higher grade (P 5 0.0136) andadvanced stage (P 5 0.0444) using the Cochran-Armitage trend test. There was no evidence of asignificant difference in the frequency of Xq LOHbetween serous and nonserous-type tumors (datanot shown).

To determine whether there is a significantrelationship between LOH at Xq and LOH atTP53, we examined LOH at the TP53 locus in 123cases. Thirty-three of 65 (51%) informative casesshowed LOH at the TP53 locus. LOH at TP53 wasalso associated with higher grades (P 5 0.0136) andmore advanced stages (P 5 0.0444) (the Cochran-

TABLE 1. LOH on X Chromosome of Ovarian Tumora

Locussymbol

Cytogeneticband

No. of LOH cases/No. of informative cases (%)

PcNCb NY KOR AUS Total

DXS996 Xp22.3 1/15 2/11 1/9 6/37 10/72 (13.9)DXS237 Xp22.3 1/12 2/10 1/6 N.D. 4/28 (14.3)KAL Xp22.3 3/20 2/13 2/11 N.D. 7/44 (15.9)DXS987 Xp22.2 2/17 2/10 1/6 6/42 11/75 (14.7)DXS207 Xp22.2 1/16 1/10 1/8 N.D. 3/34 (8.8)DXS1053 Xp22.2 1/16 1/8 2/10 N.D. 4/34 (11.8)DXS1195 Xp22.1 1/14 2/9 0/2 3/38 6/63 (9.5)DXS999 Xp22.1 2/16 1/8 0/5 10/50 13/79 (16.5)DXS1229 Xp22.1 1/9 1/4 0/2 N.D. 2/15 (13.3)DXS1226 Xp22.1 1/8 2/13 3/11 N.D. 6/32 (18.8)DXS989 Xp22.1 2/21 2/11 2/12 7/38 13/82 (15.9)DMD Xp21.2 2/17 2/13 2/3 9/47 15/80 (18.8)DXS6680 Xp21.1 2/6 1/10 1/10 7/30 11/56 (19.6)DXS6679 Xp11.4 0/11 1/10 1/3 5/38 7/62 (11.3)DXS228 Xp11.4 1/14 0/7 0/4 4/29 5/54 (9.3)DXS7 Xp11.4 2/15 0/6 0/3 N.D. 2/24 (8.3)MAOA Xp11.4 3/18 0/9 1/10 N.D. 4/37 (10.8)PFC Xp11.2 3/12 1/10 0/3 N.D. 4/25 (16.0)AR Xq12 3/20 2/13 0/13 7/56 12/102 (11.8)DXS458 Xq21.3 1/10 2/10 1/4 1/5 5/29 (17.2)DXS1206 Xq25 5/13 2/6 2/5 10/13 19/55 (34.5) ,0.001DXS1047 Xq25 8/18 2/13 3/13 15/57 28/101 (27.7) 0.002HPRT Xp26.1 5/13 2/9 4/12 10/53 21/87 (24.1) 0.046DXS1062 Xq26.2 2/5 2/8 2/5 N.D. 6/18 (33.3)DXS1227 Xq27.1 4/12 0/8 1/9 N.D. 5/29 (17.2)DXS1193 Xq28 2/19 1/14 0/9 N.D. 3/42 (7.1)

aA total of 123 cases were examined.bTumors were sampled in North Carolina (NC), New York (NY), South Korea (KOR).cStatistical difference in the frequency of LOH among 26 loci was determined by chi-square analysis.

237CHROMOSOME ARM Xq LOH IN OVARIAN CARCINOMA

Armitage trend test) (Table 3). As shown in Table 4,tumors that gained LOH at the TP53 locus weremore likely to gain Xq LOH than were tumors thatdid not have TP53 LOH (P , 0.001). In addition,tumors that showed LOH at Xq were more likely togain TP53 LOH than were tumors that did not haveXq LOH (P , 0.001), and most Xq LOH tumorsshowed LOH at TP53. Taken together, these re-sults suggest that loss of the TP53 function or loss ofother gene(s) near TP53 on chromosome 17 may bea prerequisite for loss of the Xq gene. These lossescontribute to progression from a well-differentiatedto a more poorly differentiated state, or to aggressive-ness.

