- Email: [email protected]
Progressive loss of selenium-binding protein 1 expression correlates with increasing epithelial proliferation and papillary complexity in ovarian serous borderline tumor and low-grade serous carcinoma
Human Pathology (2010) 41, 255–261
www.elsevier.com/locate/humpath
Original contribution
Progressive loss of selenium-binding protein 1 expression correlates with increasing epithelial proliferation and papillary complexity in ovarian serous borderline tumor and low-grade serous carcinoma Cunxian Zhang MD, PhD ⁎, Yu Edmond Wang, Peng Zhang MD, Fang Liu MD, PhD, C. James Sung MD, Margaret M. Steinhoff MD, M. Ruhul Quddus MD, W. Dwayne Lawrence MD Department of Pathology, Women and Infants Hospital of Rhode Island, Providence, RI 02905, USA Warren Alpert Medical School of Brown University, Providence, RI 02913, USA Received 23 April 2009; revised 23 July 2009; accepted 30 July 2009
Keyword: Ovarian serous borderline tumor
Summary Ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma often show a spectrum of histologic components with increasing epithelial proliferation and papillary complexity from flat cyst wall, hierarchical structures (with primary papillae branching into secondary papillae), micropapillae, and invasive carcinoma. Although tremendous research has been carried out to elucidate the causes of these tumors, the pathogenesis remains unclear. Literature has described a relationship between insufficient selenium intake and increased risk of cancer. The anticancer action of selenium has been suggested to be mediated by selenium-binding protein 1 as selenium-binding protein 1 is decreased in several cancers. The aim of the study was to examine by immunohistochemistry the expression of selenium-binding protein 1 in the various histologic components within ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma. Our study consisted of 62 cases of ovarian serous borderline tumor, 11 of micropapillary serous borderline tumor, and 7 of low-grade serous carcinoma. Review of archival slides showed flat cyst wall in 69 cases, primary and secondary papillae of hierarchical structures in 75 cases, micropapillae in 26 cases, microinvasion in 1 case, and frankly invasive carcinoma in 7 cases. The strongest immunoreactivity of selenium-binding protein 1 was seen in epithelial cells of flat cyst wall and primary papillae, followed by secondary papillae of the hierarchical structures. Micropapillae and invasive carcinoma (including microinvasion) exhibited a near complete loss of selenium-binding protein 1 expression. Selenium-binding protein 1 immunoreactivity remained the same regardless of the size of the micropapillae. Similar selenium-binding protein 1 expression was seen in the same histologic components from either ovarian serous borderline tumor or micropapillary serous borderline tumor. The gradual loss of selenium-binding protein 1 associated with increasing epithelial proliferation and papillary complexity indicates that selenium-binding protein 1 is involved in tumorigenesis of ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma. Our findings may provide a basis for future studies concerning the molecular mechanisms of selenium-
⁎ Corresponding author. Department of Pathology, Women and Infants Hospital of Rhode Island, Providence, RI 02905, USA. E-mail address: [email protected] (C. Zhang). 0046-8177/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.humpath.2009.07.019
256
C. Zhang et al. binding protein 1 in tumorigenesis as well as a possible role of selenium in chemoprevention and treatment of ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Serous borderline tumor of the ovary has been a subject of debate because of its uncertain biological behavior. Most participants at the Borderline Ovarian Tumor Workshop held in Bethesda, MD, in 2003 agreed on the following definitions [1]. Serous borderline tumor (SBOT) is characterized by the formation of hierarchical structures consisting of primary and secondary papillae. The designation of SBOT requires the presence of hierarchical papillae in greater than 10% of a cystic neoplasm. Tumors with less hierarchical papillae are categorized as cystadenoma. Micropapillary SBOT (M-SBOT) refers to an SBOT showing at least one 5mm uninterrupted focus of micropapillary or cribriform pattern. Tumors with smaller micropapillae are classified as SBOT. Destructive invasion in a borderline tumor indicates the presence of a carcinoma. The invasive carcinoma associated with SBOT is usually low-grade serous carcinoma (LGSC). SBOT, M-SBOT, and LGSC frequently show a spectrum of histologic changes ranging from flat cyst wall, hierarchical structures (with primary papillae branching into secondary papillae), micropapillae, and invasive carcinoma. The wide range of histologic changes provides an ideal system to study their biological relationship. Recent genetic and molecular studies suggest a stepwise progression from serous cystadenoma, SBOT, M-SBOT, to LGSC; this pathway is characterized by mutations in BRAF and KRAS [2,3]. In general, the development of tumor requires alterations of multiple growth control genes. Up to now, however, only a limited number of molecular changes have been known to occur in the pathway of SBOT, MSBOT, and LGSC. Selenium-binding protein 1 (SELENBP1) has been shown to be expressed differentially between normal and neoplastic tissues [4]. SELENBP1 is a cellular protein that binds to exogenously administered selenium. Numerous reports have described a relationship between insufficient selenium intake and increased risk of cancer [5-7]. SELENBP1 has been suggested to mediate the anticancer action of selenium as it is decreased in cancers of the prostate, the lung, and the colon [8-10]. By membrane proteome profiling analysis, SELENBP1 was found to be the most significantly down-regulated protein in ovarian cancer cells [4]. Down-regulation of SELENBP1 was also demonstrated in ovarian carcinomas in the experimental laying hens [11]. To date, however, no systematic studies have been carried out concerning the expressions of SELENBP1 in SBOT, M-SBOT, and LGSC.
In this study, we examined the expressions of SELENBP1 in a wide spectrum of histologic components within SBOT, M-SBOT, and LGSC.
