characterization of the molecular differences between ovarian endometrioid carcinoma and ovarian...

Journal of PathologyJ Pathol 2010; 220: 392–400Published online 27 October 2009 in Wiley InterScience(www.interscience.wiley.com) DOI: 10.1002/path.2659

Original Paper

Characterization of the molecular differences betweenovarian endometrioid carcinoma and ovarian serouscarcinomaJason Madore,1 Fengge Ren,1,2 Ali Filali-Mouhim,1 Lilia Sanchez,1,3 Martin Kobel,4 Patricia N Tonin,5,6,7

David Huntsman,4† Diane M Provencher1,8,9 and Anne-Marie Mes-Masson1,8,9*1Centre de Recherche du Centre Hospitalier de l’Universite de Montreal (CHUM)/Institut du Cancer de Montreal, Montreal, Canada2Department of Gynaecology and Obstetrics, 1st Hospital of Shanxi Medical University, Taiyuan, China3Departement de Pathologie, Universite de Montreal, Hopital Hotel-Dieu du CHUM, Montreal4Department of Pathology and Genetic Pathology Evaluation Centre of the Prostate Research Centre, Department of Pathology and LaboratoryMedicine, University of British Columbia, Vancouver General Hospital, and British Columbia Cancer Agency, Vancouver5The Research Institute of McGill University Health Centre, Montreal, Canada6Department of Human Genetics, McGill University, Montreal, Canada7Department of Medicine, McGill University, Montreal, Canada8Departement de Medicine, Universite de Montreal, Montreal, Quebec, Canada9Departement d’Obstetriques et Gynecologie, Division de Gynecologie Oncologie, Universite de Montreal, Montreal, Canada

*Correspondence to:Anne-Marie Mes-Masson,CR-CHUM/ICM, 1560, RueSherbrooke Est, Montreal,Quebec, Canada H2L 4M1.E-mail: [email protected]

†This article was published onlineon 27 October 2009. Atypographical error wassubsequently identified in theauthor name David Huntsman.This notice is included in theonline and print versions toindicate that both have beencorrected, 10 December 2009.

No conflicts of interest weredeclared.

Received: 27 July 2009Revised: 16 October 2009Accepted: 19 October 2009

AbstractThe histopathological diagnosis of high-grade endometrioid and serous carcinoma of theovary is poorly reproducible under the current morphology based classification system,especially for anaplastic, high-grade tumours. The transcription factor Wilms’ tumour-1(WT1 ) is differentially expressed among the gynaecological epithelia from which epithelialovarian cancers (EOCs) are believed to originate. In EOCs, WT1 protein is observed inthe majority of serous carcinomas and in up to 30% of endometrioid carcinomas. It isunclear whether the latter is a reflection of the actual incidence of WT1 protein expression inendometrioid carcinomas, or whether a significant number of high-grade serous carcinomashave been misclassified as endometrioid carcinoma. Several genetic aberrations are reportedto occur in EOCs. These include mutation of the TP53 gene, aberrant activation of β-cateninsignalling and loss of PTEN protein expression, among others. It is unclear whether theseaberrations are histotype-specific. The aim of this study was to better define the molecularcharacteristics of serous and endometrioid carcinomas in an attempt to address the problemswith the current histopathological classification methods. Gene expression profiles wereanalysed to identify reproducible gene expression phenotypes for endometrioid and serouscarcinomas. Tissue microarrays (TMA) were used to assess the incidence of TP53, β-catenin and PTEN aberrations in order to correlate their occurrence with WT1 as animmunohistochemistry based biomarker of serous histotype. It was found that nuclearWT1 protein expression can identify misclassified high-grade endometrioid carcinomas andthese tumours should be reassigned to serous histotype. Although low-grade endometrioidcarcinomas rarely progress to high-grade carcinomas, a combined WT1-negative, TP53-positive immunophenotype may identify an uncommon high-grade subtype of ovarianendometrioid carcinoma. GEO database: array data accession number GSE6008.Copyright 2009 Pathological Society of Great Britain and Ireland. Published by JohnWiley & Sons, Ltd.

Keywords: epithelial ovarian cancer; DNA microarray; tissue microarray; immunohisto-chemistry; endometrioid carcinoma; serous carcinoma; histotype differential diagnosis

Introduction

Epithelial ovarian cancer (EOC) is often described asa lethal gynaecological malignancy; however, EOCis a heterogeneous group of cancers with remark-able differences in pathobiology and tumour behaviour[1–3]. Molecular biology has demonstrated that the

diversity among EOCs is a result of alternative modesof molecular pathogenesis between different EOC his-totypes [4,5]. There are five main EOC histotypes, twoof which, serous carcinoma and endometrioid carci-noma, account for the majority of malignant ovar-ian tumours (>75%) [6]. However, for high-gradetumours the differential diagnosis of endometrioid and

Copyright 2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.www.pathsoc.org.uk

Molecular differences between endometrioid and serous ovarian carcinoma 393

serous carcinoma is challenging and there is a lack ofreproducibility using the current classification system[7]. Recently, updated architectural and cytologicalcriteria has been proposed to permit a more precisehistopathological diagnosis, although some interob-server variation persists [8].

