SET complex in serous epithelial ovarian cancer

SET complex in serous epithelial ovarian cancer

V�eronique Ouellet1, C�ecile Le Page

1, Marie-Claude Guyot

1, Christian Lussier

2, Patricia N. Tonin

3–5,

Diane M. Provencher1,6,7 and Anne-Marie Mes-Masson1,7*

1Centre de Recherche du Centre Hospitalier de l’Universit�e de Montr�eal (CHUM), Institut du cancer de Montr�eal, Montreal, Canada2Department of Pathology, Centre Hospitalier de l’Universit�e de Montr�eal (CHUM), Montreal, Canada3Department of Human Genetics, McGill University, Montreal, Canada4Research Institute of McGill University Health Centre, Montreal, Canada5Department of Medicine, McGill University, Montreal, Canada6Division of Gynecologic Oncology, Universit�e de Montr�eal, Montreal, Canada7Department of Medicine, Universit�e de Montr�eal, Montreal, Canada

With low cure rates but increasing diverse treatment options thatprovide variable remission times, ovarian cancer is increasinglybeing recognized as a chronic disease. This reality indicates theneed for a better understanding of factors influencing disease pro-gression. In a previous global analysis of gene expression, we iden-tified genes differentially expressed when comparing serous epi-thelial ovarian tumors of low and high malignant potential (grade0 vs grade 3). In this analysis, 4 out of 5 members of the SET com-plex, SET, APE1, NM23 and HMGB2, were highly expressed ininvasive grade 3 tumors. To further investigate the expression ofthese genes and the fifth member of the SET complex (pp32), weperformed immunohistochemistry, on a tissue array composed of235 serous tumors of different grades and disease stages. A signifi-cant correlation between expression of all SET complex proteinsand the tumor differentiation was observed (p < 0.05). When com-bining all tumors, overexpression of Nm23 (p 5 0.04), Set (p 50.004) and Ape1 (p 5 0.004) was associated with the clinical stageof the disease. No marker by itself was associated with prognosis.The combination of a high level of Nm23 in the context of a lowlevel of Set compared to all other combinations of these markersdid confer a better prognosis (p 5 0.03). When combined, highexpression of Hmgb2 and low expression of Ape1 was also associ-ated with patient prognosis (p 5 0.05). These findings suggest thata strategy that sums the activities of different partners within apathway may be more appropriate in designing nomograms forpatient stratification.' 2006 Wiley-Liss, Inc.

Key words: serous epithelial ovarian cancer; low malignant potential/borderline tumors; diagnostic and prognositic markers; SET complex;immunohistochemistry

Among the gynecologic malignancies, epithelial ovarian cancer(EOC) is infrequent but it is the most lethal mainly because of theabsence of symptoms at the earliest stages of the disease.1 The sur-vival at 5 years of patient with an advanced stage disease reachesonly 30%. EOC is a complex disease and tumors can be subdi-vided as low malignant potential (LMP) or borderline and invasive(TOV) tumors. Both types of tumors present multilayer prolifera-tion but TOVs show higher level of cellular atypia than LMPtumors. LMP tumors do not possess the capacity to invade thestroma although microinvasion can be observed.2–4 In EOC, 4main histopathology types can be observed (serous, endometrioid,mucinous and clear cell) and the most common type is serous. Ma-lignant tumors of EOC can present different degrees of differentia-tion. LMPs are the most differentiated tumors and referred to aseither grade 0 (G0) or B, and TOVs can be well (G1), moderately(G2) or poorly (G3) differentiated. Clinical staging in EOC variesfrom stage I to IV, where stage I represents disease limited to 1 orboth ovaries, stage II is associated with pelvic extension, stage IIIcorresponds to spreading within the abdominal cavity and patientswith stage IV tumors present liver or distant metastasis.5,6

We have previously defined, using Affymetrix gene expressionmicroarray analysis, a molecular signature of serous LMPs andTOVs of G3 either in primary culture or in tumor tissue modelsystem.7 This signature was established using 3 different statisticalmethods (Signal-to-noise ratio, Mann-Whitney U test and Signifi-

cance Analysis of Microarray) and a subset of genes were vali-dated by quantitative-PCR.7 Based on this analysis, we found 4out of the 5 members of the SET complex, composed of NM23(also known as NME1, NDPKA, GAAD, PP2A), pp32 (PHAP1,LANP, ANP32, I1PP2a), Set (I2PP2A, IGAAD, PHAP2), Hmgb2(HMG2) and Ape1 (APEX, REF1, HAP1). This complex is impli-cated in apoptosis induced by granzyme-A triggered by immunecells (cytotoxic T lymphocytes or NK cells)8 that recognize eitherviral infected or tumor cells. Following lymphocyte attack, thetryptase granzyme-A enters cells using membrane holes createdby the perforin and cleaves specific substrates including 3 mem-bers of the SET complex, Set, Hmgb2 and Ape1. The cleavage ofthese proteins leads to release of the nuclease activity of Nm23which in turn results in single-stranded cleavage of DNA and cas-pase-independent apoptosis.9 The complex is also implicated inresponse to oxidative stress and DNA repair.10–12

In this report, we focused on protein expression of the SETcomplex proteins in EOC. For this purpose, we used a tissue arraycomposed of 235 serous tumor samples of different grades to eval-uate protein expression. The staining intensity was related to tu-mor grade, disease stage and patient prognosis in order to evaluatetheir usefulness in the stratification of EOC tumors.

Material and methods

Patients and tissue specimens

Following appropriate consent, tumor samples were collectedthrough the Division of Gynecologic Oncology at the Centre hos-pitalier de l’Universit�e de Montr�eal (Hopital Notre-Dame). An in-dependent pathologist reviewed and scored tumor samples accord-ing to the Federation International of Gynecology and Obstetrics(FIGO) criteria for histopathology, grade and stage of the disease.6

Clinical data including diagnosis, treatment and clinical outcomessuch as disease free interval and survival rate were extracted fromthe Systeme d’Archivage des Donn�ees en Oncologie (SARDO).For the study, we focused on samples of serous histopathologyobtained from chemotherapy na€ıve patients.

