Vascular Endothelial Growth Factor Expression in Ovarian Cancer: A Model for Targeted Use of Novel Therapies? Free

Ovarian cancer is the leading cause of death from gynecologic malignancies. As symptoms are vague and nonspecific, these tumors often present at an advanced stage. Developments in chemotherapy and surgical techniques have had minimal effect on patient survival over the last few decades (1).

Tumor stage and residual tumor mass, following primary cytoreductive surgery, have been shown to most reliably predict outcomes in patients with ovarian cancer (2) but offer no information with regard to the potential sensitivity to molecular “targeted” therapy. Investigation of novel prognostic markers offers an insight into the mechanisms of tumor development and suggests potential avenues for the development of new therapeutic agents particularly through the use of monoclonal antibody therapies.

Angiogenesis has been established as a vital component in the mechanisms involved in tumor growth and metastasis (3, 4). The angiogenic potential of tumors can be assessed by microvessel density. Earlier studies illustrated the importance of angiogenesis in tumor development, with microvessel density directly correlating with a poor prognosis in ovarian (5) and other tumors (6, 7).

Vascular endothelial growth factor (VEGF) is a multifunctional cytokine that stimulates angiogenesis and increases microvascular permeability through binding to specific receptors expressed on vascular endothelial cells (8, 9). Although VEGF is produced by several tumors and hypoxic tissues (10), its receptors are expressed primarily by endothelial cells. It has been shown to have a crucial role in neovascular formation in tumors, providing nourishment for the highly metabolic tumor cells as well as providing access to the host vasculature (11).

Studies have suggested a specific role for VEGF in various phases of ovarian carcinogenesis, with effects on tumor growth and neovascularization seen in animal models and in humans (12, 13). Higher levels of VEGF are shown in ovarian carcinomas when compared with normal ovaries (14, 15).

Recent interest has focused on the use of antiangiogenic drugs in an attempt to inhibit the protumor effects of VEGF and other such cytokines. These studies have shown some antitumor effects but have shown significant side effects. The future role of such therapies is yet to be established (16–18).

Previous research has produced inconsistent evidence with regard to the importance of VEGF in ovarian cancer and its relation to prognosis. These studies often suffered from low patient numbers and the use of subset analysis (19–21). This study was designed to determine VEGF status in patients with ovarian carcinoma and investigate its relation to prognosis. This was done through the assessment of over 350 consecutive patients using tissue microarray technology. Second, elucidation of the prognostic role of VEGF in patients with ovarian cancer might enable more individualized adjuvant treatments using developing novel therapies to inhibit tumor angiogenesis; these therapies are likely to be most effective in tumors expressing high levels of VEGF.

Patients. A total of 358 patients with ovarian cancer were entered in this study, and these consisted of patients undergoing a laparotomy for primary ovarian cancer. Information on cancer size, stage, presence or absence of residual disease after surgery, histologic type and grade, age at diagnosis, and type of adjuvant treatment were collected for all patients (Table 1). From this original population, histologic material was available for analysis in 339 cases. The paraffin-embedded tissue blocks from these patients dated back from January 1, 1984 until December 31, 1997. Disease-specific survival was calculated from the operation date until November 31, 2005 when any remaining survivors were censored. The database was audited to ensure validity; there were no major discrepancies with >97% of data available.

From each of the 339 patients with histologic material available, two tumor samples were used for analysis. In total, 320 carcinomas were analyzed for VEGF following the omission of the cores, which were “uninterpretable” due to tissue loss during processing (6% core loss).

During the study period, patients with high-grade stage I and stage II to IV disease received chemotherapy. The specific chemotherapy varied but reflected the best current practice; most recently, this treatment was platinum based. Sixty-two patients participated in the International Collaborative Group for Ovarian Neoplasia trials I to IV during which the allocated chemotherapy was randomized.

Although the study spans a 14-y period, there was no significant change in the survival of patients treated in the earlier or latter part of the study. This is in line with the unaltered survival of ovarian cancer patients over the last 30 y (22).

