Cyclin D1, P53, and P21waf1/Cip1 Expression Is Predictive of Poor Clinical Outcome in Serous Epithelial Ovarian Cancer

5168 Vol. 10, 5168–5177, August 1, 2004 Clinical Cancer Research

Cyclin D1, p53, and p21Waf1/Cip1 Expression Is Predictive of Poor Clinical Outcome in Serous Epithelial Ovarian Cancer

Anish Bali,1 Philippa M. O’Brien,1 p27Kip1 and p21Waf1/Cip1 were significantly associated with Lyndal S. Edwards,2 Robert L. Sutherland,1 increasing tumor grade. Conclusions: This study confirms that dysregulation of Neville F. Hacker,3 and Susan M. Henshall1 cell cycle genes is common in EOC, and that aberrant ex- 1 Cancer Research Program, Garvan Institute of Medical Research, pression of critical cell cycle regulatory proteins can predict Darlinghurst, New South Wales; 2Department of Anatomical Pathology, Prince of Wales Hospital, Randwick, New South Wales; patient outcome in serous EOC. and 3Gynaecological Cancer Centre, Royal Hospital for Women, Randwick, New South Wales, Australia INTRODUCTION In developed countries, ovarian cancer is the leading cause ABSTRACT of death from gynecological malignancies. Typically, this can- Purpose: Dysregulation of cell cycle control, in partic- cer has an insidious onset, and consequently, 70% of women

ular G1-S-phase transition, is implicated in the pathogenesis present with disease that has spread beyond the ovary, resulting of most human cancers, including epithelial ovarian cancer in a high mortality rate despite optimal surgery and aggressive (EOC). However, the prognostic significance of aberrant cell chemotherapy (1). Epithelial ovarian cancer (EOC) constitutes cycle gene expression in EOC remains unclear. at least five different histological subtypes, of which, serous Experimental Design: The expression of selected genes cystadenocarcinoma is the most prevalent (60% of all EOCs). The remainder include mucinous, endometrioid, clear cell, from the pRb pathway that regulates G1-S-phase progression, including cyclin D1, p16Ink4a, cyclin E, p27Kip1, Brenner, and mixed phenotype tumors. In addition to their p21Waf1/Cip1, and p53, was examined in a consecutive series distinct morphological appearance and subtle clinical differ- of 134 serous EOC using immunohistochemistry and the ences, there is molecular evidence for heterogeneity between results correlated to disease outcome. different EOC subtypes (2, 3). Various clinical and pathological Results: Molecular markers predictive of reduced over- features of ovarian cancer are used as predictors of clinical all survival in univariate analysis were overexpression of outcome, of which, the volume of postoperative residual disease and reduced and International Federation of Gynecologists and Obstetricians (0.03 ؍ and p53 (P (0.03 ؍ cyclin D1 (P .(FIGO) stage are the most important (4, 5) ,(0.02 ؍ and p21Waf1/Cip1 (P (0.05 ؍ expression of p27Kip1 (P Dysregulation of normal cell cycle control has been impli- with the latter three also being prognostic for a shorter cated in the pathogenesis of most human cancers. In particular, progression-free interval. In addition, patients displaying abnormal expression of regulatory proteins that control G -S- overexpression of p53 with concurrent loss of p21Waf1/Cip1 1 phase transition, a critical rate-limiting step in cell cycle pro- -and progres (0.0008 ؍ had a significantly shorter overall (P gression, are frequently observed. G -S transition requires phos- On multivariate analysis, 1 .(0.0001 ؍ sion-free survival (P phorylation of the retinoblastoma protein pRb, which results in overexpression of cyclin D1 and combined loss of p21Waf1/Cip1 the release of the E2F family of transcription factors that in turn in the presence of p53 overexpression were independent activate genes essential for entry into S phase (6). Phosphoryl- predictors of overall survival. Similarly, the combination of ation of pRb is initiated by cyclin D1/(CDK)4-6 complexes and p21Waf1/Cip1 loss and p53 overexpression was independently completed by cyclin E/CDK2 in late G1. Alterations in cyclin predictive of a shorter progression-free interval. Overex- and/or cyclin-dependent kinase (CDK) expression results in pression of p53 and cyclin E and reduced expression of increased cell proliferation and are thought to contribute to malignancy. Furthermore, down-regulation or inactivation of the CDK inhibitors, including p21Waf1/Cip1, p27Kip1, and Ink4a p16 , which normally cause G1 arrest by binding to cyclin- Received 12/16/03; revised 4/6/04; accepted 4/22/04. CDK complexes, are often observed in diverse human tumors, Grant support: Gynaecological Oncology Fund of the Royal Hospital further rendering the cell susceptible to uncontrolled extracel- for Women, Sydney, Australia, and the R. T. Hall Trust. R. Sutherland lular proliferation signals (7). Frequently mutated in a wide and S. Henshall are also supported by the National Health & Medical range of human cancers, p53 is a negative regulator of cell cycle Research Council of Australia, The Cancer Council New South Wales, control, which inhibits cell cycle progression in part by activat- and the Prostate Cancer Foundation of Australia (S. Henshall). Waf1/Cip1 The costs of publication of this article were defrayed in part by the ing p21 expression, and also controls the exit of cells payment of page charges. This article must therefore be hereby marked from the cell cycle into programmed cell death. advertisement in accordance with 18 U.S.C. Section 1734 solely to Several studies have determined expression of these critical indicate this fact. cell cycle regulatory proteins in EOC. Although changes in Requests for reprints: Philippa O’Brien, Cancer Research Program, Garvan Institute of Medical Research, 384 Victoria Street, Darlinghurst, expression levels are frequently detected, there are many con- NSW 2010, Australia. Phone: 61-2-9295-8337; Fax: 61-2-9295-8321; flicting findings, making it difficult to delineate the role of E-mail: [email protected]. individual genes in EOC development and progression (8).