The molecular features and occurrence of high-grade endometrioid and serous carcinoma remainpoorly defined and may be confounding the studyof EOC. By definition, high-grade tumours lackhistotype-specific differentiation features. It is not sur-prising, then, that DNA microarray studies show thathigh-grade endometrioid and serous carcinomas candisplay similar gene expression profiles [9,10]. Thismay simply be an artefact of an inadequate patho-logical classification system that is unable to discrimi-nate between poorly differentiated high-grade tumours.Alternatively, it might suggest a histogenetic relation-ship between morphologically distinctive tumours ofthe same histotype.

The pathogenesis of serous and endometrioid ovar-ian cancers remains poorly understood. Serous carci-nomas are conventionally thought to arise from neo-plastic ovarian surface epithelium [11]. Alternatively,recent studies suggest that a portion of serous carci-nomas arise as peritoneal or ovarian implants froman undetected primary intraepithelial tubal carcinoma[12–15]. Endometrioid carcinomas, on the other hand,are thought to arise through transformation of ovarianendometriosis [16,17], a common and typically benigndisorder, defined as the occurrence of ectopic endome-trial tissue (stroma and epithelium) found aberrantly aspart of the ovary.

The transcription factor Wilms’ tumour-1 (WT1) isa putative gynaecological biomarker that is differen-tially expressed among gynaecological epithelia [18].Nuclear WT1 protein is observed in tubal and ovariansurface epithelia [18]. In contrast, endometrial epithe-lium does not express nuclear WT1 protein [18]. Thus,WT1 may be a useful biomarker, based on its differ-ential expression in the normal tissues, from whichEOC tumours are alleged to originate. In ovarian can-cers, it is currently reported that nuclear WT1 proteinis observed in the majority of serous carcinomas andalso 0–30% of endometrioid carcinomas [19–21]. Itis unclear whether this is a reflection of the actualincidence of WT1 protein expression in these two his-totypes, or whether WT1’s efficacy as a biomarkeris confounded by the misclassification of high-gradeserous carcinomas.

Several genetic aberrations are reported to occurwith significant frequency in EOC and it is not clearwhether these aberrations might be specific to dif-ferent EOC histotypes. The most well-documentedinclude mutation in the tumour protein-53 gene(TP53 ), resulting in nuclear TP53 protein accumu-lation [22,23], aberrant activation of β-catenin pro-tein signalling, resulting in nuclear localization ofβ-catenin protein [24–26], and loss of heterozygos-ity for the tumour suppressor phosphatase and tensin

homologue (PTEN), leading to loss of PTEN pro-tein expression [27,28]. As these genetic alterationscan lead to detectable changes at the protein level,their incidence can be analysed by immunohisto-chemistry on tissue microarrays (TMA) [24,28,29].However, it should be noted that immunohistochem-istry and gene mutation analyses are not perfectlyconcordant [30,31].

The aim of this study was to better define the molec-ular characteristics of serous and endometrioid carci-noma to address the inherent problems with the currenthistopathological classification system. We analyseda published DNA microarray dataset (GSE6008 ) toidentify gene expression phenotypes for endometri-oid and serous carcinoma. Nuclear WT1 immunola-belling was investigated as a putative gynaecologicalbiomarker of serous histotype in EOC. TMAs wereused to assess the incidence of TP53, β-catenin andPTEN aberrations and correlate their occurrence withWT1. We further validated the results in an inde-pendent ovarian tumour cohort from the VancouverHospital Health Sciences Centre (VHHSC cohort).Importantly, for the VHHSC cohort, data was avail-able for both the original diagnosis at time of surgery(sDDx) and a revised diagnosis (rDDx) employingnew architectural and cytological criteria for histotypeassignment [8].

Methods

Gene expression analysis of GSE6008 EOC dataset

DNA microarray dataset accession number GSE6008[32] was downloaded from the Gene ExpressionOmnibus public database repository. This dataset wasgenerated on an Affymetrix Human Genome U133Aplatform at the University of Michigan [32]. TheGSE6008 dataset contains both endometrioid (n = 34)and serous carcinomas (n = 40). Only endometrioidand serous carcinomas were selected for this analysis.Data were pre-processed using Bioconductor [33] andR computing language. Affymetrix .CEL files under-went background subtraction, normalization and log2 transformation, using Bioconductor’s gcrma pack-age. Bioconductor’s genefilter package was used tofilter out genes with insufficient variation. Expres-sion values retained in the analyses required estimatedintensities >200 fluorescence units in at least foursamples, and a log base 2 scale of at least 0.3 forthe interquartile range. This filtering step yielded 8315probe sets. Principal component analysis (PCA), avariable reduction procedure, was preformed to reducethe gene expression data in such a way so as toemphasize similarities and differences between geneexpression profiles, using R’s stats package. Sam-ples in the PCA were labelled as WT1-positive if theexpression value was >4.5 for WT1-specific probe-set216 953 s at.

J Pathol 2010; 220: 392–400 DOI: 10.1002/pathCopyright 2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

394 J Madore et al

Table 1. Grade, stage and IHC biomarker frequency by pathology determined histotype diagnosis and WT1 determined subgroupsof endometrioid carcinoma (CHUM-ND tissue microarray cohort)

Subtype Stage I Stage II Stage III Stage IV Grade 1 Grade 2 Grade 3 WT1 TP53 β-Catenin PTEN n

Serous carcinoma 9 3 100 12 10 38 77 120 63 0 17 125Endometrioid carcinoma 21 2 14 3 18 14 8 16 12 11 12 40Endometrioid (WT1+) 2 0 10 3 4 6 6 16 8 0 4 16Endometrioid (WT1−) 18 2 4 0 14 8 2 0 4 11 8 24

Tissue microarray construction

Valid consent was obtained from the Ethics Committeeat the Centre Hospitalier de l’Universite de Montreal.Cases were selected from consenting patients to con-struct TMAs of 40 endometrioid carcinomas and 124serous carcinomas (Table 1). Representative tumourareas were annotated by a pathologist and two coreswere arrayed per case (Beecher Instruments). TMAswere cut at 4 µm thickness onto Superfrost+ glassslides.