Tissue microarray construction. All tumors received following resection in the operating theater were incised, fixed immediately in 10% neutral buffered formalin overnight, and then embedded in paraffin wax, ensuring optimal tissue fixation and preservation for histologic examination.

Tissue microarrays were constructed as described previously (23). For each tumor, 5-μm section slides stained with H&E were first used to locate representative areas of viable tumor tissue. Needle core biopsies (0.6 mm) from the corresponding areas on the paraffin-embedded tumor blocks were then placed at prespecified coordinates in recipient paraffin array blocks using a manual tissue arrayer (Beecher Instruments). Array blocks were constructed with between 76 and 133 cores in each, and five copies of the array were assembled using different points within the representative tumor area. Fresh 5-μm sections were obtained from each tissue microarray block and placed on coated glass slides to allow the immunohistochemical procedures to be done, preserving maximum tissue antigenicity.

Immunohistochemical staining. A prediluted rabbit anti-human VEGF antibody was used (SP28, Abcam). Optimization of the staining was done on ovarian cancer whole-section mounts using a range of incubation times. Two-hour incubation was chosen to stain the arrays as it showed the best results with minimal background staining. Two copies of the ovarian tissue microarrays were cut from the paraffin blocks (4 μm thickness), transferred to extra-adhesive glass slide, and stained in a single run using a routine streptavidin-biotin peroxidase technique. Briefly, slides were dewaxed in xylene (2 × 10 min) and rehydrated in three grades of ethanol (99%, 90%, and then 70%) for 1 min each. Endogenous peroxidase activity was blocked by immersing the slides in a 0.03% solution of H₂O₂ in methanol for 25 min. Heat-induced epitope retrieval was done using an 800-W rotary microwave oven. Slides were put in a plastic vessel containing 10 mmol/L sodium citrate buffer (pH 6.0) and treated for 10 min on high power and then for 10 min on low power. Slides were then taken out and cooled down under running tap water for 20 min and washed with TBS (Dako) for 10 min more. To reduce nonspecific adsorption of antibodies to tissue, the slides were incubated with normal swine serum (Dako) diluted 1:20 in TBS for 10 min. The test sections were then left to incubate with the anti-VEGF antibody for 1 h at room temperature. Negative control sections were incubated with normal swine serum under the same conditions. Following a thorough wash with TBS, 100 μL of the biotinylated anti-mouse antibody, diluted 1:100 (Dako), were applied for 30 min, followed by another TBS wash and then 100 μL of the streptavidin-biotin/horseradish peroxidase complex solution (prepared 30 min in advance). This was left to incubate with the sections for 60 min. The color was developed using 3,3′-diaminobenzidine (Dako) and enhanced with 0.5% CuSO₄ solution, and the sections were then counterstained with Mayer's hematoxylin solution (Dako) for 1 min.

Scoring of cores. The intensity of the staining was estimated on a four-tiered scale, encoded as 0 (absent), 1 (weak), 2 (moderate), and 3 (strong). The pathologist and researcher who reviewed the immunostaining of the tissue samples were blinded to the clinicopathologic data of the patients.

Statistical analysis. The correlation between VEGF expression levels and other prognostic variables was statistically analyzed by means of Pearson's χ² test.

Survival rates were examined using Kaplan-Meier plots for analysis of censored data. The statistical significance of differences between the survival rates of groups with different VEGF expression was assessed using the log-rank test. The independent prognostic significance of variables was assessed in multivariate analysis by means of a multivariant Cox regression model and the −2 log likelihood test (omnibus test). P values ≤0.05 were assumed statistically significant. All statistical analysis was done using the computer statistical program Statistical Package for the Social Sciences 15.0 (SPSS).