Table 1 Clinical and pathological characteristics of the serous EOC serous EOC was identified from the case records of the Gynae- patient cohort (n ϭ 134) cological Cancer Centre at the Royal Hospital for Women, Clinicopathological parameter No. of patients % of patients Sydney, Australia. The age at diagnosis, preoperative CA125 Patient age level, performance status, volume of postoperative residual dis- Յ65 years 84 62.7 ease, presence of intraoperative ascites, FIGO stage, and tumor Ͼ65 years 50 37.3 grade were obtained retrospectively from patient records and are FIGO stage shown in Table 1. Patient outcome was obtained from medical I 10 7.5 II 4 3.0 records and the New South Wales State Register of Births, III 102 76.1 Deaths and Marriages. Median follow-up time for the cohort IV 18 13.4 was 29.8 months (2.6–135.7 months) and 2- and 5-year disease- Tumor grade specific survival rates were 67 and 31%, respectively. All ex- Well (G1) 7 5.2 Moderately (G2) 64 47.8 perimental procedures were approved by the Research Ethics Poorly (G3) 63 47 Committee of the South Eastern Sydney Area Health Service Residual disease (00/115). Յ 2 cm 102 76.1 Immunohistochemistry. Immunohistochemistry was Ͼ2 cm 32 23.9 Presence of ascites* performed on formalin-fixed, paraffin embedded tissue samples No 45 33.8 sectioned at 4 ␮m and mounted on SuperFrost slides (Menzel- Yes 88 66.2 Glaser, Braunschweig, Germany). The tissue sections were Performance status† deparaffinized in xylene (twice for 5 min each) and rehydrated 0 86 70.5 1,2,3 36 29.5 through graded ethanol. Antigen retrieval was achieved by heat- Adjuvant chemotherapy ing the samples using a microwave (p53, p21Waf1/Cip1), pressure No 2 1.5 cooker (pRb), or boiling water bath (p27Kip1, p16Ink4a, cyclin Yes 132 98.5 D1, and cyclin E). Endogenous peroxidase activity was Disease progression No 22 16.4 quenched using 3% hydrogen peroxide (H2O2) in methanol for Yes 112 83.6 5 min. The sections were then blocked with normal horse serum * n ϭ 133. followed by incubation with mouse monoclonal antibodies to † n ϭ 122. either p53 (clone DO-7; DAKO Corporation, Carpinteria, CA), p21Waf1/Cip1 (clone 70; Transduction Laboratories, Lexington, KY), pRb (clone G3-245; PharMingen, San Diego, CA), Similarly, the prognostic significance of changes in cell cycle p27Kip1 (clone 57; Transduction Laboratories), p16Ink4a (clone regulatory gene expression in EOC is unclear. The results of 16PO4; Neomarkers, Fremont, CA), cyclin D1 (clone DCS-6; such studies may be confounded by several factors, including DAKO Corporation), or cyclin E (HE12; PharMingen). The small patient sample sizes, differing therapeutic treatments be- Vectastain Elite Avidin-Biotin Complex kit (Vector Laborato- tween institutions that influence patient outcome, and the con- ries, Burlingame, CA) was used as the detection system and siderable disease heterogeneity of EOC. Indeed, it is likely that color development was obtained using 3,3-diaminobenzidine different mechanisms contribute to loss of cell cycle control in (Vector Laboratories). Counterstaining was undertaken with different histological subtypes of EOC (8). Therefore, the aim of Whitlock’s hematoxylin, Scott’s blueing solution with or with- this study was to determine the expression levels of key proteins out light green before dehydration through graded ethanol and regulating G1-S-phase progression in a large consecutive series xylene. Formalin-fixed, paraffin-embedded cell lines, known for of 134 patients sourced from a single referral center, all of their status of the specific genes studied, were used as positive whom had been diagnosed with serous EOC and treated using and negative controls. similar surgical and adjuvant chemotherapy regimes. The prog- Scoring. Two independent observers (A. Bali and L. Ed- nostic significance of gene expression was determined using wards, a specialist gynecological pathologist) assessed the pat- comprehensive clinicopathological follow-up data for each tern of staining of molecular markers for each tissue sample. patient. Standardization of scoring was achieved by comparison of scores between observers and any discrepancies were resolved MATERIALS AND METHODS by consensus. Scores were given as a percentage of positive Patients and Tissue Samples. A total of 134 patients nuclear staining within representative areas of the tumor sample. treated by primary laparotomy between 1988 and 1998 for The percentage score above which staining was representative

Table 2 Immunohistochemical analysis of cell cycle gene expression in serous EOC Cyclin Cyclin D1 p16Ink4a E p27Kip1 p53 p21Waf1/Cip1 pRb Range of staining (% positive) 0–60 0–100 0–95 0–95 0–100 0–60 0–100 Median staining 2 15 20 40 27 4 70 Cutoff value (positive staining) Ͼ10 Ͼ15 Ͼ10 Ͼ65 Ͼ10 Ͼ10 Ͼ20 Positive tumors (%) 25 (19) 74 (55) 92 (68) 45 (33) 79 (59) 31 (23) 106 (79)

to the date of disease progression as specified by a rise in CA125 or radiological or surgical evidence of relapse. The length of overall survival (OS) was defined from the date of primary laparotomy to the date of patient death or to the date of last follow-up. The relationship between gene expression and the clinical and pathological parameters was determined using ␹2 and Fisher’s exact test where appropriate. P Յ 0.05 was accepted as statistically significant.

