p53 Expression in Ovarian Borderline Tumors and Stage I Carcinomas

Transcription

1 ANATOMIC PATHOLOGY p5 Expression in Ovarian Borderline Tumors and Stage I Carcinomas JOLANTA KUPRYJANCZYK, MD, PHD, DEBRA A. BELL, MD, DAVID W. YANDELL, ScD, ' ROBERT E. SCULLY, MD, AND ANN D. THOR, MD Seventynine ovarian serous and mucinous borderline tumors, 6 stage I carcinomas and 9 stage II IV carcinomas were studied for p5 protein accumulation with monoclonal antibody PAbl8./>5 protein was expressed in 4% of borderline tumors, 6% of stage I carcinomas, and 64% of higher stage carcinomas. All immunopositive carcinomas accumulated p5 protein in the primary tumor, and 95% of them showed concordance in staining among different tissue blocks. A difference in frequency of p5 protein accumulation between stage I and higher stage serous carcinomas was not statistically significant. ps positivity was associated with microinvasion, microcarcinoma and coexistent carci noma in mucinous borderline tumors (P =.5). An association between ps protein expression and poor tumor differentiation in Stage I carcinomas was statistically significant (P =.). p5 positivity was observed in a poorly differentiated endometrioid carcinoma as well as in adjacent benign endometriotic tissue. These results suggest that p5 abnormalities may be early events in ovarian cancer, possibly contributing to malignant transformation of some borderline tumors, endometriosis and other carcinoma precursors. (Key words: Endometriosis; Ovarian borderline tumors; Ovarian carcinomas; p5 protein accumulation) Am J Clin Pathol 994;: The p5 tumor suppressor gene encodes a 9amino acid nuclear phosphoprotein with very low physiologic levels that is thought to be a negative regulator of the cell cycle. ' Stabilization and accumulation of the p5 protein may result from p5 gene alterations, p5 protein complex formation with other cellular or viral proteins, or other posttranslational mechanisms. " 5 These events may abolish normal p5 protein function. Mutant p5 protein may be characterized by loss of antiproliferative activity, a dominant negative effect on the wild type protein, and active oncogenic activity.,6 ' 7 We and others have shown a high concordance between p5 gene missense mutations and positivity with the monoclonal antibody PAbl8 (nonsense and splicesite mutations, homozygous deletions and some DNA rearrangements may be missed by immunohistochemical method). 8 ' 9 Although gene alterations are the most frequent cause ofp5 protein accumulation, positive immunostaining has been observed in tumors without detectable mutations and a cancerprone family has been described with ubiquitous accumulation of the wild type protein in normal tissues. 5,8,9 Immunohistochemical detection of p5 accumulation is a rapid and efficient method to screen for p5 From the 'Department of Pathology. Harvard Medical School and the James Homer Wright Laboratories of Massachusetts General Hospital. Boston, Massachusetts; the ^Department of Cancer Biology, Harvard School of Public Health, Boston, Massachusetts; and the Molecular Genetics Research Laboratory, Massachusetts Eye and Ear Infirmary, Boston, Massachusetts. Supported in part by NIH P CA Manuscript received August 5, 99; revision accepted October 5, 99. Address reprint requests to Dr. Thor: Department of Pathology, University of Vermont College of Medicine, A49 Medical Alumni Building, Burlington, VT 545. abnormalities, and detects the majority of mutations that occur in ovarian tumors. 9 p5 gene alterations in ovarian cancer are common, having been demonstrated in 8% of carcinomas, and the majority of them are missense mutations in a conserved region of the gene. 9 However, most tumors studied to date have been late stage ovarian cancers; the role of p5 alterations in malignant transformation remains unclear. 