DISCUSSION

In this study, we demonstrated that ovariancarcinomas display a significant frequency of LOHat the Xq25-26.1 region. LOH at a specific chromo-somal region is a hallmark of the presence of arecessive tumor suppressor gene. Since both alleles

of tumor suppressor genes on autosomal chromo-somes are expected to be expressed in normal cells,two events which inactivate both alleles of the geneare required for tumor formation (Knudson, 1971).In contrast, most genes on the inactive X chromo-some in female somatic cells are transcriptionallysilent. Therefore, one event which inactivates corre-sponding genes on the active X chromosome shouldbe sufficient for the loss of the gene function.However, there are genes on the inactive X chromo-some which are not subjected to X inactivation andare expressed from both allele on active and inac-tive X chromosomes. To identify the gene atXq25-Xq26.1, it is important to determine whetheractive or inactive X chromosome is involved inLOH. The methylation status of the AR gene intumors with Xq25-26.1 LOH suggests that the lossoccurs on the inactive X chromosome. If the geneaffected by LOH is a recessive tumor suppressorgene, that is, if it has been subjected to Knudson’s‘‘two-hit’’ hypothesis (Knudson, 1971), the gene on

Figure 1. Allelic loss patterns of ovarian carcinomas for the Xchromosome. For each carcinoma, all informative loci are shown. Opensquare, constitutional heterozygosity with LOH; solid square, constitu-tional heterozygosity with no LOH; circle, constitutional heterozygositywith AI; blank space, uninformative. With the assumption that alleles inall regions between loci exhibiting allelic loss are lost, solid lines indicate

retained region of X chromosome and open space represent region ofallelic loss. The broken line represents the regions that are uncertain forsome loci. The minimum region of loss at Xq is 5 Mb and is locatedbetween DXS1206 and HPRT. Another LOH overlapping region ismapped to Xp 21.1 between DMD and DXS6679.

238 CHOI ET AL.

the inactive X must escape from X inactivation, andloss should be found in both the inactive and activeX chromosomes. An explanation for the preferen-tial loss of inactive Xq may be that a large deletionof the corresponding region of active X wouldeliminate gene(s) that are crucial to cell survival andcells that have lost the active gene on the inactive Xare selected. Thus far, no genes at Xq25-26.1 areknown to escape X inactivation; however, growingnumbers of expressed genes located outside of thepseudoautosomal region of inactive X have beenreported (Miller et al., 1995). Searching for theexpressed genes from the 5 Mb LOH overlappingregion on an inactivated X chromosome might be afeasible approach to identify the gene. Alterna-tively, the suppressor gene might act in a dose-dependent manner, while the function of the geneis eradicated by one hit. In this case, subtle muta-tions or microdeletion of the gene on either aninactive or active X, or a large deletion containingthe Xq25-26.1 region on the inactive X, may contrib-

ute to loss of the suppressive function. If this is thecase, the putative suppressor gene on Xq25-26.1may be involved more frequently in ovarian carcino-genesis than we observed through LOH analysis.

Previous studies have shown a high frequency(41–60%) of LOH at the Xp21.1 region in ovariancarcinoma (Yang-Feng et al., 1992, 1993; Osborneand Leech, 1994). Another study using comparativegenomic hybridization showed frequent loss of theXp region in ovarian cancer (Iwabuchi et al., 1995).In our study, we found a lower frequency of XpLOH at the Xp21.1 region than in previous studies(Table 1). We found that four tumors had LOH onlyat the Xp region, and, in 23 of 123 cases, we foundLOH in chromosome arm Xp containing Xp21.1.The common overlapping region defined by twotumors was mapped to the Xp21.1 region (Fig. 1).These results do not exclude the presence of asecond tumor suppressor gene on Xp. The hypoth-esis of tumor suppressor genes on both Xp and Xqis consistent with the findings of many cases of total

Figure 2. LOH at chromosome Xq25-26.1 loci in sporadic ovariancarcinoma. A partial ideogram of chromosome arm Xq and thecytogenetic position of three loci encompassing Xq25-26.1 are shown tothe left of the diagram. Representative microsatellite analysis of normal

(N) and tumor (T) DNAs from patients K10, A24, and D35 was chosento illustrate three informative loci encompassing the common LOHregion. The arrows indicate alleles showing LOH.

239CHROMOSOME ARM Xq LOH IN OVARIAN CARCINOMA

chromosome X loss in ovarian carcinoma in thisstudy and others (Gallion et al., 1990; Jenkins et al.,1993). Further examination of more cases of ovarian

tumors may clarify the presence of specific LOH atthe Xp21 region.