RESULTS Correlation between Patterns of Gene Expression and Clinicopathological Parameters. Increased expression of p53 (P ϭ 0.03) and cyclin E (P ϭ 0.03) and reduced levels of p21Waf1/Cip1 (P ϭ 0.002) and p27Kip1 (P ϭ 0.04) were all significantly associated with increasing tumor grade (Table 3). FIGO stage, volume of postoperative residual disease, and presence of ascites did not correlate with the expression pattern of any of the proteins studied. The expression of p53 appeared to be negatively correlated with that of p21Waf1/Cip1 (Table 3). Of the cancers demonstrat- ing p53 expression of Ͻ10%, 40% displayed p21Waf1/Cip1 expression Ͼ 10% in contrast to those with p53 overexpression in which only 14% demonstrated a concurrent increase in p21Waf1/Cip1 levels (P ϭ 0.007). An inverse relationship between p53 and p16Ink4a expression (P ϭ 0.004) was also observed. Of the p53-positive tumors, 37% demonstrated p16Ink4a expression, compared with 62% expressing p16Ink4a in the p53- negative cancers. We also determined a weak positive correlation between overexpression of p53 and cyclin E expression. Sixty-seven percent of p53-positive tumors were also positive for cyclin E expression in contrast to p53-negative tumors where 50% were cyclin E positive (P ϭ 0.05). The expression of p27Kip1 was positively correlated with Fig. 1 Immunohistochemistry on primary tumor tissue. A, p53 over- that of p21Waf1/Cip1 (P ϭ 0.006). In 55% of tumors that dem- expression in a poorly differentiated serous EOC. B, p53-negative Waf1/Cip1 Ͼ Waf1/Cip1 onstrated p21 expression, 10% also showed increased staining in a poorly differentiated (G3) serous EOC. C, p21 Kip1 Waf1/Cip1 expression Ͼ10% in a moderately differentiated (G2) serous EOC. D, levels of p27 , as compared with p21 -negative can- p21Waf1/Cip1-negative staining in a moderately differentiated (G2) serous cers (Ͻ10%), in which increased p27Kip1 staining was observed EOC. E, cyclin D1 overexpression in a poorly differentiated (G3) serous in only 28% of cases. We also determined a negative correlation EOC. F, cyclin D1-negative staining in a poorly differentiated (G3) between p27Kip1 and cyclin D1 expression. Of the p27Kip1- serous EOC. G, high p27Kip1 expression in a poorly differentiated (G3) serous EOC. H, low p27Kip1 expression in a poorly differentiated (G3) positive tumors, 9% demonstrated cyclin D1 expression, Kip1 serous EOC. Magnification, ϫ40. whereas 24% of the p27 -negative cancers were positive for cyclin D1 expression (P ϭ 0.03). No other significant associations between gene expression were observed. Survival Analysis. On univariate analysis, clinicopatho- of overexpression was based on reports in the published litera- logical determinants of reduced OS and progression-free inter- ture. The median percentage of immunostaining, percentage val included the presence of ascites, a volume of residual dis- values deemed as overexpression, and number of specimens ease Ն 2 cm, advanced FIGO stage (III/IV), increasing tumor displaying positive staining are shown in Table 2. Representa- grade (grade 2/3), and poor performance status (Ն1; Table 4). tive photomicrographs of tumor tissue showing positive and Molecular markers predictive of reduced OS on univariate anal- negative staining for the specific antigens are presented in yses were overexpression of p53 (P ϭ 0.03), loss of p21Waf1/Cip1 Fig. 1. (P ϭ 0.02), increased levels of cyclin D1 (P ϭ 0.03), and Survival Analysis. Survival analysis was performed us- reduced expression of p27Kip1 (P ϭ 0.05; Table 4). Aberrant ing a Kaplan-Meier analysis and Cox proportional hazards expression of p53 (P ϭ 0.004), p21Waf1/Cip1 (P ϭ 0.01), and model. Predictive variables of disease outcome on univariate p27Kip1 (P ϭ 0.05) were also significant determinants of re- analysis were incorporated into multivariate analysis to identify duced progression-free interval. The simultaneous assessment of factors that predict clinical outcome independently of other p53 and p21Waf1/Cip1 expression revealed that patients whose clinicopathological or molecular parameters. The progression- tumors displayed p53 overexpression with concurrent loss of free interval was calculated from the date of primary laparotomy p21Waf1/Cip1 had a poorer prognosis for progression-free interval

Table 3 Association of gene expression and clinicopathological parameters in serous EOC Cyclin D1 p16Ink4a Cyclin E p27Kip1 p53 p21Waf/Cip1 pRb Stage 0.86 0.28 0.45 0.33 0.70 0.08 0.44 Grade 0.21 0.27 0.03 0.04* 0.03 0.002* 0.31 Residual disease 0.68 0.46 0.53 0.48 0.81 0.69 0.18 Ascites 0.49 0.54 0.73 0.39 0.17 0.28 0.29

Cyclin D1 0.19 0.19 0.03* 0.60 0.42 0.49 p16Ink4a 0.33 0.71 0.004* 0.62 0.48 Cyclin E 0.35 0.05 0.30 0.49 p27Kip1 0.44 0.006 0.33 p53 0.007* 0.77 p21Waf/Cip1 0.32 NOTE. Significant P values are in bold type. All associations are positive unless marked with an asterisk (*), designating a negative correlation.

(P ϭ 0.0001) and OS (P ϭ 0.0008) in comparison to patient disease remained the most important determinant of progres- tumors with other combinations of expression of these two sion-free interval and OS (Table 5), similar to other cohorts (4, proteins. 5). The presence of ascites and performance status became In a multivariate analysis using a Cox proportional hazards insignificant in both multivariate models. Tumor grade and model incorporating residual disease, presence or absence of clinical stage were not included for purposes of multivariate ascites, and patient performance status, the volume of residual analysis. Consistent with population-based studies of serous