9 " Studies of p5 gene mutations in multiple tumor samples from different anatomic sites have shown constant results, suggesting a unifocal origin of ovarian cancer with mutation that occurs before tumor spread. * 5 Our previous findings of two p5 germline mutations and p5 immunopositivity in a borderline component of one carcinoma with p5 mutation suggested that in some cases, p5 gene alterations are early events in ovarian cancer. 9 p5 positivity in both components of a carcinosarcoma also suggested that the mutation occurred early, before differentiation into epithelial and mesenchymallike components. 9 Benign or borderline epithelium has been identified in the majority of mucinous and serous carcinomas suggesting malignant progression from altered benign epithelium. 6 Borderline tumors (BTs), by definition, lack the stromal invasion and severe atypia characteristic of carcinoma. They generally have a good prognosis, although a minority of them behave aggressively. 7,8 Unfortunately, there are no histopathologic, immunohistochemical, or flow cytometric criteria to predict which stage I BTs will spread. 9 Features suggestive of early tumor progression in BTs have been identified, including microscopic foci of noninvasive carcinoma or microinvasion. These features are not useful in predicting which BT patients will have disease progression. However, noninvasive carcinomatous foci or microinvasion provide an excellent opportunity to study the development of genetic events in early ovarian malignancy. p5 protein is commonly accumulated in ovarian carcinomas and may serve as a marker of genetic abnormality in these 67 Downloaded from on January 8

2 67 ANATOMIC PATHOLOGY tumors. 9, Therefore, we have evaluated the frequency of p5 protein accumulation in ovarian BTs (with and without above mentioned features) and in stage I carcinomas and compared the results with p5 accumulation in stage II IV carcinomas. MATERIALS AND METHODS Clinicopathologic Data We studied 54 ovarian tumors categorized according to the World Health Organization (WHO) criteria, 7 with the exception of two mucinous tumors that were diagnosed as carcinomas in the absence of destructive stromal invasion (because of epithelial proliferation and severe cytologic atypia). Eight mucinous carcinomas had a borderline component. Two mucinous borderline tumors (MBTs) with small (. cm and cm ) noninvasive foci of grade I carcinoma were diagnosed as MBTs with microcarcinoma. A borderline component was identified in four serous carcinomas. Nine cases of stromal microinvasion in serous borderline tumors (s) were identified (defined as the presence of nests of epithelial cells in the stroma that did not evoke an obvious stromal reaction and did not exceed mm in diameter ). Invasive implants of s (n = ) were characterized by an aggressiveappearing infiltration into underlying tissue, whereas noninvasive implants (n = 6) did not destroy normal tissue architecture. The criteria for microinvasion and implants in MBTs were the same as in s. The tumor types, subcategories and references for the criteria used in tumor classification are listed in Table. Thirtythree of 6 stage I carcinomas (International Federation of Gynecologists and Obstetricians 4 ) were unilateral and no cancer was present in other genital tract organs, except in one case of combined endometrioid carcinomas of the ovary and endometrium. Twenty three carcinomas were classified as stage IA or IB, were stage IC (in 7 cases the stage IC classification was based on intraoperative tumor rupture). Eight of 9 endometrioid and clear cell carcinomas appeared to arise within endometriotic cysts. Thirtytwo stage IIIV ovarian carcinomas (9 serous, endometrioid, and mucinous) previously studied forp55 gene alterations and protein accumulation, 9 together with 7 additional stage IIIV carcinomas were also included in the study (Table ). Hematoxylin and eosinstained sections from the blocks used for immunohistochemical assay were reviewed histologically (J.K. and D.A.B.) for degree of nuclear abnormality. The atypia in BTs was classified somewhat arbitrarily, based on the nuclear features and nuclear/cytoplasmic ratio, from to, similar to that used in other studies, 5 and the highest grade was recorded in each case, even if it was seen only focally. The carcinomas were assessed as nuclear grade according to the criteria of Ewing. 