INTRODUCTION

The human survivin gene spans 14.7 kb on telomeric position of chromosome 17 and is localized to band q25 (1, 2). Survivin protein has the ability to suppress apoptosis (programmed cell death) and regulate cell division (2). Although not present in most normal adult differentiated tissues, its mRNA and the protein are present in large amounts in fetal tissue. Most cancers, regardless of lineage, differentiation, and histologic type, overexpress survivin, suggesting reactivation of the gene (1) and a potential target for drug therapy (3).

Survivin is a member of the inhibitor-of-apoptosis proteins (IAP) family (4, 5). The possible mechanism whereby it blocks apoptosis is via an effect on caspase-9, which is activated via extrinsic and intrinsic pathways (5, 6). In the latter cell death pathway, upstream stimuli such as activation of p53 induce expression of pro-apoptotic bcl-2 family proteins such as bax. Anti-apoptotic proteins such as bcl-2 and bcl-XL inhibit the membrane-permeabilizing effects of bax and other pro-apoptotic proteins and also may inactivate caspase-9 by binding to, and inactivating, the adapter protein Apaf-1 (7, 8). Hence, caspase-3 and other downstream effector caspases are not activated, and apoptosis is inhibited. Survivin suppression of apoptosis is thought to be at the level of caspase 9, where it causes dissociation of the Apaf-1-caspase 9 complex. Caspase-3 is not activated, and apoptosis is inhibited (4, 9).

Survivin’s interaction with the mitotic spindle apparatus may be essential for its anti-apoptotic function. Disruption of the survivin–microtubule interactions that occur at the beginning of mitosis results in increased caspase-3 activity with loss of the anti-apoptotic function of survivin (10). Survivin expression in carcinoma nuclei or by reverse transcriptase polymerase chain reaction (RT-PCR) is shown to strongly correlate with proliferation index in hepatocellular (11), endometrial (10, 11), and ovarian (12, 13) carcinoma. It is felt that survivin promotes cell proliferation by interacting with Cdk4 with the release of p21 (11).

Survivin expression correlates with more aggressive behavior in colorectal (5, 14) and gastric (15) carcinoma and in neuroblastoma (16), with increased recurrences and metastases and reduced life expectancy (17). It is associated with reduced apoptosis in vivo (17). Because most anticancer chemotherapy kills cancer cells by activating apoptosis, increased survivin expression may decrease effectiveness of such therapy. Thus, survivin antagonists may be beneficial in treatment of such cancers (lymphoma, malignant melanoma, breast, cervix, prostate, lung, colorectal, and basal and squamous cell skin cancers) (3, 5, 18, 19, 20, 21, 22, 23, 24, 25, 26).

We studied the frequency of survivin expression in human ovarian carcinoma to assess the possible future role for therapies that inactivate this protein and, hence, increase apoptosis. Survivin expression also was correlated with expression of bcl-2, bcl-x, and mutant p53 (anti-apoptotic markers) and of bax (a pro-apoptotic marker), with clinicopathologic prognostic parameters, and with outcome.

MATERIALS AND METHODS

Forty-nine ovarian carcinoma diagnosed consecutively in the Department of Pathology, Emory University Hospital, Atlanta, Georgia between 1984 and 1995 were studied, provided paraffin blocks were available. They were graded as 1 to 3 (well, moderately, and poorly differentiated, respectively) on the basis of cellular atypia, architectural complexity, and invasive propensity (27). FIGO staging was based on findings at clinical examination and by surgical exploration. Stage I represents growth limited to the ovaries, Stage II involves one or both ovaries with pelvic extension, Stage III includes tumors that involve one or both ovaries with peritoneal implants outside the pelvis and/or positive retroperitoneal or inguinal nodes, and Stage IV represents tumors with distant metastases (27). Clinicopathologic parameters and follow-up were obtained from surgical pathology reports and from the Oncology Data Bank, Winship Cancer Center, Atlanta, Georgia.