Immunohistochemistry

The following antibodies were used: monoclonalmouse anti-human Wilms’ tumour 1 protein, 6F-H2, dilution 1 : 100; monoclonal mouse anti-humantumour-protein-53 protein, DO-7, dilution 1 : 100 (bothDakoCytomation); polyclonal rabbit anti-human β-catenin protein, Ab2982, dilution 1 : 50 (Abcam); mon-oclonal rabbit anti-human phosphatase and tensinhomologue deleted on chromosome X protein,138G6/mAb#9559, dilution 1 : 50 (Cell SignalingTechnology).

The TMA slides were heated at 55 ◦C for 15 min,deparaffinized in xylenes and rehydrated in an ethanolgradient, followed by phosphate-buffered saline (PBS)wash. Antigen retrieval, 20 min at high temperatureunder high pressure in buffer (10 mM sodium cit-rate, 0.05% Tween 20, pH 6.0) for TP53, or cit-rate–EDTA (1 mM EDTA, 0.05% Tween 20, pH 8.0),followed by 3% hydrogen peroxide in PBS, followedby 3 × 5 min PBS. Protein blocking using a serum-free reagent (X0909, DakoCytomation) for 20 min andincubated with primary antibody overnight in a humidchamber at 4 ◦C. The slides were then washed inPBS, incubated with either goat anti-mouse IgG–HRP(1 : 300) (sc-2031, Santa-Cruz Biotechnology) or rab-bit anti-goat biotin-conjugated antibody (1 : 300) (sc-2774, Santa-Cruz Biotechnology) for 30 min, followedby incubation with a streptavidin–peroxidase com-plex (DakoCytomation) for 20 min at room temper-ature. Reaction products were developed using 3-3-diaminobenzidine containing 0.3%-H2O2 for 3 min.Nuclei were counterstained with diluted haematoxylin.TMAs were observed by brightfield microscopy anddigitally imaged (Aperio ScanScope). Substitution ofthe primary antibody with IgG served as a negativecontrol.

Protein expression was assessed according to thepercentage of epithelial cells and the intensity of

staining, based on visual observation (0, negative;1, low; 2, moderate; and 3, high). All TMAs wereanalysed by two independent observers. The data werereported in a dichotomous mode; a given sample wasconsidered positive if >10% of epithelial tumour cellsshowed nuclear localization of the respective protein.For PTEN, a tumour was considered negative if at least10% of epithelial cells showed zero immunolabellingin both nuclear and cytoplasmic compartments.

Independent TMA cohort data analysis

The Genetic Pathology Evaluation Centre (GPEC;www.gpec.ubc.ca) provided the independent TMAdataset from cases at the Vancouver Hospital HealthScience Centre (VHHSC), Vancouver, BC, Canada.Consult C. B. Gilks et al [8] for GPEC Pathology,immunohistochemistry and TMA protocols.

Results

Unsupervised class discovery analysis of GSE6008gene expression

PCA of the GSE6008 [32] gene expression datasetillustrates that endometrioid tumours expressing theWT1 gene distribute closely with serous carcino-mas. Endometrioid carcinomas not expressing the WT1gene are distinguished from serous carcinomas basedon global gene expression patterns Three outlier sam-ples, which were positive for WT1 gene expressionyet grouped with the endometrioid tumours, wereobserved. (Figure 1; p = 6.03E-7, Fisher’s exact test).

WT1, TP53, β-catenin and PTENimmunohistochemistry

Using TMAs, histotype diagnosis was correlatedto immunohistochemistry data for WT1, TP53, β-cateninand PTEN. Table 1 shows a summary of theCHUM-ND cohort; Table 2 shows endometrioid car-cinomas only. Nuclear WT1 protein immunolabellingwas observed in 97% of serous carcinomas and 40% ofendometrioid carcinomas. For endometrioid tumours,WT1 immunolabelling correlated with tumour stage(Figure 2A; p = 0.009, Fisher’s exact test) and cellu-lar grade (Figure 2B; p = 0.038, Fisher’s exact test).Nuclear WT1 immunolabelling was further observedin all (n = 50) low-malignant-potential serous tumoursbut not in clear cell or mucinous EOC histotypes(data not shown). Normal ovarian surface epithelium



Figure 1. Principle component analysis of endometrioidcarcinoma and serous carcinoma. Visual representation ofthe underlying gene expression similarity between WT1+endometrioid carcinoma (triangles, �, n = 16) and serouscarcinoma (filled squares, �, n = 41) versus the characteristicgene expression phenotype of WT1− endometrioid carcinoma(circles, °, n = 18). x axis, principle component 1; y axis,principle component 2

and normal tubal epithelium were also found to beWT1-positive (data not shown). For micrographs, seeFigure 3A, B.