6 These data are shown in Table. Immunohistochemical Methods and Scoring Formalin fixed paraffinembedded sections of ovarian tumor and additional tissues in some cases were stained (Table ) (96 blocks total with a mean of.7 blocks/case in BTs,.5 in stage I carcinomas, and. in higherstage carcinomas). A monoclonal antibody to human p5 (PAbl8, Cambridge Research Biochemicals, Wilmington, DE), which recognizes TABLE. HISTOLOGIC TUMOR CLASSIFICATION AND P5 ACCUMULATION Tumor Type Serous borderline tumors With noninvasive implants With invasive implants With microinvasion With lymph node involvement 7,8 Serous stage I carcinomas With borderline component Serous stage IIIV carcinomas With borderline component Mucinous borderline tumors Intestinal type 7 Endocervicallike type 9 Mixed epithelial miillerian type 4 With noninvasive implants With pseudomyxoma peritonei With microinvasion With microcarcinoma Mucinous stage I carcinomas With borderline component Endometrioid stage I carcinomas Clear cell type stage I carcinomas Other types of stage IIIV carcinoma^ Mucinous Endometrioid No. of Cases 4 ( bilateral) ( bilateral) 5$ 6 ( bilateral) 4 7 ( bilateral) 9 Cases with PAB8() Staining* 6(4%) ot (4%) (66%) 5(4%) 4 4 (4%) 4 (4%) (%) * Results from the ovarian tumors except one case of with positive implants. t Single positive cells were observed in of 5 serial sections stained. X Includes tumors studied previously. 9 an epitope between amino acids and 79 of the p5 protein 7 was used. It was applied at a dilution of :8 overnight at 4 C. Biotinylated antimouse Ig (Vector Laboratories, Burlingame, CA) (:5) and horseradish peroxidase(hrp) streptavidin (Zymed Laboratories, San Francisco, CA) (:) were used as bridging antibody and linking complex, as previously described. 8 Control cell lines were used as described. 9 The staining was microscopically evaluated, and categorized TABLE. NUCLEAR ATYPIA/GRADE IN CASES STUDIED AND P5 IMMUNOPOSITIVITY Atypia/p5 positive cases MBT Grade/p5 positive cases Stage ICa Stage IIIV Ca 7 8 p5() () () () 8 p5() (5) () (5) () 9 6* 8 p5() () () (7) (5) = serous borderline tumor; MBT = mucinous borderline tumor; Ca = carcinoma. * Includes two MBTs with a microcarcinoma. A.J.CP. November 994 Downloaded from on January 8

3 KUPRYJANCZYK ET AL. 67 p5 in Ovarian Cancer Case No TABLE. P5 ACCUMULATION IN CASES WITH ADDITIONAL TISSUE STAINED Histologic Tumor Type ECa ECa Tissue Stained Noninvasive implants Noninvasive implants Tumor in the lymph node ot Recurrent tumor, noninvasive Recurrent tumor, invasive Adjacent endometriotic tissue Proliferative phase endometrium Endometrial carcinoma Hyperplastic mildly atypical endometrial glands lmmunostaining _ = serous borderline tumor; E = endometrioid: Ca = carcinoma; = ovarian tumor. accumulated in 4% of 7 serous, 4% of mucinous (all of the positive tumors were obviously invasive), 4% of endometrioid, and % of 9 clear cell stage I carcinomas (Table ). The difference in p5 positivity by histologic subtype was not statistically significant. In three p5 positive mucinous carcinomas the borderline component was also positive (Fig. B). The only serous stage I carcinoma with a borderline component was negative for p5 staining. p5 protein was expressed by both endometrioid carcinoma and adjacent mildly atypical, focally hyperplastic endometriosis in one case (Table and Fig. C and D). Proliferative uterine endometrium from this patient was negative forpji accumulation. In one case of synchronous ovarian and endometrial endometrioid carcinomas, focal p5 positivity was present in the ovarian tumor and in hyperplastic mildly atypical uterine endometrial glands, but the endometrial carcinoma was negative (Table ). In these stage I carcinomas, a significant difference in p5 positivity by tumor