Table 4 Univariate analysis associating clinicopathological parameters and gene expression to clinical outcome in serous EOC. Progression-free interval OS Parameter Relative risk 95% CI P Relative risk 95% CI P Age (yrs) Յ65 versus Ͼ65 0.96 0.65–1.43 0.84 1.04 0.67–1.61 0.87 Stage I–II versus III–IV 3.33 1.54–7.18 0.002 4.35 1.59–11.90 0.004 Grade 1 versus 2–3 3.34 1.06–10.56 0.04 5.54 1.32–23.26 0.02 2 versus 3† 1.07 0.73–1.57 0.72 1.06 0.69–1.64 0.78 Residual disease Յ2cmversus Ͼ2 cm 2.83 1.82–4.40 <0.0001 3.06 1.92-4.86 <0.0001 Ascites No versus yes 2.05 1.34–3.12 0.0009 2.28 1.39–3.74 0.001 Performance status 1,2,3 versus 0 1.58 1.03–2.43 0.03 2.15 1.35–3.44 0.001 CA125 Յ500 versus Ͼ500 0.88 0.54–1.43 0.60 0.77 0.45–1.34 0.36 Cyclin D1 Յ10% versus Ͼ10% 1.39 0.88–2.21 0.16 1.72 1.06–2.78 0.03 p16Ink4a Յ15% versus Ͼ15% 1.29 0.89–1.88 0.18 1.20 0.79–1.83 0.40 Cyclin E Յ10% versus Ͼ10% 1.00 0.69–1.47 0.99 1.32 0.86–2.05 0.21 p27Kip1 Յ65% versus Ͼ65% 1.52 1.00–2.29 0.05* 1.60 1.00–2.56 0.05* p53 Յ10% versus Ͼ10% 1.77 1.20–2.61 0.004 1.62 1.05–2.50 0.03 p21Waf1/Cip1 Յ10% versus Ͼ10% 1.84 1.13–3.00 0.01* 1.90 1.09–3.33 0.02* pRb Յ20% versus Ͼ20% 0.89 0.57–1.39 0.61 1.06 0.64–1.74 0.83 p53 Ͼ 10% and p21Waf1/Cip1 Ͻ 10% 2.10 1.43–3.09 0.0001 2.08 1.36–3.19 0.0008 versus other combinations of p53 and p21Waf1/kip1 (see text). NOTE. See “Materials and Methods” for definitions of OS and progression-free interval. Data census was at 5 years. Significant P values are in boldface type. Abbreviation: CI, confidence interval. * Negative correlation (see text). † n ϭ 127.

Table 5 Multivariate Cox regression analysis of progression-free survival and OS of patients with serous EOC Progression-free interval OS Clinical Parameter Relative risk 95% CI P Relative risk 95% CI P Residual disease, Յ2cmversus Ͼ2 cm 2.76 1.64–4.64 0.0001 2.91 1.68–5.03 0.0001 Ascites, no versus yes 1.45 0.91–2.30 0.12 1.29 0.75–2.22 0.35 Performance status, 1,2,3 versus 0 1.17 0.75–1.83 0.49 1.55 0.96–2.52 0.073 Cyclin D1, 10% versus Յ10% NT 1.75 1.03–2.96 0.038 p53 Ͼ 10% and p21 Ͻ 10% versus 2.20 1.44–3.37 0.0003 2.35 1.44–3.83 0.0006 other combinations of p53 and p21waf1/cip1 (see text). NOTE: Significant P values are in boldface type. Abbreviations: CI, confidence interval; NT, not tested.

ovarian cancer, which show that only ϳ15% of these cancers are DISCUSSION classified as stage I/II (9), only 14 of 134 tumors (10.4%) in this The past 5 years have seen a number of studies reporting cohort were classified as stage I/II, and thus, the numbers were conflicting results on the prognostic value of aberrant cell cycle too small to be included in a multivariate analysis. Similarly, gene expression in determining the survival of patients with only seven patients were classified as having well-differentiated EOC. One reason for this discrepancy is the considerable dis- tumors (G1). Moreover, we found no survival advantage for ease heterogeneity of EOC because different alterations in cell patients with moderately differentiated tumors (G2, n ϭ 64) as cycle regulatory genes are likely to underlie the varied presen- compared with poorly differentiated ones (G3, n ϭ 63; Table 4), tations of EOC. Therefore, we confined our study to patients reflected when only those patients with G2 or G3 tumors were diagnosed with a single histological subtype, serous ovarian included in a multivariate model of progression-free interval cancer, the most frequently diagnosed form of EOC. In addition, (hazard ratio 0.90, 95% confidence interval 0.58–1.39, P ϭ the patients were recruited from a single referral institution, 0.64) or OS (hazard ratio 0.79, 95% confidence interval 0.48– which is reflected in the high proportion of advanced FIGO 1.29, P ϭ 0.34) along with residual disease, ascites, and per- stage III/IV tumors in our cohort and is consistent with the formance status. frequent late-stage presentation of women with serous EOC. Waf1/Cip1 Overexpression of p53 and low p21 expression Nevertheless, examination of traditional clinicopathological remained significant when incorporated into the multivariate variables confirms that this cohort behaves in accordance with analysis (data not shown) and when combined had a stronger other published series of EOC with regard to disease-specific effect at predicting OS (P ϭ 0.0006) and progression-free and progression-free survival (4, 5). Therefore, we examined the interval (P ϭ 0.0003; Fig. 2 and Table 5). Overexpression of prognostic value of selected cell cycle regulatory genes in this cyclin D1 (P ϭ 0.038) was also identified as an independent cohort of 134 patients. predictor of OS when incorporated into the multivariate model. The formation of cyclin D1/CDK4–6 complexes is an Hence, cyclin D1 and combined p53 and p21Waf1/Cip1 expres- early critical step regulating G1-S-phase progression, and ab- sion are stronger predictors of patient outcome than tumor normally high levels of cyclin D1 are found in many tumor classification as moderately (G2) or poorly differentiated (G3) types. Although mutation or amplification of the cyclin D1 gene tumors, commonly used as prognostic indicators in the clinic. is rare, increased levels of mRNA and protein have been re- Kip1 Overexpression of p27 did not retain its significance when ported in 14–59% of invasive ovarian cancers (10–14). In our considered together with residual disease, presence of ascites, cohort of serous EOC, 19% of tumors demonstrated Ͼ10% and patient performance status (data not shown). nuclear accumulation of cyclin D1, consistent with these stud- Stratification of p53 Expression and Survival. The ies. In addition, we found that overexpression of cyclin D1 is an staining pattern of p53 across the sample set revealed a bimodal independent prognostic indicator of OS in patients with serous distribution (Fig. 3). p53 expression of Ͻ5% was observed in 55 EOC. Overexpression of cyclin D1 with or without gene ampli- (41%) tumors, Ͼ50% in 53 (40%) tumors, with the remainder fication has been identified as an adverse prognostic biomarker (n ϭ 26, 19%) being equally distributed between 5 and 50%. in lung, pancreas, and tongue carcinoma (15–17); however, this Only two (8%) patients that had p53 expression of between 5 is the first study of an association with OS for EOC. Only one and 50% were alive 5 years after initial diagnosis, in contrast to other study has reported a significant relationship between cy- 23 patients (42%) who were still alive at 5 years after diagnosis clin D1 overexpression (as determined by mRNA expression) in whom p53 expression was Ͻ5% (P Ͻ 0.0001). The 5-year and a shorter progression-free interval, performed in a small survival rate of patients with p53 expression Ͼ 50% was 36%, series of 24 patients with invasive ovarian malignancies (12). which is similar to those with Ͻ5% expression. A similar These authors also reported a significant association between association between expression of p53 and progression-free cyclin D1 overexpression with serous carcinomas and with interval was also demonstrated (P Ͻ 0.0001). Kaplan-Meier poorly differentiated tumors. With regards to the latter finding, survival curves of progression-free interval and OS and log-rank we observed a trend but not a significant association between P values are shown in Fig. 3. cyclin D1 overexpression and increasing tumor grade (likely