Serous epithelial ovarian cancer tissue array

Using a hematoxilin-eosin stained slide to guide appropriate sam-ple selection, we arrayed 2 cores (0.6 mm diameter) from each tis-sue sample. An experienced pathologist reviewed samples. The se-rous tissue array composed of 56 LMPs, and 179 invasive tumors ofG1 (n 5 11), G2 (n 5 53) and G3 (n 5 115) was sectioned, stainedwith hematoxilin-eosin and received a final pathology review. For

Grant sponsor: Canadian Institutes for Health Research; Grant number:MOP-36056.*Correspondence to: CR-CHUM/ICM, 1560, rue Sherbrooke est, Mon-

treal, Quebec, Canada H2L 4M1. Fax:1514-412-7703.E-mail: [email protected] 6 February 2006; Accepted 28 March 2006DOI 10.1002/ijc.22054Published online 5 July 2006 in Wiley InterScience (www.interscience.

wiley.com).

Int. J. Cancer: 119, 2119–2126 (2006)' 2006 Wiley-Liss, Inc.

Publication of the International Union Against Cancer

immunohistochemistry, the tissue array was cut in 5 mm sectionsand stained with appropriate antibodies and controls.

Antibodies

For immunohistochemistry analysis, the following antibodieswere used: anti-I1PP2A goat polyclonal (sc-5652), anti-I2PP2Agoat polyclonal (sc-5655) and anti-Ref-1 mouse monoclonal (sc-17774) were purchased from Santa Cruz Biotechnology (SantaCruz, CA). The antibody anti-nm23 rabbit polyclonal antibody(RB-116) was purchased from Medicorp (Montreal, Quebec, Can-ada). The anti-Hmgb2 rabbit polyclonal (556529) was purchasedfrom BD PharMingen (San Diego, CA).

Immunohistochemistry

Tissue arrays were stained by an immunoperoxidase method asdescribed previously.13 Briefly, tissue sections were heated to60�C for 30 min, deparaffinized in toluene and rehydrated in anethanol gradient. Antigen retrieval was performed using a micro-wave by boiling in citrate buffer (0.01 M citric acid adjusted to pH6.0). A protein blocking serum-free reagent (DakoCytomation,Mississauga, ON, Canada) was used followed by incubation withthe primary antibody (60 min.) at room temperature. Tissues wereincubated with either a secondary biotinylated antibody (DakoCytomation Inc) or with a rabbit anti-goat biotin-conjugated anti-body (1:300) (sc-2774, Santa Cruz Biotechnology) for 20 min fol-lowed by incubation with a streptavidin–peroxidase complex(DakoCytomation) for 20 min at room temperature. Liquid diami-nobenzidine was used to visualize the reaction (DakoCytomation)and nuclei were counterstained with hematoxylin. Negative controlswere performed using phosphate buffered saline instead of the pri-mary antibody. Protein expression was scored according to theextent (as a percentage of total malignant cells) and intensity (valueof 0 for absence, 1 for low, 2 for moderate and 3 for high intensity)of staining based on visualization. All slides were visualized bylight microscopy at 203 magnification and scored in a blind studyby 2 independent observers with a correlation of >90%.

Statistical analysis

A Mann-Whitney U test was performed to evaluate the associa-tion between immunohistochemistry staining intensity and clinicalvariables (grades and stages). The clinical stages were divided into2 groups: early (stages I–II) and advanced (stages III–IV). Spear-man test was performed to evaluate a correlation between proteinexpression and clinical parameters. Significance of markers to pre-dict relapse and survival of EOC patients was performed usingKaplan-Meier survival curve coupled to a log rank test. Thethreshold of the intensity used in the log rank test to discriminatethe 2 groups of low or high intensity were fixed based on the re-ceiver operating characteristics (ROC) curves. ROC curves wereused to establish the optimal threshold of staining intensity thatyielded the best possible sensitivity and specificity values of eachmarker to predict the invasive potential (G0 compared to G1, G2and G3). Statistical significance was set at p < 0.05. Statisticalanalyses were performed using the SPSS software version 11(SPSS, Chicago, IL).

Results

Selection of SET complex members

In our previously reported gene expression microarray analyses,we compared tumor tissues from 6 LMPs against 12 TOVs of G3,or primary cultures from 6 LMPs against 8 G3 TOVs.7 This studyallowed us to observe the differential expression of 4 (SET, APE1,NM23 and HMGB2) out of 5 genes included in the SET complex.7

The fifth member of the complex (pp32) could not be assessed bythis method since it fell below the sensitivity of the microarray. Intumor tissues, HMGB2 and APE1, and to a lesser extent NM23and SET, where more highly expressed in TOVs than LMPs.7

Staining of SET complex proteins in tumor tissuesand association with histological grade

We evaluated protein expression of SET complex members todetermine their usefulness as diagnostic and/or prognostic markersin EOC. To this end, we used a tissue array composed of 235 se-rous EOC tumors of different grades and stages (Table I). In Fig-ure 1, pictures of a core from different grades stained with differ-ent antibodies are shown. The staining of all proteins wasobserved in the nucleus and the cytoplasm. Protein expression ofthe SET complex members was assessed to evaluate their associa-tion with tumor differentiation. All markers, except pp32 (p 50.09), showed a significant differential expression between LMPscompared to all other TOVs (p < 0.001 for Hmgb2, Set and Ape1and p 5 0.05 for Nm23) (Table II). Two markers, Set and Ape1,showed differential expression even within highly differentiatedtypes of tumors (G0 vs G1) (p 5 0.004 and p5 0.02, respectively)(Table II). All proteins presented differential expression betweentumors of G0 and those of G3 representing the 2 extremes in tu-mor differentiation (p < 0.001 for Set, Hmgb2 and Ape1, p 50.02 for Nm23 and p 5 0.04 for pp32) (Table II). Differentialexpression of Set, Hmgb2 and Ape1 was also observed betweentumors of G0 and those of G2 (p < 0.001, all) (Table II). Tumorsof G1 and G2 differentially expressed Nm23 (p 5 0.04), pp32(p 5 0.03) and Ape1 (p 5 0.04) (Table II). In addition to theseproteins, expression of Hmgb2 (p 5 0.01) was also differentiallyexpressed between tumors of G1 and those of G3 (p 5 0.04 forNm23, p5 0.02 for pp32 and p 5 0.02 for Ape1) (Table II).