Nuclear accumulation of TP53 protein was observedin 55% of serous carcinomas. For endometrioid car-cinoma, 32% were positive for TP53 nuclear pro-tein accumulation (Table 2). There was a statisticalassociation between WT1 and TP53 positivity (p =0.038, Fisher’s exact test) for endometrioid carci-nomas. Briefly, 50% of WT1-positive endometrioidcarcinomas were also positive for TP53, comparedto 13% of WT1-negative endometrioid carcinomas(Table 2). Only four of 24 WT1-negative endometrioidcarcinomas exhibited TP53 protein accumulation. Twoof these three tumours represented the only grade 3,WT1-negative, endometrioid carcinomas in the cohort.

No serous carcinomas were found to have nuclearlocalized β-catenin protein. For endometrioid carci-noma, nuclear localization of β-catenin protein wasobserved in 28% of tumours by TMA. Nuclearβ-catenin protein was detected in 11/24 WT1-negativeendometrioid carcinomas and in 0/16 WT1-positiveendometrioid carcinomas (Table 2). Thus, nuclearlocalization of β-catenin was negatively associatedwith WT1 biomarker expression (Figure 4; p = 0.006,Fisher’s exact test). This suggests that β-catenin sig-nalling aberrations are exclusive to WT1-negativeendometrioid carcinomas and are not observed inWT1-positive endometrioid carcinoma or serous car-cinoma. For micrographs, see Figure 3C, D.

Loss of PTEN protein expression was observedin 30% (12/40) of endometrioid carcinomas and in13% (13/111) of serous carcinomas. PTEN loss did

Table 2. Summary of IHC-TMA data for the CHUM-NDendometrioid cohort

Sample WT1 TP53 β-Catenin PTEN∗ Stage Grade

1014 − − − − I 21180 − − − − I 11826 − − − − II 12007 − − − + III 11271 − − − + I 11276 − + − + I 21154 − − − + I 2805 − + − + I 1676 − − − + I 21012 − − − + II 21587 − − − + I 21686 − − − + I 12120 − − − + III 21049 − − + − I 21147 − − + − I 1290 − − + + I 11338 − − + + I 137 − + + − III 31697 − − + + I 1213 − − + − I 12154 − − + + I 1112 − + + − III 3627 − + + I 11328 − − + + I 1283 + − − − IV 1962 + − − − III 32255 + − − + I 1921 + − − + III 3891 + − − + III 31149 + − − − − 22108 + − − + I 1280 + − − + III 3858 + + − + III 2864 + + − + IV 2869 + + − − III 3837 + + − + III 21412 + + − + III 1893 + + − + IV 2851 + + − + III 2863 + + − + III 3

(−) tumour sample is negative for that IHC marker; (+) tumour positivefor that IHC marker.∗ For PTEN a (−) indicates loss of PTEN protein expression. (n = 40).

not correlate with any of the other biomarkers exam-ined. However, for serous carcinoma there was astrong association between loss of PTEN and betteroverall survival (p = 0.028, log rank; Figure 5). Thisassociation was not seen for endometrioid carcinoma(p = 0.319, log rank). However, for WT1-determinedsubgroups of endometrioid carcinoma, the associationbetween prognosis and PTEN loss improved for theWT1-positive group, although this did not reach sta-tistical significance (p = 0.111, log rank).

Independent cohort analysis for WT1, TP53,β-catenin and PTEN

TMA data were evaluated in an independent TMAcohort generated from EOC cases at the VancouverHospital Health Sciences Centre (VHHSC). Impor-tantly, histotype diagnosis was re-evaluated for the


396 J Madore et al

VHHSC cohort, using updated criteria for differentialhistotype diagnosis. Previously published TMA datafor WT1 and TP53, as well as recent data for β-cateninand PTEN, were analysed in the VHHSC cohort. Theoriginal sDDx and the updated rDDx was comparedto the TMA data (Table 3).

There was a significantly lower incidence of WT1protein expression in VHHSC serous carcinoma cohortthan observed for the CHUM-ND cohort (79% ver-sus 97%). For TP53, 21/26 TP53-positive sDDxendometrioid tumours were reclassed as another EOChistotype, typically serous (18/21). The remainingfive TP53-positive endometrioid tumours were WT1-negative and two of these cases were β-catenin posi-tive. Aberrant activation of β-catenin was observed in4/183 (2%) rDDx serous carcinomas and in 39/111(35%) rDDx endometrioid carcinomas. There wasno significant change in PTEN loss after rDDx(Table 3).

Discussion

Principal component analysis of the GSE6008 geneexpression dataset by Wu et al identified two dis-tinctive subgroups of endometrioid carcinoma, basedon variance in their global gene expression pat-terns [32]. One of these subgroups was highly sim-ilar to serous carcinoma and tended to be of higher

A

B

Figure 2. Clinical parameters in sDDx endometrioidcarcinoma vs. WT1 biomarker expression, as determinedby immunohistochemistry on CHUM-ND tissue microarrays.(A) sDDx endometrioid carcinoma vs. pathoclinical stage.(B) sDDx endometrioid carcinoma vs. pathoclinical grade.CHUM-ND cohort (n = 40)

tumour grade [32]. Genetic annotation of the GSE6008dataset also revealed that p53 mutation was com-mon among those endometrioid carcinomas with aserous-like gene expression profile [32]. Moreover,deregulated β-catenin signalling and defects in thePI3K–Pten pathway were shown to be typical amongthose endometrioid carcinomas that did not share geneexpression homology to serous carcinoma and whichalso tended to be low-grade [32]. In a further anal-ysis of the GSE6008 dataset, WT1 gene expressionis demonstrably associated to those endometrioid car-cinomas with a serous-like gene expression profile(Figure 1; p = 6.03E-7, Fisher’s exact test).