2. Materials and methods We retrieved a total of 62 cases of SBOT, 11 of M-SBOT, and 7 of LGSC from our institutional pathology archives during a period between 1993 and 2005. Of these cases, 68 were the result of total hysterectomy with bilateral salpingooophorectomy, and 12 were from salpingo-oophorectomy or cystectomy alone. The original pathology reports were reviewed for gross characteristics of the tumors. Archival hematoxylin and eosin slides were evaluated to confirm the diagnosis. SBOT, M-SBOT, and LGSC were defined based on the consensus opinion at the Borderline Ovarian Tumor Workshop held in 2003 [1]. Grossly, all 62 cases of SBOT showed complex cysts with papillary projections. The 11 cases of M-SBOT exhibited largely complex cysts with focal solid component. All 7 cases of LGSC showed partly solid and partly cystic components. Under microscopic examination, each lesion showed many of the histologic components in the spectrum of flat cyst wall, primary and secondary papillae of the hierarchical structures, micropapillarity, and invasive carcinoma as in LGSC (Table 1). Flat cyst wall was lined by a single layer of bland-appearing serous epithelium within a cystic tumor (Fig. 1A). Primary papillae protruded from cystic wall and contained broad fibrovascular cores lined by serous epithelium (Fig. 1B). Secondary papillae arose from the primary papillae and were composed of tiny papillae lined by serous epithelium (Fig. 1B). Micropapillae were elongate filiform papillary structures that arose directly in a nonhierarchical fashion from cyst walls or from large fibrous or edematous papillae and contained cores with little or no connective tissue (Fig. 1C). Cribriform morphology with punched-out lumens was classified under micropapillarity [1]. Frankly invasive carcinoma exhibited areas of stromal invasion greater than 10 mm2 (Fig. 1D). Stromal invasion of smaller than 10 mm2 was defined as microinvasion [12]. One case of M-SBOT showed microinvasion. Representative sections of various histologic components of tumor, as well as normal fallopian tube and ovary when present, were cut at 5 μm from paraffin blocks for SELENBP1 immunohistochemical studies. The sections were routinely deparaffinized and hydrated through a gradient of ethanol. Antigen retrieval was achieved by
Progressive loss of SELENBP1 expression Table 1
257
Histologic characteristics of tumors
Tumor types
SBOT (62 cases) M-SBOT (11 cases) LGSC (7 cases) Total
No. of cases with particular histologic components Flat cyst wall
Hierarchical structures Primary papillae
Secondary papillae
62 6 1 69
62 10 3 75
62 10 3 75
Micropapillae
Invasive carcinoma
9 11 6 26
0 1 7 8
NOTE. The area of micropapillae in SBOT was smaller than 5 mm. Each case of M-SBOT showed at least one 5-mm uninterrupted focus of micropapillae, although smaller foci of micropapillae were also seen. The invasive carcinoma seen in one case of M-SBOT measured 4 mm2. Each case of LGSC showed a greater than 10 mm2 area of stromal invasion. In the 26 cases with micropapillae, 22 were of the micropapillary type, and 4 mixed micropapillary and cribriform type.
incubating the sections with citrate buffer (pH 6.1) at 95°C. Sections were immersed in 3% H2O2 at room temperature for 10 minutes to block any endogenous peroxidase activity. They were incubated with a monoclonal mouse antibody against human SELENBP1 (MBL International, Woburn, MA) at a dilution of 1:200 at room temperature for 35 minutes.
The sections were then incubated with a secondary antibody previously conjugated to horseradish peroxidase–labeled polymer at room temperature for 35 minutes. After the sections were incubated with diaminobenzidine and H2O2 at room temperature for 8 minutes, they were counterstained with hematoxylin and cover-slipped.
Fig. 1 Illustration of various histologic components (hematoxylin and eosin stain, original magnification ×100). Flat cyst wall is lined by a single layer of serous epithelial cells (A), primary papillae protrude from cyst wall and are lined by serous epithelium (long arrow in B), secondary papillae (short arrow in B) arise from primary papillae, micropapillae are elongate filiform papillary structures that arise directly in a nonhierarchical fashion from cyst walls or from large fibrous or edematous papillae and contained cores with little or no connective tissue (C), and invasive carcinoma exhibits stromal invasion (D).
258 In each run of the staining, we included several controls: (1) a negative reagent control (MBL International) used to substitute the primary antibody, (2) a positive tissue control with a section of normal fallopian tube known to be positive for SELENBP1, and (3) a negative tissue control with a section of high-grade ovarian serous carcinoma known to be negative for SELENBP1. The specificity of immunostaining was confirmed by a positive stain in the mucosal epithelial
C. Zhang et al. cells of fallopian tube but a negative stain in high-grade ovarian serous carcinoma and on the sections that were stained with a negative reagent control substituting the primary antibody. Based on our preliminary results, we scored SELENBP1 immunostains using a 4-point scale (0 to 3+) system; staining in fewer than 5% of cells was recorded as “0,” staining in 6% to 30% of cells as 1+, staining in 31% to
Fig. 2 Immunohistochemical evaluation of SELENBP1 expressions. Diffuse and strong SELENBP1 immunostain was seen in the mucosal epithelial cells of normal fallopian tube (A), flat cyst wall (B), and primary papillae (long arrow in C). The immunostain was decreased in secondary papillae (short arrow in C). Almost complete loss of SELENBP1 immunoreactivity was seen in micropapillae (D) and invasive carcinoma (E and F). Original magnifications ×100 in A, B, C, D, and E; ×400 in F.
Progressive loss of SELENBP1 expression Table 2
259
SELENBP1 immunohistochemical characteristics in various histologic components Normal fallopian Normal tube ovary
No. of cases 70 SELENBP1 immnostaining 2.7 ± 0.4 scores
Flat cyst Hierarchical structure Micropapillae Invasive wall carcinoma Primary papillae Secondary papillae
36 69 75 2.7 ± 0.3 2.8 ± 0.3 2.4 ± 0.4
75 1.3 ± 0.4
26 0.2 ± 0.4
7 0.2 ± 0.4
NOTE. Statistical analysis of SELENBP1 scores are as follows. (1) Among normal fallopian tube, normal ovary, flat cyst wall, and primary papillae, P = .631. (2) Between micropapillae and invasive carcinoma, P = .602. (3) Secondary papillae compared to normal fallopian tube, P = .015; secondary papillae compared to normal ovary, P = .018; secondary papillae compared to flat cyst wall, P = .012; secondary papillae compared to primary papillae, P = .019. (4) Secondary papillae compared to micropapillae, P = .031; secondary papillae compared to invasive carcinoma, P = .033. The one case of microinvasion was not included for statistical analysis.