Fig. 2 Kaplan-Meier curves and log-rank P values for 5-year progression-free interval (PFI), and OS according to cyclin D1 expression (A); p53 expression (B); p21Waf1/Cip1 expression (C); and combined p53 and p21Waf1/Cip1 expression (D). Numbers in brackets following parameter stratification indicate the number of patients in each group.

relationship between cyclin D1 expression and tumor differen- tiation in EOC. The activity of the cyclin D1/CDK4-6 complex is predominantly regulated by p16Ink4a, a ubiquitous tumor suppressor gene inactivated in many human cancers (18). Although loss of heterozygosity of p16Ink4a occurs in 30–40% of ovarian tumors, mutations in the remaining allele are rare (19). In addition, promoter methylation of p16Ink4a also appears to be uncommon (20, 21). Regardless of the mechanism, down-regulation of p16Ink4a is often observed in EOC (21–23). However, it appears that levels of p16Ink4a may vary between histological subtypes. In particular, low p16Ink4a expression may be more common in mucinous and endometrioid EOC, and overexpression of p16Ink4a is more common in high-grade serous cancers (24). Our results fit with the latter observation, with 55% of serous cancers expressing p16Ink4a in Ͼ15% of tumor cells. Loss of p16Ink4a has been associated with poor clinical outcome in several cancers, including head and neck and prostate cancer (17, 25). There is no evidence to suggest that aberrant expression of p16Ink4a is associated with clinicopathological parameters or patient outcome in ovarian cancer, and our results would support these findings (19, 22). However, we did observe a significant negative correlation between the expression of p16Ink4a and p53. p53 expression is indirectly regulated by a second tumor suppressor gene encoded at the INK4A/ARF locus, p14ARF. There is thought to be extensive cross-talk between the p16Ink4a/Rb and the p14ARF/p53 pathways, therefore, raising the possibility that levels of p16Ink4a and p53 may be coregulated (18). The second pathway involved in cell cycle progression to S phase involves cyclin E/CDK2 complexes and is primarily regulated by the CDK inhibitors p27Kip1 and p21Waf1/Cip1.Cy- clin E mRNA overexpression has been detected in 22–45% of ovarian cancers and, correspondingly, gene amplification frequently detected (26–28). High cyclin E expression has been associated predominantly with serous and clear cell carcinomas (24, 29), consistent with our results where 68% of patients exhibited overexpression of cyclin E. Only one study has reported a significant association of high cyclin E expression with patient outcome (28); our results agree with earlier studies where such an association was not found (24, 29). We did, however, note an association of cyclin E expression with increasing tumor grade, not previously reported in EOC. We also found that cyclin E levels appeared to correlate with p53 expression. Mutation of the p53 gene has been implicated in the development of Ͼ50% of all human cancers. Gene mutations and/or overexpression of p53 have been detected in Fig. 3 A, distribution of p53 expression in serous EOC patient cohort. 30–80% of EOC, particularly in serous EOC, the majority of B, Kaplan-Meier curves and log-rank P values for 5-year progression- free interval (PFI), and (C) 5-year OS, according to p53 stratification. which are missense mutations that closely correlate with accu- Patients were assigned three subgroups according to their percentage of mulation of mutated p53 protein (30). However, the role of p53 p53 nuclear staining: p53 Ͻ 5% [n ϭ 55 (41%)]; p53 Ͼ 5 Ͻ 51% [n ϭ mutation and cellular accumulation in determining the outcome 26, (19%)]; p53 Ͼ 50% [n ϭ 53 (40%)]. of patients with EOC is far from conclusive (31–34). In the present study of serous EOC, p53 overexpression (Ͼ10% nuclear staining) was able to predict patient outcome independent of the strongest clinicopathological prognostic markers in our due to the small number of well-differentiated tumors). In con- cohort. A closer examination of p53 expression revealed that trast, early studies of cyclin D1 overexpression in ovarian cancer only 8% of patients with intermediate expression (5–50%) were reported significant co-segregation with well-differentiated tu- alive at 5 years after diagnosis, which was significantly poorer mors (10, 14). A larger study may be warranted to define the than those with low (Ͻ5%: 23% survival) or high (Ͼ50%: 36%