Three WT1-positive outlier tumours are observed byPCA, two of which grouped directly with the WT1-negative endometrioid tumours. A singular pitfall ofsome microarray analysis is contamination by multi-ple tissue compartments. In the case of WT1, bothstromal and endothelial cells have been demonstratedto express WT1 gene products and may explain theconfounding outliers [34,35]. To address this issue, welooked at epithelial expression and nuclear localizationof WT1 protein by IHC. Expression of cytoplasmic orstromal WT1 protein was considered artefactual to theutility of WT1 as an IHC biomarker of serous celltype.

We correlated the IHC results for epithelial–nuclearWT1 immunolabelling with TMA-based immunohis-tochemical data for TP53, PTEN and β-catenin. TheTMA data corroborated those trends observed by geneexpression analysis. Endometrioid carcinomas posi-tive for nuclear WT1 immunolabelling tended to behigh-grade and to show nuclear TP53 protein accu-mulation, a characteristic of serous carcinoma, anddid not show aberrant β-catenin activation, a charac-teristic of low-grade endometrioid carcinoma. WT1-negative endometrioid carcinomas tended to be low-grade and TP53-negative and to frequently acquireactivated β-catenin signalling. Loss of PTEN occurredat low frequency in both endometrioid and seroustumours and did not associate with WT1 determinedsubgroups of endometrioid carcinoma.

The gene expression and TMA data suggest thatWT1-positive endometrioid carcinomas tend not toresemble their low-grade, WT1-negative counterparts.Furthermore, given their molecular similarity to serouscarcinomas, it is possible that they are misclassifiedserous carcinomas or morphological variants sharinga common somatic background. The divergence alsosuggests that low-grade endometrioid carcinomas area molecularly distinctive disease with an alternativepathobiology or somatic origin.

Alternative somatic origins for low-grade and high-grade endometrioid tumours is also corroborated byWT1 biomarker expression patterns in the puta-tive normal gynaecological epithelia from whichEOCs are thought to arise. The epithelial cells ofbenign endometriosis are WT1-negative, leading tothe conjecture that WT1-negative EOCs arise from



Figure 3. Sample immunohistochemical microphotographs of ovarian endometrioid carcinoma. (A) WT1−; (B) WT1+;(C) β-catenin+; (D) β-catenin−; (E) TP53−; (F) TP53+ (×20 objective lens)

Figure 4. Detection of aberrant β-catenin signalling, aberrantTP53 signalling and loss of PTEN protein expression, in WT1determined subgroups of ovarian endometrioid carcinoma byIHC-TMA. CHUM-ND cohort (n = 40)

WT1-negative endometriosis lesions. Likewise, WT1-positive EOC tumours may arise from WT1-positivegynaecological epithelia, such as the ovarian surfaceepithelia or Fallopian tube epithelia. However, it isunclear whether epigenetic loss or gain of WT1 geneexpression might confound WT1’s usefulness as abiomarker capable of differentiating somatic origin orhistotype diagnosis for EOC [36–38].

Typically, WT1-negative endometrioid carcinomasare low-grade and low-stage, while WT1-positiveendometrioid carcinomas are high-grade and high-stage (Figure 2; p = 0.038 and 0.009, Fisher’s exacttest). However, two high-grade, high-stage, WT1-negative endometrioid carcinomas were identified inthe CHUM-ND cohort. These two samples wereamong four of 24 WT1-negative endometrioid car-cinomas identified in the CHUM-ND cohort as hav-ing TP53 aberrations, and the only two sampleswith concurrent β-catenin and TP53 aberrations. Oneof these two samples was spontaneously immortal-ized in cell culture and gave rise to an aggres-sive and well-characterized endometrioid carcinomacell line, TOV112D [39–48]. Thus, while WT1-negative endometrioid carcinomas do not character-istically acquire TP53 defects, they do occur and mayaffect a more aggressive phenotype from these typi-cally indolent tumours. Importantly, this implies theexistence of high-grade endometrioid tumours that arenot misclassified serous carcinomas.

To further validate these observations, data from anindependent cohort at the VHHSC was analysed. Inthe VHHSC cohort there was a significantly lowerincidence of WT1 positivity among serous carcino-mas when compared to the CHUM-ND cohort (79%


398 J Madore et al

Figure 5. For serous carcinoma, loss of PTEN protein expression was associated with better prognosis in the CHUM-ND cohortbut not the VHHSC cohort. (A) CHUM-ND serous carcinoma (p = 0.028 , log rank; n = 125). (B) VHHSC serous carcinoma(p = 0.770, log rank; n = 200)

Table 3. Distribution of IHC biomarkers in VHHSC tissuemicroarray cohort for endometrioid and serous carcinomas atoriginal diagnosis (sDDx) and at revised diagnosis (rDDx)

HistotypeDDx WT1+ TP53+ β-Catenin+ PTEN−

sDDx serous (124/180) (76/178) (5/173) (24/180)n = 180 69% 43% 3% 13%rDDx serous (149/189) (88/187) (4/183) (26/200)n = 200 79% 47% 2% 13%sDDx endometrioid (23/152) (26/184) (39/152) (42/150)n = 184 21% 18% 26% 28%rDDx endometrioid (3/111) (7/107) (39/111) (39/117)n = 117 3% 7% 35% 33%

vs. 97%). This could be explained by alternative pro-tocols for antigen retrieval, antibody titre or immuno-histochemistry cut-point calculation [49]. The discrep-ancy may also be caused by differences in the fixa-tion and archiving of surgical samples between insti-tutions, leading to differences in antigen preserva-tion. Regardless, the disagreement in WT1 immunola-belling between the cohorts raises technical questions,as well as biobanking questions, that need further con-sideration for the study of WT1 as biomarker in EOC.