60% of cells as 2+, and staining in more than 60% of cells as 3+. In a preliminary study, we also divided the intensity of immunostaining into 4 groups: absent, weak, moderate, and strong. By Spearman test, we found a strong positive correlation between the intensity of immunostaining and the percentage of positive cells (R = 0.96) in 10 preliminary samples. Thus, scoring of immunostains was based only on the percentage of positive cells in the final study. Also, because the nuclear and cytoplasmic stains paralleled, total cellular (instead of separate nuclear and cytoplasmic) scores were assessed. In cases showing separate foci of the same histologic pattern, an average score of staining for the same histologic pattern was used. Kruskal-Wallis test was used to compare the scores of immunostaining among different histologic components. The same test was also used to compare the patients' age or tumor size among SBOT, M-SBOT, and LGSC. MannWhitney test was used to compare the immunostaining scores in the same histologic components, such as flat cyst wall versus flat cyst wall, primary papillae versus primary papillae, secondary papillae versus secondary papillae, or micropapillae versus micropapillae, between SBOT and M-SBOT. The difference was considered significant when the P value was smaller than .05.
3. Results The patients' age ranged from 27 to 82 years (mean, 48.5 years) in SBOT, 37 to 72 years (mean, 51.6 years) in M-SBOT, and 40 to 76 years (mean, 52.5 years) in LGSC.
Table 3
Comparison of SELENBP1 immunostaining scores in the same histologic components between SBOT and M-SBOT Flat cyst wall
SELENBP1 immunostaining scores P
Although there was a tendency for an average older age from SBOT, M-SBOT, to LGSC, the difference did not reach a statistical significance (P = .242). A significant difference in tumor size was seen in the following order: SBOT (from 4.8 to 15.4 cm; mean, 7.2 cm) b M-SBOT (from 8.0 to 21.5 cm; mean, 10.8 cm) b LGSC (from 10.5 to 26.5 cm; mean, 15.5 cm) (P = .041). Immunohistochemistry showed strong and diffuse SELENBP1 staining in mucosal epithelial cells of normal fallopian tube (the staining was somewhat more intense in the ciliated cells than in the secretory and intercalated cells) (Fig. 2A). Diffuse and strong SELENBP1 immunostaining was also seen in normal surface epithelium of ovary, epithelial cells lining flat cyst wall (Fig. 2B), and epithelial cells lining primary papillae of the hierarchical structures (Fig. 2C). The immunoreactivity in the epithelial cells of secondary papillae in the hierarchical structures was significantly lower than that in normal fallopian tube, normal ovary, flat cyst wall, or primary papillae (Table 2). The immunoreactivity in micropapillae (Fig. 2D) was similar to that in invasive carcinoma (Fig. 2E and F), but both were significantly lower than that in secondary papillae of the hierarchical structures. Because the immunoreactivity remained the same regardless of the size of micropapillae, micropapillae of different sizes were not separated in this report. No difference in SELENBP1 expression was seen in the same histologic components between SBOT and M-SBOT (Table 3). The immunostaining in the focus of microinvasion in one case of M-SBOT was similar to that in micropapillae or frankly invasive carcinoma, although statistic comparison was not performed because of the small sample size concerning microinvasion.
SBOT M-SBOT
2.8 ± 0.2 (62) 2.9 ± 0.3 (6) .576
NOTE. The value in parenthesis indicates the number of cases.
Hierarchical structures Primary papillae
Secondary papillae
2.4 ± 0.3 (62) 2.5 ± 0.4 (10) .554
1.3 ± 0.4 (62) 1.3 ± 0.5 (10) .521
Micropapillae
0.3 ± 0.5 (9) 0.2 ± 0.4 (11) .502
260
4. Discussion Within SBOT, M-SBOT, and LGSC, the degree of epithelial proliferation and papillary complexity in an increasing order ranged from flat cyst wall, primary papillae, secondary papillae, micropapillae, and invasive components. Our study showed a gradual loss of SELENBP1 expression associated with increasing epithelial proliferation and papillary complexity within SBOT, M-SBOT, and LGSC. The highest level of SELENBP1 expression was seen in flat cyst wall and primary papillae of the hierarchical structures, similar to that in normal fallopian tube and ovary. An intermediate level of SELENBP1 expression was observed in secondary papillae of the hierarchical structures. A near complete loss of SELENBP1 expression was seen in micropapillae and invasive carcinoma. This is the first systematic study to examine SELENBP1 expression in SBOT, M-SBOT, and LGSC. The expression of SELENBP1 in normal fallopian tube and ovary indicates a normal biologic function of SELENBP1 in these tissues. The expression of SELEBP1 in flat cyst wall and primary papillae of the hierarchical structures within the tumor at the same level as in normal fallopian tube and ovary indicates that the functions of SELENBP1 are retained in these histologic components. The exact functions of SELENBP1 are unclear, but a gradual decrease of SELENBP1 associated with increasing epithelial proliferation and papillary complexity in SBOT, M-SBOT, and LGSC suggests that SELENBP1 is related to the development of these tumors. Although loss of SELENBP1 expression may be an effect of tumor development, the growth inhibitory roles of SELENBP1 in liver and prostate cells [8,13] may suggest that the protein may be involved in tumorigenesis of SBOT and LGSC of the ovary. The near complete loss of SELENBP1 in both micropapillae and invasive carcinoma suggests that the biological features of malignancy may have been established in micropapillae. Although it is likely that additional genetic events contribute to tumor development, reduction of SELENBP1 expression may be a key step in the transition from the more quiescent (normal tissue, flat cyst wall, and primary papillae of the hierarchical structures), to the more proliferative (secondary papillae of the hierarchical structures), and finally to the most proliferative (micropapillae and invasive carcinoma) tissue. Provided that various histologic components within the same neoplasm are derived from the same clone, differential expressions of SELENBP1 indicate that the various histologic components represent a spectrum of cellular differentiation. Of special interest was our finding showing a more severe loss of SELENBP1 expression in secondary than primary papillae of the hierarchical structures. This may seem difficult to understand if one argues that the epithelial cells lining the secondary papillae are probably better differentiated than those lining the primary papillae based on the frequent observation of smaller epithelial cells in the