Association of staining intensity and clinical stage

The clinical stage is an important factor in EOC to evaluatepatient prognosis. We therefore assessed the protein expression ofthe SET complex members to determine their association withclinical stage using a Mann-Withney U test. High expression ofNm23 (p 5 0.04), Set (p 5 0.004) and Ape1 (p 5 0.004) wasassociated with advanced stage when all tumors were included inthe analysis (Table III). However, when the 2 classes (LMPs andTOVs) were separated, this differential expression was notobserved although a trend was noted for Nm23 (p 5 0.06) inLMPs (Table III).

Association of SET complex proteins and patient prognosis

Finally, we evaluated whether the SET complex proteins wereassociated with patient outcome. We assessed the staining intensityto predict overall survival of the patients with a minimum of 18months of follow up. Despite their association with the differentia-tion of the tumor, none of the SET complex proteins alone wasassociated with prognosis of patient with invasive EOC (p > 0.05)(Table IV). However, when we tried all possible two-way combina-

TABLE I – DESCRIPTION OF THE SEROUS TISSUE ARRAY

Serous array samplesTumor grade

Grade 0 Grade 1 Grade 2 Grade 3

Number of tumors 56 11 53 115Mean age 6 S.D. (years) 496 14 41 6 15 606 12 62 6 10Disease staging

Stage I–II (early) 21 1 3 10Stage III–IV (advanced) 35 10 50 114

Risk of LMPsLow 31 – – –High 23 – – –

Tumor residuum>2 cm – 6 19 47<2 cm – 5 23 44

Disease free intervalMean 6 S.D. (months) 526 36 51 6 49 136 10 16 6 12Censured patient 7/36 3/6 24/30 26/48

SurvivalMean 6 S.D. (months) 586 33 56 6 47 286 14 28 6 16Censured patient 0/41 2/7 12/37 50/62

2120 OUELLET ET AL.

tion of SET complex proteins we were able to find combinationsthat were associated with prognosis. In particular, we noted thathigh expression of Nm23 with low expression of Set was associated

with a better prognosis when compared to all other possible combi-nations of these 2 markers (p 5 0.03) (Fig. 2a). Furthermore, weobserved that high expression of Hmgb2 and low expression of

FIGURE 1 – Expression of the SET complex proteins on epithelial ovarian tumor tissues of different grades varying from G0 to G3. Represen-tative images of immunoperoxidase-stained cores are shown for each protein and grade (320 magnification). Nuclear and/or cytoplasmic stain-ing was observed for all proteins. Nm23 and pp32 presented differential expression between G0/G3 and G1/G2-G3. Hmgb2 showed differentialexpression between G3/G0–G1–G2 and G0/G2. Finally, expression of Set is differentially expressed between G0/G1–G2–G3 and Ape1 betweenG0/G1–G2–G3 and G1/G2–G3.

2121SET COMPLEX IN SEROUS OVARIAN TUMORS

Ape1 was also associated with a better prognosis (p5 0.05) (Fig. 2b).Of the 21 patients that had pair-wise combinations of markers and abetter prognosis, only 4 had high Nm23, high Hmgb2, low Set andlow Ape1. We also evaluated the association of staining intensity withdisease-free interval representing the time between initial treatmentand the first disease relapse defined by the RECIST criteria.14 Whenthe relapse cut off was set at 18 months, no significant association wasfound although Nm23 (p5 0.08) tended to associate with the disease-free interval (Table IV).

Discussion

Gene expression microarray analysis based on the comparisonof LMPs and G3 TOVs allowed us to identify candidates that aredifferentially expressed in EOC.7 Identification of members of

SET complex led to the analysis of protein expression of complexmembers in tumor tissues and their correlation to clinical parame-ters in serous EOC patients.

The SET complex proteins are ubiquitously expressed eitherassociated with the endoplasmic reticulum or in the nucleus.While cytoplasmic localization rather than specific endoplasmicreticulum localization was assessed here, the cellular localizationof the SET complex proteins found in this study is consistent withthe literature. The role of the SET complex in cancer can be due tothe oxidative stress presents in tumors. Following oxidative stress,SET members are found in the nucleus where cooperation be-tween them was proposed. Set and pp32 favor DNA accessibil-ity15–18 allow recognition of the DNA distortions by Hmgb2.9

Nm23 by its glycosylase/lyase activity favors the base excisionrepair9 that is executed by Ape1.12

TABLE II – SET COMPLEX PROTEIN EXPRESSIONS DETERMINED BY IMMUNOHISTOCHEMISTRYAND ASSOCIATION TO TUMOR GRADE

ProteinTumorgrade

Meanintensity of

staining6 S.D.

Tumor grade comparison (p value of a Mann–Whitney U test)

0–123 0–1 0–2 0–3 1–2 1–3 2–3

Nm23 0 556 14 0.051 0.18 0.09 0.021 0.041 0.041 0.981 486 162 626 193 616 18

pp32 0 636 16 0.09 0.16 0.17 0.041 0.031 0.021 0.841 716 202 606 153 586 14

Set 0 566 13 <0.0011 0.0041 <0.0011 <0.0011 0.94 0.94 0.981 726 202 686 173 686 14

Hmgb2 0 446 14 <0.0011 0.21 <0.0011 <0.0011 0.09 0.0011 0.0051

1 446 92 566 203 666 20

Ape1 0 566 15 <0.0011 0.021 <0.0011 <0.0011 0.041 0.021 0.701 686 152 806 163 816 16

1Indicates that statistical significance is reached.