In this study, it was found that Xq LOH coin-cides with LOH at the TP53 locus and is present inhigh grade and advanced tumors. This suggests thatthe losses of this gene and TP53, or another genenear TP53, might contribute to the progressionfrom a well-differentiated to a poorly differentiatedstate, or to metastatic potential. Recently, a highincidence of LOH at Xq near the AR locus has beenreported in LMP tumors (Cheng et al., 1996;Chenevix-Trench et al., 1997). These losses are alsofound in inactive X chromosomes (Cheng et al.,1996). The same authors have described substantialLOH at the Xq region in advanced ovarian carcino-mas, suggesting that the same gene is involved indifferent stages of tumorigenesis (Cheng et al.,1996). However, the location of LOH on the Xchromosome in carcinoma, as determined by thisstudy, is different from that of the LMP region.This suggests that there may be two differenttumor suppressor genes on the X chromosome that

Figure 3. Allelic loss patterns of two LMP tumors for the Xchromosome. Open square, constitutional heterozygosity with LOH;solid square, constitutional heterozygosity with no LOH; X, uninforma-tive.

Figure 4. Analysis of X chromosome inactivation. HpaII-digested(1) and nondigested (-) DNA from either lymphocytes (N) or tumortissues (T) from case N1 were subjected to PCR amplification of the ARgene. The upper allele (white arrow) is lost in a tumor. The bottom allele(black arrow) is retained in a tumor and is sensitive to HpaII digestion,indicating that this allele is on the active X chromosome and that the lostallele in a tumor is on the inactive X chromosome. The same resultswere obtained in six other cases.

TABLE 2. Histologic Grade/Clinical Stage and Xq LOH

No. of Xq LOH cases/no.of cases examined (%) Pa

Histologic gradeb

1 0/92 & 2/3 8/32 (25) 0.01363 19/48 (39.6)

FIGO Stagec

1 1/11 (9.1)2 2/11 (18.2) 0.04443 21/58 (36.2)4 4/10 (40)

aStatistical significance of the association of Xq LOH with tumor gradeand stage was determined by the Cochran-Armitage trend test.bA total of 89 carcinoma cases were analyzed.cA total of 90 carcinoma cases were analyzed.

TABLE 3. Histologic Grade/Clinical Stage and TP53 LOH

No. of TP53 LOH cases/no.of cases examined (%) Pa

Histologic gradeb

1 1/4 (25)2 & 2/3 11/18 (61.1) 0.01363 17/22 (77.3)

FIGO Stagec

1 2/10 (20)2 3/6 (50) 0.04443 20/28 (71.4)4 7/9 (77.8)

aStatistical significance of the association of TP53 LOH with tumor gradeand stage was determined by the Cochran-Armitage trend test.bA total of 44 carcinoma cases were analyzed.cA total of 53 carcinoma cases were analyzed.

240 CHOI ET AL.

are not subjected to X inactivation, one responsiblefor the genesis of LMP and another responsible forcarcinoma progression.

Because few studies of LOH on the X chromo-some have been performed for other tumor types, itis not yet known whether Xq25-26.1 LOH iscommon to other tumors. A significant LOH (26%)at the DXS3 locus (Xq21.3) has been reported forcervical carcinoma (Mitra et al., 1994). Xq LOH hasalso been demonstrated in renal oncocytomas(Thrash-Bingham et al., 1996). More detailed stud-ies of LOH in these tumors and the examination ofXq LOH in tumors from other tissues will showwhether this loss is common or specific to ovariancarcinomas.

The transfer of the X chromosome to rodent orhuman tumor cells has been known to inducesenescence-like tumor cell growth arrest (Klein etal., 1991; Wang et al., 1992). These results suggestthat the X chromosome contains genes that areimportant to mammalian carcinogenesis. It wouldbe interesting to determine whether introduction ofthe X chromosome via microcell chromosome trans-fer has any effect on cell growth in vitro or vivo oron the degree of differentiation of ovarian carci-noma cells that have lost the Xq25-26.1 region. Acombined approach, using a genetic complementa-tion study and positional cloning, may facilitate theisolation of this gene.

Finally, this study reveals that 12 carcinomas(10.7% of carcinomas examined), and no LMP orbenign tumors exhibited microsatellite instability.This finding is compatible with a previous study

(Osborne and Leech, 1994), suggesting that micro-satellite mutations are rare in ovarian cancer.