C. Zhang et al. secondary than in the primary papillae. The smaller size of the epithelial cells in the secondary papillae, however, may be the result of degenerative changes rather than their better biologic differentiation compared to the cells in the primary papillae. Indeed, the presence of secondary papillae indicates more exuberant proliferation of ovarian serous neoplasms and is required to distinguish borderline tumors from cystadenomas. It then becomes logical to find a decrease in SELENBP1 expression associated with proliferation from primary to secondary papillae of the hierarchical structures. The current definition of M-SBOT requires the presence of at least one 5-mm continuous focus of micropapillary or cribriform morphology in an otherwise SBOT [1]. Compared to SBOT, M-SBOT is more often associated with bilaterality, surface ovarian involvement, extraovarian diseases, invasive implants, and recurrence [12]. The biologic significance of lesions with micropapillarity smaller than 5 mm is less clear. Our study showing the same degree of SELENBP1 reduction in the micropapillae regardless of size suggests that smaller micropapillae may be similar to the larger ones at the molecular level. Our findings warrant further clinicopathologic studies on the biological behavior of SBOT with smaller than 5-mm focus of micropapillae. Loss of SELENBP1 is a late event in colorectal carcinogenesis as severe reduction of SELENBP1 is frequent in colorectal carcinomas but uncommon in adenomas [10]. In contrast, marked reduction of SELENBP1 appears to be an early event in tumorigenesis of SBOT, M-SBOT, and LGSC as it is shown by our study to occur in the secondary papillae of the hierarchical structures within the tumors. Yet, loss of SELENBP1 expression is a later event in this pathway compared to mutations of BRAF and KRAS that take place as early as in flat cyst wall of serous cystadenoma [14]. Decreased expression of SELENBP1 was previously reported in borderline ovarian tumors, but it was not known whether the reduced SELENBP1 expression was uniform in different histologic components within borderline tumors [4]. We were able to demonstrate a differential expression of SELENBP1 in various histologic components within SBOT. Although a decreased expression of SELENBP1 was previously described in high-grade ovarian serous carcinoma of the human subjects [4] and of the experimental hens [11], our study showed a reduction of SELENBP1 expression in LGSC of the human ovary. SBOTs can be seen in women of all ages, but many occur during reproductive age. In perimenopausal and menopausal women, hysterectomy and bilateral salpingo-oophorectomy are generally the treatment of choice. More conservative surgical procedure, however, may be a better option for younger women who wish to preserve their fertility. Effective chemotherapy may be necessary to prevent tumor recurrence in these patients. However, SBOTs and LGSCs do not respond to conventional cytotoxic chemotherapy because of their indolent nature. Thus, searching for novel target-based therapy is needed. Selenium has been implicated as an important chemopreventive and chemotherapeutic
Progressive loss of SELENBP1 expression agent for several epithelial tumors, including cancers of the prostate and colon [15-17]. Our study showing a differential expression of SELENBP1 in the various components of SBOT, M-SBOT, and LGSC not only indicates a role of SELENBP1 in the development of these tumors but also suggests the potential utility of selenium in prevention and treatment of SBOT and LGSC. Because the effects of selenium are mediated by SELENBP1, loss of SELENBP1 expression in tumors may have a negative impact on the ability of selenium to control tumor cell growth. It has been reported, however, that treating ovarian tumor cells with a selenium compound increases SELENBP1 expression [4]. The increased expression of SELENBP1 secondary to selenium treatment may facilitate the anticancer effect of selenium in ovarian tumors. In summary, our study revealed a differential expression of SELENBP1 in various histologic components of SBOT, M-SBOT, and LGSC. The gradual loss of SELENBP1 expression associated with increasing epithelial proliferation and papillary complexity indicates that SELENBP1 is involved in tumorigenesis. Our findings may provide a basis for future studies concerning the molecular mechanisms of SELENBP1 in tumorigenesis as well as a possible role of selenium in chemoprevention and treatment of SBOT, M-SBOT, and LGSC.
References [1] Seidman JD, Soslow RA, Vang R, et al. Borderline ovarian tumors: diverse contemporary viewpoints on terminology and diagnostic criteria with illustrative images. HUM PATHOL 2004;35:918-33. [2] Shih IM, Kurman RJ. Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. Am J Pathol 2004; 164:1511-8.
261 [3] Shih IM, Kurman RJ. Molecular pathogenesis of ovarian borderline tumors: new insights and old challenges. Hum Cancer Res 2005;11: 7273-9. [4] Huang KC, Park DC, Ng SK, et al. Selenium binding protein 1 in ovarian cancer. Int J Cancer 2006;118:2433-40. [5] Clark LC, Combs GF, Turnbull BW, et al. Effects of selenium supplementation for cancer prevention in patients with carcinoma of the skin. A randomized controlled trial. Nutritional Prevention of Cancer Study Group. JAMA 1996;276:1957-63. [6] Fleming J, Ghose A, Harrison PR. Molecular mechanisms of cancer prevention by selenium compounds. Nutr Cancer 2001;40:42-9. [7] Rayman MP. The importance of selenium to human health. Lancet 2000;356:233-41. [8] Yang M, Sytkowski AJ. Differential expression and androgen regulation of the human selenium-binding protein gene hSP56 in prostate cancer cells. Cancer Res 1998;58:3150-3. [9] Chen G, Wang H, Miller CT, et al. Reduced selenium-binding protein 1 expression is associated with poor outcome in lung adenocarcinomas. J Pathol 2004;202:321-9. [10] Kim H, Kang HJ, You KT, et al. Suppression of human seleniumbinding protein 1 is a late event in colorectal carcinogenesis and is associated with poor survival. Proteomics 2006;6:3466-76. [11] Stammer K, Edassery SL, Barua A, et al. Selenium-binding protein 1 expression in ovaries and ovarian tumors in the laying hen, a spontaneous model of human ovarian cancer. Gynecol Oncol 2008; 109:115-21. [12] Silverberg SG, Bell DA, Kurman RJ, et al. Borderline ovarian tumors: key points and workshop summary. HUM PATHOL 2004;35:920-7. [13] Bansal MP, Mukhopadhyay T, Scott J, Cook RG, Mukhopadhyay R, Medina D. DNA sequencing of a mouse liver protein that binds selenium: implications for selenium's mechanism of action in cancer prevention. Carcinogenesis 1990;11:2071-3. [14] Ho CL, Kurman RJ, Dehari R, Wang TL, Shih IM. Mutations of BRAF and KRAS precede the development of ovarian serous borderline tumors. Cancer Res 2004;64:6915-8. [15] Combs GF, Gray WP. Chemopreventive agents: selenium. Pharmacol Ther 1998;79:179-92. [16] Ip C. Lessons from basic research in selenium and cancer prevention. J Nutr 1998;128:1845-54. [17] Medina D. Mechanisms of selenium inhibition of tumorigenesis. Adv Exp Med Biol 1986;206:465-72.