TABLE III – SET COMPLEX PROTEIN EXPRESSIONS DETERMINED BY IMMUNOHISTOCHEMISTRY AND ASSOCIATION TO DISEASE STAGING

Protein Tumor stage

Staining intensity p value of the Mann–Whitney U test

All tumors LMPs only TOVs only Alltumors

LMPsonly

TOVsonly

Mean S.D. Mean S.D. Mean S.D.

Nm23 Early 53 15 50 12 58 18 0.041 0.06 0.58Advanced 60 18 58 15 61 19

Pp32 Early 55 16 57 17 53 15 0.20 0.14 0.24Advanced 61 15 66 15 60 15

Set Early 60 15 55 10 67 19 0.0041 0.62 0.49Advanced 67 15 58 15 68 14

Hmgb2 Early 52 18 44 9 64 21 0.09 0.53 0.62Advanced 58 21 45 17 61 20

Ape1 Early 65 20 55 16 80 17 0.0041 0.69 0.97Advanced 76 18 56 14 80 16

1Indicates that statistical significance is reached.

TABLE IV – SURVIVAL AND DISEASE FREE INTERVAL ANALYSIS PERFORMED USING A KAPLAN–MEIER COUPLED TO A LOG RANK TEST

ProteinsSurvival Disease-free interval

G0123 G123 G2 G3 G123 G2 G3 G1231

Nm23 0.62 0.67 0.47 0.65 0.23 0.58 0.40 0.30 0.34pp32 0.38 0.56 0.78 0.83 0.16 0.40 0.51 0.08 0.18Set 0.062 0.65 0.76 0.80 0.88 0.33 0.58 0.56 0.91Hmgb2 0.25 0.44 0.68 0.49 0.88 0.56 0.57 0.58 0.77Ape1 0.032 0.72 0.59 0.80 0.50 0.33 0.44 0.81 0.72

1Censure was set either at 18 (left) or 24 (right) months.–2Indicates that statistical significance is reached.

2122 OUELLET ET AL.

Independently, all proteins included in the SET complex wereshown to have clinical significance in different types of cancer.However, to our knowledge, no study to date has evaluatedexpression of the whole protein complex within a specific cancer.Spearman test correlations revealed that protein expression of allmembers are correlated to the expression of Nm23 and Set (p <0.01, R > 0.21, data not shown) and are overexpressed in TOVscompared to LMPs suggesting their implication in malignancy ofEOC. By their granzyme-A attack, natural killer and cytotoxic T

lymphocytes can be implicated in tumor cells recognition andelimination. Granzyme-A is known to cleave, among others, the2 Nm23 inhibitors, Set and pp32 thus activating the DNAse ac-tivity of Nm2312. It would have been interesting to correlate theexpression of the proteins with the apoptosis status of the tumors.However, we were not able to evaluate the level of apoptosisinduced by Nm23 as single-stranded nicks in the DNA cannot bevisualized in the tissue arrays used in this study by conventionalIHC.

FIGURE 2 – Relation between combinationof Set complex markers and cumulative sur-vival of patients with EOC. Kaplan-Meiergraphical representation of survival curvesdemonstrated a better survival associated with(a) high expression of Nm23 and low expres-sion of Set (p 5 0.03) or (b) high expressionof Hmgb2 and low expression of Ape1 (p 50.05) compared all other two-way combinationof these markers. Results are graphically repre-sented as Kaplan-Meier curves. Patients with afollow up of less than 18 months were not usedfor this analysis. Log rank test was used to ver-ify the significance of the difference in survival.

2123SET COMPLEX IN SEROUS OVARIAN TUMORS

Nm23 was shown to be implicated in the formation of nucleo-tide triphosphate from the nucleotide diphosphate (NDP) by itsNDP kinase activity (reviewed in Refs. 19,20) and its expressionwas reported to be associated with higher cell proliferation.21 Asit was reported that TOVs had higher proliferation rate thanLMPs, these results are consistent with ours where TOVs overex-pressed Nm23 compared to LMPs and G1 tumors.22,23 The dualand controversial role of Nm23 in cancer progression is unre-solved to date. For example, in several cancers includingbreast,24 melanoma25 and liver,26 Nm23 was demonstrated to bea metastasis suppressor (and reviewed in Refs. 27,28). However,Nm23 act as an oncogene in neuroblastoma,29–31 leukemia andlymphoma32,33 but also in tumors of lung34 and pancreas,35 co-lon,36 where its overexpression is associated with aggressivenessand/or poor patient survival (reviewed in Refs. 27,28). Nm23also appears to have a role in cell differentiation, although thedirection of altered expression is dependent on the cell type(reviewed in Refs. 28,37). In a breast cancer cell line38 or neuralprogenitors,39 overexpression of Nm23 leads to the recovery ofnormal differentiation functions. In contrast, in a colon cancercell line28 as well as in hematopoietic progenitors,40 lymphoma41

or leukemia cells,42 the differentiation process is associated witha decrease of Nm23. In our study, expression of Nm23 was sig-nificantly overexpressed in tumors of moderate (G2) and highgrade (G3) compared to the low grade (G0 or G1) consistent withthe results of other groups where TOVs (G2 and G3) presentedhigher expression of Nm23 in comparison to LMPs43 or benignlesions.44–46

The only other member of the SET complex described in EOCis the DNA binding protein Hmgb2. In a DNA microarray analysisstudy, Hmgb2 was found to be overexpressed in serous TOVs re-sistant to chemotherapy.47 In our study, where all samples arefrom chemotherapy na€ıve patients, Hmgb2 appeared to be associ-ated with tumor differentiation since it was overexpressed inTOVs of G2 and G3 compared to those of G0 and G1. Our resultsare consistent with a previously published study where Hmgb2was also reported to be associated with malignant tumors com-pared to benign lesions in gastrointestinal tumors.48