grade ( and versus ) was observed {P =.). p5 was positive in 5 (64%) of 9 stage II IV carcinomas. This included (66%) of 5 serous and of 4 other stage II IV carcinomas. Of three serous carcinomas with a borderline component, one carcinoma was positive, and in this case p5 positivity was focally present in both borderline and carcinomatous components (Table ). as negative (); focal positivity, with fewer than 5% tumor cells positive in one or multiple foci; low positivity, with 5% to 4% of tumor cells positive; or high positivity, with 4% to % of tumor cells positive. The highest category observed in any tumor fragment of each case was recorded. Statistical analysis was done using twotailed Fisher's exact test, 9 and a P value of.5 was regarded as the upper limit of significance. RESULTS Borderline Tumors Six (4%) of 4 s showed p5 positivity in either the ovarian or extraovarian tumor, or both (Table ). The positive cases included three cases with noninvasive implants, one case with BT in the parenchyma of a lymph node (Fig. A), and two stage I, otherwise unremarkable cases (Tables and ). Among the negative s were two with invasive implants, and all nine with stromal microinvasion. There was no association between the immunopositivity of s and the degree of nuclear atypia, the presence of implants, microinvasion and invasiveness of implants. p5 immunopositivity in the borderlineappearing epithelium was seen in 5 (4%) of 6 ovarian MBTs (Table ). The positive cases include of with stromal microinvasion, of others with microcarcinomas, and another with noninvasive implants. There were no associations between p5 positivity of the ovarian tumor and the degree of nuclear atypia, the epithelial subtype of the tumor, or the presence of pseudomyxoma peritonei. Carcinomas The frequency of p5 accumulation was 6% in stage IA and IB carcinomas and 54% in stage IC tumors. p5 protein was p5 Accumulation in Different Tumor Groups The distribution of p5 immunoreactivity differed among tumor groups. Borderline tumors with p5 positivity were always focally stained (<5% of the epithelial cells) with one exception of low positivity. In contrast, the majority of carcinomas were more diffusely positive (4% to % of positive cells) in at least one section. This difference in the staining distribution category between borderline and malignant tumors was statistically significant (P <.). Variation in staining (ie, positivity and negativity from one tissue block to another) in ovarian tumors was seen in of 44 carcinomas and in 5 of 8 BTs, but the difference was not statistically significant. A difference between p5 positivity in s and MBTs versus stage I carcinomas (including carcinomas with a borderline component) was observed and statistically significant {P =.). However, this significance disappeared if the p5 staining data from borderline components adjacent to carcinomas was added into the BT data. In mucinous tumors with histologic evidence of tumor progression (microinvasion, microcarcinoma, and coexistent BT and carcinoma) 5 (4%) of tumors were positive (Table ). An association between p5 positivity and tumor progression so defined was statistically significant, when this tumor group was compared with MBTs without evidence of progression (P =.5). A similar comparison in the s with evidence of tumor progression (9 with microinvasion, with invasive implants, and 4 with carcinoma) versus other s was not statistically significant (Table ). Stage IIV serous carcinomas were more frequently p5 positive than s (P <.). The frequency ofp5 positive tumors in stage IIIV versus stage I was not statistically significant. p5 Protein Accumulation and Survival Followup data were available for patients with stage I carcinomas (mean, 5.8 years; range, 8 months to 9 years). Vol. No. 5 Downloaded from on January 8