In summary, we demonstrated that a high propor-tion (,40%) of ovarian carcinomas of high gradeand advanced stages show LOH at a 5 Mb region ofinactive Xq25-26.1, suggesting the presence of atumor suppressor gene that may contribute to theprogression of ovarian carcinoma.

ACKNOWLEDGMENTS

We thank Drs. J. Boyd and P. Vojta for criticalcomments on this manuscript.

REFERENCES

Agresti A (1990) Models for binary response variables. In: CategorialData Analysis. New York: John Wiley & Sons, pp 79–129.

Allen RC, Zoghbi HY, Moseley AB, Rosenblatt HM, Belmont JW(1992) Methylation of HpaII and HhaI sites near the polymorphicCAG repeat in the human androgen-receptor gene correlates withX chromosome inactivation. Am J Hum Genet 51:1229–1239.

Borrensen AL (1992) Oncogenesis in ovarian cancer. Acta ObstetGynecol Scand 155:25–30.

Bouloux PMG, Hardelin JP, Munroe P, Kirk JMW, Legouis R,Levilliers J, Hazan J, Weissenbach J, Petit C (1991) A dinucleotiderepeat polymorphism at the Kallmann locus (Xp22.3). Nucl AcidsRes 19:5453.

Carvalho MRS, Medindl A, Strom TM, Muller B, Wittwer B, RossM, Meitinger T (1994) X-chromosomal retinitis pigmentosa:Genetic mapping and cloning approaches. Cytogenet Cell Genet67:343.

Chenevix-Trench, Leary J, Kerr J, Michel J, Kefford R, Hurst T,Parsons, PG, Friedlander M, Khoo SK (1992) Frequent loss ofheterozygosity on chromosome 18 in ovarian adenocarcinomawhich does not always include the DCC locus. Oncogene 7:1059–1065

Chenevix-Trench G, Kerr J, Friedlander M, Hurst T, Sanderson B,Coglan M, Ward B, Leary J, Khoo S-K (1994) Homozygousdeletions on the short arm of chromosome 9 in ovarian adenocarci-noma cell lines and loss of heterozygosity in sporadic tumors. Am JHum Genet 55:143–149.

Chenevix-Trench G, Kerr J, Hurst T, Shih Y-C, Purdie D, BergmanL, Friedlander M, Sanderson B, Zournazi A, Coombs T, Leary JA,Crawford E, Shelling AN, Cooke I, Ganesan TS, Searle J, Choi C,Barrett JC, Khoo S-K, Ward B (1997) Analysis of loss of heterozy-gosity and K-ras mutations and in ovarian neoplasms: Clinicopatho-logical correlations. Genes Chromosomes Cancer 18:75–83.

Cheng PC, Gosewehr JA, Kim TM, Velicescu M, Wan M, Zheng J,Felix JC, Cofer KF, Luo P, Biela BH, Godorov G, Dubeau L (1996)Potential role of the inactivated X chromosome in ovarian epithe-lial tumor development. J Natl Cancer Inst 88:510–518.

Clemens PR, Fenwick RG, Chamberlain JS, Gibbs RA, de AndradeM, Chakraborty R (1991) Carrier detection and prenatal diagnosisin Duchenne and Becker muscular dystrophy families, usingdinucleotide repeat polymorphism. Am J Hum Genet 49:951–960.

Cliby W, Ritland S, Hartmann L, Dodson M, Halling KC, Keeney G,Podratz KC, Jenkins RB (1993) Human epithelial ovarian cancerallelotype. Cancer Res 53:2393–2398.

Coleman MP, Murray JC, Willard HF, Nolan KF, Reid KBM, BlakeDJ, Lindsay S, Bhattacharya SS, Wright A, Davies KE (1991)Genetic and physical mapping around the properdin P gene.Genomics 11:991–996.

Edwards A, Hammond HA, Jin L, Caskey T, Chakraborty R (1992)Genetic variation at five trimeric and tetrameric tandem repeatloci in four human population groups. Genomics 12:241–253.

Futreal PA, Barrett JC, Wiseman RW (1991) An Alu polymorphismintragenic to the TP53 gene. Nucl Acids Res 19:6977.

Gallion HH, Powell DE, Smith LW, Morrow JK, Martin AW, vanNagell JR, Donaldson ES (1990) Chromosome abnormalities inhuman epithelial ovarian malignancies. Gynecol Oncol 38:473–477.