www.elsevier.com/locate/humpath
Original contribution
Progressive loss of selenium-binding protein 1 expression correlates with increasing epithelial proliferation and papillary complexity in ovarian serous borderline tumor and low-grade serous carcinoma Cunxian Zhang MD, PhD ⁎, Yu Edmond Wang, Peng Zhang MD, Fang Liu MD, PhD, C. James Sung MD, Margaret M. Steinhoff MD, M. Ruhul Quddus MD, W. Dwayne Lawrence MD Department of Pathology, Women and Infants Hospital of Rhode Island, Providence, RI 02905, USA Warren Alpert Medical School of Brown University, Providence, RI 02913, USA Received 23 April 2009; revised 23 July 2009; accepted 30 July 2009
Keyword: Ovarian serous borderline tumor
Summary Ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma often show a spectrum of histologic components with increasing epithelial proliferation and papillary complexity from flat cyst wall, hierarchical structures (with primary papillae branching into secondary papillae), micropapillae, and invasive carcinoma. Although tremendous research has been carried out to elucidate the causes of these tumors, the pathogenesis remains unclear. Literature has described a relationship between insufficient selenium intake and increased risk of cancer. The anticancer action of selenium has been suggested to be mediated by selenium-binding protein 1 as selenium-binding protein 1 is decreased in several cancers. The aim of the study was to examine by immunohistochemistry the expression of selenium-binding protein 1 in the various histologic components within ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma. Our study consisted of 62 cases of ovarian serous borderline tumor, 11 of micropapillary serous borderline tumor, and 7 of low-grade serous carcinoma. Review of archival slides showed flat cyst wall in 69 cases, primary and secondary papillae of hierarchical structures in 75 cases, micropapillae in 26 cases, microinvasion in 1 case, and frankly invasive carcinoma in 7 cases. The strongest immunoreactivity of selenium-binding protein 1 was seen in epithelial cells of flat cyst wall and primary papillae, followed by secondary papillae of the hierarchical structures. Micropapillae and invasive carcinoma (including microinvasion) exhibited a near complete loss of selenium-binding protein 1 expression. Selenium-binding protein 1 immunoreactivity remained the same regardless of the size of the micropapillae. Similar selenium-binding protein 1 expression was seen in the same histologic components from either ovarian serous borderline tumor or micropapillary serous borderline tumor. The gradual loss of selenium-binding protein 1 associated with increasing epithelial proliferation and papillary complexity indicates that selenium-binding protein 1 is involved in tumorigenesis of ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma. Our findings may provide a basis for future studies concerning the molecular mechanisms of selenium-
⁎ Corresponding author. Department of Pathology, Women and Infants Hospital of Rhode Island, Providence, RI 02905, USA. E-mail address: [email protected] (C. Zhang). 0046-8177/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.humpath.2009.07.019
256
C. Zhang et al. binding protein 1 in tumorigenesis as well as a possible role of selenium in chemoprevention and treatment of ovarian serous borderline tumor, micropapillary serous borderline tumor, and low-grade serous carcinoma. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Serous borderline tumor of the ovary has been a subject of debate because of its uncertain biological behavior. Most participants at the Borderline Ovarian Tumor Workshop held in Bethesda, MD, in 2003 agreed on the following definitions [1]. Serous borderline tumor (SBOT) is characterized by the formation of hierarchical structures consisting of primary and secondary papillae. The designation of SBOT requires the presence of hierarchical papillae in greater than 10% of a cystic neoplasm. Tumors with less hierarchical papillae are categorized as cystadenoma. Micropapillary SBOT (M-SBOT) refers to an SBOT showing at least one 5mm uninterrupted focus of micropapillary or cribriform pattern. Tumors with smaller micropapillae are classified as SBOT. Destructive invasion in a borderline tumor indicates the presence of a carcinoma. The invasive carcinoma associated with SBOT is usually low-grade serous carcinoma (LGSC). SBOT, M-SBOT, and LGSC frequently show a spectrum of histologic changes ranging from flat cyst wall, hierarchical structures (with primary papillae branching into secondary papillae), micropapillae, and invasive carcinoma. The wide range of histologic changes provides an ideal system to study their biological relationship. Recent genetic and molecular studies suggest a stepwise progression from serous cystadenoma, SBOT, M-SBOT, to LGSC; this pathway is characterized by mutations in BRAF and KRAS [2,3]. In general, the development of tumor requires alterations of multiple growth control genes. Up to now, however, only a limited number of molecular changes have been known to occur in the pathway of SBOT, MSBOT, and LGSC. Selenium-binding protein 1 (SELENBP1) has been shown to be expressed differentially between normal and neoplastic tissues [4]. SELENBP1 is a cellular protein that binds to exogenously administered selenium. Numerous reports have described a relationship between insufficient selenium intake and increased risk of cancer [5-7]. SELENBP1 has been suggested to mediate the anticancer action of selenium as it is decreased in cancers of the prostate, the lung, and the colon [8-10]. By membrane proteome profiling analysis, SELENBP1 was found to be the most significantly down-regulated protein in ovarian cancer cells [4]. Down-regulation of SELENBP1 was also demonstrated in ovarian carcinomas in the experimental laying hens [11]. To date, however, no systematic studies have been carried out concerning the expressions of SELENBP1 in SBOT, M-SBOT, and LGSC.
In this study, we examined the expressions of SELENBP1 in a wide spectrum of histologic components within SBOT, M-SBOT, and LGSC.