Set, a nucleosome assembly protein, and pp32 were reportedto be the specific inhibitors of Nm23.9 Set and pp32 were shownto play a role in the modification of chromatin18 and the regula-tion of the half-life of some mRNA (reviewed in Ref. 8). Set pre-vents DNA demethylation, histone acetylation and is implicatedin DNA accessibility to favor DNA repair or activation of tran-scription.15–18 High levels of Set are found in Wilms tumors butnot in benign renal disease49 and was reported to be highlyexpressed in transformed human cell lines compared to normalquiescent or differentiated cells.49 In our study, the lowest Setexpression was associated with the most differentiated low gradetumors and this is consistent with its reported expression in othercancers.

pp32 is a nuclear phosphoprotein expressed in self-renewingpopulation of cells.50–52 This protein was also shown to promoteactivation of caspase-9 after involvement of the mitochondriaduring apoptosis53 and to be involved in tumor differentiation.54

In breast, pp32 was present but not differentially expressed whenmalignant and benign tumors were compared.55 However, inprostate, overexpression of pp32 was reported in moderate andhigh grade tumors compared to benign lesions52 and reduction ofits expression leads to cell differentiation and reduction in prolif-eration.54 Its potent role as a tumor suppressor was also reportedas constitutive expression of pp32 leads to suppression of tumori-genicity caused by oncogenes in vitro and in vivo50,56 whilesilencing its endogenous expression sensitize the cells to onco-genic transformation.50 In our analysis, we showed overexpres-sion in low grade tumors (G0 and G1) compared to high grade(G2 or G3) which would be consistent with a suppressive role ofpp32 in EOCs.

The last member of the SET complex, Ape1, is a nucleotideexcision repair factor known to be translocated to the nucleus fol-

lowing an oxidative stress where, probably in cooperation with theother SET complex members, it repairs DNA damages using itsbase excision repair capability. In cancer, Ape1 activity and pro-tein were reported to be elevated in gliomas57 while in osteosar-coma, high expression of Ape1 is associated with poor prognosisof the patient.58 Furthermore, Ape1 repair activity confers resist-ance to chemotherapy or radiotherapy in gliomas,57 osteosar-coma,58 cervival cancer59 and germ cell cancer.60 In our analysis,highest Ape1 expression was also associated with the more malig-nant disease, which is consistent with observations made in othercancers.

In previous studies, 2 members of the SET complex wereshown to be associated with patient outcome. In invasive EOCs,high expression of Nm23 was associated with poor patient prog-nosis,61 which is in contrast to the observations made by otherswhere low expression of Nm23 is related to a better prognosis ofEOC patients although this trend did not reach significance in allanalyses.62–64 However, other reports, including ours, did notshow an association between Nm23 expression alone and prog-nosis.65,66 High level of Ape1 is associated with a better progno-sis of patients receiving radical radiotherapy in bladder can-cer.67 In our study, individually, members of the SET complexwere not associated with patient outcome (survival or disease-free interval). In contrast, specific two-way combinations ofmarkers did significantly stratify patients based on prognosis. Inparticular, tumors expressing high level of Nm23 and low levelof Set, or low levels of Ape1 and high levels of Hmgb2, weresignificantly associated with a better prognosis as compared toall other tumors. It is noteworthy that these analyses allowed usto identify a set of 21 patients with the specified expression pro-files associated with a better prognosis, of which only 4 werefound to be in common between the 2 significant pair-wise anal-yses. These results suggest that these combinations (high Nm23/low Set and high Hmgb2/low Ape1) act as independent markersof prognosis.

In conclusion, we provide here the first simultaneous study ofall members of the SET complex in a cancer context and we dem-onstrated that protein expression of Nm23, pp32, Set, Hmgb2 andApe1 are deregulated and can be related to the differentiation levelof the tumors in EOC. This study highlights the advantage of look-ing at all members of a pathway and the importance of holisticapproach when dealing with a specific pathway since the sum ofthe deregulation of each member is what generates the down-stream effect of the pathway. In particular, this appeared to bereflected in the survival analysis as the combination of 2 markers,instead of a protein alone, was required to significantly definepatient prognosis. Not only does this study suggest that furtherinvestigation of the SET complex in EOC is warranted, but theyalso point to the usefulness of multiple parameters when definingpatient stratification nomograms.

Acknowledgements

We are grateful to Louise Champoux, Lise Portelance, Manonde Ladurantaye, Marise Roy and Jason Madore for technical as-sistance. We would like to thank Drs. Drouin, Gauthier, Dubuc-Lissoir and Sauthier, as well as the members of the gynecology–oncology and pathology services of the Hopital Notre-Dame(CHUM) for tumor procurement. We are grateful to the labora-tory members for thoughtful discussions. This work was sup-ported by a grant from the Canadian Institutes of HealthResearch (CIHR) to Drs. A.-M.M.-M., P.N.T. and D.M.P. Tumorbanking was supported by the Banque de tissus et de donn�ees ofthe R�eseau de recherche sur le cancer of the Fonds de la recher-che en sant�e du Qu�ebec (FRSQ). V.O. was supported by student-ships from the CIHR and Canderel fund of the Institut du Cancerde Montr�eal. Dr. D.M.P. is a recipient of a Chercheur-ClinicienSenior and Dr. A.-M.M-M. is a recipient of a Chercheur National,both fellowships provided by the FRSQ.

2124 OUELLET ET AL.

References

1. Auersperg N, Wong AS, Choi KC, Kang SK, Leung PC. Ovarian sur-face epithelium: biology, endocrinology, and pathology. Endocr Rev2001;22:255–88.

2. Tavassoli FA. Serous tumor of low malignant potential with earlystromal invasion (serous LMP with microinvasion). Mod Pathol1988;1:407–14.