Immunohistochemistry

Five-micrometer sections of the formalin-fixed, paraffin-embedded tumors were immunostained, after steam heat–induced epitope retrieval, using polyclonal survivin (1:80; Santa Cruz Biotechnology, Santa Cruz, CA), monoclonal bcl-2 (1:160; Onc 124; DAKO, Carpinteria, CA), polyclonal bcl-x (1:1500; PharMingen, San Diego, CA), polyclonal bax (1:1500; PharMingen), and monoclonal p53 (1:20; DO-7; Novocastra, Newcastle upon Tyne, UK). For survivin, an avidin biotin-complex kit (LSAB + System; DAKO) was used with the DAKO Autostainer. With the other four antisera, an avidin biotin enzyme-complex kit (Signet Laboratories Inc., Dedham, MA) was used in combination with the automated TechMate 1000 immunostaining system (Biotek Solutions Inc., Santa Barbara, CA; 28). Hematoxylin was used as counterstain for survivin and p53, and hematoxylin and eosin, as counterstains for the other three antigens. Sections of colonic adenocarcinoma were used as positive controls for survivin (Fig. 1) and p53, tonsil for bcl-2, and normal breast ducts and lobules for bcl-x and bax. Negative controls had primary antibody replaced by buffer. The three polyclonal antibodies acted as isotype-specific controls for each other. The survivin antibody used has been shown to react specifically with survivin of human origin by Western blotting and immunohistochemistry (IHC). Immunostain was recorded as 0–3+ according to stain intensity, distribution in cytoplasm and/or nucleus, and percentage of cancerous cells that stained positively. Tumors were scored on a four-tiered system, with <10% of carcinoma cells staining called negative; 10–25% positivity scored as 1, 26–50% scored as 2, 51–75% scored as 3, and 76–100% scored as 4.

FIGURE 1
figure 1

Survivin expression that is absent in normal colon (A); in nuclei of colon cancer (B); and in nuclei of two papillary serous carcinomas of the ovary (C and D). Magnification, 400×.

Full size image

Statistics

Survivin expression was compared with age, using a t test and compared with other categorical factors using Fisher’s exact test. Kaplan-Meier methods were used to calculate overall survival rates and disease-free survival rates among different groups. These groups were then compared using log-rank methods. Because of small sample size and the fact that few univariate results were significant, no multivariate methods were used.

RESULTS

Table 1 indicates the frequency in ovarian carcinoma of survivin and apoptotic marker expression as 74%, 37%, 94%, 78%, and 60% for survivin (Figs. 1 and 2), bcl-2, bcl-x, bax, and p53 (Fig. 3), respectively.

TABLE 1 Frequency of Expression of Survivin and Apoptotic Markers in 49 Ovarian Carcinoma
Full size table
FIGURE 2
figure 2

A, lack of nuclear survivin expression in an ovarian mucinous tumor. B–D, nuclear survivin expression in an ovarian poorly differentiated carcinoma and in two papillary serous carcinomas, respectively. Magnification, 400×.

Full size image
FIGURE 3
figure 3

Serous carcinomas of the ovary with expression of cytoplasmic bcl-2, bax, and bcl-x and nuclear mutant p53. Magnification, 400×.

Full size image

Table 2 documents the clinical and pathologic features of the 49 ovarian carcinoma studied, according to whether survivin expression was present or not. Histologic grade (Grades 2 and 3; P = .0014), histologic type (serous and poorly differentiated; P = .0376), and p53 mutation (P = .0414) were significantly associated with survivin expression. The number of endometrioid and mucinous carcinomas was too small for assessment of association. Although patient age tended to be older in patients with survivin expression in their carcinoma, none of the other parameters (clinical stage; expression of bcl-2, bcl-x, or bax) were significantly associated with survivin expression.

TABLE 2 Clinicopathologic Features in 49 Ovarian Carcinoma According to Survivin Expression
Full size table

Table 3 correlates overall and disease-free survival in 44 ovarian carcinoma patients according to whether survivin or the other apoptotic markers studied were present or absent. The mean overall survival obtained for these 44 patients was 4.4 years (range, 0.2–13.5 y). Mean disease-free survival was 3.8 years (range, 0.2–13.5 y). Expression of none of these markers was predictive or prognostic of improved overall or disease-free survival in ovarian carcinoma patients.

TABLE 3 Correlation of Survival with Expression of Survivin and Apoptotic Markers in 44 Ovarian Carcinoma Patients with Follow-Up
Full size table

DISCUSSION

Strong survivin expression has been demonstrated in solid malignancies of the lung (29), colon (30), pancreas (31), stomach (15, 32), esophagus (33), breast (34), prostate gland (35), bladder (36, 37), endometrium (10, 38), uterine cervix (38), and ovary (12, 13, 39, 40) and in hematopoietic malignancies (1), melanoma and non-melanoma skin cancers (19, 21, 41), and neuroblastoma (16). The percentage of survivin-positive cases varies from 35% in gastric cancer (15) to 93% in primary and metastatic malignant melanoma (19), with expression in 51–86% of ovarian cancers (12, 13, 40, 42). Reported as a cytoplasmic stain (2, 14, 15, 16, 32, 43), survivin expression is seen by us and others (11, 32, 39, 44) predominantly in nuclei. Survivin is also assayed by RT-PCR, with 76% correlation between mRNA and nuclear survivin expression by IHC; results are 23% negative and 77% positive with both techniques (45).