After pathological review using updated cytologi-cal and architectural criteria, 90% of WT1-positiveendometrioid carcinomas from the VHHSC cohortwere reassigned to serous histotype [8]. Thus, con-clusions from the current CHUM-ND study, that WT1is a useful biomarker that can aid in the determinationof histotype, are in keeping with conclusions from theprevious VHHSC study [8]. However, the corollaryargument, that negative WT1 biomarker labelling indi-cates a non-serous EOC histotype, was not validated,given the lower incidence of WT1-positive serous car-cinomas in the VHHSC cohort. Thus, the differentialdiagnosis of true high-grade endometrioid and serouscarcinoma remains problematic.

In the CHUM-ND cohort, loss of PTEN waslinked to a subset of serous carcinomas that had afavourable prognosis (Figure 5A). However, the asso-ciation between loss of PTEN and prognosis was

not validated using the VHHSC cohort (Figure 5B).In this instance the disparity is due to differencesin the composition of the two cohorts. The VHHSCcohort included only high-grade, optimally-debulked,serous carcinomas. In contrast, the CHUM-ND cohortincluded all serous carcinomas, regardless of clinicalparameters, including a subset of low-grade serous car-cinomas. It has been suggested that low-grade seroustumours are a unique disease entity and not a con-tinuum of disease progression to their more commonhigh-grade counterparts [50,51]. Loss of PTEN, then,is a molecular lesion that occurs irregardless of histo-type and is not a lesion that is specific to low-gradeendometrioid carcinomas, but also to low-grade serouscarcinomas.

Currently all ovarian carcinomas are treated withsimilar therapies, therefore accurate subtyping,although intellectually satisfying, is arguably not clin-ically relevant. However, it may be that subtype-specific therapeutic options for EOC are needed, mak-ing accurate subtyping an important goal. These dataprovide a genomics-based and immunohistochemistry-supported rationale to reclassify high-grade WT1-positive endometrioid ovarian carcinomas as high-grade serous carcinomas. Emerging trends in themolecular description of ovarian cancer point to adiverse group of tumours with molecular and clinicaldifferences, not only between tumours of different his-totypes but also between definable molecular subtypeswithin tumours of the same histotype.

Acknowledgements

The authors are very grateful to the staff and patients of theGynaecological Oncology Service at the Hopital Notre-Dameand at the Vancouver Hospital Health Sciences Centre. Wethank Manon de Ladurantaye and Louise Champoux for theircontributions. This work was supported by a grant from theCanadian Institutes of Health Research (CIHR) to A-MM-M,PNT and DMP. Tumour banking was supported by the Banquede Tissus et de Donnees of the FRSQ, affiliated with theCTRNet. DH and the Ovarian Cancer Research Programmeof British Columbia (OvCaRe) and the Genetic Pathology



Evaluation Centre (GPEC) research units have received supportfrom the Michael Smith Foundation for Health Research.

References

1. Gilks CB. Subclassification of ovarian surface epithelial tumorsbased on correlation of histologic and molecular pathologic data.Int J Gynecol Pathol 2004;23(3):200–205.

2. Pieretti M, Hopenhayn-Rich C, Khattar NH, Cao Y, Huang B,Tucker TC. Heterogeneity of ovarian cancer: relationships amonghistological group, stage of disease, tumor markers, patientcharacteristics, and survival. Cancer Invest 2002;20(1):11–23.

3. Wang V, Li C, Lin M, Welch W, Bell D, Wong YF, et al. Ovariancancer is a heterogeneous disease. Cancer Genet Cytogenet2005;161(2):170–173.

4. Bell DA. Origins and molecular pathology of ovarian cancer. ModPathol 2005;18(suppl 2):S19–32.

5. Caduff RF, Svoboda-Newman SM, Ferguson AW, Johnston CM,Frank TS. Comparison of mutations of Ki-RAS and p53immunoreactivity in borderline and malignant epithelial ovariantumors. Am J Surg Pathol 1999;23(3):323–328.

6. Seidman JD, Kurman RJ. Pathology of ovarian carcinoma.Hematol Oncol Clin N Am 2003;17(4):909–925, vii.

7. Cramer SF, Roth LM, Ulbright TM, Mazur MT, Nunez CA,Gersell DJ, et al. Evaluation of the reproducibility of theWorld Health Organization classification of common ovariancancers. With emphasis on methodology. Arch Pathol Lab Med1987;111(9):819–829.

8. Gilks CB, Ionescu DN, Kalloger SE, Kobel M, Irving J, Clarke B,et al. Tumor cell type can be reproducibly diagnosed and is ofindependent prognostic significance in patients with maximallydebulked ovarian carcinoma. Hum Pathol 2008;39(8):1239–1251.