Gedeon AK, Holman K, Richards RI, Mulley JC (1992) Characteriza-tion of new PCR based markers for mapping and diagnosis: AC

TABLE 4. Xq LOH and TP53 LOH

No. of Xq LOHcases/no. of

cases examined Pa

Xq LOH cases among TP53 LOHcases 14/23 ,0.001a

Xq LOH cases among casesretaining heterozygosity atTP53 2/32

No. of TP53 LOHcases/no. of

cases examined Pb

TP53 LOH cases among Xq LOHcases 16/18 ,0.001b

TP53 LOH cases among casesretaining heterozygosity at Xq 17/48

aSignificant difference in frequency of Xq LOH between TP53 LOH-positive cases and the cases retaining heterozygosity at TP53 wasdetermined by chi-square test.bSignificant difference in frequency of TP53 LOH between Xq LOH-positive cases and the cases retaining heterozygosity at the Xq regionwas determined by chi-square test.

241CHROMOSOME ARM Xq LOH IN OVARIAN CARCINOMA

repeat markers at the DXS237 (GMGX9) and DXS102 (cX 38.1).Am J Med Genet 43:255–260.

Gyapay G, Morissette J, Vignal A, Dib C, Fizames C, Millasseau P,Marc S, Bernardi G, Lathrop M, Weissenbach J (1994) The1993–94 Genethon human genetic linkage map. Nat Genet7:246–339.

Hinds HL, Hendriks RW, Craig IW, Chen ZY (1992) Characteriza-tion of a highly polymorphic region near the first exon of thehuman MAOA gene containing a GT dinucleotide and a novelVNTR motif. Genomics 13:896–897.

Ignar-Trowbridge D, Risinger JI, Dent GA, Kohler M, Becrchuck A,McLachlan JA, Boyd J (1992) Mutation of the Ki-ras oncogene inendometrial carcinoma. Am J Obstet Gynecol 176:227–232.

Iwabuchi H, Sakamoto M, Sakunaga H, Ma Y-Y, Carcangiv ML,Pinkel D, Yang-Feng TL, Gray JW (1995) Genetic analysis ofbenign, low-grade, and high grade ovarian tumors. Cancer Res55:6172–6180.

Jenkins RB, Bartelt D, Stalboerger P, Persons D, Dahl RJ, Podratz K,Keeney G, Hartmann L (1993) Cytogenetic studies of epithelialovarian carcinoma. Cancer Genet Cytogenet 71:76–86.

Klein CB, Conway K, Wang XW, Bhamra RK, Lin X, Cohen MD,Annab L, Barrett JC, Costa M (1991) Senescence of nickel-transformed cells by an X chromosome: Possible epigeneticcontrol. Science 251:796–799.

Knudson AG (1971) Mutations and cancer: Statistical study ofretinoblastoma. Proc Natl Acad Sci USA 68:820–823.

Lee JH, Kavanagh JJ, Wildrick DM, Wharton JT, Blick M (1990)Frequent loss of heterozygosity on chromosomes 6q, 11, and 17 inhuman ovarian carcinomas. Cancer Res 50:2724–2728.

Mazars R, Pujol P, Maudelonde T, Jeanteur P, Theillet C (1991) p53mutations in ovarian cancer, a late event? Oncogene 6:1685–1690.

Miller AP, Gustashaw K, Wolff DJ, Rider SH, Monaco AP, Eble B,Schlessinger D, Gorski JL, van Ommen G-J, Weissenbach J,Willard HF (1995) Three genes that escape X chromosomeinactivation are clustered within 6 Mb YAC contig and STS map inXp11.21-p11.22. Hum Mol Genet 4:731–739.

Milner B, Allan L, Eccles D, Kitchener H, Leonard R, Kelly K,Parkin D, Haites N (1993) p53 mutation is a common geneticevent in ovarian carcinoma. Cancer Res 53:2128–2132.

Mitra AB, Murty VVVS, Li RG, Pratap M, Luthra UK, Chaganti RSK(1994) Allelotype analysis of cervical carcinoma. Cancer Res54:4481–4487

Moore BJ, Kwan S.P, Bech-Hansen NT (1992) A polymorphicdinucleotide repeat at the DXS7 locus. Nucl Acids Res 20:929.

Osborne RJ, Leech V (1994) Polymerase chain reaction allelotypingof human ovarian cancer. Br J Cancer 69:429–438.

Oudet C, Weber C, Kaplan J, Segues B, Croquette MF, Roman, EO,Hanauer A (1993) Characterization of a highly polymorphicmicrosatellite at the DXS207 locus: Confirmation of very closelinkage to the retinoschisis disease gene. J Med Genet 30:300–303.