survival) levels of p53. A similar association was also observed cancers completely lacked p27Kip1 expression (51). In our study, for progression-free survival. The reason for this observation is we found reduced p27Kip1 expression was a marker of poor unclear. First, it is possible that some of the tumors are over- clinical outcome (shorter OS and progression-free interval) in expressing wild-type functional p53. Expression levels of p53 in univariate analysis, but this did not retain significance as an normal cells are very low, increasing transiently in response to independent prognostic factor when adjusted for other clinico- cellular stresses. However, levels of p53 can be stabilized by pathological and molecular variables of disease outcome. interactions with other intracellular proteins such as mdm-2, We found a significant positive association of p27Kip1 leading to positive immunostaining (35, 36). Second, certain expression with both p21Waf1/Cip1 and cyclin D1 expression. p53 mutations may result in very high levels of p53 protein that Correlation of expression of p27Kip1 and p21Waf1/Cip1 fits with may have some residual activity in regulating cellular functions their similar inhibitory roles in G1 checkpoint regulation and is (37, 38). Third, p53 point mutations resulting in a stop codon, consistent with other reports (51), although incongruous with frameshift, or nonsense mutations may result in altered or trun- the different mechanisms that regulate expression of each gene, cated proteins that are not detected by immunohistochemistry being predominantly posttranslational and –transcriptional, (39). Ovarian cancers with p53 gene mutations, resulting in respectively. truncated proteins, develop distant metastases more rapidly than A correlation between loss of p27Kip1 expression and cy- tumors with missense or no mutations, and consequently, sur- clin D1 overexpression, both of which were associated with vival in these patients is significantly shorter (40). More com- increasing tumor grade in our cohort, is more difficult to ex- prehensive mutation and functional studies are required to de- plain. A positive correlation between cyclin D1 and p27Kip1 has cipher the role of mutant and wild-type p53 in ovarian been reported in several cancers, including EOC (13). More- carcinogenesis. over, in vitro studies have shown that embryonic fibroblasts One of the roles of p53 in cell cycle control is regulation of lacking genes encoding p27Kip1 and p21Waf1/Cip1 have reduced expression of the CDK inhibitor p21Waf1/Cip1, an inhibitor of levels of cyclin D1, which are restored by reintroduction of cyclinE/CDK2 complexes. Inactivation of p53 function leads to p27Kip1 (52), suggesting a positive association between expres- loss of p21Waf1/Cip1 induction and impairment of cyclin/CDK sion of the two genes. However, an inverse correlation between complex inhibition. Indeed, we found that 77% of serous EOC p27Kip1 and cyclin D1 expression has been reported in a mouse in our cohort had low or no expression of p21Waf1/Cip1, which model of mammary carcinogenesis (53). It is thus likely that was significantly associated with poor survival and a shorter these results reflect our incomplete understanding of the effects progression-free interval, as previously demonstrated by others of p27Kip1 expression on cell cycle regulation, including the Waf1/Cip1 (41–44). In addition, we found significant loss of p21 promotion of G1 progression via binding to cyclin D1/CDK expression in tumors that overexpressed p53, although this did complexes (7), particularly in the presence of altered levels of not seem to be the case in all tumors. Again, conflicting data cyclins and CDKs as observed in tumors. have been reported regarding an association between p53 and The interaction of cell cycle regulatory proteins leads to p21Waf1/Cip1 expression in EOC (32, 41, 42), presumably re- phosphorylation and inactivation of pRb and subsequent entry flecting both p53-dependent and -independent mechanisms of of the cells into S-phase. The RB gene is a well-defined tumor p21Waf1/Cip1 regulation. We analyzed the prognostic relevance suppressor and mutations have been described in a wide variety of combining p53 and p21Waf1/Cip1 expression and found loss of of human cancers (54). Loss of heterozygosity at the RB locus p21Waf1/Cip1 in conjunction with p53 overexpression was a is common in EOC; however, the majority of ovarian carcino- stronger predictor of reduced survival and progression-free in- mas exhibit pRb expression (8). Reduced expression of pRb has terval than either molecule alone, suggesting p53-dependent been previously associated with increasing tumor grade, stage, mechanisms dominate p21Waf1/Cip1 expression in serous EOC. and residual disease volume in EOC but not patient outcome These results are consistent with findings in other studies (55). In contrast, a recent study determined that high expression (41, 45). of pRb was an independent prognostic marker of poor OS (24). p21Waf1/Cip1 also plays a functional role in cell differenti- We found that 79% of serous EOC expressed pRb (Ͼ20%) but ation (7). We found a significant correlation of elevated did not find any associations with clinicopathological parame- p21Waf1/Cip1 protein with well-differentiated serous EOC com- ters, patient outcome, or other molecular markers. pared with a marked loss of p21Waf1/Cip1 expression in moder- In conclusion, this study confirms that deregulation of ately and poorly differentiated cancers. A larger series of serous expression of critical cell cycle regulatory proteins, in particular EOC is required to confirm this result; however, a similar cyclin D1, cyclin E, p21Waf1/Cip1 and p27Kip1, are frequent finding has been reported in a study incorporating EOC of events in serous EOC. Furthermore, cyclin D1 expression and a varying histological subtypes (41). combination of p53 and p21Waf1/Cip1 expression are independent We also examined expression of the CDK inhibitor prognostic markers of disease progression. p27Kip1, which is also frequently down-regulated in human cancers, and its loss correlated with poor prognosis in several ACKNOWLEDGMENTS tumor types, including breast and prostate cancer (46, 47). Three Kip1 We thank Professor Donald Marsden and Dr. Greg Robertson, studies support an independent prognostic role of p27 in Consultant Gynaecological Oncologists at the Gynecological Cancer determining the clinical outcome of patients with EOC, includ- Centre, Royal Hospital for Women, and Dr. Catherine Camaris, Depart- ing serous cancer (48–50). A further study of 185 patients ment of Pathology, Royal Hospital for Women for their professional reported a trend (which did not reach significance) of reduced assistance in carrying out this study. We also thank Patricia Vanden OS in a small subgroup of patients (n ϭ 11) whose ovarian Bergh and Therese Malone for coordinating the clinical and tissue