9. Schwartz DR, Kardia SL, Shedden KA, Kuick R, Michailidis G,Taylor JM, et al. Gene expression in ovarian cancer reflectsboth morphology and biological behavior, distinguishing clearcell from other poor-prognosis ovarian carcinomas. Cancer Res2002;62(16):4722–4729.

10. Hilton JL, Geisler JP, Rathe JA, Hattermann-Zogg MA, DeY-oung B, Buller RE. Inactivation of BRCA1 and BRCA2 in ovariancancer. J Natl Cancer Inst 2002;94(18):1396–1406.

11. Scully RE. Pathology of ovarian cancer precursors. J Cell BiochemSuppl 1995;23:208–218.

12. Piek JM, Kenemans P, Verheijen RH. Intraperitoneal serousadenocarcinoma: a critical appraisal of three hypotheses on itscause. Am J Obstet Gynecol 2004;191(3):718–732.

13. Medeiros F, Muto MG, Lee Y, Elvin JA, Callahan MJ, Felt-mate C, et al. The tubal fimbria is a preferred site for early adeno-carcinoma in women with familial ovarian cancer syndrome. AmJ Surg Pathol 2006;30(2):230–236.

14. Kindelberger DW, Lee Y, Miron A, Hirsch MS, Feltmate C,Medeiros F, et al. Intraepithelial carcinoma of the fimbria andpelvic serous carcinoma: evidence for a causal relationship. AmJ Surg Pathol 2007;31(2):161–169.

15. Lee Y, Miron A, Drapkin R, Nucci MR, Medeiros F, Saleemud-din A, et al. A candidate precursor to serous carcinoma that orig-inates in the distal fallopian tube. J Pathol 2007;211(1):26–35.

16. Bell KA, Kurman RJ. A clinicopathologic analysis of atyp-ical proliferative (borderline) tumors and well-differentiatedendometrioid adenocarcinomas of the ovary. Am J Surg Pathol2000;24(11):1465–1479.

17. Dinulescu DM, Ince TA, Quade BJ, Shafer SA, Crowley D,Jacks T. Role of K-ras and Pten in the development of mousemodels of endometriosis and endometrioid ovarian cancer. NatMed 2005;11(1):63–70.

18. Shimizu M, Toki T, Takagi Y, Konishi I, Fujii S. Immunohisto-chemical detection of the Wilms’ tumor gene (WT1 ) in epithelialovarian tumors. Int J Gynecol Pathol 2000;19(2):158–163.

19. Acs G, Pasha T, Zhang PJ. WT1 is differentially expressed inserous, endometrioid, clear cell, and mucinous carcinomas of theperitoneum, fallopian tube, ovary, and endometrium. Int J GynecolPathol 2004;23(2):110–118.

20. Waldstrom M, Grove A. Immunohistochemical expression ofWilms’ tumor gene protein in different histologic subtypes ofovarian carcinomas. Arch Pathol Lab Med 2005;129(1):85–88.

21. Cathro HP, Stoler MH. The utility of calretinin, inhibin, andWT1 immunohistochemical staining in the differential diagnosisof ovarian tumors. Hum Pathol 2005;36(2):195–201.

22. Berchuck A, Kohler MF, Marks JR, Wiseman R, Boyd J, Bast RCJr. The p53 tumor suppressor gene frequently is altered ingynecologic cancers. Am J Obstet Gynecol 1994;170(1, pt1):246–252.

23. Buller RE, Lallas TA, Shahin MS, Sood AK, Hatterman-Zogg M,Anderson B, et al. The p53 mutational spectrum associatedwith BRCA1 mutant ovarian cancer. Clin Cancer Res2001;7(4):831–838.

24. Palacios J, Gamallo C. Mutations in the β-catenin gene(CTNNB1 ) in endometrioid ovarian carcinomas. Cancer Res1998;58(7):1344–1347.

25. Gamallo C, Palacios J, Moreno G, Calvo de Mora J, Suarez A,Armas A. β-Catenin expression pattern in stage I and IIovarian carcinomas : relationship with β-catenin gene mutations,clinicopathological features, and clinical outcome. Am J Pathol1999;155(2):527–536.

26. Willner J, Wurz K, Allison KH, Galic V, Garcia RL, Goff BA,et al. Alternate molecular genetic pathways in ovarian carcinomasof common histological types. Hum Pathol 2007;38(4):607–613.

27. Obata K, Hoshiai H. Common genetic changes betweenendometriosis and ovarian cancer. Gynecol Obstet Invest2000;50(suppl 1):39–43.

28. Kurose K, Zhou XP, Araki T, Cannistra SA, Maher ER, Eng C.Frequent loss of PTEN expression is linked to elevatedphosphorylated Akt levels, but not associated with p27 and cyclinD1 expression, in primary epithelial ovarian carcinomas. Am JPathol 2001;158(6):2097–2106.

29. Chang K, Ding I, Kern FG, Willingham MC. Immunohistochem-ical analysis of p53 and HER-2/neu proteins in human tumors.J Histochem Cytochem 1991;39(9):1281–1287.

30. Katabuchi H, Tashiro H, Cho KR, Kurman RJ, Hedrick Ellen-son L. Micropapillary serous carcinoma of the ovary: an immuno-histochemical and mutational analysis of p53 . Int J Gynecol Pathol1998;17(1):54–60.