Pejovic T, Heim S, Mandahl N, Elmfors B, Floderus L-M, FurgyikS, Helm G, Willen H, Mitelman F (1989) Consistent occurrence ofa 19p1 marker chromosome and loss of 11p material in ovarianseropapillary cystadenocarcinomas. Genes Chromosomes Cancer1:167–171.

Pejovic T, Heim S, Mandahl N, Elmfors B, Furgyik S, FloderusL-M, Helm G, Willen H, Mitelman F (1991) Bilateral ovarian

carcinoma: Cytogenetic evidence of unicentral origin. Int J Cancer47:358–361.

Pejovic T, Heim S, Mandahl N, Baldetorp B, Elmfors B, FloderusU-M, Furgyik S, Helm G, Himmelmann A, Willen H, Mitelman F(1992) Chromosome aberrations in 35 primary ovarian carcinomas.Genes Chromsomes Cancer 4:58–68.

Sato T, Saito H, Morita R, Koi S, Lee JH, Nakamura Y (1991)Allelotype of human ovarian cancer. Cancer Res 51:5118–5122.

Schwab M, Amler LC (1990) Amplification of cellular oncogenes: Apredictor of clinical outcome in human cancer. Genes Chromo-somes Cancer 1:181–193.

Slamon DJ, Godolphin W, Jones LA, Holt JA, Wong SG, Keith DE,Levin WJ, Stuart SG, Udove J, Ullrich A, Press MF (1989) Studiesof the HER-2/neu proto-oncogene in human breast and ovariancancer. Science 244:707–712.

Teneriello MG, Ebina M, Linnoila RI, Henry M, Nash JD, Park RC,Birrer MJ (1993) p53 and Ki-ras gene mutations in epithelialovarian neoplasms. Cancer Res 53:3103–3108.

Thompson FH, Emerson J, Albert D, Liu Y, Guan XY, Burgess A,Fox S, Taetle R, Weinstein R, Makar R, Powel D, Trent T (1994)Clonal chromosome abnormalities in 54 cases of ovarian carci-noma. Cancer Genet Cytogenet 73:33–45.

Thrash-Bingham CA, Salazar H, Greenberg RE, Tartof KD (1996)Loss of heterozygosity studies indicate that chromosome arm 1pharbors a tumor suppressor gene for renal oncocytomas. GenesChromosomes Cancer 16:64–67.

Vandamme B, Lissens W, Amfo K, Sutter PD, Bourgain C, Vamos E,Greve JD (1992) Deletion of chromosome 11p13-11p15.5 se-quences in invasive human ovarian cancer is a subclonal progres-sion factor. Cancer Res 52:6646–66542.

Wang XW, Lin X, Klein CB, Bhamra RK, Lee Y-W, Costa MA (1992)A conserved region in human and Chinese hamster X chromo-somes can induce cellular senescence of nickel-transformed Chi-nese hamster cell lines. Carcinogenesis 13:555–561.

Weber JL, Kwitek AE, May PE, Polymeropoulos MH, Ledbetter S(1990) Dinucleotide repeat polymorphisms at the DXS453 andDXS458 loci. Nucl Acids Res 18:4037.

Weissenbach J, Gyapay G, Dib C, Vignal A, Morissette J, MillasseauP, Vaysseix G, Lathrop M (1992) A second-generation linkage mapof the human genome. Nature 359:794–801.

Willard HF, Cremers JL, Mandel JL, Monaco AP, Nelson DL,Schlessinger D (1994) Report of the fifth international workshopon human X chromosome mapping 1994. Cytogenet Cell Genet67:296–327.

Yang-Feng TL, Li S, Han H, Schwartz PE (1992) Frequent loss ofheterozygosity on chromosomes Xp and 13q in human ovariancancer. Int J Cancer 52:575–580.

Yang-Feng TL, Han H, Chen K-C, Li S, Claus EB, Carcangiu ML,Chambers SK, Chambers JT, Schwartz PE (1993) Allelic loss inovarian cancer. Int J Cancer 54:546–551.

Zheng J, Robinson WR, Ehlen T, Yu MC, Dubeau L (1991)Distinction of low grade from high grade human ovarian carcino-mas on the basis of losses of heterozygosity on chromosomes 3, 6,and 11 and HER-2/neu gene amplification. Cancer Res 51:4045–4051.

242 CHOI ET AL.