Survivin expression may be a useful diagnostic, prognostic, and predictive marker in certain malignancies. The presence of serum anti-survivin antibodies in patients with lung and colorectal cancer may prove a novel diagnostic marker (25, 46). Survivin assay in urine may be easier and cheaper for diagnosis of urothelial carcinoma than cystoscopy (47).

Survivin may also prove to be a useful prognostic marker (2). Table 4 demonstrates the prognostic impact of survivin in neoplasms according to whether it is demonstrated by IHC in nuclei (IHCn), by IHC in cytoplasm (IHCc), by RT-PCR, by Western blot, or by RNase protection assay. In general, the presence of survivin indicates a carcinoma of unfavorable histology; advanced stage; decreased 5-year, overall, and disease-free survival; and increased recurrence rate as compared with the case of carcinomas without survivin expression. This is most frequently seen with expression of survivin by IHCn, RT-PCR, Western blot, and/or RNase protection assay. However, in some studies of neuroblastoma (16), laryngeal squamous cell (48), and colorectal carcinoma (2, 5, 14, 30), poor prognosis was demonstrated associated with IHCc. In prostate cancer, neuroendocrine cells express survivin (35). Progression occurs in 63% of cancers with neuroendocrine differentiation, compared with 40% without. This suggests a trend to poorer prognosis of prostate cancers with neuroendocrine differentiation and survivin expression (35).

TABLE 4 Prognostic Impact of Survivin in Neoplasms According to Method of Survivin Assay
Full size table

Overall survival in gastric cancer (32) and disease-free survival in urothelial bladder carcinoma (56) are improved in the presence of nuclear survivin expression (IHCn). In multivariate analysis, survivin was shown to be an independent favorable prognosticator in gastric cancer (32). Although not statistically significant, disease-free survival was 27.2 months in urothelial carcinomas with survivin expression, versus 9.9 months without survivin (56).

Pancreatic (57), gastric (32, 58), esophageal (51), and urothelial (37) carcinomas show no association between survivin IHCc expression and patient survival. The gastric carcinomas were, in addition, studied with RT-PCR. The presence of cytoplasmic survivin (IHCc) in colorectal (14) and gastric (15) carcinoma was unrelated to histologic grade.

We show, in ovarian cancer, a 74% frequency of nuclear survivin expression that correlates with the 51–81% reported frequency by IHCn (12, 13), 86% by RT-PCR (40), and 73% by IHCc (42). Although we noted no effect of survivin expression on overall or disease-free survival, there was a correlation between IHCn and poor prognostic markers (grade, histologic type, and p53 mutation). Yoshida et al. (39) show significantly shorter disease-free survival, and Takai et al. (12) show a significant association with proliferation, clinical stage, histologic grade, clinical outcome, and survival rate in ovarian cancers with nuclear survivin expression. IHCc-expressing ovarian cancers showed clinical resistance to taxol/platinum-based therapeutic regimes, with a lower clinical or pathological complete remission rate (42).

Other studies have related apoptotic markers to prognosis in ovarian carcinoma. Lohmann et al. (59), showed increased bcl-2 expression in patients with shortened overall and disease-free survival and showed increased bcl-x and bax expression with increased overall and disease-free survival, although results were not statistically significant. For patients whose carcinomas showed bcl-2–bax and bcl-2:bcl-x ratios of <1, there was a trend toward survival advantage. Mano et al. (60) showed significantly poorer survival in patients with Stage III–IV ovarian serous or endometrioid carcinomas expressing bcl-2. Initial response to chemotherapy was inversely related to bcl-2 expression, which was also an independent prognostic factor (60). In several other studies (61, 62, 63, 64), bcl-2 expression was associated with favorable prognosis and low-grade carcinomas in ovarian cancer patients. Bax expression, particularly in bcl-2–negative cancers, correlated with a bad clinical outcome (63), or with improved prognosis, of independent prognostic significance (65). p53 mutation, on the other hand, appeared significantly more in high-grade cancers of advanced stage, with residual disease after surgery, and with decreased survival, being an independent prognostic indicator (61, 62, 66). Munakata et al. (67) showed no relationship between bax expression and prognosis. Apoptotic index is shown to increase with grade, being low in tumors of low malignant potential and mucinous carcinomas (62, 66). High apoptotic index has significantly correlated with high mitotic index, high-grade histology, and decreased overall survival (61), whereas an inverse relationship was noted between apoptotic index and bcl-2 expression.