31. Nenutil R, Smardova J, Pavlova S, Hanzelkova Z, Muller P,Fabian P, et al. Discriminating functional and non-functional p53in human tumours by p53 and MDM2 immunohistochemistry.J Pathol 2005;207(3):251–259.

32. Wu R, Hendrix-Lucas N, Kuick R, Zhai Y, Schwartz DR,Akyol A, et al. Mouse model of human ovarian endometrioidadenocarcinoma based on somatic defects in the Wnt/β-catenin andPI3K/Pten signaling pathways. Cancer Cell 2007;11(4):321–333.

33. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M,Dudoit S, et al. Bioconductor: open software developmentfor computational biology and bioinformatics. Genome Biol2004;5(10):R80.

34. Goldstein NS, Uzieblo A. WT1 immunoreactivity in uterinepapillary serous carcinomas is different from ovarian serouscarcinomas. Am J Clin Pathol 2002;117(4):541–545.

35. Sumathi VP, Al-Hussaini M, Connolly LE, Fullerton L, McClug-gage WG. Endometrial stromal neoplasms are immunoreactivewith WT-1 antibody. Int J Gynecol Pathol 2004;23(3):241–247.

36. Kobel M, Kalloger SE, Boyd N, McKinney S, Mehl E, Palmer C,et al. Ovarian carcinoma subtypes are different diseases:implications for biomarker studies. PLoS Med 2008;5(12):e232.

37. Barbolina MV, Adley BP, Shea LD, Stack MS. Wilms tumorgene protein 1 is associated with ovarian cancer metastasis andmodulates cell invasion. Cancer 2008;112(7):1632–1641.

38. Hogdall EV, Christensen L, Kjaer SK, Blaakaer J, Christensen IJ,Gayther S, et al. Expression level of Wilms’ tumor 1 (WT1)protein has limited prognostic value in epithelial ovarian cancer:from the Danish ‘MALOVA’ ovarian cancer study. Gynecol Oncol2007;106(2):318–324.

39. Tonin PN, Hudson TJ, Rodier F, Bossolasco M, Lee PD, Novak J,et al. Microarray analysis of gene expression mirrors the biologyof an ovarian cancer model. Oncogene 2001;20(45):6617–6626.


400 J Madore et al

40. Wu R, Zhai Y, Fearon ER, Cho KR. Diverse mechanisms of β-catenin deregulation in ovarian endometrioid adenocarcinomas.Cancer Res 2001;61(22):8247–8255.

41. Manderson EN, Mes-Masson AM, Novak J, Lee PD, ProvencherD, Hudson TJ, et al. Expression profiles of 290 ESTs mappedto chromosome 3 in human epithelial ovarian cancer cell linesusing DNA expression oligonucleotide microarrays. Genome Res2002;12(1):112–121.

42. Presneau N, Mes-Masson AM, Ge B, Provencher D, Hudson TJ,Tonin PN. Patterns of expression of chromosome 17 genes inprimary cultures of normal ovarian surface epithelia and epithelialovarian cancer cell lines. Oncogene 2003;22(10):1568–1579.

43. Samouelian V, Maugard CM, Jolicoeur M, Bertrand R, ArcandSL, Tonin PN, et al. Chemosensitivity and radiosensitivity profilesof four new human epithelial ovarian cancer cell lines exhibitinggenetic alterations in BRCA2, TGFβ-RII, KRAS2, TP53 and/orCDNK2A. Cancer Chemother Pharmacol 2004;54(6):497–504.

44. Gagne JP, Gagne P, Hunter JM, Bonicalzi ME, Lemay JF,Kelly I, et al. Proteome profiling of human epithelial ovarian can-cer cell line TOV-112D. Mol Cell Biochem 2005;275(1–2):25–55.

45. Smith JA, Ngo H, Martin MC, Wolf JK. An evaluation ofcytotoxicity of the taxane and platinum agents combinationtreatment in a panel of human ovarian carcinoma cell lines.Gynecol Oncol 2005;98(1):141–145.

46. Popadiuk CM, Xiong J, Wells MG, Andrews PG, Dankwa K,Hirasawa K, et al. Antisense suppression of pygopus2 resultsin growth arrest of epithelial ovarian cancer. Clin Cancer Res2006;12(7, pt 1):2216–2223.

47. Birch AH, Quinn MC, Filali-Mouhim A, Provencher DM, Mes-Masson AM, Tonin PN. Transcriptome analysis of serous ovariancancers identifies differentially expressed chromosome 3 genes.Mol Carcinog 2007;47:56–65.

48. Zietarska M, Maugard CM, Filali-Mouhim A, Alam-Fahmy M,Tonin PN, Provencher DM, et al. Molecular description of a 3Din vitro model for the study of epithelial ovarian cancer (EOC).Mol Carcinog 2007;46(10):872–885.

49. McCabe A, Dolled-Filhart M, Camp RL, Rimm DL. Automatedquantitative analysis (AQUA) of in situ protein expression,antibody concentration, and prognosis. J Natl Cancer Inst2005;97(24):1808–1815.

50. Shih Ie M, Kurman RJ. Ovarian tumorigenesis: a proposed modelbased on morphological and molecular genetic analysis. Am JPathol 2004;164(5):1511–1518.

51. Schmeler KM, Gershenson DM. Low-grade serous ovarian cancer:a unique disease. Curr Oncol Rep 2008;10(6):519–523.


characterization of the molecular differences between ovarian endometrioid carcinoma and ovarian...

Documents