4 674 ANATOMIC PATHOLOGY >'tlf** *' " *'«**^i v t v.» {'*.'" s > t * A., :?' V «^ w* W?$'t;. ^ S * *f: >., sf^r :,»J± ^*r ^?*. %l *.^ S$ «r r* "*.:J.i,*. FIG.. A, Lymph node focus of a serous borderline tumor with several p5 positive cells (PAb 8, methyl green counterstaining, X). B, Benignappearing epithelium with mild atypia and strong p5 positivity, a component of a mucinous invasive carcinoma (PAb 8, methyl green counterstaining, XI6). C and D, Endometriosis adjacent to a carcinoma with p5 positivity in the surface epithelium (PAb 8, methyl green counterstaining, X). Two of patients with p5 positive tumors died of ovarian cancer. The other patients had no evidence of disease at last followup (mean 6.5 years). Among the 8 patients whose tumors were p5 negative, 7 had no evidence of disease (mean followup, 5.5 years) and one patient had a recurrence after seven years. Although negative for p5 by immunohistochemistry, the tumors from this patient had an insertion mutation in exon 7 of the p5 gene (Kupryjanczyk et al, Hum Pathol 994 [in press]). The difference in the recurrence rate between p5 positive and negative stage I carcinomas was not statistically significant. None of 8 patients with p5 positive BTs and clinical followup (mean, 5 years) (including patients whose tumors had microinvasion or microcarcinoma) had evidence of recurrent disease at last followup. AJ.C.P. November 994 Downloaded from on January 8

5 KUPRYJANCZYK ET AL. 675 p5 in Ovarian Cancer DISCUSSION The literature suggests that somatic p5 gene alterations may occur at different stages of malignant progression. Studies of ovarian cancer show a higher percentage of p5 protein accumulation in late rather than in early stage carcinomas. However, we have shown widespread concordance in p5 mutation or overexpression between primary and metastatic ovarian tumors. 9 Similar results have now been published by several groups investigating the clonality of ovarian cancer. " 5 To date, p5 analyses of early stage ovarian cancer and borderline lesions are limited. Our data suggests that p5 alterations occur early in disease progression. Immunohistochemical analysis of stage I carcinomas showed p5 accumulation in many cases, and this abnormality was present in every tumor block studied. Although the percentage of p5 positive stage I carcinomas was lower than higher stage carcinomas, this difference was not statistically significant. We report a significant association between p5 protein accumulation and tumor progression in MBTs, as well as immunopositivity in borderline component of mucinous carcinomas. This suggests that p5 abnormalities occur relatively early in tumor progression and may play a role in the transformation from borderline to malignant disease. However, this hypothesis is based on subjective histologic criteria for the discrimination of MBTs from mucinous carcinomas. Furthermore, we cannot exclude the possibility that in some cases the coexisting benignappearing or borderlineappearing component may be malignant epithelium with cellular differentiation. This has been observed in metastatic mucinous carcinomas. This study suggests that in some cases p5 abnormalities may play a role in the malignant transformation of ovarian endometriosis. In the one case in which endometriosis was stained, p5 protein accumulated in the mildly atypical endometriotic epithelium adjacent to a p5 positive endometrioid carcinoma. Proliferative uterine endometrium from the same patient was negative, similar to the findings of Kohler and colleagues, who used the same antibody and reported no staining of normal endometrium. A comparison of the number of p5 positive tumors among different types of ovarian serous neoplasia showed a higher percentage of p5 abnormalities in carcinomas than in BTs. However, unlike mucinous carcinomas, serous carcinomas only infrequently develop in BTs, and majority are de novo tumors, for which precursor lesions are inapparent. It is believed that many serous carcinomas arise from the ovarian surface epithelium or its inclusion cysts, and this view is supported by histologic findings in cases with microscopic ovarian serous carcinomas. ' 4 Our