2. Materials and methods We retrieved a total of 62 cases of SBOT, 11 of M-SBOT, and 7 of LGSC from our institutional pathology archives during a period between 1993 and 2005. Of these cases, 68 were the result of total hysterectomy with bilateral salpingooophorectomy, and 12 were from salpingo-oophorectomy or cystectomy alone. The original pathology reports were reviewed for gross characteristics of the tumors. Archival hematoxylin and eosin slides were evaluated to confirm the diagnosis. SBOT, M-SBOT, and LGSC were defined based on the consensus opinion at the Borderline Ovarian Tumor Workshop held in 2003 [1]. Grossly, all 62 cases of SBOT showed complex cysts with papillary projections. The 11 cases of M-SBOT exhibited largely complex cysts with focal solid component. All 7 cases of LGSC showed partly solid and partly cystic components. Under microscopic examination, each lesion showed many of the histologic components in the spectrum of flat cyst wall, primary and secondary papillae of the hierarchical structures, micropapillarity, and invasive carcinoma as in LGSC (Table 1). Flat cyst wall was lined by a single layer of bland-appearing serous epithelium within a cystic tumor (Fig. 1A). Primary papillae protruded from cystic wall and contained broad fibrovascular cores lined by serous epithelium (Fig. 1B). Secondary papillae arose from the primary papillae and were composed of tiny papillae lined by serous epithelium (Fig. 1B). Micropapillae were elongate filiform papillary structures that arose directly in a nonhierarchical fashion from cyst walls or from large fibrous or edematous papillae and contained cores with little or no connective tissue (Fig. 1C). Cribriform morphology with punched-out lumens was classified under micropapillarity [1]. Frankly invasive carcinoma exhibited areas of stromal invasion greater than 10 mm2 (Fig. 1D). Stromal invasion of smaller than 10 mm2 was defined as microinvasion [12]. One case of M-SBOT showed microinvasion. Representative sections of various histologic components of tumor, as well as normal fallopian tube and ovary when present, were cut at 5 μm from paraffin blocks for SELENBP1 immunohistochemical studies. The sections were routinely deparaffinized and hydrated through a gradient of ethanol. Antigen retrieval was achieved by
Progressive loss of SELENBP1 expression Table 1
257
Histologic characteristics of tumors
Tumor types
SBOT (62 cases) M-SBOT (11 cases) LGSC (7 cases) Total
No. of cases with particular histologic components Flat cyst wall
Hierarchical structures Primary papillae
Secondary papillae
62 6 1 69
62 10 3 75
62 10 3 75
Micropapillae
Invasive carcinoma
9 11 6 26
0 1 7 8
NOTE. The area of micropapillae in SBOT was smaller than 5 mm. Each case of M-SBOT showed at least one 5-mm uninterrupted focus of micropapillae, although smaller foci of micropapillae were also seen. The invasive carcinoma seen in one case of M-SBOT measured 4 mm2. Each case of LGSC showed a greater than 10 mm2 area of stromal invasion. In the 26 cases with micropapillae, 22 were of the micropapillary type, and 4 mixed micropapillary and cribriform type.
incubating the sections with citrate buffer (pH 6.1) at 95°C. Sections were immersed in 3% H2O2 at room temperature for 10 minutes to block any endogenous peroxidase activity. They were incubated with a monoclonal mouse antibody against human SELENBP1 (MBL International, Woburn, MA) at a dilution of 1:200 at room temperature for 35 minutes.
The sections were then incubated with a secondary antibody previously conjugated to horseradish peroxidase–labeled polymer at room temperature for 35 minutes. After the sections were incubated with diaminobenzidine and H2O2 at room temperature for 8 minutes, they were counterstained with hematoxylin and cover-slipped.
Fig. 1 Illustration of various histologic components (hematoxylin and eosin stain, original magnification ×100). Flat cyst wall is lined by a single layer of serous epithelial cells (A), primary papillae protrude from cyst wall and are lined by serous epithelium (long arrow in B), secondary papillae (short arrow in B) arise from primary papillae, micropapillae are elongate filiform papillary structures that arise directly in a nonhierarchical fashion from cyst walls or from large fibrous or edematous papillae and contained cores with little or no connective tissue (C), and invasive carcinoma exhibits stromal invasion (D).
258 In each run of the staining, we included several controls: (1) a negative reagent control (MBL International) used to substitute the primary antibody, (2) a positive tissue control with a section of normal fallopian tube known to be positive for SELENBP1, and (3) a negative tissue control with a section of high-grade ovarian serous carcinoma known to be negative for SELENBP1. The specificity of immunostaining was confirmed by a positive stain in the mucosal epithelial
C. Zhang et al. cells of fallopian tube but a negative stain in high-grade ovarian serous carcinoma and on the sections that were stained with a negative reagent control substituting the primary antibody. Based on our preliminary results, we scored SELENBP1 immunostains using a 4-point scale (0 to 3+) system; staining in fewer than 5% of cells was recorded as “0,” staining in 6% to 30% of cells as 1+, staining in 31% to
Fig. 2 Immunohistochemical evaluation of SELENBP1 expressions. Diffuse and strong SELENBP1 immunostain was seen in the mucosal epithelial cells of normal fallopian tube (A), flat cyst wall (B), and primary papillae (long arrow in C). The immunostain was decreased in secondary papillae (short arrow in C). Almost complete loss of SELENBP1 immunoreactivity was seen in micropapillae (D) and invasive carcinoma (E and F). Original magnifications ×100 in A, B, C, D, and E; ×400 in F.
Progressive loss of SELENBP1 expression Table 2
259
SELENBP1 immunohistochemical characteristics in various histologic components Normal fallopian Normal tube ovary
No. of cases 70 SELENBP1 immnostaining 2.7 ± 0.4 scores
Flat cyst Hierarchical structure Micropapillae Invasive wall carcinoma Primary papillae Secondary papillae
36 69 75 2.7 ± 0.3 2.8 ± 0.3 2.4 ± 0.4
75 1.3 ± 0.4
26 0.2 ± 0.4
7 0.2 ± 0.4
NOTE. Statistical analysis of SELENBP1 scores are as follows. (1) Among normal fallopian tube, normal ovary, flat cyst wall, and primary papillae, P = .631. (2) Between micropapillae and invasive carcinoma, P = .602. (3) Secondary papillae compared to normal fallopian tube, P = .015; secondary papillae compared to normal ovary, P = .018; secondary papillae compared to flat cyst wall, P = .012; secondary papillae compared to primary papillae, P = .019. (4) Secondary papillae compared to micropapillae, P = .031; secondary papillae compared to invasive carcinoma, P = .033. The one case of microinvasion was not included for statistical analysis.