3. Burger CW, Prinssen HM, Baak JP, Wagenaar N, Kenemans P. Themanagement of borderline epithelial tumors of the ovary. Int J Gyne-col Cancer 2000;10:181–97.

4. Dietel M, Hauptmann S. Serous tumors of low malignant potential ofthe ovary. 1. Diagnostic pathology. Virchows Arch 2000;436:403–12.

5. de Souza PL, Friedlander ML. Prognostic factors in ovarian cancer.Hematol Oncol Clin North Am 1992;6:761–82.

6. Heintz AP, Odicino F, Maisonneuve P, Beller U, Benedet JL, Creas-man WT, Ngan HY, Pecorelli S. Carcinoma of the ovary. Int J Gynae-col Obstet 2003;83 (Suppl 1):135–66.

7. Ouellet V, Provencher DM, Maugard CM, Page CL, Ren F, Lussier C,Novak J, Ge B, Hudson TJ, Tonin PN, Mes-Masson AM. Discrimina-tion between serous low malignant potential and invasive epithelialovarian tumors using molecular profiling. Oncogene 2005;24:4672–87.

8. Lieberman J, Fan Z. Nuclear war: the granzyme A-bomb. Curr OpinImmunol 2003;15:553–9.

9. Fan Z, Beresford PJ, Oh DY, Zhang D, Lieberman J. Tumor suppres-sor NM23-H1 is a granzyme A-activated DNase during CTL-medi-ated apoptosis, and the nucleosome assembly protein SET is its inhibi-tor. Cell 2003;112:659–72.

10. Nakamura J, Swenberg JA. Endogenous apurinic/apyrimidinic sites ingenomic DNA of mammalian tissues. Cancer Res 1999;59:2522–6.

11. Bennett RA, Wilson DM, III, Wong D, Demple B. Interaction ofhuman apurinic endonuclease and DNA polymerase b in the baseexcision repair pathway. Proc Natl Acad Sci USA 1997;94:7166–9.

12. Fan Z, Beresford PJ, Zhang D, Xu Z, Novina CD, Yoshida A, Pom-mier Y, Lieberman J. Cleaving the oxidative repair protein Ape1enhances cell death mediated by granzyme A. Nat Immunol 2003;4:145–53.

13. Ouellet V, Guyot M-C, Le Page C, Filali-Mouhim A, Lussier C,Tonin PN, Provencher D, Mes-Masson AM. Tissue array validation ofmicroarray candidates identifies new molecular markers that corre-lated with tumor grade and outcomes in serous epithelial ovarian can-cer. Int J Cancer 2006;119:599–607.

14. Therasse P, Arbuck SG, Eisenhauer EA, Wanders J, Kaplan RS, Rubin-stein L, Verweij J, Van Glabbeke M, van Oosterom AT, Christian MC,Gwyther SG. New guidelines to evaluate the response to treatment insolid tumors. European Organization for Research and Treatment ofCancer, National Cancer Institute of the United States, National CancerInstitute of Canada. J Natl Cancer Inst 2000;92:205–16.

15. Seo SB, McNamara P, Heo S, Turner A, Lane WS, Chakravarti D. Reg-ulation of histone acetylation and transcription by INHAT, a humancellular complex containing the set oncoprotein. Cell 2001;104:119–30.

16. Shikama N, Chan HM, Krstic-Demonacos M, Smith L, Lee CW,Cairns W, La Thangue NB. Functional interaction between nucleo-some assembly proteins and p300/CREB-binding protein family coac-tivators. Mol Cell Biol 2000;20:8933–43.

17. Cervoni N, Detich N, Seo SB, Chakravarti D, Szyf M. The oncopro-tein Set/TAF-1b, an inhibitor of histone acetyltransferase, inhibitsactive demethylation of DNA, integrating DNA methylation and tran-scriptional silencing. J Biol Chem 2002;277:25026–31.

18. Seo SB, Macfarlan T, McNamara P, Hong R, Mukai Y, Heo S, Chak-ravarti D. Regulation of histone acetylation and transcription by nu-clear protein pp32, a subunit of the INHAT complex. J Biol Chem2002;277:14005–10.

19. Kimura N, Shimada N, Fukuda M, Ishijima Y, Miyazaki H, Ishii A,Takagi Y, Ishikawa N. Regulation of cellular functions by nucleosidediphosphate kinases in mammals. J Bioenerg Biomembr 2000;32:309–15.

20. Roymans D, Willems R, Van Blockstaele DR, Slegers H. Nucleosidediphosphate kinase (NDPK/NM23) and the waltz with multiple part-ners: possible consequences in tumor metastasis. Clin Exp Metastasis2002;19:465–76.

21. Caligo MA, Cipollini G, Fiore L, Calvo S, Basolo F, Collecchi P, Ciar-diello F, Pepe S, Petrini M, Bevilacqua G. NM23 gene expression cor-relates with cell growth rate and S-phase. Int J Cancer 1995;60:837–42.

22. Halperin R, Zehavi S, Dar P, Habler L, Hadas E, Bukovsky I, Schnei-der D. Clinical and molecular comparison between borderline serousovarian tumors and advanced serous papillary ovarian carcinomas.Eur J Gynaecol Oncol 2001;22:292–6.

23. Kitakata H, Nemoto-Sasaki Y, Takahashi Y, Kondo T, Mai M, MukaidaN. Essential roles of tumor necrosis factor receptor p55 in liver metasta-sis of intrasplenic administration of colon 26 cells. Cancer Res 2002;62:6682–7.

24. Leone A, Flatow U, VanHoutte K, Steeg PS. Transfection of humannm23-H1 into the human MDA-MB-435 breast carcinoma cell line:

effects on tumor metastatic potential, colonization and enzymatic ac-tivity. Oncogene 1993;8:2325–33.