results do not exclude the possibility that in the serous type neoplasia, as well, p5 alterations may first occur in premalignant epithelium, in this type of precursor lesions, leading to malignant progression. The number of serous carcinomas associated with a BT was too small to evaluate p5 positivity in the borderline component, but p5 positivity was not associated with stromal microinvasion and invasiveness of implants in the few cases of studied. These data, along with the very good prognosis of these lesions, raise the possibility that stromal microinvasion may not be a precursor of frankly invasive carcinoma. There may be several possible explanations for focal p5 immunopositivity in BTs. The difference in staining pattern (category) between BTs and carcinomas may be secondary to cell cycle effects and differences in the proliferation rate between both tumor groups. 5 Alternatively, p5 gene alterations in BTs may be initially focal. Finally, p5 immunopositivity in BTs may also reflect wild type p5 protein accumulation, without genetic alterations. 6 Genetic analyses of these tumors will be required to discern which of these hypotheses are correct. An interesting finding was p5 protein accumulation in an ovarian and in a lymph node focus of tumor from the same case. According to the current view, borderline foci in lymph nodes may represent either metastasis (if the tumor cells lie exclusively within sinusoids) or a separate primary neoplasm (if the tumor is located in the lymph node parenchyma and benign miillerian inclusion glands are also present in the lymph node), 7 ' 8 as in the above case. The rarity of p5 abnormalities in BTs may speak in favor an ovarian origin of the lymph node foci in the above case. In summary, our findings suggest that p5 abnormalities may be early events in ovarian cancer, possibly contributing to malignant transformation of some BTs, endometriosis, or other epithelial precursor lesions. Our data do not support a prognostic significance forp5 immunopositivity in borderline tumors or stage I carcinomas. However, we recognize that the number of such cases was small, and separation by tumor grade, histologic type, and clinical stage was not possible. Acknowledgments. J.K.. gratefully acknowledges support from the Kosciuszko Foundation (while she was affiliated with Medical Academy, Warsaw, Poland) and from funds of the Department of Pathology, Massachusetts General Hospital. The authors thank Dr. J. Southern for help with statistical analysis, S. Edgerton for assistance with immunohistochemical staining, and S. Conley for the photographs. REFERENCES. Levine AJ, Momand J, Finlay CA. The p5 tumor suppresor gene. Nature 99;5: Reich NC, Ohren M, Levine AJ. Two distinct mechanisms regulate the levels of a cellular tumor antigen, p5. Mol Cell Biol 98;:45.. Shay JW, Werbin H, Funk WD, Wright WE. Cellular and molecular advances in elucidating p5 function. Mutation Res 99;77: Momand J, Zambetti GP, Olson DC, et al. The mdm oncogene product forms a complex with the p5 protein and inhibits p5. mediated transactivation. Cell 99;69: Barnes DM, Hanby AM, Gillet CE, et al. Abnormal expression of wild type p5 protein in normal cells of a cancer family patient. Lancet 99;4: Marshall CHJ. Tumor suppressor genes. Cell 99;64:6. 7. Dittmer D, Pati S, Zambetti G, et al. Gain of function mutations in p5. Nature Gen 99;4: Bodner SM, Minna JD, Jensen SM, et al. Expression of mutant p5 proteins in lung cancer correlates with the class of p5 gene mutation. Oncogene 99;7: Kupryjanczyk J, Thor A, Beauchamp R, et al. p5 gene mutations and protein accumulation in human ovarian cancer. Proc Natl AcadSciUSA 99;9: Mazars R, Pujol P, Maudelonde T, et al. p5 mutations in ovarian cancer: A late event? Oncogene 99;6: Okamoto A, Sameshima Y, Yokoyama S, et al. Frequent allelic losses and mutations of the p5 gene in human ovarian cancer. Cancer Res 99;5: Mok ChH, Tsao SW, Knapp RC, et al. Unifocal origin of advanced human epithelial ovarian cancers. Cancer Res 99;5:595. Vol. No. 5 Downloaded from on January 8