60% of cells as 2+, and staining in more than 60% of cells as 3+. In a preliminary study, we also divided the intensity of immunostaining into 4 groups: absent, weak, moderate, and strong. By Spearman test, we found a strong positive correlation between the intensity of immunostaining and the percentage of positive cells (R = 0.96) in 10 preliminary samples. Thus, scoring of immunostains was based only on the percentage of positive cells in the final study. Also, because the nuclear and cytoplasmic stains paralleled, total cellular (instead of separate nuclear and cytoplasmic) scores were assessed. In cases showing separate foci of the same histologic pattern, an average score of staining for the same histologic pattern was used. Kruskal-Wallis test was used to compare the scores of immunostaining among different histologic components. The same test was also used to compare the patients' age or tumor size among SBOT, M-SBOT, and LGSC. MannWhitney test was used to compare the immunostaining scores in the same histologic components, such as flat cyst wall versus flat cyst wall, primary papillae versus primary papillae, secondary papillae versus secondary papillae, or micropapillae versus micropapillae, between SBOT and M-SBOT. The difference was considered significant when the P value was smaller than .05.
3. Results The patients' age ranged from 27 to 82 years (mean, 48.5 years) in SBOT, 37 to 72 years (mean, 51.6 years) in M-SBOT, and 40 to 76 years (mean, 52.5 years) in LGSC.
Table 3
Comparison of SELENBP1 immunostaining scores in the same histologic components between SBOT and M-SBOT Flat cyst wall
SELENBP1 immunostaining scores P
Although there was a tendency for an average older age from SBOT, M-SBOT, to LGSC, the difference did not reach a statistical significance (P = .242). A significant difference in tumor size was seen in the following order: SBOT (from 4.8 to 15.4 cm; mean, 7.2 cm) b M-SBOT (from 8.0 to 21.5 cm; mean, 10.8 cm) b LGSC (from 10.5 to 26.5 cm; mean, 15.5 cm) (P = .041). Immunohistochemistry showed strong and diffuse SELENBP1 staining in mucosal epithelial cells of normal fallopian tube (the staining was somewhat more intense in the ciliated cells than in the secretory and intercalated cells) (Fig. 2A). Diffuse and strong SELENBP1 immunostaining was also seen in normal surface epithelium of ovary, epithelial cells lining flat cyst wall (Fig. 2B), and epithelial cells lining primary papillae of the hierarchical structures (Fig. 2C). The immunoreactivity in the epithelial cells of secondary papillae in the hierarchical structures was significantly lower than that in normal fallopian tube, normal ovary, flat cyst wall, or primary papillae (Table 2). The immunoreactivity in micropapillae (Fig. 2D) was similar to that in invasive carcinoma (Fig. 2E and F), but both were significantly lower than that in secondary papillae of the hierarchical structures. Because the immunoreactivity remained the same regardless of the size of micropapillae, micropapillae of different sizes were not separated in this report. No difference in SELENBP1 expression was seen in the same histologic components between SBOT and M-SBOT (Table 3). The immunostaining in the focus of microinvasion in one case of M-SBOT was similar to that in micropapillae or frankly invasive carcinoma, although statistic comparison was not performed because of the small sample size concerning microinvasion.
SBOT M-SBOT
2.8 ± 0.2 (62) 2.9 ± 0.3 (6) .576
NOTE. The value in parenthesis indicates the number of cases.
Hierarchical structures Primary papillae
Secondary papillae
2.4 ± 0.3 (62) 2.5 ± 0.4 (10) .554
1.3 ± 0.4 (62) 1.3 ± 0.5 (10) .521
Micropapillae
0.3 ± 0.5 (9) 0.2 ± 0.4 (11) .502
260
4. Discussion Within SBOT, M-SBOT, and LGSC, the degree of epithelial proliferation and papillary complexity in an increasing order ranged from flat cyst wall, primary papillae, secondary papillae, micropapillae, and invasive components. Our study showed a gradual loss of SELENBP1 expression associated with increasing epithelial proliferation and papillary complexity within SBOT, M-SBOT, and LGSC. The highest level of SELENBP1 expression was seen in flat cyst wall and primary papillae of the hierarchical structures, similar to that in normal fallopian tube and ovary. An intermediate level of SELENBP1 expression was observed in secondary papillae of the hierarchical structures. A near complete loss of SELENBP1 expression was seen in micropapillae and invasive carcinoma. This is the first systematic study to examine SELENBP1 expression in SBOT, M-SBOT, and LGSC. The expression of SELENBP1 in normal fallopian tube and ovary indicates a normal biologic function of SELENBP1 in these tissues. The expression of SELEBP1 in flat cyst wall and primary papillae of the hierarchical structures within the tumor at the same level as in normal fallopian tube and ovary indicates that the functions of SELENBP1 are retained in these histologic components. The exact functions of SELENBP1 are unclear, but a gradual decrease of SELENBP1 associated with increasing epithelial proliferation and papillary complexity in SBOT, M-SBOT, and LGSC suggests that SELENBP1 is related to the development of these tumors. Although loss of SELENBP1 expression may be an effect of tumor development, the growth inhibitory roles of SELENBP1 in liver and prostate cells [8,13] may suggest that the protein may be involved in tumorigenesis of SBOT and LGSC of the ovary. The near complete loss of SELENBP1 in both micropapillae and invasive carcinoma suggests that the biological features of malignancy may have been established in micropapillae. Although it is likely that additional genetic events contribute to tumor development, reduction of SELENBP1 expression may be a key step in the transition from the more quiescent (normal tissue, flat cyst wall, and primary papillae of the hierarchical structures), to the more proliferative (secondary papillae of the hierarchical structures), and finally to the most proliferative (micropapillae and invasive carcinoma) tissue. Provided that various histologic components within the same neoplasm are derived from the same clone, differential expressions of SELENBP1 indicate that the various histologic components represent a spectrum of cellular differentiation. Of special interest was our finding showing a more severe loss of SELENBP1 expression in secondary than primary papillae of the hierarchical structures. This may seem difficult to understand if one argues that the epithelial cells lining the secondary papillae are probably better differentiated than those lining the primary papillae based on the frequent observation of smaller epithelial cells in the