25. Leone A, Flatow U, King CR, Sandeen MA, Margulies IM, LiottaLA, Steeg PS. Reduced tumor incidence, metastatic potential, andcytokine responsiveness of nm23-transfected melanoma cells. Cell1991;65:25–35.

26. Suzuki E, Ota T, Tsukuda K, Okita A, Matsuoka K, Murakami M,Doihara H, Shimizu N. nm23-H1 reduces in vitro cell migration andthe liver metastatic potential of colon cancer cells by regulating myo-sin light chain phosphorylation. Int J Cancer 2004;108:207–11.

27. Hartsough MT, Steeg PS. Nm23/nucleoside diphosphate kinase inhuman cancers. J Bioenerg Biomembr 2000;32:301–8.

28. Lombardi D, Lacombe ML, Paggi MG. nm23: unraveling its biologi-cal function in cell differentiation. J Cell Physiol 2000;182:144–9.

29. Hailat N, Keim DR, Melhem RF, Zhu XX, Eckerskorn C, BrodeurGM, Reynolds CP, Seeger RC, Lottspeich F, Strahler JR, Hanash SM.High levels of p19/nm23 protein in neuroblastoma are associated withadvanced stage disease and with N-myc gene amplification. J ClinInvest 1991;88:341–5.

30. Chang CL, Zhu XX, Thoraval DH, Ungar D, Rawwas J, Hora N,Strahler JR, Hanash SM, Radany E. Nm23-H1 mutation in neuroblas-toma [letter]. Nature 1994;370:335–6.

31. Leone A, Seeger RC, Hong CM, Hu YY, Arboleda MJ, Brodeur GM,Stram D, Slamon DJ, Steeg PS. Evidence for nm23 RNA overexpres-sion, DNA amplification and mutation in aggressive childhood neuro-blastomas. Oncogene 1993;8:855–65.

32. Niitsu N, Okabe-Kado J, Okamoto M, Takagi T, Yoshida T, Aoki S,Hirano M, Honma Y. Serum nm23-H1 protein as a prognostic factorin aggressive non-Hodgkin lymphoma. Blood 2001;97:1202–10.

33. Niitsu N, Okabe-Kado J, Kasukabe T, Yamamoto-Yamaguchi Y,Umeda M, Honma Y. Prognostic implications of the differentiationinhibitory factor nm23-H1 protein in the plasma of aggressive non-Hodgkin’s lymphoma. Blood 1999;94:3541–50.

34. Ozeki Y, Takishima K, Mamiya G. Immunohistochemical analysis ofnm23/NDP kinase expression in human lung adenocarcinoma: associationwith tumor progression in Clara cell type. Jpn J Cancer Res 1994;85:840–6.

35. Nakamori S, Ishikawa O, Ohhigashi H, Kameyama M, Furukawa H,Sasaki Y, Inaji H, Higashiyama M, Imaoka S, Iwanaga T, Funai H,Wada A, Kimura N. Expression of nucleoside diphosphate kinase/nm23 gene product in human pancreatic cancer: an association withlymph node metastasis and tumor invasion. Clin Exp Metastasis 1993;11:151–8.

36. Royds JA, Cross SS, Silcocks PB, Scholefield JH, Rees RC, Stephen-son TJ. Nm23 �anti-metastatic� gene product expression in colorectalcarcinoma. J Pathol 1994;172:261–6.

37. Lacombe ML, Milon L, Munier A, Mehus JG, Lambeth DO. Thehuman Nm23/nucleoside diphosphate kinases. J Bioenerg Biomembr2000;32:247–58.

38. Howlett AR, Petersen OW, Steeg PS, Bissell MJ. A novel function forthe nm23-H1 gene: overexpression in human breast carcinoma cellsleads to the formation of basement membrane and growth arrest. J NatlCancer Inst 1994;86:1838–44.

39. Gervasi F, D’Agnano I, Vossio S, Zupi G, Sacchi A, Lombardi D.nm23 influences proliferation and differentiation of PC12 cells inresponse to nerve growth factor. Cell Growth Differ 1996;7:1689–95.

40. Willems R, Van Bockstaele DR, Lardon F, Lenjou M, Nijs G, SnoeckHW, Berneman ZN, Slegers H. Decrease in nucleoside diphosphate ki-nase (NDPK/nm23) expression during hematopoietic maturation. J BiolChem 1998;273:13663–8.

41. Aryee DN, Simonitsch I, Mosberger I, Kos K, Mann G, Schlogl E,Potschger U, Gadner H, Radaszkiewicz T, Kovar H. Variability ofnm23-H1/NDPK-A expression in human lymphomas and its relationto tumour aggressiveness. Br J Cancer 1996;74:1693–8.

42. Yamashiro S, Urano T, Shiku H, Furukawa K. Alteration of nm23gene expression during the induced differentiation of human leukemiacell lines. Oncogene 1994;9:2461–8.

43. Schneider J, Romero H, Ruiz R, Centeno MM, Rodriguez-EscuderoFJ. nm23 expression in advanced and borderline ovarian carcinoma.Anticancer Res 1996;16(3A):1197–202.

44. Kaya H, Evans MF, Ekicioglu G, Tuzlali S, Kullu S, Foster C. Anti-metastasis gene expression and numerical chromosomal abnormalitiesof chromosomes 1 & 17 in serous tumours of the ovary. Eur J Gynae-col Oncol 2001;22:240–2.

45. Alaiya AA, Franzen B, Fujioka K, Moberger B, Schedvins K, Silf-versvard C, Linder S, Auer G. Phenotypic analysis of ovarian carci-noma: polypeptide expression in benign, borderline and malignanttumors. Int J Cancer 1997;73:678–83.

46. Harlozinska A, Bar JK, Gerber J. Nm23 expression in tissue sections andtumor effusion cells of ovarian neoplasms. Int J Cancer 1996;69:415–19.

47. Bernardini M, Lee CH, Beheshti B, Prasad M, Albert M, Marrano P,Begley H, Shaw P, Covens A, Murphy J, Rosen B, Minkin S, et al.