databases held at the Garvan Institute of Medical Research and the 24. Milde-Langosh K, Hangan M, Bamberger A-M, Loning T. Expres- Royal Hospital for Women and Darren Head for assistance with immu- sion and prognostic value of the cell-cycle regulatory proteins, Rb, nohistochemistry. p16MTS2, p21WAF1, p27KIP1, cyclin E, and cyclin D2, in ovarian cancer. Int J Gynecol Pathol 2003;22:168–74. 25. Chakravarti A, Heydon K, Wu CL, et al. Loss of p16 expression is REFERENCES of prognostic significance in locally advanced prostate cancer: an anal- 1. Holschneider CH, Berek JS. Ovarian cancer: epidemiology, biology, ysis from the Radiation Therapy Oncology Group protocol 86-10. J Clin and prognostic factors. Semin Surg Oncol 2000;19:3–10. Oncol 2003;21:3328–34. 2. Scully RE. Pathology of ovarian cancer precursors. J Cell Biol 26. Courjal F, Louason G, Speiser P, Katsaros D, Zeillinger R, Theillet 1995;23 (Suppl):208–18. C. Cyclin gene amplification and overexpression in breast and ovarian 3. Ono K, Tanaka T, Tsunoda T. Identification by cDNA microarray of cancers: evidence for the selection of cyclin D1 in breast and cyclin E genes involved in ovarian carcinogenesis. Cancer Res 2000;60:5007–11. in ovarian tumors. Int J Cancer 1996;69:247–53. 4. Eisenhauer EA, Gore M, Neijt JP. Ovarian cancer: should we be 27. Marone M, Scambia G, Giannitelli C, et al. Analysis of cyclin E and managing patients with good and bad prognostic factors in the same CDK2 in ovarian cancer: gene amplification and RNA overexpression. manner? Ann Oncol 1999;10 (Suppl 1):9–15. Int J Cancer 1998;75:34–9. 5. Clark TG, Stewart ME, Altman DG, Gabra H, Smyth JF. A prog- 28. Farley J, Smith LM, Darcy KM, et al. Cyclin E expression is a nostic model for ovarian cancer. Br J Cancer 2001;85:944–52. significant predictor of survival in advanced, suboptimally debulked 6. Weinberg R. The retinoblastoma protein and cell cycle control. Cell ovarian epithelial cancers: a Gynecol Oncol Group study. Cancer Res 1995;81:323–31. 2003;63:1235–41. 7. Sherr CJ, Roberts JM. CDK inhibitors: positive and negative regu- 29. Rosenberg E, Demopoulos RI, Zeleniuch-Jacquotte A, et al. Ex-

lators of G1-phase progression. Genes Dev 1999;13:1501–12. pression of cell cycle regulators p57(KIP2), cyclin D1, and cyclin E in 8. Milde-Langosch K, Riethdorf S. Role of cell-cycle regulatory pro- epithelial ovarian tumors and survival. Hum Pathol 2001;32:808–13. teins in gynecological cancer. J Cell Physiol 2003;196:224–44. 30. Kmet LM, Cook LS, Magliocco AM. A review of p53 expression 9. Sherman ME, Mink PJ, Curtis R, et al. Survival among women with and mutation in human benign, low malignant potential, and invasive borderline ovarian tumors and ovarian carcinoma. Cancer (Phila.) 2004; epithelial ovarian tumors. Cancer (Phila.) 2003;97:389–404. 100:1045–52. 31. Anttila MA, Ji H, Juhola MT, Saarikoski SV, Syrjanen KJ. The 10. Masciullo V, Scambia G, Marone M, et al. Altered expression of prognostic significance of p53 expression quantitated by computerized cyclin D1 and CDK4 genes in ovarian carcinomas. Int J Cancer 74: image analysis in epithelial ovarian cancer. Int J Gynecol Pathol 1999; 390–5. 18:42–51. 11. Worsley SD, Ponder BA, Davies BR. Overexpression of cyclin D1 32. Werness BA, Freedman AN, Piver MS, Romero-Gutierrez M, in epithelial ovarian cancers. Gynecol Oncol 1997;64:189–95. Petrow E. Prognostic significance of p53 and p21(waf1/cip1) immuno- 12. Barbieri F, Cagnoli M, Ragni N, et al. Increased cyclin D1 expres- reactivity in epithelial cancers of the ovary. Gynecol Oncol 1999;75: sion is associated with features of malignancy and disease recurrence in 413–8. ovarian tumors. Clin Cancer Res 1999;5:1837–42. 33. Sagarra RA, Andrade LA, Martinez EZ, Pinto GA, Syrjanen KJ, 13. Dhar KK, Branigan K, Parkes J, et al. Expression and subcellular Derchain SF. p53 and Bcl-2 as prognostic predictors in epithelial ovar- localization of cyclin D1 protein in epithelial ovarian tumour cells. Br J ian cancer. Int J Gynecol Cancer 2002;12:720–7. Cancer 1999;81:1174–81. 34. Havrilesky L, Darcy KM, Hamdan H, et al. Prognostic significance 14. Sui L, Tokuda M, Ohno M, Hatase O, Hando T. The concurrent of p53 mutation and p53 overexpression in advanced epithelial ovarian expression of p27(kip1) and cyclin D1 in epithelial ovarian tumors. cancer: a Gynecol. Oncol Group study. J Clin Oncol 2003;20:3814–25. Gynecol Oncol 1999;73:202–9. 35. Ashcroft M, Vousden KH. Regulation of p53 stability. Oncogene 15. Keum JS, Kong G, Yang SC, et al. Cyclin D1 overexpression is an 1999;18:7637–43. indicator of poor prognosis in resectable non-small cell lung cancer. Br J Cancer 1999;81:127–32. 36. Ashcroft M, Kubbutat MH, Vousden KH. Regulation of p53 function and stability by phosphorylation. Mol Cell Biol 1999;19:1751–8. 16. Gansauge S, Gansauge F, Ramadani M, et al. Overexpression of cyclin D1 in human pancreatic carcinoma is associated with poor 37. Bradford CR, Zhu S, Wolf GT, et al. Overexpression of p53 prognosis. Cancer Res 1997;57:1634–7. predicts organ preservation using induction chemotherapy and radiation in patients with advanced laryngeal cancer. Otolaryngol Head Neck 17. Bova RJ, Quinn DI, Nankeruis JS, et al. Cyclin D1 and p16INK4A Surg 1995;113:408–12. expression predict reduced survival in carcinoma of the anterior tongue. Clin Cancer Res 1999;5:2810–9. 38. Sengelov L, Horn T, Steven K. p53 nuclear immunoreactivity as a predictor of response and outcome following chemotherapy for meta- 18. Sherr CJ. The INK4a/ARF network in tumour suppression. Nat Rev Mol Cell Biol 2001;2:731–7. static bladder cancer. J Cancer Res Clin Oncol 1997;123:565–70. 19. Marchini S, Codegoni A, Bonazzi C. Absence of deletions but 39. Meek DW. Post-translational modification of p53. Semin Cancer frequent loss of expression of p16INK4 in human ovarian tumours. Br J Biol 1994;5:203–10. Cancer 1997;76:146–9. 40. Sood AK, Sorosky JI, Dolan M, Anderson B, Buller RE. Distant 20. Brown I, Milner BJ, Rooney PH, Haites NE. Inactivation of the metastases in ovarian cancer: association with p53 mutations. Clin p16INK4A gene by methylation is not a frequent event in sporadic Cancer Res 1999;5:2485–90. ovarian carcinoma. Oncol Rep 2001;8:1359–62. 41. Antilla MA, Kosma VM, Hongxiu J, et al. p21/WAF expression as 21. McCluskey LL, Chen C, Delgadillo E, Felix JC, Muderspach LI, related to p53, cell proliferation and prognosis in epithelial ovarian Dubeau L. Differences in p16 gene methylation and expression in cancer. Br J Cancer 1999;79:1870–8. benign and malignant ovarian tumours. Gynecol Oncol 1999;72:87–92. 42. Schmider A, Friedmann W, Gee C. p21(WAF1/CIP1) protein 22. Sui L, Dong Y, Ohno M, et al. Inverse expression of Cdk4 and p16 expression is associated with prolonged survival but not with p53 in epithelial ovarian tumors. Gynecol Oncol 2000;79:230–7. expression in epithelial ovarian carcinoma. Gynecol Oncol 2000;77: 23. Havrilesky LJ, Alvarez AA, Whitaker RS, Marks JR, Berchuck A. 237– 42. Loss of expression of the p16 tumor suppressor gene is more frequent in 43. Plisiecka-Halasa J, Karpinska G, Szymanska T, et al. p21WAF1, advanced ovarian cancers lacking p53 mutations. Gynecol Oncol 2001; p27KIP1, TP53 and C-MYC analysis in 204 ovarian carcinomas treated 83:491–500. with platinum-based regimens. Ann Oncol 2003;14:1078–85.