Clinicopathologic characteristics. In the current cohort of ovarian cancer patients, the mean age at diagnosis was 61 years (range, 24-90). Using the current Surveillance, Epidemiology, and End Results Program age categorization system, 59% of the patients were in group 3 (>60 years at diagnosis), 40% were 30 to 60 years at diagnosis, and only 2 of 357 were <30 years. Serous cystadenocarcinoma was the commonest histologic type (49%) followed by undifferentiated (15%), endometrioid (12%), mucinous cystadenocarcinoma (10%), clear cell (7%), and other types (7%). All patients were treated surgically, of which 42% had their masses optimally debulked with no macroscopic disease left. Clinicopathologic staging showed that the majority of patients (50%) were stage III followed by 27% in stage I, 11% in stage IV, and 11% in stage III. When histologic grading was applicable, almost two thirds of the tumors were poorly differentiated (grade 3). Twenty-two percent were moderately differentiated, and only 11 were deemed well differentiated. All patients' characteristics are summarized in Table 1.

VEGF analysis. VEGF analysis was done on 320 primary ovarian tumors, with the remaining 19 tumor cores not being interpretable due to tissue loss during the immunohistochemical processing. When staining was positive, it was primarily of cytoplasmic location, and its pattern was uniform among the cancer cells within each core. Only 22 tumors (6.9%) were strongly positive, whereas the remaining tumors (298 of 320) exhibited either weak (135 of 320, 42.2%) or moderate (126 of 320, 39.4%) staining. Thirty-seven tumor cores (11.6%) failed to show any noteworthy staining. Photomicrographs of each category are shown in Fig. 1A to D. The scoring system was designed to identify potentially sensitive tumors to anti-VEGF therapies, and as such, cases were categorized as either “high” expressers (represented by the strongly stained group) or “low” expressers (composed of the negative, weak, and moderate groups).

VEGF correlations. Using χ² test, VEGF expression did not correlate with the patients' age, tumor grade, stage, or histologic type, nor was it associated with the presence or absence of residual disease or with the administration of adjuvant chemotherapy.

Survival analysis revealed that patients who had tumors with low levels of VEGF had a median survival time of 24.1 months, whereas that of high VEGF expression was 13.7 months (Table 2A). The 12- and 24-month survival rates for high VEGF expression were 56.0% and 38.1%, respectively, compared with 66.4% and 50.0% for low expression (Table 2B). This difference in survival was shown to be statistically significant with the log-rank test (test statistic = 4.2; P = 0.04), and a Kaplan-Meier survival plot is shown in Fig. 2.

Using the Cochrane-Armitage test, VEGF did not correlate with any of the established clinicopathologic prognostic variables (e.g., age, stage, and grade). On its own, high VEGF expression correlated with poor prognosis, yielding an unadjusted hazard ratio (HR) of 1.59 [95% confidence interval (95% CI), 1.02-2.49; P = 0.042]. However, to ensure the findings were truly independent of other clinicopathologic variables, the Cox model was expanded to incorporate stage, residual disease, and the receipt of adjuvant therapy. The adjusted HR for high VEGF expression was 1.78 (95% CI, 1.08-2.94; P = 0.023; Table 3). The independent effect of VEGF was reinforced by doing the Cox model excluding VEGF as a variable; the HRs for stage, residual disease, and adjuvant therapy were unaltered. Using the −2 log likelihood test (omnibus test), the effect of VEGF on our multivariate model was shown to be significant (P = 0.036). This suggests that the model as a whole is significant and the covariate effect of VEGF is different from zero.

Despite the assumed importance of VEGF in the progression of cancer, relatively few studies of the expression of this protein in ovarian cancer have been linked to comprehensive clinical data, particularly data on survival.

The distribution of histologic subtypes is consistent with established literature, with serous carcinoma being the most prevalent followed by endometrioid and mucinous (1). The typical distribution of ovarian cancer in relation to stage and differentiation is seen with the majority presenting with poorly differentiated, advanced (stage III and IV) disease (24). The clinicopathologic features correlate with established characteristics of ovarian cancer, suggesting that the study contains a representative cohort of patients, and hence, the study findings will be of relevance to a general population.

The expression of VEGF in our patients exhibited a similar pattern to that of Goodheart et al. (19), with a strong expression seen in 6.9% of the tumors. Lesser expression ranged from moderate (39.4%) to weak (42.2%) to no expression in 11.6% of the patients. Other groups reported varying levels of VEGF using different antibodies to stain the tissue, different scoring systems, and different cutoff points.