bcl-2 is reported to be strongly expressed in surface epithelium of normal ovaries and in benign and borderline ovarian tumors, but weakly in carcinomas (69, 70, 71). bcl-xL and bax expression were higher in carcinoma than normal ovarian tissue (71). DNA demethylase (dMTase) is reported in almost 90% of ovarian carcinomas but only in 9% of non-cancerous ovaries (40). In dMTase-negative cancers (40%) and benign ovary (91%), survivin gene exon 1 was found to be methylated, compared with 79% unmethylated in dMTase-positive ovarian cancers (35). Exon 1 was methylated in 83% of survivin-negative cancers (86% were RT-PCR positive for survivin expression) (40). Thus, survivin gene exon 1 may be a target for demethylation by dMTase.

Inhibition of apoptosis is a common property of cancer cells, enabling them to increase their survival and facilitate their escape from immune surveillance and cytotoxic therapies. Survivin, as an inhibitor of apoptosis, appears to have a role in cancer progression or drug resistance (5, 42, 72). Survivin has also been shown to be an inducible radioresistance factor in pancreatic cancer cells (67) and possibly also in nonsmall cell lung carcinoma (72).

Survivin phosphorylation on threonine (Thr34) may be required to preserve cell viability at cell division. Manipulation of this pathway to regulate apoptosis may help in removal of cancer cells at mitosis (74). A phosphorylation-defective survivin mutant with alanine replacing threonine (Thr34→Ala) triggered apoptosis in several melanoma cell lines and increased cell death induced by cisplatinum in vitro (20). In established melanomas in mice, this survivin mutant resulted in increased apoptosis, decreased proliferation, and inhibition of tumor growth by 60–70% (20). Thus, survivin may be a new target for antimelanoma therapeutic strategies. A survivin (Thr34→Ala mutant adenovirus initiated the mitochondrial apoptotic pathway when used to infect cell lines of cervical, lung, colorectal, prostate, and breast carcinoma (22). It was as effective as taxol and more effective than adriamycin in inducing tumor cell apoptosis, and it increased taxol-induced cell death (22). In SCID mice injected with MCF-7 breast carcinoma cell lines infected with this mutant adenovirus, de novo tumor formation was suppressed, tumor growth was inhibited by 40%, and intraperitoneal spread decreased, with an associated increased apoptosis and decreased proliferation (22). Thus, adenoviral targeting of the survivin pathway may prove to be a novel cancer gene therapy for certain tumors (22). Inhibitors of cyclin-dependent kinases such as Flavopiridol also prevent Thr34 phosphorylation and have antitumor effect in renal, prostate, and colon cancer and in non-Hodgkins lymphoma (75).

Blockage of survivin expression may also be used therapeutically in cancers. Intratumoral injection of plasmids encoding antisense survivin blocks survivin expression, stimulates generation of tumor-specific cytotoxic T lymphocytes, decreases tumor growth, and may be a beneficial treatment for large cell lymphomas (18). Antisense oligonucleotides that reduce survivin expression in tumor cells induce apoptosis, decrease colony formation in soft agar, and sensitize tumor cells to chemotherapy (18, 19, 21, 76). Survivin antisense or survivin dominant negative mutant (Cystein 84→Ala) transfection of model keratinocyte and malignant melanoma cell lines resulted in decreased endogenous survivin, increased apoptotic cells, and decreased proliferation (19, 21). This therapeutic targeting of survivin may be useful in management of recurrent or advanced basal and squamous cell carcinoma, and melanoma, the majority of which (81%, 92%, and 93%, respectively) express survivin (19, 21). In a lung adenocarcinoma cell line, antisense oligonucleotides targeting different regions of survivin mRNA downregulated survivin mRNA by 70%, induced apoptosis (increasing caspase-3 activity, nuclear condensation and fragmentation, and trypan blue uptake), and sensitized tumor cells to the chemotherapeutic agent etoposide (76). Normal blood leukocytes, with minimal survivin mRNA, were unaffected. The molecular relationship between effector cell protease receptor-1 (EPR-1) and survivin, whereby ZnSO4 induction of an EPR-1 mRNA suppresses endogenous survivin expression, suggests a potential treatment of cancer with EPR-1 (77). Apoptosis is increased and cell proliferation inhibited.