2125SET COMPLEX IN SEROUS OVARIAN TUMORS

High-resolution mapping of genomic imbalance and identification ofgene expression profiles associated with differential chemotherapyresponse in serous epithelial ovarian cancer. Neoplasia 2005;7:603–13.

48. Koon N, Schneider-Stock R, Sarlomo-Rikala M, Lasota J, SmolkinM, Petroni G, Zaika A, Boltze C, Meyer F, Andersson L, Knuutila S,Miettinen M et al. Molecular targets for tumour progression in gastro-intestinal stromal tumours. Gut 2004;53:235–40.

49. Carlson SG, Eng E, Kim EG, Perlman EJ, Copeland TD, BallermannBJ. Expression of SET, an inhibitor of protein phosphatase 2A, in renaldevelopment and Wilms’ tumor. J Am Soc Nephrol 1998;9:1873–80.

50. Bai J, Brody JR, Kadkol SS, Pasternack GR. Tumor suppression andpotentiation by manipulation of pp32 expression. Oncogene 2001;20:2153–60.

51. Brody JR, Kadkol SS, Mahmoud MA, Rebel JM, Pasternack GR.Identification of sequences required for inhibition of oncogene-medi-ated transformation by pp32. J Biol Chem 1999;274:20053–5.

52. Kadkol SS, Brody JR, Epstein JI, Kuhajda FP, Pasternack GR. Novelnuclear phosphoprotein pp32 is highly expressed in intermediate- andhigh-grade prostate cancer. Prostate 1998;34:231–7.

53. Jiang X, Kim HE, Shu H, Zhao Y, Zhang H, Kofron J, Donnelly J,Burns D, Ng SC, Rosenberg S, Wang X. Distinctive roles of PHAPproteins and prothymosin-a in a death regulatory pathway. Science2003;299:223–6.

54. Brody JR, Kadkol SS, Hauer MC, Rajaii F, Lee J, Pasternack GR.pp32 reduction induces differentiation of TSU-Pr1 cells. Am J Pathol2004;164:273–83.

55. Kadkol SS, El Naga GA, Brody JR, Bai J, Gusev Y, Dooley WC, Pas-ternack GR. Expression of pp32 gene family members in breast can-cer. Breast Cancer Res Treat 2001;68:65–73.

56. Chen TH, Brody JR, Romantsev FE, Yu JG, Kayler AE, Voneiff E,Kuhajda FP, Pasternack GR. Structure of pp32, an acidic nuclear pro-tein which inhibits oncogene-induced formation of transformed foci.Mol Biol Cell 1996;7:2045–56.

57. Bobola MS, Blank A, Berger MS, Stevens BA, Silber JR. Apurinic/apyrimidinic endonuclease activity is elevated in human adult glio-mas. Clin Cancer Res 2001;7:3510–18.

58. Wang D, Luo M, Kelley MR. Human apurinic endonuclease 1 (APE1)expression and prognostic significance in osteosarcoma: enhanced sen-

sitivity of osteosarcoma to DNA damaging agents using silencing RNAAPE1 expression inhibition. Mol Cancer Ther 2004;3:679–86.

59. Herring CJ, West CM, Wilks DP, Davidson SE, Hunter RD, Berry P,Forster G, MacKinnon J, Rafferty JA, Elder RH, Hendry JH, MargisonGP. Levels of the DNA repair enzyme human apurinic/apyrimidinic en-donuclease (APE1, APEX, Ref-1) are associated with the intrinsicradiosensitivity of cervical cancers. Br J Cancer 1998;78:1128–33.

60. Robertson KA, Bullock HA, Xu Y, Tritt R, Zimmerman E, UlbrightTM, Foster RS, Einhorn LH, Kelley MR. Altered expression of Ape1/ref-1 in germ cell tumors and overexpression in NT2 cells confers re-sistance to bleomycin and radiation. Cancer Res 2001;61:2220–5.

61. Schneider J, Pollan M, Jimenez E, Marenbach K, Martinez N, VolmM, Marx D, Meden H. nm23-H1 expression defines a high-risk subpo-pulation of patients with early-stage epithelial ovarian carcinoma. BrJ Cancer 2000;82:1662–70.

62. Srivatsa PJ, Cliby WA, Keeney GL, Dodson MK, Suman VJ, RochePC, Podratz KC. Elevated nm23 protein expression is correlated withdiminished progression-free survival in patients with epithelial ovar-ian carcinoma. Gynecol Oncol 1996;60:363–72.

63. Tas F, Tuzlali S, Aydiner A, Saip P, Salihoglu Y, Iplikci A, Topuz E.Prognostic role of nm23 gene expression in patients with ovarian can-cer. Am J Clin Oncol 2002;25:164–7.

64. Scambia G, Ferrandina G, Marone M, Benedetti Panici P, GiannitelliC, Piantelli M, Leone A, Mancuso S. nm23 in ovarian cancer: correla-tion with clinical outcome and other clinicopathologic and biochemi-cal prognostic parameters. J Clin Oncol 1996;14:334–42.

65. Baekelandt M, Holm R, Trope CG, Nesland JM, Kristensen GB. Thesignificance of metastasis-related factors cathepsin-D and nm23 inadvanced ovarian cancer. Ann Oncol 1999;10:1335–41.

66. Leary JA, Kerr J, Chenevix-Trench G, Doris CP, Hurst T, HoughtonCR, Friedlander ML. Increased expression of the NME1 gene is asso-ciated with metastasis in epithelial ovarian cancer. Int J Cancer 1995;64:189–95.

67. Sak SC, Harnden P, Johnston CF, Paul AB, Kiltie AE. APE1 andXRCC1 protein expression levels predict cancer-specific survival fol-lowing radical radiotherapy in bladder cancer. Clin Cancer Res 2005;11:6205–11.

2126 OUELLET ET AL.