Our results indicate that patients with tumors that express high levels of VEGF have worse survival rates compared with those with medium, low, or no VEGF. A median survival advantage of ∼10 months is seen among the latter group when compared with the former (P = 0.04; Fig. 2). This result is further substantiated when a multivariant Cox regression model was constructed in which established prognostic factors were accounted for. VEGF maintained statistical significance with regard to patient survival (P = 0.023; Table 3). The HR indicates that patients with high VEGF have ∼75% higher risk of dying than their low VEGF counterparts.

The role of angiogenesis in ovarian carcinoma development remains unclear; there are contradictory studies with regard to the influence of microvessel density in ovarian cancer prognosis (5, 25, 26). Immunohistochemical assessment of VEGF within a tumor offers further information about the potential for angiogenic activity and its effects on tumor behavior and subsequent prognosis for the patient. The literature has been equally controversial with regard to VEGF expression within ovarian tumors, some authors showing no independent relationship with prognosis (27, 28), whereas other studies have shown a significant independent prognostic influence. Patients with early-stage disease (International Federation of Gynecologists and Obstetricians stages I and II) showed poorer prognosis with increased VEGF expression within the tumor (29). Shen et al. (15) showed elevated expression of VEGF (using mRNA) to be predictive of a poor prognosis; interestingly, there was no correlation with microvessel density, which contradicts previous work (27). Raspollini et al. (21) illustrated that VEGF and microvessel density were both independent predictors of survival in advanced disease (International Federation of Gynecologists and Obstetricians stage III) and also correlated with the likelihood of response to chemotherapy; similar findings were also seen when including early-stage disease (20). Because we have such a large number of cases, we can clearly show that the prognostic effects of VEGF are seen at all disease stages and that these effects are independent of confounding variables such as stage, grade, and residual disease.

Our study illustrated no clear associations between VEGF and any of the clinicopathologic variables, including stage and grade of tumor; this agrees with some studies, most of which had a typical distribution of disease according to stage (5, 20, 27, 30, 31). Some studies have suggested that stage and grade are associated with VEGF, although these studies tended to have an unusually high proportion of early-stage disease (up to 58%; refs. 15, 32, 33).

Zhang et al. (34), in a study of infiltrating T cells in ovarian cancer, showed that the absence of intratumoral T cells was associated with higher levels of VEGF. This group of patients had early recurrence rates and short survival. It is therefore thought that VEGF further affects the behavior of ovarian tumors by reducing the number of T cells in the tumor milieu, suppressing the defenses of the immune system against ovarian cancer.

There is early evidence that VEGF-mediated angiogenesis may be used as a novel pathway in the treatment of ovarian cancer as has been witnessed in colon (35), breast (36), and lung (37) carcinomas. Monk et al. (38) have shown some clinical benefit from using a monoclonal antibody against VEGF (bevacizumab) in recurrent ovarian cancer; this has also been used successfully in the palliative treatment of ascites in refractory disease (39). Another strategy devised to suppress angiogenesis in ovarian cancer is the interception of VEGF with receptor decoys, such as VEGF-Trap, which has shown encouraging results in early-phase trials (16, 40).

Our subgroup of high VEGF-expressing tumors accounts for <10% of patients, and hence, we are identifying a small group who seem to have a much worse prognosis. The small proportion of tumors expressing high levels of VEGF may explain why smaller studies failed to find significant prognostic associations (27, 28). The clinical value of identifying these patients and adapting treatment will therefore have a limited effect on overall population survival; however, we may have identified a specific group of patients who are highly sensitive to antiangiogenic drugs. VEGF status is independent of stage in chemotherapy-naive patients, suggesting that there may be a role for bevacizumab as first-line treatment in addition to standard chemotherapy in selected patients. This would represent a significant new role for such agents, as most studies have looked at use in recurrent platinum-resistant disease (17, 18).

We conclude that expression of VEGF is an independent prognostic indicator in a large series of patients with all stages of ovarian cancer. High VEGF expression only occurs in a small proportion of ovarian cancers but may denote a specific group in which antiangiogenic therapy is more effective.