C. Zhang et al. secondary than in the primary papillae. The smaller size of the epithelial cells in the secondary papillae, however, may be the result of degenerative changes rather than their better biologic differentiation compared to the cells in the primary papillae. Indeed, the presence of secondary papillae indicates more exuberant proliferation of ovarian serous neoplasms and is required to distinguish borderline tumors from cystadenomas. It then becomes logical to find a decrease in SELENBP1 expression associated with proliferation from primary to secondary papillae of the hierarchical structures. The current definition of M-SBOT requires the presence of at least one 5-mm continuous focus of micropapillary or cribriform morphology in an otherwise SBOT [1]. Compared to SBOT, M-SBOT is more often associated with bilaterality, surface ovarian involvement, extraovarian diseases, invasive implants, and recurrence [12]. The biologic significance of lesions with micropapillarity smaller than 5 mm is less clear. Our study showing the same degree of SELENBP1 reduction in the micropapillae regardless of size suggests that smaller micropapillae may be similar to the larger ones at the molecular level. Our findings warrant further clinicopathologic studies on the biological behavior of SBOT with smaller than 5-mm focus of micropapillae. Loss of SELENBP1 is a late event in colorectal carcinogenesis as severe reduction of SELENBP1 is frequent in colorectal carcinomas but uncommon in adenomas [10]. In contrast, marked reduction of SELENBP1 appears to be an early event in tumorigenesis of SBOT, M-SBOT, and LGSC as it is shown by our study to occur in the secondary papillae of the hierarchical structures within the tumors. Yet, loss of SELENBP1 expression is a later event in this pathway compared to mutations of BRAF and KRAS that take place as early as in flat cyst wall of serous cystadenoma [14]. Decreased expression of SELENBP1 was previously reported in borderline ovarian tumors, but it was not known whether the reduced SELENBP1 expression was uniform in different histologic components within borderline tumors [4]. We were able to demonstrate a differential expression of SELENBP1 in various histologic components within SBOT. Although a decreased expression of SELENBP1 was previously described in high-grade ovarian serous carcinoma of the human subjects [4] and of the experimental hens [11], our study showed a reduction of SELENBP1 expression in LGSC of the human ovary. SBOTs can be seen in women of all ages, but many occur during reproductive age. In perimenopausal and menopausal women, hysterectomy and bilateral salpingo-oophorectomy are generally the treatment of choice. More conservative surgical procedure, however, may be a better option for younger women who wish to preserve their fertility. Effective chemotherapy may be necessary to prevent tumor recurrence in these patients. However, SBOTs and LGSCs do not respond to conventional cytotoxic chemotherapy because of their indolent nature. Thus, searching for novel target-based therapy is needed. Selenium has been implicated as an important chemopreventive and chemotherapeutic
Progressive loss of SELENBP1 expression agent for several epithelial tumors, including cancers of the prostate and colon [15-17]. Our study showing a differential expression of SELENBP1 in the various components of SBOT, M-SBOT, and LGSC not only indicates a role of SELENBP1 in the development of these tumors but also suggests the potential utility of selenium in prevention and treatment of SBOT and LGSC. Because the effects of selenium are mediated by SELENBP1, loss of SELENBP1 expression in tumors may have a negative impact on the ability of selenium to control tumor cell growth. It has been reported, however, that treating ovarian tumor cells with a selenium compound increases SELENBP1 expression [4]. The increased expression of SELENBP1 secondary to selenium treatment may facilitate the anticancer effect of selenium in ovarian tumors. In summary, our study revealed a differential expression of SELENBP1 in various histologic components of SBOT, M-SBOT, and LGSC. The gradual loss of SELENBP1 expression associated with increasing epithelial proliferation and papillary complexity indicates that SELENBP1 is involved in tumorigenesis. Our findings may provide a basis for future studies concerning the molecular mechanisms of SELENBP1 in tumorigenesis as well as a possible role of selenium in chemoprevention and treatment of SBOT, M-SBOT, and LGSC.
References [1] Seidman JD, Soslow RA, Vang R, et al. Borderline ovarian tumors: diverse contemporary viewpoints on terminology and diagnostic criteria with illustrative images. HUM PATHOL 2004;35:918-33. [2] Shih IM, Kurman RJ. Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. Am J Pathol 2004; 164:1511-8.
261 [3] Shih IM, Kurman RJ. Molecular pathogenesis of ovarian borderline tumors: new insights and old challenges. Hum Cancer Res 2005;11: 7273-9. [4] Huang KC, Park DC, Ng SK, et al. Selenium binding protein 1 in ovarian cancer. Int J Cancer 2006;118:2433-40. [5] Clark LC, Combs GF, Turnbull BW, et al. Effects of selenium supplementation for cancer prevention in patients with carcinoma of the skin. A randomized controlled trial. Nutritional Prevention of Cancer Study Group. JAMA 1996;276:1957-63. [6] Fleming J, Ghose A, Harrison PR. Molecular mechanisms of cancer prevention by selenium compounds. Nutr Cancer 2001;40:42-9. [7] Rayman MP. The importance of selenium to human health. Lancet 2000;356:233-41. [8] Yang M, Sytkowski AJ. Differential expression and androgen regulation of the human selenium-binding protein gene hSP56 in prostate cancer cells. Cancer Res 1998;58:3150-3. [9] Chen G, Wang H, Miller CT, et al. Reduced selenium-binding protein 1 expression is associated with poor outcome in lung adenocarcinomas. J Pathol 2004;202:321-9. [10] Kim H, Kang HJ, You KT, et al. Suppression of human seleniumbinding protein 1 is a late event in colorectal carcinogenesis and is associated with poor survival. Proteomics 2006;6:3466-76. [11] Stammer K, Edassery SL, Barua A, et al. Selenium-binding protein 1 expression in ovaries and ovarian tumors in the laying hen, a spontaneous model of human ovarian cancer. Gynecol Oncol 2008; 109:115-21. [12] Silverberg SG, Bell DA, Kurman RJ, et al. Borderline ovarian tumors: key points and workshop summary. HUM PATHOL 2004;35:920-7. [13] Bansal MP, Mukhopadhyay T, Scott J, Cook RG, Mukhopadhyay R, Medina D. DNA sequencing of a mouse liver protein that binds selenium: implications for selenium's mechanism of action in cancer prevention. Carcinogenesis 1990;11:2071-3. [14] Ho CL, Kurman RJ, Dehari R, Wang TL, Shih IM. Mutations of BRAF and KRAS precede the development of ovarian serous borderline tumors. Cancer Res 2004;64:6915-8. [15] Combs GF, Gray WP. Chemopreventive agents: selenium. Pharmacol Ther 1998;79:179-92. [16] Ip C. Lessons from basic research in selenium and cancer prevention. J Nutr 1998;128:1845-54. [17] Medina D. Mechanisms of selenium inhibition of tumorigenesis. Adv Exp Med Biol 1986;206:465-72.