Survivin may be required for tumor angiogenesis because it is highly expressed in newly formed blood vessels in response to vascular endothelial growth factor (VEGF) and basic fibroblast growth factor and mediates angiopoietin inhibition of endothelial cell apoptosis (78, 79, 80, 81). Hence, therapies inhibiting survivin expression or function may prevent tumors from acquiring an adequate blood supply (18). Antisense oligonucleotide to survivin is shown to suppress de novo expression of survivin in endothelial cells by VEGF (23).

Cytotoxic T lymphocyte responses against two survivin-deduced peptide epitopes are described in peripheral blood from chronic lymphatic leukemia patients and in lymph nodes with metastatic melanoma (24). Circulating antibodies to survivin are detected in serum of patients with colonic and lung cancer (25, 46). In vitro, survivin induced specific CD8+ effector T cells (82). Thus, survivin may be a target for anticancer immunotherapeutic strategies (24).

We show a significant correlation between survivin expression (IHCn) and p53 mutation (both with anti-apoptotic functions) in ovarian cancers. This relationship is explained by the recent demonstration of wild-type p53 repressing survivin expression at both mRNA and protein levels, binding the survivin promoter (83, 84). This association, however, is not always the case in ovarian cancer with cytoplasmic survivin expression (42). In colorectal carcinoma, p53 abnormalities were not shown to correlate with IHCc survivin expression (14), although the latter was strongly associated with bcl-2 expression and reduced apoptotic index (14). Survivin expression (IHCc), even in bcl-2–negative colorectal cancers, also resulted in reduced apoptotic index (14). Similarly, in gastric (15) and breast carcinoma (34), cytoplasmic survivin was significantly associated with low apoptotic index, which, in breast carcinoma (34), was an independent prognostic factor associated with poor survival. In ovarian cancers, we fail to show a correlation between nuclear survivin expression and bcl-2 or bcl-x, all apoptotic inhibitors. On the other hand, Lu et al. (15) showed that cytoplasmic survivin expression significantly segregated gastric cancers with p53 and bcl-2 positivity. Apoptotic index was significantly decreased compared with that in survivin-negative tumors (15). Thus, correlation between different apoptotic markers varies from negative (bcl-2, p53, and apoptotic index [30, 57, 62]; survivin and p53 [14, 30]; survivin and bcl-2, bcl-x, bax, this study; survivin mRNA and apoptotic index [58]) to inverse (bcl-2 and p53 [64, 66]; survivin and apoptotic index [14, 31]) to positive (survivin and bcl-2 [14, 15]; survivin and apoptotic index [57]; survivin and mutant p53, Lu et al. [15], Sarela et al. [57], and this study).

Grabowski et al. (51) report the translocation of survivin during carcinogenesis, with cytoplasmic localization in basal layers of the normal esophageal squamous mucosa, in high-grade dysplasia and in carcinoma, with, in addition, nuclear localization in high-grade dysplasia and the associated carcinomas. In hepatocellular carcinoma, the predominant function of survivin is shown to be its cell cycle nuclear distribution, not the cytoplasmic caspase-3–dependent antiapoptotic effect (11). The differences of the amino-acid sequence of the carboxy-terminal domain of survivin are shown to be responsible for its cytoplasmic localization and the nuclear localization of its splicing variant, survivin Δ Ex 3 (51). Thus, several reports (Table 4) show nuclear survivin immunostain to have prognostic significance, possibly related to its cell cycle effect, whereas cytoplasmic survivin, with its antiapoptotic effect, has no prognostic significance (32, 37, 51, 57, 58). Nuclear survivin expression has not been shown by us to be an independent prognostic or predictive factor to identify ovarian cancer patients at increased risk of recurrence or of poor overall and/or disease-free survival. Nevertheless, nuclear survivin expression, present in the majority of the ovarian carcinomas we studied, did correlate with poor prognostic factors (high histologic grade, mutant p53, poor histologic type) and may predict response to antisurvivin therapies that induce apoptosis reduction as a result of survivin expression (14).