44. Rose SL, Goodheart MJ, DeYoung BR, Smith BJ, Buller RE. p21 50. Shigemasa K, Shiroyama Y, Sawasaki T, et al. Underexpression of expression predicts outcome in p53-null ovarian carcinoma. Clin Cancer cyclin-dependent kinase inhibitor p27 is associated with poor prognosis Res 2003;9:1028–32. in serous ovarian carcinomas. Int J Oncol 2001;18:953–8. 45. Geisler HE, Geisler JP, Miller GA, et al. p21 and p53 in ovarian 51. Baekelandt M, Holm R, Trope CG, Nesland JM, Kristensen GB. carcinoma: their combined staining is more valuable than either alone. Lack of independent prognostic significance of p21 and p27 expression Cancer (Phila.) 2001;92:781–6. in advanced ovarian cancer: an immunohistochemical study. Clin Can- cer Res 1999;5:2848–53. 46. Chiarle R, Pagano M, Inghirami G. The cyclin dependent kinase inhibitor p27 and its prognostic role in breast cancer. Breast Cancer Res 52. Cheng M, Olivier P, Diehl JA, et al. The p21(Cip1) and p27(Kip1) 2001;3:91–4. CDK ’inhibitors’ are essential activators of cyclin D-dependent kinases in murine fibroblasts. EMBO J 1999;18:1571–83. 47. Tsihlias J, Kapusta LR, DeBoer G, et al. Loss of cyclin-dependent 53. Jang TJ, Kang MS, Kim H, Kim DH, Lee JI, Kim JR. Increased kinase inhibitor p27Kip1 is a novel prognostic factor in localized human expression of cyclin D1, cyclin E and p21(Cip1) associated with de- prostate adenocarcinoma. Cancer Res 1998;58:542–8. creased expression of p27(Kip1) in chemically induced rat mammary 48. Newcomb EW, Sosnow M, Demopoulos RI, Zeleniuch-Jacquotte carcinogenesis. Jpn J Cancer Res 2000;91:1222–32. A, Sorich J, Speyer JL. Expression of the cell cycle inhibitor p27KIP1 54. Sellers WR, Kaelin WG Jr. Role of the retinoblastoma protein in the is a new prognostic marker associated with survival in epithelial ovarian pathogenesis of human cancer. J Clin Oncol 1997;15:3301–12. tumors. Am J Pathol 1999;154:119–25. 55. Konstantinidou AE, Korkolopoulou P, Vassilopoulos I, et al. Re- 49. Mascuillo V, Ferrandina G, Pucci B, et al. p27Kip1 expression is duced retinoblastoma gene protein to Ki-67 ratio is an adverse prognos- associated with clinical outcome in advanced epithelial ovarian cancer: tic indicator for ovarian adenocarcinoma patients. Gynecol Oncol 2003; multivariate analysis. Clin Cancer Res 2000;6:4816–22. 88:369–78.

Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2004 American Association for Cancer Research. Cyclin D1, p53, and p21Waf1/Cip1 Expression Is Predictive of Poor Clinical Outcome in Serous Epithelial Ovarian Cancer

Anish Bali, Philippa M. O'Brien, Lyndal S. Edwards, et al.

Clin Cancer Res 2004;10:5168-5177.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/10/15/5168

Cited articles This article cites 49 articles, 15 of which you can access for free at: http://clincancerres.aacrjournals.org/content/10/15/5168.full#ref-list-1

Citing articles This article has been cited by 10 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/10/15/5168.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/10/15/5168. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.