6 676 ANATOMIC PATHOLOGY. Jacobs IJ, Kohler MF, Wiseman RW, et al. Clonal origin of epithelial ovarian carcinoma: analysis by loss of heterozygosity, p5 mutation, and Xchromosome inactivation. J Natl Cancer Inst 99;84: Tsao SW, Mok CH, Knapp RC, et al. Molecular genetic evidence of a unifocal origin for human serous ovarian carcinomas. Gynecol Oncol 99;48:5. 5. Kupryjanczyk J, Thor AD, Beauchamp R, et al. Ovarian, peritoneal and endometrial serous carcinoma: p5 analysis supports a clonal origin of multifocal disease. Modern Pathol 994 (in press). 6. Puis LE, Powell DE, DePriest PD, et al. Transition from benign to malignant epithelium in mucinous and serous ovarian cystadenocarcinoma. Gynecol Oncol 99;47: Serov SF, Scully RE. Histological typing of ovarian tumors. In: International Histological Classification of Tumors. No. 9. Geneva: World Health Organization, Leake JF, Currie JL, Rosenhein NB, Woodruff JD. Longterm followup of serous ovarian tumors of low malignant potential. Gynecol Oncol 99;47: de Nictolis M, Montironi R, Tommasoni S, et al. Serous borderline tumor of the ovary. Cancer 99;7:56.. Bell DA, Scully RE. Ovarian serous borderline tumors with stromal microinvasion: A report of cases. Hum Pathol 99; :974.. Marks JR, Davidoff AM, Kerns BJ, et al. Overexpression and mutation of p5 in epithelial ovarian cancer. Cancer Res 99;5: Hart WR, Norris HJ. Borderline and malignant mucinous tumors of the ovary: Histologic criteria and clinical behaviour. Cancer 97;:45.. Bell DA, Weinstock MA, Scully RE. Peritoneal implants of ovarian serous borderline tumors: Histologic features and prognosis. Ca/?av 988;6:. 4. Kottmeier LH, ed. Annual Report on the Treatment of Gynecologic Cancer. Vol. 7. Stockholm: International Federation of Gynecologists and Obstetricians, Koern J, Trope CC, Abeler VM, et al. A retrospective study of 7 borderline tumors of the ovary treated at the Norwegian Radium Hospital from 97 to 98. Cancer 99;7: Ewing J. Neoplastic diseases. Philadelphia: WB Saunders, Banks L, Matlashewski G, Crawford L: Isolation of humanp5specific monoclonal antibodies and their use in the studies of human p5 expression. Eur J Biochem 986; 59: Thor AD, Moore DH, Edgerton SM, et al. Accumulation of p5 tumor suppresor gene protein is an independent marker of prognosis in breast cancers. J Natl Cancer Inst 99;84: Mehta CR, Patel NR. A network algorithm for the exact treatment of Fisher's Exact Test in RxC contingency tables. Journal of the American Statistics Association 98;78: Kohler MF, Kerns BJM, Humphrey PA, et al. Mutation and overexpression of p5 in earlystage epithelial ovarian cancer. Obstet Gynecol 99;8: Young RH, Hart WR. Metastases from carcinomas of the pancreas simulating primary mucinous tumors of the ovary. Am J Surg Pathol 989; : Kohler, MF, Berchuck A, DavidofFAM, et al. Overexpression and mutation of p5 in endometrial carcinoma. Cancer Res 99;5:667.. Bell DA, Scully RE. Early de novo ovarian carcinoma: A study of fourteen cases. Cancer 994;7: Graham RM, Schueller EF, Graham JB. Detection of ovarian cancer at an early stage. Obstet Gynecol 965;6: Reich NC, Levine AJ: Growth regulation of a cellular tumour antigen, p5, in nontransformed cells. Nature 984;8: Fritsche M, Haessler C, Brandner G. Induction of nuclear accumulation of the tumorsuppressor protein p5 by DNAdamaging agents. Oncogene 99;8: Ehrmann RL, Federschneider JM, Knapp RC. Distinguishing lymph node metastases from benign glandular inclusions in lowgrade ovarian carcinoma. Am J Obstet Gvnecol 98; 6: Farhi DC, Silverberg SG. Pseudometastases in female genital cancer. Pathol Annu 98; 7: Rutgers JL, Scully RE. Ovarian mullerian mucinous papillary cystadenoma of borderline malignancy: A clinicopathologic analysis. Cancer 988;6: Rutgers JL, Scully RE. Ovarian mixedepithelial papillary cystadenomas of borderline malignancy of mullerian type: A clinicopathologic analysis. Cancer 988;6: A.J.C.P. November 994 Downloaded from on January 8